Event Structure and the Left Periphery: Studies on Hungarian

E V E N T S T R U C T U R E A N D T H E L E F T P E R I P H E RY S T U D I E S I N N AT U R A L L A N G U A G E A N D...

Author: Katalin E. Kiss

43 downloads 978 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

E V E N T S T R U C T U R E A N D T H E L E F T P E R I P H E RY

S T U D I E S I N N AT U R A L L A N G U A G E A N D L I N G U I S T I C T H E O RY

VOLUME 68 Managing Editors Marcel den Dikken, City University of New York Liliane Haegeman, University of Lille Joan Maling, Brandeis University

Editorial Board Guglielmo Cinque, University of Venice Carol Georgopoulos, University of Utah Jane Grimshaw, Rutgers University Michael Kenstowicz, Massachusetts Institute of Technology Hilda Koopman, University of California, Los Angeles Howard Lasnik, University of Maryland Alec Marantz, Massachusetts Institute of Technology John J. McCarthy, University of Massachusetts, Amherst Ian Roberts, University of Cambridge

The titles published in this series are listed at the end of this volume.

EVENT STRUCTURE AND THE LEFT PERIPHERY Studies on Hungarian Edited by KATALIN E´. KISS Budapest, Hungary

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-1-4020-4754-1 (PB) ISBN 978-1-4020-4753-4 (HB) ISBN 978-1-4020-4755-8 (ebook) Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands. www.springer.com

Printed on acid-free paper

All Rights Reserved 2008 Springer No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

TA B L E O F C O N T E N T S

ACKNOWLEDGEMENTS

vii

KATALIN E´. KISS / Aims and Background KATALIN E´. KISS / The Function and the Syntax of the Verbal Particle BOLDIZSA´R ESZES / Verbal Particles Telicizing Stative Psych Verbs CHRISTOPHER PINÕŃ / Definiteness Effect Verbs CHRISTOPHER PINÕŃ / Weak and Strong Accomplishments ANIKO CSIRMAZ / Particles and a Two Component Theory of Aspect KATALIN E´. KISS / From the Grammaticalization of Viewpoint Aspect to the Grammaticalization of Situation Aspect ANIKO´ CSIRMAZ / Accusative Case and Aspect KATALIN E´. KISS / Apparent or Real? On the Complementary Distribution of Identificational Focus and the Verbal Particle ANIKO´ CSIRMAZ / Aspect, Negation and Quantifiers BALA´ZS SURAŃYI / Predicates, Negative Quantifiers and Focus: Specificity and Quantificationality of N-Words

1 17 57 75 91 107

REFERENCES INDEX

287 297

v

129 159 201 225 255

ACKNOWLEDGEMENTS

This volume summarizes the results of a three-year project taking place at the Research Institute for Linguistics of the Hungarian Academy of Sciences, financed by OTKA, the Hungarian Scientific Research Fund, under grant TS 040 705. We wish to acknowledge both the financial support of OTKA, and the logistic and intellectual support of the Research Institute for Linguistics, including the intellectual contribution of two PhD students associated with our project, Annama´ria Bene and Csaba Olsvay. We have also greatly benefited by the comments of two anonymous reviewers.

vii

K ATA L I N E´ . K I S S

AIMS AND BACKGROUND

1.

AIMS

In the Minimalist framework, properties of syntax beyond those representing general properties of computational efficiency are believed to be derivable from the requirements of external systems accessing syntax. In line with these assumptions, the studies of this volume examine the hypothesis that the structure of the left periphery of the Hungarian sentence – displaying varying orders of the verb, the verbal particle, the focus constituent, the negative particle, and quantifiers, associated with different interpretations and aspectual values – is at least partially motivated by the structure of the event described; that is, event type correlates with word order, with the presence or absence of the verbal particle, with the presence or absence of an internal argument, with its [þ/specific] feature, with the aspectual interpretation of the sentence, with the interpretation of focus, and with the interpretation of pronouns participating in negative concord, among others. Although the data to be analyzed are mainly Hungarian, the questions to be answered are of crosslinguistic relevance. They include questions like what role the verbal particle plays in determining aspect; whether it is a telicizer contributing to situation aspect, or a perfectivizer contributing to viewpoint aspect; and if the socalled aspectual argument, undergoing a delimited change and thereby measuring out the event, is invariably the internal argument. It is also questioned why verbs denoting a delimited change of state, and those denoting creation or coming into being project different syntactic structures, and how their syntactic differences correlate with the specificity feature of their internal argument. It is tested if the lack of morphological viewpoint aspect marking in a language means the absence of viewpoint aspect, and if the simultaneous marking of viewpoint aspect and situation aspect is redundant. The old question of what the source of the ‘exhaustive identification’ function of structural focus is will be given a new answer. The questions concerning negation include, among others, in what way negation affects aspect; what types of quantifiers – if any – pronominal elements participating in negative concord represent; and in what way their distribution is constrained by event structure. This introductory chapter is devoted to discussing some of the background notions that the studies of the volume adopt, and to outlining the basic claims that each of the papers defends.

1 ´ Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 1–15. Katalin E. 2008 Springer.


2 2.

BACKGROUND NOTIONS

2.1.

Event structure

The field of research referred to by the term ‘event structure’ in current linguistic theory has grown out of a seminal study of Zeno Vendler (1967).1 Vendler classified verbs according to the ‘‘time schemata they presuppose’’ – or, in a more linguistic terminology, according to the temporal properties of the predicates in which they can occur. These properties, for example, the features [þ/durative], [þ/bounded], and [þ/homogeneous], have come to be understood to represent the properties of the events (or, in other terminologies, eventualities (Bach 1981), or situations (Smith 1991)) that the predicates denote. Vendler distinguished four major verb classes: states, activities, accomplishments, and achievements. States are durative and unbounded, involving no internal change – e.g. John knows Mary. Activities are also durative, with no inherent end point; however, unlike states, they express an ongoing change – e.g. John talked during the class. Accomplishments are non-homogeneous durative events, containing an ongoing activity with a necessary endpoint, a culmination – e.g. John mowed the lawn. Achievements lack duration; they involve a change with an immediate culmination – e.g. John found the ring. Vendler’s theory has been developed in various directions in lexical semantics, propositional semantics, and syntax alike. Of these developments, it is the theories of Krifka (1989, 1992), and Dowty (1991), introducing the notion of homomorphism between the theme argument denotation and the domain of the event, and the notion of incremental theme, that have had greatest influence on the studies of this volume, particularly on their interpretation of the notion of telicity (or inherent boundedness/ inherent delimitation), the property distinguishing accomplishments and achievements from states and activities. These theories have shown that in the case of telic predicates, the ‘part-of’ relation defined on the theme argument is also preserved in the domain of the event. That is, if x is part of y, then if a telic predicate maps y as theme onto event e, it must map x onto an event e¢ which is part of e. For example, in the case of the event of John eating an apple, a look at the apple will tell us if the event has not begun yet, or has taken place halfway, or has been completed altogether. This approach has helped the understanding of the role of the Hungarian verbal particle, an element predicating the resulting state or resulting location of the theme undergoing a delimited change. In the terminology of Dowty (1991), all themes displaying a theme-to-event homomorphism are incremental themes. The class of incremental themes contains two major subclasses: effected objects (e.g. in build a house, write a letter), and objects entailed to undergo a definite change of state (e.g. in paint a house, polish a shoe). Tenny (1994), on the other hand, categorizes these two types of themes separately, distinguishing incremental-theme predicates (i.e., essentially verbs of creation, consumption, and coming into being) from change-of-state predicates. Tenny also identifies a third major type of verbs in the case of which the changing of the theme is homomorphic with the passing of the event: verbs denoting motion along a path. In their case it is the path – whether explicit (e.g. in John played the

AIMS AND BACKGROUND

3

sonata) or implicit, with only its terminal point named (e.g. in John pushed the bike into the corner) – that measures out the event. The studies of this volume support the adequacy of Tenny’s classification. Incremental theme verbs, denoting creation or coming into being, taking a non-presupposed theme, and verbs denoting a delimited change of state or location, taking a presupposed theme, are syntactically encoded in radically different ways in Hungarian, and the syntactic encoding of the latter two subtypes, those denoting a delimited change of state, and those denoting a delimited change of location, also display systematic differences. What is more, the differences manifest in Hungarian shed light on similar differences in other languages, e.g. in English, as well (see chapters 4 and 5). 2.2.

Situation aspect versus viewpoint aspect

Event structure and aspect are overlapping notions; both concern the temporal organization of events/situations. In the theories of Vendler (1967), Comrie (1976), Dahl (1985), or Kiefer (1994), aspects are different ways of viewing the internal temporal constituency of situations; and situation types are related to aspect by constraining the aspectual possibilities of the verbs describing them. In the aspectual theory of Smith (1991), on the other hand, event structure, identified as the internal temporal structure of situations, represents one of the two ingredients of aspectual meaning. For Smith, the aspectual meaning of a sentence results from the interaction of situation type and viewpoint. Borik (2002) and Borik & Reinhart (2004) also put forth a two-component aspectual theory, calling the internal temporal organization of situations semantic aspect, as opposed to morpho-syntactic tense and viewpoint aspect. In Smith’s theory, situations differ with respect to duration, boundedness/telicity, and dynamism. Borik (2002) and Borik & Reinhart (2004), on the other hand, attribute a distinguished role only to the feature [þ/telic] in categorizing types of semantic aspect. The content of the notion of telicity has been subject to much controversy. A family of interpretations – see Verkuyl (1972), Smith (1991), and Tenny (1994), among others – identify the telicity of a predicate with the inherent boundedness, or delimitation, of the event denoted. In a different approach, developing Vendler’s (1967) homogeneity property – represented by Dowty (1986), Krifka (1998), Borik (2002), and Borik & Reinhart (2004), etc. – atelic and telic predicates differ in whether they are homogeneous or non-homogeneous. The homogeneity of predicates is based on their subinterval property. In the formulation of Borik & Reinhart (2002), if a telic event holds at a given interval I, there can be no other interval distinct from I at which it also holds. That is, if the telic proposition John wrote his term paper is true of a two-week interval, it cannot also be true of a ten-day subinterval of the situation. For an atelic event, if it holds at an interval I, there is always some other interval at which it also holds. That is, if the atelic proposition John was playing tennis from 2 to 5 is true, then the truth of the proposition John was playing tennis from 3 to 4 also follows.

4


Whereas situation type determines the inherent temporal properties of a situation, viewpoint determines if the sentence provides a complete or partial view of it. A situation represented in its entirety, with its initial point and end point included, exemplifies the perfective viewpoint – e.g. John ate the soup. A situation represented partially, without its initial point and end point, on the other hand, exemplifies the imperfective viewpoint – e.g. John was eating the soup. (Smith (1991) actually also allows a third viewpoint, the neutral one, associated with aspectually vague sentences.) The literature on aspect proposes several tests to distinguish telic predicates from atelic ones, and perfective predicates from imperfective ones. A more difficult question is how telicity can be distinguished from perfectivity; when the tests setting delimited predicates apart from non-delimited ones test telicity (i.e., inherent delimitation) and when they test perfectivity (i.e. delimitation imposed on the situation by the viewpoint). Delimited predicates can cooccur with time-span adverbials such as in an hour (e.g. John read the article in an hour), whereas non-delimited predicates accept durative adverbials such as for an hour (e.g. John was reading the article for an hour). Linguists identifying aspect with viewpoint aspect, e.g. Kiefer (1994), use this test to check perfectivity; Verkuyl (1972), Dowty (1986), Smith (1991), Borik (2000), and Borik & Reinhart (2004), on the other hand, regard it as a test of telicity. In the opinion of de Swart (1998, 2000) and Csirmaz (in chapters 6 and 10), what durative and completive time adverbials test is the þ or value of the subinterval property. Both situation aspect and viewpoint aspect are marked for this property, and they can be marked for it differently – e.g. (1)

For two days, John was writing the essay in a week. ‘For two weeks John was engaged in an activity which would have led him to write the essay in a week.’

A generally accepted test of atelicity is the so-called progressive entailment; that is, in languages distinguishing the past progressive and the simple past, an atelic sentence in the past progressive entails the truth of the corresponding sentence in the simple past (see Dowty 1979). In the case of telic sentences, the entailment does not hold. This test shows the predicate watch TV to be atelic, and the predicate write an essay to be (potentially) telic: (2)

a. John was watching TV fi John watched TV. b. John was writing an essay fi / John wrote an essay.

In a version of this test proposed by Borik (2002), an atelic sentence in the present progressive entails the truth of the corresponding sentence in the present

AIMS AND BACKGROUND

5

perfect: (3)

a. John is watching TV fi John has watched TV. b. John is writing an essay fi / John has written an essay.

According to a telicity test proposed by Verkuyl (1972), a telic predicate associated with two conjoined temporal phrases is understood to denote two independent situations – see (4a). In the case of atelics, on the other hand, the sentence is ambiguous; it can denote either a single extended situation, or two distinct situations – see (4b): (4)

a. John went on a trip on Monday and on Tuesday. b. John was travelling on Monday and on Tuesday.

When we test (im)perfectivity, we test whether or not the situation denoted by the predicate is viewed as a whole, with its initial and end points included. When a perfective predicate, representing a complete event, is combined with a when-clause, the situations denoted by the two predicates are understood consecutively – see (5a). In the case of an imperfective main predicate, representing an event in progress, the interpretation is simultaneous – see (5b). (5)

a. When I entered, John closed the window. b. When I entered, John was closing the window.

A temporal clause introduced by after presupposes that the situation denoted by the main clause begins after the event in the temporal clause has ended, hence it only accepts a perfective main predicate. Compare: (6)

a. After I entered, John closed the window. b. *After I entered, John was closing the window.

A perfective predicate in the present tense has future reference – see (7a); the present tense form refers to the actual present only in the case of an imperfective verb – see (7b): (7)

a. John goes abroad. b. John lives abroad. 2.3.

The left periphery

It is evident that event type is determined primarily by the lexical core of sentence structure. It is less obvious that event structure also interacts with the

6


functional layers. The studies of this volume focus on the correlation of event structure and the structure and interpretation of the left periphery of the Hungarian sentence. Hungarian sentence structure represents a version (in fact, a prototype) of the articulated universal sentence structure emerging from the work of Rizzi (1997) and Cinque (1999), consisting of a lexical layer and a series of inflectional and complementizer projections. The highest functional heads of the Hungarian sentence – listed in (8b) – are nearly identical with those that Rizzi (1997, 2002) attributes to universal sentence structure – see (8a). (8)

a. Force > Top* > Int > Top* > Focus > Top* b. Force > Top* > (Int) > Dist > (Neg) > Focus > (Neg)

Like in universal sentence structure, the topmost projection of the Hungarian sentence is a ForceP, harboring the complementizer. It dominates a – possibly iterated – TopP projection, with a topic constituent in its specifier. TopP may dominate a rarely used interrogative projection (IntP), hosting a non-focussed mie´rt ‘why’ wh-phrase. Whereas in universal sentence structure this interrogative element can be followed by further topics, in Hungarian, it can dominate one or more DistP projections, the landing sites of distributive quantifiers. (If distributive quantifiers are preposed into scope positions by adjunction (Quantifier-Raising) instead of substitution, then no DistPs need to be assumed in the Hungarian sentence, either.) The next lower projection in both languages is a Foc(us)P, containing a focus in its specifier. FP can subsume further TopP projections in Italian, but not in Hungarian. Here is a Hungarian sentence realizing the projections represented in (8b): (9)

Nem tudom, [CP hogy [TopP Jańos [IntP mie´rt [DistP mindenkit [FocP not know-I that John why everyone-ACC EBE´DRE [hı´vott meg]]]]]]] lunch-for invited PRT ‘I wonder why John invited everyone FOR LUNCH.’

Negation is allowed by Rizzi (1997) (and also by Zanuttini (1994)) to appear between more than one functional projections, so the possibility of generating a NegP both above FP and below it is not a unique property of the Hungarian sentence, either, e.g.: (10)

[TopP Jańost [NegP nem [FocP TEGNAP [NegP nem [hı´vtam fel]]]]] John not yesterday not called-I up ‘It wasn’t yesterday that I didn’t call up John.’

AIMS AND BACKGROUND

7

Despite appearances, pre-focus negation is not constituent negation; it triggers the same negative concord that predicate negation does. Compare: (11)

a. [DistP Senkit [NegP nem [FocP EŃ I nobody-ACC not ‘Nobody was invited BY ME.’ b. [TopP Eń [DistP senkit [NegP nem I nobody-ACC not ‘I did not invite anybody.’

[hı´vtam meg]]]] invited PRT [hı´vtam meg]]]] invited PRT

The lower section of the left periphery of the Hungarian sentence involves some open questions. As is clear from examples (9), (10), and (11), the focus, or the lower negative particle is immediately followed by the V. The problem is that, whereas in neutral sentences, the verb is preceded by its particle or bare nominal complement, in the presence of a focus and/or negation, the ‘particle/bare nominal, V’ order is reversed. Various explanations have been proposed for this phenomenon. For example, Brassai (1863–65), and E´. Kiss (1987, 1994) assume that the focus is in complementary distribution with the verbal particle/bare nominal complement, representing an alternative filler of the same preverbal slot, as shown in (12a). According to Brody (1990, 1995), the focus occupies the specifier of a higher projection (a FocusP) than that dominating the particle, and its immediately preverbal position is due to V-to-F movement. ((12b) is an updated version of a Brody-style structure.) Olsvay (2000b), on the other hand, assumes that in the presence of a negative particle and/or a focus the V moves into the head position of a Non-Neutral Phrase (NNP) (rather than into F or Neg); that is, the ‘V, particle’ order attested in the case of focus and/or negation arises from V movement to NN – see (12c). Chapter 9 of this book will adopt Olsvay’s proposal.

(12) a.

TopP Spec

Top’ Top

PredP

Spec Particle/Focus

Pred’

Pred

VP V


8

b.

TopP

Spec

TopP

c. Top’

Top

Spec

FocP

Spec Focus

Top

Foc’

Foc Vi

Top’ FocP

Spec Focus PredP

Spec Particle

Foc

Pred’

Pred ti

Foc’

VP

V ti

NNP

NN Vi

PredP

Spec Particle

Pred’

Pred ti

VP

V ti A problem that has so far raised doubts concerning focus theories involving V-movement is that they predict a fixed word order for the postverbal section of the sentence crossed by the V (e.g. the particle is expected to immediately follow the V in (12b,c)); in fact, however, the postverbal word order in the Hungarian sentence is free. Chapter 2 proposes a novel solution of this problem, claiming that V-movement to Pred in neutral sentences, and V movement across Pred to Non-Neutral in sentences involving negation and/or focussing creates a phase which has the V in its head, and the postverbal section in its domain. In the domain of this phase, to be handed over to Spell-Out, the projection which has its head removed is flattened at the syntax-phonology interface, and its constituents are reordered according to their phonological weight in PF. Observe what happens to the hierarchical left periphery of the PredP projection in (13a) when it becomes the domain of a higher phase (13b). If the different postverbal elements are replaced by shorter or longer expressions of the same category, the optimal postverbal order also changes, in accordance with Behaghel’s (1932) Law of the Growing Constituents.

AIMS AND BACKGROUND

(13)

9

a. [PredP Nagyon [PredP meg se´rtette [vP Boldizsa´r Lilit]]] PRT offended Boldizsa´r Lili-ACC very.much ‘Boldizsa´r offended Lili very much.’ nagyon Boldizsa´r ]]] b. [NegP Nem [NNP se´rtette [PredP meg Lilit ´ r-NOM PRT Lili-ACC very Boldizsa not offended {meg nagyon Jańos Piroska´t} PRT very.much John-NOM Piroska-ACC {meg Jańos Piroska´t olyan rettenetesen} PRT John-NOM Piroska-ACC so terribly

The category of the projection harboring the verbal particle is also a controversial issue of Hungarian syntax. Some analyses place the preverbal particle (and the bare nominal complement, which is in complementary distribution with the particle) in the lexical layer of the sentence. Horvath (1986), Brody (1990), and Szendro´í (2003) assume that the particle is the left sister of the V under a V’ node. E´. Kiss (1994) places the particle into the specifier of VP. Nowadays, the preverbal particle position is more often analyzed as an A-bar position, the specifier of a functional projection, because it can also be filled by long movement from a lower clause, e.g.: (14) Fel kellene, hogy hı´vjuk Pe´tert. up needs that call-we Peter-ACC ‘It is necessary that we call up Peter.’ In view of examples like (14), the verbal particle is placed into the specifier of an AspP projection by Pinõn (1995), den Dikken (2004), and E´. Kiss (2002a), and into the specifier of PredP in Csirmaz (2004) and in the studies of this volume. In chapter 6, Csirmaz argues for an AspP projection different from PredP, whose perfective/imperfective head is responsible for the viewpoint aspect of the sentence. AspP dominates PredP, and it triggers movement from the specifier or head of PredP. The inflectional layer of the Hungarian sentence involves a ModalityP, a TenseP, a MoodP, an AgrOP, and an AgrSP projection, with the modality, tense, mood, object agreement and subject agreement suffixes realized on the V in the opposite order, in accordance with the Mirror Principle (see Bartos 2000). The question whether the inflectional layer dominates PredP and AspP or is dominated by them cannot be decided empirically, but the universal hierarchy of Cinque (1999) suggests the former order. Thus the following heads are hypothesized in the lower section of the left periphery of the Hungarian sentence: (15)

AgrS > AgrO > Mood > Tense > Modality > Asp > Pred


10 For example: (16)

[AgrSP[AgrOP[MoodP[TP[ModP[AspPmegi[Asp’PERF[PredPti i -hat -0 -n-a´-tok]]]]]]]] PRT drink-POSS-PRES-COND-3-2PL ‘you would be allowed to drink it up’

The different types of adverbials are interspersed among the constituents of the left periphery basically as predicted by Cinque (1999) and, in a different framework, by Ernst (2002). The effect of event structure on the left periphery is manifested, among others, in the interaction of the verbal particle and viewpoint aspect – a question touched upon in chapter 2, and discussed in detail in chapter 6. Whereas the presence or absence, and the type of the verbal particle is a key ingredient of situation aspect, the position of the verbal particle will be shown to correlate with viewpoint aspect (which will be derived from the assumption that the Asp head imposes certain requirements on the type of element admitted to its specifier). The formal aspect of the interaction of event structure with focussing and negation, the (apparent?) complementary distribution of the focus constituent, the negative particle, and the verbal particle in the immediately preverbal position, is taken up in chapter 9. On the level of interpretation, the problem is whether focussing and negation neutralize situation aspect; whether negation indeed turns all types of events into states. These questions will be examined in chapter 10. Both situation aspect and viewpoint aspect are dependent on the specificity of the internal argument of the predicate. This correlation is a recurring theme of the chapters of this book, discussed in chapters 2, 4, 5, 8, 9, and 11. Telic predicates expressing a change of state or location take a specific internal argument and a verbal particle; whereas telic predicates expressing coming into being, appearance, or creation take a non-specific internal argument and no particle. Telic predicates of the latter type fail some of the standard tests of telicity – they are to be identified as weak accomplishments in chapter 5. The fact that event type correlates with the specificity feature of the internal argument has an interesting reflex in negative sentences: the universal or bound indefinite interpretation of negative pronominal elements participating in negative concord depends on the type of the event described in the sentence – as shown in chapter 11. Event structure also interacts with topic selection. This question, and, in general, the effect of event structure upon the discourse organization of the sentence was analyzed by E´. Kiss (2005a).

3.

THE PROBLEMS EXAMINED

The chapters of this volume study phenomena related to event structure in syntax and the syntax-semantics interface on the basis of Hungarian data; nevertheless, the generalizations arrived at promise to be of cross-linguistic relevance.

AIMS AND BACKGROUND

11

The primary indicator of event type in Hungarian is the presence or absence, the type, and the position of the verbal particle. Chapter 2, entitled The function and the syntax of the verbal particle, aims to provide a theory of the verbal particle which can predict both the functional interpretation and the structural position of the major types of Hungarian verbal particles. Particles are shown to fall into three major types, corresponding to three major types of events. Resultative particles mark telic sentences denoting a delimited change of state, terminative particles mark telic sentences denoting a delimited change of location, whereas locative particles mark sentences denoting existence or spatial configuration in a given location. Contrary to widespread assumptions, the verbal particle is not a perfectivizer; it marks situation aspect, and it has an indirect role in determining viewpoint aspect. The clarification of the function of the verbal particle leads to a syntactic analysis which treats the particle as a secondary predicate predicated of the theme argument, and identifies its canonical preverbal position as the specifier of a PredP projection. The proposed syntactic analysis correctly predicts that particles accompany only transitive and unaccusative verbs; unergatives can accept a particle only if they are complemented by a non-thematic object. It also falls out that telic predicates expressing the creation, or the coming into being, of their theme can involve no particle – because their [– specific] theme cannot function as the subject of predication. In the case of certain types of psychological verbs, some predictions of the theory of the verbal particle presented in chapter 2 are not borne out. Chapter 3, entitled Verbal particles telicizing stative psych verbs, examines in what way particle plus psychological verb combinations, taking an Experiencer subject and Theme or Stimulus object, expressing a change in the subject’s mental (emotional or perceptual) state, are special. These complex verbs, like meg-kedvel ‘come to like’ and meg-sajna´l ‘come to feel sorry for’ on the one hand, and meg-la´t ‘catch sight of’ and meg-hall ‘come to hear’, on the other hand, are derived from simple stative verbs (kedvel ‘like’, la´t ‘see’ etc.) by means of the verbal particle meg. The particle is identified as an eventuality description modifier mapping a state to an event, which brings about a change of situation aspect type by providing a dynamic meaning component. The semantic structure of these particle plus verb complexes is analyzed by introducing the notion of a mental change of state, using a scale-based degree semantics, originally developed for the semantic analysis of adjectives and certain English verb classes involving an incremental theme. These verbs also contradict generalizations concerning the mapping of lexical structure onto syntactic structure, hence they pose a problem for several current theories of the lexicon-syntax interface. For example, they necessitate a relaxation of the linking theory and the Measuring-Out Constraint of Tenny (1994). Whereas telic change-of-location and change-of-state predicates, including telic psych-verbs, take a verbal particle in Hungarian, telic incremental-theme predicates, i.e., verbs of creation, coming into being, and appearance, take none. Chapters 4 and 5 are devoted to the analysis of the latter type of predicates. Chapter 4 deals with definiteness effect verbs, which is a much larger set in Hungarian than has been

12


identified in English. Chapter 4 also explains why; because definiteness effect verbs are systematically polysemous in that their so-called definiteness effect meaning is simply one of two or more meanings that they have. In Hungarian, the reading with the definiteness effect is easy to set apart because – unlike the other meaning(s), it is associated with the bare verb plus theme argument with no particle involved. In English, where the two readings can be realized by one and the same string, the systematic polysemy is less transparent. Concerning the definiteness effect meaning, it is claimed that there is both uniformity and diversity in the semantics of these verbs. The uniformity is seen in the idea that every definiteness effect verb introduces a novel discourse referent corresponding to its direct internal argument. The diversity is revealed in the idea that although every definiteness effect verb specifies an ‘end condition’ that comes to hold of its internal argument, the exact value of this end condition varies across subclasses of definiteness effect verbs, hence there is no single end condition that does duty for all definiteness effect verbs. Chapter 5 on Weak and strong accomplishments argues that the fact that telic change-of-state and change-of-location predicates, on the one hand, and predicates expressing creation, coming into being, and appearance, on the other hand, are encoded in different syntactic structures is not a syntactic idiosyncrasy; the different syntactic structures reflect semantic differences. Verbs of creation, coming into being, and appearance are not diagnosed as accomplishments by every traditional test – therefore, the category ‘weak accomplishment’ is introduced. It is argued that the essential difference between weak and strong accomplishments lies in the fact that the latter are presuppositional in a way that the former are not, and this difference is captured by means of a notion of finishing that figures in the analysis of strong accomplishments but not in that of weak accomplishments. Chapter 6, Particles and a two-component theory of aspect, adopts a version of the two-component aspect theory elaborated by Smith (1991) and Olsen (1997). It argues for the independence of situation aspect and viewpoint aspect, for example, by analyzing telic imperfective events. The chapter examines how Hungarian particles and other types of verbal modifiers affect the two kinds of aspect. Resultative and terminative particles, denoting an endpoint, can alter situation aspect properties by telicizing the event. Particles interact with the viewpoint specification of the event through their position relative to the verb. The telicizing effect of particles follows from a compositional approach to situation aspect. The interaction of particle position and viewpoint aspect is derived from the local relation between the two elements and the properties of viewpoint aspect. The viewpoint aspect of an event is encoded in the functional head Asp. It is argued that for all finite event descriptions in Hungarian, Asp is present and is marked as either perfective or imperfective. In order to derive the interaction of verbal modifiers and aspect, it is assumed that the Asp head has an EPP feature, which must be checked by the closest possible constituent, whether it be a particle to be moved from Spec,PredP to Spec,AspP, or the V to be moved from Pred to Asp. A terminative or resultative verbal modifier in Spec,AspP delimits the reference time just as it delimits the event time, and yields an output in which the reference time includes

AIMS AND BACKGROUND

13

the endpoint of the event. Since this is excluded by the definition of imperfective aspect, the EPP feature of an imperfective Asp cannot be checked by a terminative/ resultative particle; it is checked by the V moved to Asp, or by a verbal modifier which does not denote an endpoint. Chapter 7, entitled From the grammaticalization of viewpoint aspect to the grammaticalization of situation aspect, discusses a change in the history of Hungarian in the course of which the language has lost its complex tense system also encoding viewpoint aspect, and parallel with that process, it has developed a system of marking telicity – primarily by means of verbal particles. The chapter raises the question if the near simultaneity of these two processes was a mere coincidence. Although it does not challenge the claim that situation aspect and viewpoint aspect represent two separate subsystems of aspect, it argues that the overt morphological encoding of both systems would be redundant to a large extent, because viewpoint aspect correlates with situation aspect in a predictable way in the majority of cases. In the unmarked case, predicates marked as telic have the perfective viewpoint, and transitive and unaccusative atelic predicates have the imperfective viewpoint. In the case of unergative atelics, the viewpoint is identified on the basis of tense, time adverbials, and context. For the encoding of the occasional mismatches, the language adopts particular lexical, morphological, and syntactic means. Chapter 8, Accusative case and aspect, focusses on the function and the licensing conditions of non-thematic objects in Hungarian. The distribution of non-theta marked reflexives and body parts is accounted for by means of the requirement that the argument structure of the predicate contain at least one argument per subevent (the Argument per subevent condition of Rappaport Hovav and Levin (2001)). Other types of non-theta marked constituents are analyzed as situation delimiters, directly affecting the aspectual properties of the events they modify. Accusative marked situation delimiters include temporal and spatial measure adverbials, multiplicatives, and adjectives. It is shown that even though they all serve to delimit events, they do so in different ways. While temporal and measure adverbials merely denote the endpoint of the event, multiplicatives and adjectives introduce and modify a result. The syntax of the Hungarian verbal particle is intertwined with the syntax of identificational focus – as was briefly discussed above, in connection with structures (12a–c). Chapter 9, entitled Apparent or real? On the complementary distribution of identificational focus and the verbal particle, aims to clarify the structural positions of these two types of immediately preverbal constituents, disentangling and explaining the similarities and the differences in their syntactic behavior. It is claimed that what underlies the shared properties of foci and verbal particles/bare nominals is a functional and a phase-structural similarity. Not only the verbal particle and the bare nominal are predicates, predicated of the (explicit or implicit) theme argument of the verb, but the identificational focus is also a predicate, expressing specification, predicated of the open proposition constituted by the VP. The exhaustivity associated with structural focus is claimed to be an implicature of specificational predication. The similarities in the syntactic behavior of the verbal particle, the bare nominal, and the focus are derived from the fact that they play similar roles in phase-structure.

14


PredP, with the verbal particle/bare nominal in Spec,PredP, and the functionally more extended FocP, with the identificational focus in Spec,FocP, represent different realizations of the v*P phase of the Hungarian sentence. Both of them are headed by the V (owing to V movement), and both of them have a postverbal domain that is reordered in the phonological component according to the phonological weight of the constituents, satisfying Behagel’s Law of the Growing Constituents (1932). The last two chapters of this book, chapters 10 and 11, examine how event structure interacts with negation. Chapter 10, entitled Aspect, negation, and quantifiers, explores the effect of clausal negation on aspect. Since negation allows modification by durative adverbials, it has been suggested that it is an aspectual operator, yielding derived states from all event predicates. This chapter argues that negation does not alter the aspectual properties of the event predicate. Rather, it yields a higher predicate of times which has the subinterval property, the property which adverbial modification is sensitive to. With negation present, the predicate of times modified by the durative adverbial applies to the reference time and not to the event time. It is shown that in addition to negation, the modification of reference time by durative adverbials is also licensed by downward entailing quantifiers and arguments modified by only. These data provide further arguments against the aspectual operatorhood of negation. It is noted that structurally case-marked adverbials are restricted to modifying the event time. It is argued that this restriction arises from locality constraints on case checking and adverbial modification. The chapter on Predicates, negative quantifiers and focus: specificity and quantificationality of n-words examines the consequences of the event-structure-motivated restrictions on the [þ/ specificity] feature of the theme argument in negative sentences. As discussed in chapters 4 and 5, verbs expressing creation, coming into being, and existence require their theme argument to be non-specific; their prefixed counterparts, on the other hand, only allow a theme represented by a specific, presuppositional noun phrase. In negative sentences in which the theme is a pronominal element participating in negative concord, the (non-)specificity requirement associated with the theme argument is reflected in the (non-)specific reading of the negative pronominal (called ‘n-word’) – e.g. (16)

a. Nem e´rkezett senki. not arrived nobody ‘There did not arrive anybody.’ b. Nem e´rkezett meg senki. not arrived PRT anybody ‘None of the persons has arrived.’

It is contended in this chapter that this specificity–non-specificity contrast is to be analyzed in terms of an ambiguity in the interpretation of n-words themselves, between a quantificational and a non-quantificational lexical meaning. (16b) involves an n-word that is universally quantified and scopes above negation, while (16a)

AIMS AND BACKGROUND

15

involves a Heimian indefinite n-word, which is existentially closed in the scope of negation (cf. Heim 1982). Although the two resulting interpretations are logical equivalents, they are compositionally different and they behave differently in natural language. The chapter also investigates and derives the distribution of the two readings in various domains of the clause. Focussing affects the interpretation of n-words in an interesting way: their (morphosyntactically optionally realized) EVEN component (see Lee and Horn (1994), Lahiri (1998), and Horn (2000)) operates on the bare predicative restriction of n-words, treating it as an extreme element on a predicate scale (cf. Krifka 1995). NOTE 1

Vendler’s work was not without precedents; some of its basic ideas can be traced back to Ryle (1949).


T H E FU N C T I O N A N D T H E S Y N TA X O F T H E V E R B A L PA RT I C L E

1.

INTRODUCTION

Perhaps the hardest descriptive problem of Hungarian syntax is how to analyze the verbal particle. Is it a head or a phrase? What structural position does it occupy? Does it form a constituent with the V? Is its default preverbal position a base-generated or a derived position? Is its complementary distribution with the focus in the preverbal slot real or apparent? Does the reverse, V-particle order of negated sentences result from V-movement across the particle, or from the blocking of particle movement? How is the interaction of particle position and aspectual interpretation to be represented? Many of these questions also arise in the better known Indo-European languages – e.g. in the Germanic languages (even if they are exempt from the problem of the interaction of the particle with focussing), or in the Slavic languages (even if the Slavic equivalents of the Hungarian verbal particles are non-separable verbal prefixes, raising less word order problems). The questions arising have been given many different answers in the literature, and the proposed analyses seem to represent similar levels of descriptive adequacy, and seem to rely on principles of Universal Grammar to similar extents. The work summarized in this chapter has been motivated by the the assumption that the understanding of the role that the verbal particle plays in the conceptual-intentional interpretation of the sentence and in its prosodic realization may facilitate the selection of the most adequate syntactic representation. It is widely held that the verbal particle is a perfectivizing operator. This study will also confirm that the presence or absence, the type, and the position of the Hungarian verbal particle is related to the aspectual interpretation of the sentence. The view that the verbal particle denotes perfective aspect, however, will be refuted by demonstrating that not all perfective sentences involve a verbal particle, and not all sentences containing a particle are perfective. It will be claimed that particles contribute to aspectual interpretation by indicating situation aspect. Verbal particles will be shown to fall into three main classes. Resultative particles mark telic sentences describing an inherently delimited change of state, by denoting the resultant state of the individual undergoing the change. Terminative particles mark telic sentences describing an inherently delimited change of location, by denoting the end location of the moving individual. Locative particles appear in atelic sentences expressing existence or spatial configuration; they denote the location of the argument whose existence or spatial configuration is asserted. The interaction of particles with viewpoint aspect is indirect, the details of which will be examined in chapter 6. 17 ´ Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 17–55. Katalin E. 2008 Springer.


18

The clarification of the functions of the verbal particle leads to a syntactic analysis which treats the particle as a secondary predicate predicated of the theme argument, and identifies its canonical preverbal position as the specifier of a PredP projection. The proposed syntactic analysis correctly predicts that particles accompany only transitive and unaccusative verbs; unergatives can accept a particle only if they are complemented by a pseudo-theme. It also falls out that telic predicates expressing the creation, or the coming into being, of their theme can involve no particle – because their [–specific] theme cannot function as the subject of predication. The effect of focussing and negation upon the proposed structure will be taken up in chapter 9 of this book. This chapter is organized as follows: Section 2 examines the distribution of verbal particles across situation types, as well as their functions and their syntactic licensing conditions. The three main types of verbal particles identified will be discussed in separate subsections. The resultative particles of telic change-of-state sentences are analyzed in section 2.1, and the terminative particles of telic change-oflocation sentences are examined in section 2.2. Section 2.3 discusses why telic sentences expressing creation or coming into being – unlike those expressing a change of state or a change of location – involve no verbal particle. The locative particles of atelic sentences expressing existence or spatial configuration in a particular location are investigated in section 2.4. Section 3 is devoted to the question whether resultative and terminative particles mark viewpoint aspect or situation aspect. Section 4 discusses the syntax of the verbal particle. 2.

T Y P E S O F V E R B A L PA RT I C L E S A N D S I T U AT I O N T Y P E S

As will be argued below, the presence or absence, and the type of the Hungarian verbal particle is determined by the type of the situation described in the given sentence.1 The classification of situations that underlies the distribution of verbal particles is a version of the typology of situations emerging from the work of Vendler (1967), Dowty (1991), Smith (1991), Tenny (1994), and others. The primary criterion of the classification of situations – and of the sentences describing them – is (a)telicity. Telic situations contain a change with a conclusion, an outcome, and, accordingly, telic sentences describe a process leading to a result. Whereas the traditional classification of telic sentences into accomplishments and achievements is based on the [þ/durative] feature of the event (see Smith 1991:30), the classification to be adopted here, following Tenny (1994) in crucial respects, is based on the quality of the change denoted. Three types of telic sentences are distinguished: those describing a change of state, those describing a change of location, and those describing creation or coming into being. As will be shown (relying on E´. Kiss (2005b) in many respects), sentences describing a change of state or a change of location contain a resultative or a terminative verbal particle, respectively; those describing creation or coming into being, on the other hand, necessarily lack a particle. A third type of verbal particles marks a subtype of atelic sentences: those expressing existence or spatial configuration in a location.

S Y N TA X O F V E R B A L PA RT I C L E

2.1.

19

Resultative particles in telic change-of-state sentences

Telic change-of-state sentences describe a change of their theme argument which leads to its new state. The resulting state of the theme is sometimes expressed by an adjective phrase or a noun phrase – see (1a–c); in most cases, however, the resultative element is a verbal particle with little or no descriptive content – see (2a–c). a. E´va szo´´ke´re festette a haja´t. Eve blond-to dyed the hair-her ‘Eve dyed her hair blond.’ b. Eszter tı´z szeletre va´gta a torta´t. Esther ten slices-into cut the cake ‘Esther cut the cake into ten slices.’ c. A hu´s puha´ra fo´´tt. the meat tender -to cooked ‘The meet cooked tender.’ (2) a. E´va be-festette a haja´t. Eve in dyed the hair-her ‘Eve dyed her hair.’ b. Eszter fel-va´gta a torta´t. Esther up cut the cake ‘Esther cut up the cake.’ c. A hu´s meg-fo´´tt. the meat PRT cooked ‘The meat cooked.’

(1)

(The resultative or terminative verbal particle plus V complex is spelled as one word. For expository purposes, I separate them by a hyphen.) The particle be ‘in’ in (2a) plays the same role as szo´´ke´re ‘blond-to’ in (1a); it expresses that the object of dying, Eve’s hair, has assumed a new color as a result of the dying process. The function of the particle fel ‘up’ in (2b) is also similar to that of the noun phrase tı´z szeletre ‘ten slices-into’ in (1b); it shows that the cake, originally undivided, has been cut into pieces. The particle meg in (2c), too, has essentially the same function as the casemarked adjective in (1c): it means that the meat has attained the required (tender) state as a consequence of cooking.2 What the particles differ in from their phrasal counterparts is that they lack descriptive content; they merely mean that the individual affected by the given change has been totally affected, and it has attained the new state following from the given change. Both the sentences in (1a–c) and those in (2a–c) express that the events described have been completed; however, the completion is not marked by a particular tense of the verb (Hungarian has only two tenses: a past and a present); it is indicated by the resultative element predicating the resultant state of the theme. In the case of accomplishment predicates, the verb and the resultative element clearly correspond to the process component and the resultant state component of the complex event, respectively.3 Thus in (2a,b), festette ‘dyed’, va´gta ‘cut’, and fo´´tt


20

‘cooked’ denote homogeneous processes affecting the theme; be ‘in’, fel ‘up’, and meg, on the other hand, denote the totally changed, new state attained by the theme after it has been affected by the given process in its entirety. Although the particles have no (or little) descriptive content in themselves, their meanings can be reconstructed from the meanings of their verbs. Each particle refers to the particular state resulting from the process specified by the given verb.4 The omission of the particle yields a regular process sentence: (3)

a. E´va festette a haja´t. Eve dyed the hair-her ‘Eve was dying her hair.’ b. Eszter va´gta a torta´t. Esther cut the cake ‘Esther was cutting the cake.’ c. A hu´s fo´´tt. the meat cooked ‘The meat was cooking.’

In the case of achievement predicates, the two components of the event are practically simultaneous, inseparable, therefore, the correspondence between the verb and the process phase, and between the resultative particle and the resultant state is less transparent. Observe: (4)

a. Jańos meg-tala´lta a gy u˝ r u˝ t. John PRT found the ring ‘John (has) found the ring.’ b. E´va fel-e´bredt. Eve up woke ‘Eve woke up/has woken up.’ c. Zoltań el -e´rte a csućsot. Zoltan PRT reached the top ‘Zoltan has reached the top.’

The verbs tala´l ‘find’, e´bred ‘wake’, and e´r ‘reach’ denote momentary changes affecting the theme, and the particles meg, fel ‘up’, and el ‘off’ refer to the resultant states of the theme. (4a,b) imply the resultant states asserted in (5a,b). ((4c) seems more idiosyncratic; the result state meaning component of el-e´r ‘reach’ cannot be lexicalized separately.) (5)

a. A the ‘The b. E´va Eve ‘Eve

gy u˝ r u˝ ring ring is fent up is up.’

meg-van. PRT is here.’ van. is

21


Nevertheless, the removal of the particle from an achievement predicate usually does not yield an acceptable semelfactive predicate expressing a momentary process: (6)

a. *Jańos ’tala´lta a ’gy u˝ r u˝ t.5 John found the ring b. *E´va ’e´bredt. Eve woke c. *Zoltań ’e´rte a ’csućsot. Zoltan reached the top

In (1a–c), the resultative phrases clearly represent secondary predicates predicated of the theme argument. Because of the similar role of the resultative particles in (2a–c), there is good reason to assume that they also function as secondary predicates predicated of the theme. The implication relations between (4a) and (5a), and between (4b) and (5b) suggest that this analysis can also be extended to achievement predicates – even if we also have to allow for non-transparent cases, in which the meaning of the particle plus verb complex cannot be divided into separate ‘momentary process’ and ‘result state’ components. Incidentally, the secondary predicate analysis of verbal particles also represents a main stream in the study of IndoEuropean verbal particles – see e.g. Kayne (1985), Hoekstra (1988), Larson (1988b), den Dikken (1995), and Winkler (1996). The assumption that the verbal particle is predicated of the theme argument is supported by the fact that the theme in a telic change-of-state sentence must be [þspecific] – whether the resultant state is expressed by a resultative phrase or by a particle. First let us consider some evidence that the theme argument of a resultative construction can only be [þ specific]. In Hungarian, there are two types of noun phrases that can only have a non-specific reading: bare singulars and bare plurals. Neither of them can represent the theme of a resultative construction:6 (7)

a. *E´va tı´z szeletre va´gott torta´t. / *E´va fel-va´gott Eve ten slices-into cut cake Eve up cut b. *Puha´ra fo´´tt tyu´k. / *Meg-fo´´tt tyu´k. PRT cooked hen tender cooked hen c. *Jańos meg-tala´lt gy u˝ r u˝ ket. John PRT found rings d. *Fel-e´bredtek lańyok. up woke girls

torta´t. cake

The indefinite article, numeral determiners, or ne´hańy ‘some, a few’ can, in principle, have either a specific or a non-specific reading. According to En¸c (1991), a test of the specificity of an indefinite noun phrase is if it is interpreted as referring to a subset of a previously introduced set. Indefinite theme arguments in a resultative construction elicit this reading. If the resultative element is dropped, the


22

‘subset’ reading is also lost. Compare: Az a´lla´sra ke´t dia´k jelentkezett. ‘To the job, two students applied.’ Tegnap be-hı´vtunk egy lańyt interju´ra. yesterday in called-we a girl-ACC interview-for ‘Yesterday, we called in a girl for an interview.’ b. Az a´lla´sra ke´t dia´k jelentkezett. ‘To the job, two students applied.’ #Tegnap hı´vtunk egy lańyt interju´ra. yesterday called-we a girl-ACC interview-for ‘Yesterday, we called a girl for an interview.’ (9) a. Pe´ter to¨bb oszta´lyta´rsa´val is besze´lni akart. ‘Peter wanted to talk with several of his classmates.’ Meg-va´rt ne´hańy lańyt. PRT waited some girls-ACC ‘He waited for some girls.’ b. Pe´ter to¨bb oszta´lyta´rsa´val is besze´lni akart. ‘Peter wanted to talk with several of his classmates.’ #Va´rt ne´hańy lańyt. waited some girls ‘He awaited some girls.’ (8)

a.

In the (a) cases, the object of the second sentence is understood to refer to a subset of the set mentioned in the first sentence. In the (b) cases, this reading is unlikely; in fact, the two sentences do not even seem to constitute a coherent text.7 The specificity requirement on themes functioning as subjects of resultative predicates is reminiscent of the specificity requirement on topics, and it may have the same source: both of them function as logical subjects of predication. Various descriptions of the logical subject–logical predicate relation share some version of the assumption that a necessary condition of logical subjecthood is that the constituent functioning as logical subject be associated with an existential presuppositon (cf. e.g. Strawson (1971), Kuroda (1972), Erteschik-Shir (1997)). Verbs denoting a change of state can be complemented with a non-specific theme argument, as well; in this case, however, they do not take a verbal particle. For example: (10)

a. Pe´ter fa´t va´gott. Peter wood-ACC cut ‘Peter cut (some) wood.’ b. Tyu´k fo´´tt a faze´kban. hen cooked the pot-in ‘A hen cooked in the pot.’

Notice that the sentences in (10) do not express accomplishments, i.e., inherently delimited, telic processes; they merely denote processes, which can be interpreted


23

either imperfectively (with the initial point and end point of the process outside the event), as in (11a) and (12a), or perfectively (with the initial point and end point of the process included in the event), as in (11b) and (12b). a. Pe´ter o´ra´k hosszat fa´t va´gott. Peter hours long wood-ACC cut ‘Peter was cutting wood for hours.’ b. Pe´ter fa´t va´gott a ta´bort u˝ zho¨z. the camp-fire-to Peter wood-ACC cut ‘Peter cut some wood for the camp-fire.’ (12) a. Tyu´k fo´´ a faze´kban. hen cooks the pot-in ‘A hen is cooking in the pot.’ b. Ebe´dre tyu´k fo´´tt. lunch-to hen cooked ‘A hen (has) cooked for lunch.’

(11)

In the case of verbs denoting a momentary process, the ‘non-specific NP plus V’ combination is more likely to be interpreted perfectively: (13)

a. A repu¨lo´´ge´p fo¨ldet e´rt. the airplane ground-ACC touched ‘The airplane touched ground.’ b. A medve embert o¨lt. the bear man-ACC killed ‘The bear killed a man.’ c. Istvań kezet ra´zott Pe´terrel. Stephen hand-ACC shook Peter-with ‘Stephen shook hands with Peter.’

According to Krifka (1992) and Tenny (1994), the fact that the internal argument of telic sentences cannot be represented by a bare nominal has a different reason; the internal argument serves to measure out, i.e., to delimit, the change described by the verb, and only a delimited, quantized noun phrase can function as a delimiter. However, as will be illustrated in section 2.4, locative particles, restricted to atelic sentences, also impose a specificity requirement on their logical subject in Hungarian. Returning to change-of-state sentences with a resultative particle, if the particle is a secondary predicate predicated of the theme, then unergative verbs, i.e., intransitive verbs with an agent argument, are predicted not to have one. This prediction is essentially borne out – cf. (14)

a. Jańos John

(*el) PRT

eńekelt. sang


24 b. E´va Eve cf. (15)

(*el) PRT

olvasott. read

a. Jańos el -eńekelte az a´ria´t. the aria John PRT sang ‘John sang/has sung the aria.’ b. E´va el -olvasta a cikket.8 Eve PRT read the article ‘Eve (has) read the article.’

Verbs like nyer ‘win’ and gyo´´z ‘win’ are also optionally transitive. As expected, they take a verbal particle only in their transitive use: La´szlo´ nyert/gyo´´zo¨tt. John won /won (17) a. La´szlo´ meg-nyerte a me´rko´´ze´st. the match La´szlo´ PRT won ‘La´szlo´ (has) won the match.’ b. Le´ko´ le -gyo´´zte Kramnyikot. Le´ko´ PRT defeated Kramnik ‘Le´ko´ (has) defeated Kramnik.’

(16)

An unergative verb can only be ‘telicized’, i.e., enabled to express an inherently delimited change, if it is supplied with both a pseudo-object and a resultative element – see in detail in chapter 8. The unmarked pseudo-object is the reflexive pronoun. The construction in (18a) expresses that the agent has attained a new state as a result of her own action. In certain idiomatic cases the pseudo-object is a bodypart of the agent – see (18b–c). (18)

a. E´va ki-dolgozta maga´t. Eve out worked herself ‘Eve worked herself tired.’ b. Zsuzsa ki-sı´rta a szeme´t. Susan out cried the eye-her ‘Susan cried her eyes out.’ c. Jańos le -ja´rta a la´ba´t John off walked the foot-his ‘John walked his feet sore.’

If we omit either the pseudo-object or the resultative particle, the sentences become sharply ungrammatical: (19)

a. *E´va ki Eve out

-dolgozott. worked


25

b. *E´va dolgozta maga´t. Eve worked herself Unergatives may also allow a temporal pseudo-object, for example: (20)

a. Pe´ter a´t -aludta a de´lutańt. Peter through slept the afternoon-ACC ‘Peter slept through the afternoon.’ b. Rozi el -besze´lgette az ido´´t. Rosy away talked the time-ACC ‘Rose talked away the time.’ c. Imre el -borozgatott ke´t o´ra´t. Imre away wined two hours-ACC ‘Imre spent two hours drinking wine.’

In a certain type of sentence the argument undergoing a delimited change of state is the experiencer. These sentences – to be discussed in detail in chapter 3 – also involve a verbal particle: (21)

´ gnes meg -szerette Jo´zsefet. a. A loved Joseph-ACC Agnes PRT ‘Agnes came to love Joseph.’ b. Jo´zsef meg -uta´lta a fo´ńo¨ke´t. hated the boss-his Joseph PRT ‘Joseph came to hate his boss.’

In such sentences, it is not the particle but the verb that denotes the resultant state of the experiencer. The particle merely serves to indicate that the state of the experiencer is a resultant state. In this sentence type, it is the experiencer-subject that must be specific – see (22a). A non-specific theme object sounds somewhat more acceptable in this context – see (22b). (22)

a. ?*A tana´rt meg-szerette´k dia´kok. students the teacher-ACC PRT loved ‘(Some) students came to like the teacher.’ b. ??A tana´r meg-szeretett dia´kokat. students the teacher PRT liked ‘The teacher came to like (some) students.’ 2.2.

Terminative particles in change-of-location sentences

A subset of telic sentences express a delimited change of location. In such cases, the culmination – and completion – of the event is the attainment of the end

26


location; the resultant state of the moving individual is its static end position. The end position attained is sometimes expressed by a preverbal noun phrase or postpositional phrase – see (23a–c), but in most cases, it is denoted by a verbal particle of a terminative role, i.e., an adverb with little descriptive content, as in (24a–c). The noun phrase or PP specifying the end location can optionally also be spelled out in the presence of a terminative verbal particle – see (25a–c). In such cases, the particle functions as a proadverbial double of the terminative noun phrase. a. Jańos az ablakhoz vitte a ta´vcso¨vet. John the window-to took the telescope ‘John took/has taken the telescope to the window.’ b. Zsuzsa a szoba´ba szaladt. Susan the room-into ran ‘Susan ran/has run into the room.’ c. Istvań az asztalra tette a ko¨nyvet. Stephen the table-on put the book ‘Stephen (has) put the book on the table.’ (24) a. Jańos oda -vitte a ta´vcso¨vet. John there took the telescope ‘John took/has taken the telescope there.’ b. Zsuzsa be -szaladt. Susan in ran ‘Susan ran/has run in.’ c. Istvań le -tette a ko¨nyvet. Stephen down put the book-ACC ‘Stephen (has) put down the book.’ (25) a. Jańos oda -vitte a ta´vcso¨vet az ablakhoz. John there took the telescope the window-to ‘John took/has taken the telescope to the window.’ b. Zsuzsa be-szaladt a szoba´ba. Susan in ran the room-in ‘Susan ran/has run into the room.’ c. Istvań le -tette a ko¨nyvet az asztalra. Stephen down put the book the table-on ‘Stephen (has) put down the book on the table.’

(23)

In this type of sentence, too, the verb, and the terminative particle (or particle plus NP/PP complex) correspond to the two components of the complex event: the verb denotes the movement, and the terminative element denotes the resulting location of the moving theme. The adverbs oda ‘there-to’, be ‘into’, and le ‘down-to’ express that the end location of the moving individual is there, in, and down, respectively.9


27

The omission of the terminative element from an accomplishment predicate yields an atelic sentence denoting a process: (26)

a. Jańos vitte a ta´vcso¨vet. John took the telescope ‘John was taking the telescope.’ b. Zsuzsa szaladt. Susan ran ‘Susan was running.’

In the case of achievements denoting a change of location, the momentary motion process and the attainment of the end location are practically simultaneous; hence cutting off the end location is usually impossible: (27)

*Istvań Stephen

tette put

a the

ko¨nyvet. book

Some approaches, e.g. Hoekstra (1988, 2005), Larson (1988b), and Winkler (1996), analyze the terminative element of change-of-location sentences as a resultative secondary predicate. In Hungarian, too, the syntactic behavior of terminative particles is comparable to that of resultative particles. Thus terminative particles are also licensed in the presence of a [þspecific] theme, which is evidence of their status as secondary predicates predicated of the theme argument. (28) cf.

a. ?*Istvań le -tett ko¨nyveket az asztalra. Stephen down put books the table-on b. Istvań ko¨nyveket tett az asztalra.10 Stephen books-ACC put the table-on ‘Stephen put (some) books on the table.’

In (29a) below, the indefinite noun phrase egy kislańyt ‘a little girl-ACC’ is nonspecific in the sense of En¸c (1991), it can only marginally refer to a member of the previously introduced set; hence the text is not coherent. In (29b), which also contains a verbal particle, the specific reading is the primary one. (29)

a.

A gyerekek sze´tsze´ledtek. ‘The children dispersed.’ #Hı´vtunk egy kislańyt. called-we a little.girl-ACC ‘We called a little girl.’ b. A gyerekek sze´tsze´ledtek. ‘The children dispersed.’ Oda-hı´vtunk egy kislańyt. PRT called-we a little.girl-ACC ‘We called in a little girl [of them].’


28

Verbs of movement in the case of which the causer and the affected theme are represented by the same individual are claimed by Levin and Rappaport (1995) to be unaccusative when expressing directed motion, and to be unergative when expressing a manner of motion. This claim is also supported by facts of Hungarian. When combined with a goal/terminus argument, Hungarian verbs of motion take a verbal particle – see (30), which is evidence of their unaccusativity, i.e., their subject functioning as a theme with the terminus predicated of it. (30)

Zsolt el-futott/el-se´ta´lt a boltba.11 Zsolt off ran /off walked the store-to ‘Zsolt (has) ran/walked off to the store.’

As manner-of-motion verbs, on the other hand, these verbs take no particle, unless they are supplemented with a pseudo-object, as in (31b) – that is, they display the behavior of unergative verbs, illustrated above under (14)–(19). (31)

a. Zsolt Zsolt ‘Zsolt b. Zsolt Zsolt ‘Zsolt

futott/se´ta´lt. ran/walked was running/walking.’ ki-futotta/ki-se´ta´lta maga´t. out-ran /out-walked himself ran/walked himself tired.’

A pseudo-object and a terminus exclude each other. This falls out from the theory of Levin and Rappaport (1995); recall that in the presence of a terminus, the subject functions as the theme, i.e., there is no room for a pseudo-theme, as well. Compare (32a), containing a terminus, with (32b), containing a mere adjunct: (32)

a. *Zsolt ki-futotta/ki-se´ta´lta maga´t a boltba. Zsolt out ran /out walked himself the store-to b. Zsolt ki-futotta /ki-se´ta´lta maga´t a boltba menet. Zsolt out ran /out walked himself the store-to going ‘Zsolt ran/walked himself tired on his way to the store.’

Many motion verbs can also take a ‘‘route’’ or ‘‘path’’ pseudo-object, which, naturally, also licenses a verbal particle: (33)

a. Csaba le -futotta a marathoni Csaba down ran the Marathon ‘Csaba (has) run the Marathon distance.’ b. Pe´ter be-ja´rta a va´rost. Peter in walked the city-ACC ‘Peter (has) walked the city.’

ta´vot. distance-ACC


29

Unexpectedly, the verbal particle also accompanies certain verbs which apparently lack a theme argument. For example: a. Zoltań ra´-va´gott az asztalra. Zoltan on struck the table-on ‘Zoltan (has) struck on the table.’ b. Be´la bele-ru´gott az ajto´ba. Bela into kicked the door-into ‘Bela (has) kicked into the door.’ (35) a. A sofo´´r ra´-kiaba´lt a gyalogosra. the driver at shouted the pedestrian-at ‘The driver (has) shouted at the pedestrian.’ b. Sa´ra ra´ -ne´zett a gyerekekre. Sarah at looked the children-at ‘Sarah (has) looked at the children.’

(34)

A way of analyzing these sentences is to claim that they contain an implicit theme. In (34a,b), it is easy to recover the theme: in (34a), it is Zoltań’s fist, and in (34b), it is Be´la’s foot. (Interestingly, if it is made explicit, it appears in the instrumental case: Zoltań ra´-va´gott az o¨kle´vel az asztalra ‘Zoltan struck with his fist on the table’; Be´la beleru´gott a la´ba´val az ajto´ba ‘Bela kicked with his foot into the door’.) In (35a,b), it must be angry words or looks that are passed on. Alternatively, we could claim that these particles represent a different type: they are directional particles rather than secondary predicates predicating the end location of the theme. Summarizing sections 2.1 and 2.2: Telic predicates expressing a bounded change of state or a bounded change of location involve complex events consisting of a (durative or momentary) process and a resultant state/resultant location. Whereas the process is denoted by a verbal predicate, the resultant state or resultant location is denoted by a resultative or terminative verbal particle, respectively. Alternatively, the resultative or terminative element can also be expressed by a preverbal AdjP, NP/DP, or PP. (In the case of change-of-location sentences, the terminus can be denoted both by a particle marking the attainment of the resulting location, and a noun phrase specifying the resulting location.) The particle functions as a secondary predicate predicated of the theme, and it is licensed only if the subject of predication is a [þspecific] argument. 2.3. Why sentences expressing creation or coming into being contain no particle Besides sentences involving a theme undergoing a delimited change of state, and sentences involving a theme undergoing a delimited change of location, there is also a third major type of telic sentences: those expressing the creation/obtainment, or the coming into being/appearance of their theme. It is this type of theme that I will call ‘incremental theme’ – narrowing down the usage of this term known from the literature. Whereas in change-of-state and change-of-location sentences, the state or location of an already existing theme changes until a new state or new location is

30


attained, in creation/coming-into-being sentences the change is from non-existence to existence, or from absence to presence. Events of this type are completed when the referent of their theme appears in its entirety; the result is the theme itself. For example: a. Jańos ke´szı´tett egy repu¨lo´´ge´pmodellt. John prepared an airplane-model b. E´va hozott su¨temeńyeket. Eve brought cookies (37) a. Szu¨letett egy gyerek. was.born a child ‘A child was born.’ b. Vende´gek e´rkeztek. guests arrived

(36)

The set of verbs capable of expressing creation/obtainment or coming into being/ appearance, and thereby occurring in the syntactic structure illustrated in (36)–(37) is not a closed set. Interestingly, in addition to verbs like szerez ‘obtain’, vesz ‘buy’, kap ‘receive’, alakul ‘formintransitive’, su¨l ‘bakeintransitive’, su¨t ‘baketransitive’, also the verbs eszik ‘eat’, iszik ‘drink’, and fogyaszt ‘consume’ display the syntactic behavior of this verb class. Perhaps it is the meaning component that they share with the verbs szerez ‘obtain’ and kap ‘receive’ that allows them to occur in this pattern. Some analyses, e.g. Tenny (1994), claim that these verbs are ‘reverse’ incremental theme verbs, expressing the reverse appearance, i.e., the disappearance, of their theme. However, other verbs expressing disappearance, e.g. meg-semmisı´t ‘annihilate’, el-tu¨ntet ‘cause to disappear’, el-t u˝ nik ‘disappear’, fel-robbant ‘explodetransitive’, fel-robban ‘explodeintransitive’, le-bont ‘dismantle’, sze´t-szed ‘take apart’, o¨ssze-te´p ‘tear up’, cannot appear in this syntactic structure. Compare: E´va evett egy alma´t. Eve ate an apple (39) a. *E´va semmisı´tett egy feńyke´pet. Eve annihilated a photo cf. b. E´va meg -semmisı´tett egy feńyke´pet. annihilated a photo Eve PRT (40) a. *Pe´ter -te´pett egy levelet. Peter tore a letter cf. b. Pe´ter o¨ssze -te´pett egy levelet. Peter up tore a letter ‘Peter tore up a letter.’ (38)

According to Szabolcsi (1986), verbs belonging to the creation/coming into being class all have an ‘exist’ component in the core of their meaning; they assert that their theme has come to exist or will come to exist in the domain of discourse. As Szabolcsi (1986) observes, since they assert the coming into being of their theme,


31

they cannot at the same time also presuppose its existence, i.e., they cannot have it realized as a definite or a [þspecific] indefinite noun phrase, associated with an existential presupposition. (Chapters 4 and 5 of this book provide a detailed semantic analysis of this verb class from a somewhat different perspective.) A verb of the ‘creation/coming into being’ class is either ungrammatical with a definite or a [þspecific] indefinite theme argument (see (41a,b)), or it only has a process reading, denoting a change affecting an already existing individual (see (42)): a. *Jańos ’ke´szı´tette ’mindegyik ’repu¨lo´´ge´pmodellt.12 John prepared every airplane-model b. *’Szu¨letett a ’gyerek. was.born the child (42) a. E´va fe´l o´ra hosszat ette az alma´t. Eve half hour long ate the apple ‘Eve was eating the apple for half an hour.’ b. *E´va fe´l o´ra alatt ette az alma´t. Eve half hour within ate the apple (41)

If creation/coming into being sentences have a non-specific theme, and, if a verbal particle functions as a predicate predicated of a specific theme, then creation/coming into being sentences are predicted to involve no verbal particle. As the above examples show, this prediction is borne out. Actually, all the above verbs can also occur with a particle and a [þspecific] theme; however, the particle þV combinations express a change of state or a change of location affecting a referent which is presupposed to exist in some form. Compare: a. Jańos ke´szı´tett egy repu¨lo´´ge´pmodellt. John prepared an airplane-model ‘John (has) prepared an airplane model.’ b. Jańos el -ke´szı´tette a repu¨lo´´ge´pmodellt. John PRT prepared the airplane-model ‘John (has) prepared the airplane model.’ c. Jańos el -ke´szı´tett egy repu¨lo´´ge´pmodellt. airplane-model John PRT prepared an ‘John (has) prepared one of the airplane models.’ (44) a. E´va tala´lt egy gy u˝ r u˝ t. Eve found a ring ‘John (has) found a ring.’ b. E´va meg -tala´lta a gy u˝ r u˝ t. found the ring Eve PRT ‘Eve (has) found the ring.’ c. E´va meg -tala´lt egy gy u˝ r u˝ t. found a ring Eve PRT ‘Eve (has) found one of the rings.’ (43)


32 (45)

a. Sa´ra Sarah ‘Sarah b. Sa´ra Sarah ‘Sarah c. Sa´ra Sarah ‘Sarah

hozott su¨temeńyt. brought cake (has) brought some cake.’ meg-hozta a su¨temeńyt. PRT bought the cake (has) brought the cake.’ meg-hozott egy su¨temeńyt. PRT brought a cake (has) brought one of the cakes.’

The (b) and (c) sentences all involve a theme argument whose referent has existed in some form previous to the event; i.e., they represent the type of specificity identified by En¸c (1991). In the case of (43c), egy repu¨lo´´ge´pmodellt corresponds to ‘one of the airplane models in question’. (44c) means that Eve has found one of the rings lost, and (45c) expresses that Sarah has brought one of the cakes previously agreed on. The meaning difference between (46a,b) is more subtle, and has been the topic of much discussion: (46)

a. Gyula ´ırt egy verset. Julian wrote a poem b. Gyula meg-ı´rt egy verset. poem Julian PRT wrote a

According to the literature, e.g. Ka´lmań (1995) and Bende-Farkas (1995), (2001), and according to my own judgment, (46b) cannot be the opening sentence of a discourse. It is felicitous if the previous discussion has already involved prospective poems, or poems in preparation, or at least reference to some writing activity, from which poems can be inferred. Chapters 4 and 5 of this book will examine predicates expressing creation and coming into being in detail. What is relevant from the present perspective is that the type of telic sentence they represent involves no verbal particle, because its necessarily non-specific theme does not license one. A verb of creation/coming into being combined with a verbal particle undergoes a ‘‘type-shift’’; instead of creation or coming into being, it denotes a delimited change of state or a delimited change of location. 2.4.

Locative particles in sentences denoting existence or spatial configuration in a given location

In the lexical semantic theory of Talmy (1985), predicates expressing a state of locatedness are analyzed as a subset of Motion predicates. Motion predicates are made up of the set of translational Move predicates and the set of locational Be predicates. Talmy claims that a state of locatedness is universally conceived and subdivided into components in the same way as an event of translational movement. Hungarian supports this claim. Just as a predicate denoting translational


33

movement to a terminus is preceded by a terminative verbal particle, a predicate denoting existence or spatial configuration in a location is preceded by a locative particle (see Deme (1959)). Even the stems of the particles are often identical, only the endings are different; locational particles often contain an obsolate locative case suffix, instead of the ancient translational case suffix of terminative particles. Compare: (47)

(48)

(49)

(50)

(51)

a. Jańos ki -a´llt (az utca´ra). John out stood the street-to ‘John (has) stood out into the street.’ b. Jańos kint a´ll (az utcań). John outside stands the street-on ‘John is standing outside in the street.’ a. A birko´zo´ le -feku¨dt (a fo¨ldre). the wrestler down lay the ground-on ‘The wrestler lay/has lain down onto the ground.’ b. A birko´zo´ lent fekszik (a fo¨ldo¨n). the wrestler down lies the ground-on ‘The wrestler is lying on the ground.’ a. Ra´ -le´pte´l (a la´bamra). onto stepped-you the foot-my-onto ‘You (have) stepped on my foot.’ b. Rajta a´llsz (a la´bamon). on stand-you the foot-my-on ‘You are standing on my foot.’ a. E´va ide -jo¨tt (az ablakhoz). Eve hereto came (the window-to) ‘Eve has come here to the window.’ b. E´va itt van (az ablakna´l). Eve here is (the window-at) ‘Eve is here at the window.’ a. A hiba be-keru¨lt (a programba). the mistake in got the program-in ‘The mistake has got into the program.’ b. A hiba bent maradt (a programban). the mistake in remained the program-in ‘The mistake has remained in the program.’

The locative particle also accompanies transitive verbs meaning ‘cause something to exist in a given location’, for example: (52)

a. Pe´ter bent hagyott egy hiba´t (a Peter in left a mistake the ‘Peter (has) left a mistake in the program.’

programban). program-in


34

b. E´va ott tartja az iratait (a Eve there keeps her papers the ‘Eve keeps her papers in the safe.’

pańce´lszekreńyben). safe-in

As is clear from these examples, the locative particle – similar to the terminative one – can function as the proadverbial double of a locative noun phrase. The locative noun phrase can also replace the locative particle in its canonical, preverbal position: (53)

a. A birko´zo´ a fo¨ldo¨n fekszik. the wrestler the ground-on lies ‘The wrestler is lying on the ground.’ b. A la´bamon a´llsz. the foot-my-on stand-you ‘You are standing on my foot.’ c. E´va az ablakna´l van. Eve the window-at is ‘Eve is at the window.’ d. A hiba a programban maradt. the mistake the program-in remained ‘The mistake remained in the program.’

The sentences in (47)–(53) describe complex situations, similar to telic sentences involving a resultative or terminative element; however, in their case, the two components of the situation are of similar types (both are states), and they take place simultaneously rather then consecutively. Thus, whereas in the change-of-location sentences in (47a), (48a), and (49a), the V denotes the movement of the subject, and the particle denotes its resultant location, in the (b) sentences, the V expresses the position/static configuration of the subject, and the particle denotes its location. These sentences comprise two predications: e.g. (47b) says that John is standing, and he is outside in the street; or the transitive (52a) expresses that Peter left a mistake, and the mistake is in the program (or perhaps: ‘Peter let there be a mistake, and the mistake is in the program’). The verbs prototypically appearing in this construction are verbs of position or spatial configuration. At the same time, process verbs expressing the activity of an agent can also combine with a locative particle: (54)

a. Jańos John ‘John b. Jańos John ‘John

fent dolgozik (az emeleten). up works the first.floor-on is upstairs working.’ kint nyı´rja a fu¨vet (az udvaron). outside cuts the grass (the courtyard-in) is outside (in the couryard) cutting the grass.’

These sentences combine the statements ‘John is upstairs; he is working’, and ‘John is outside; he is cutting the grass’, respectively. What is more, the locative element


35

represents the main assertion; these sentences would be most felicitous as answers to the question ‘Where is John?’. On the surface, such sentences represent exceptions to the claim that the verbal particle is always predicated of the theme argument; in fact, however, the addition of the locative particle makes the activities denoted by dolgozik ‘is working’ and nyı´rja a fu¨vet ‘is cutting the grass’ state-like. Or perhaps (54a,b) are not particle constructions at all. Fent and kint – unlike e.g. el or meg – can also be analyzed as full-fledged, independent adverbs; perhaps they function as focussed adverbial adjuncts in (54). If the locative particle is combined with a verb denoting a position or spatial configuration, the verb constitutes a possible predicate also without the locative element: (55)

a. Jańos a´ll (az utcań). John stands (the street-on) ‘John is standing (in the street).’ b. Pe´ter fekszik (a fo¨ldo¨n). Peter lies the ground-on ‘Peter is lying on the ground.’

If the locative noun phrase is spelled out postverbally without a proadverb double, it functions as an adjunct; it does not represent the main assertion. Verbs denoting nothing more, or little more, than existence, like van ‘be’, marad ‘remain’, or hagy ‘let’, are not substantial enough semantically to represent the predicate in themselves. They form a complex predicate either with the locative element (56), or with the theme represented by a bare nominal (57): (56)

a. A ke´p ott van a falon. the picture there is the wall-on ‘The picture is on the wall.’ b. A peńz ott maradt a fio´kban. the money there remained the drawer-in ‘The money remained in the drawer.’ c. Jańos ott hagyta az iratokat az ´ıroásztalon. John there left the papers the desk-on ‘John left the papers on the desk.’ (57) a. A falon ke´p van. the wall-on picture is ‘There is a picture on the wall.’ b. A fio´kban peńz maradt. the drawer-in money remained ‘There remained money in the drawer.’ c. Jańos iratokat hagyott az ´ıroásztalon. John papers left the desk-on ‘John left papers on the desk.’


36

Locative particles share the syntactic behavior of resultative and terminative particles; they are predicates predicated of a [þspecific] theme argument. (In (54a,b), their theme is at the same time also the agent of the verbal predicate.) A theme represented by a non-specific indefinite NP does not license a verbal particle. The following examples are ungrammatical as neutral sentences, with all major constituents but the V stressed: (58)

cf. (59)

a. *’Ott hever ’kutya a ’ku¨szo¨b there lies dog the threshold b. *’Itt a´llnak ’mellettem ’szońokok. here stand near.me speakers a. A the ‘The b. A the ‘The

elo´´tt.13 before

kutya ott hever a ku¨szo¨b elo ´´tt. dog there lies the threshold before dog is lying in front of the threshold.’ szońokok itt a´llnak mellettem. speakers here stand near.me speakers are standing near me.’

In the presence of a non-specific theme, the particle must be absent: (60)

a. A kirakatban ’ke´pek vannak. the shopwindow-in pictures are ‘In the shop-window there are pictures.’ b. A ku¨szo¨bo¨n ’kutya hever. the threshold-on dog lies ‘On the threshold, there is a dog lying.’

The specificity of the theme in constructions involving a locative verbal particle can be made clear by the following minimal pair: (61)

a. Az ´ıro´ szeret az a´llatairo´l ´ırni. ‘The writer likes to write about his animals.’ A legu´jabb regeńye´ben is van ke´t kutya. his latest novel-in too are two dog ‘There are two dogs also in his latest novel.’ b. Az ´ıro´ szeret az a´llatairo´l ´ırni. ‘The writer likes to write about his animals.’ A legu´jabb regeńye´ben is benne van ke´t kutya. his latest novel-in too in are two dog ‘There are two dogs (of his animals) also in his latest novel.’

Ke´t kutya can be identified as representing a subset of the set denoted by az a´llatai ‘his animals’ only in the presence of the verbal particle in (61b).


37

Summarizing section 2: the three major types of verbal particles are associated with three major types of situations. Locative particles appear in sentences describing existence or spatial configuration in a given location, resultative particles appear in sentences describing a delimited change of state, and terminative particles appear in sentences describing a delimited change of location. A resultative or terminative particle turns a process predicate into an accomplishment.

3.

V E R B A L PA RT I C L E S A N D A S P E C T

It has been known for a long time that verbal particles contribute to the aspectual interpretation of sentences cf. e.g. Perrot (1966), Wacha (1976, 1989), E´. Kiss (1987), Kiefer (1992, 1994), Kiefer and Ladańyi (2000). Assuming the twocomponent aspectual theory of Smith (1991), the question is if verbal particles play a role in determining situation aspect or viewpoint aspect. The conclusion at the end of section 2 suggested that the occurrence of various types of verbal particles is limited to particular types of situations. This section aims to examine whether the attested association of types of particles with types of situations means a direct correlation between them, or is merely an epiphenomenon derivable from the viewpoint-aspect marking function of particles. In other words, it aims to ascertain if the different types of particles indeed mark (a)telicity, rather than (im)perfectivity. The sentences with a resultative or terminative particle analyzed in sections 2.1 and 2.2 were both telic and perfective. Observe again some examples: a. E´va fel -va´gta a torta´t. Eve up cut the cake ‘Eve cut up the cake.’ b. Ju´lia fel -e´bredt Eve up woke ‘Eve woke up.’ (63) a. Pe´ter ki tolta a biciklit az utca´ra. Peter out pushed the bicycle the street-to ‘Peter pushed out the bicycle into the street.’ b. Az alma le -esett a fo¨ldre. the apple down fell the ground-to ‘The apple fell down onto the ground.’

(62)

These sentences are telic because they describe an inherently delimited change, and they are perfective because they represent an event with its initial point and its end point included. The question is which of these two properties is the direct consequence of the presence of the particle. Whereas a preverbal resultative particle always yields a perfective interpretation – see (62a,b) –, a postverbal resultative particle elicits an imperfective


38 reading. Compare: (64)

a. Amikor a vende´gek mege´rkeztek, E´va e´ppen va´gta fel a torta´t. when the guests arrived Eve just cut up the cake ‘When the guests arrived, Eve was just cutting up the cake.’ b. Amikor a vende´gek mege´rkeztek, E´va e´ppen fel -va´gta a torta´t. when the guests arrived Eve just up cut the cake ‘When the guests arrived, Eve had just cut up the cake.’ c. *Miutań a vende´gek mege´rkeztek, E´va ’va´gta fel a torta´t. after the guests arrived Eve cut up the cake ‘After the guests arrived, Eve was just cutting up the cake.’

The main clause of (64a), displaying a ‘V particle’ order, describes an activity in progress, without its initial and end points; i.e., the viewpoint of the clause is imperfective. This is in contrast to (64b), which is perfective, with the activity understood to be completed by the time referred to by the temporal clause. A complex sentence containing an after-clause requires the consecutive interpretation of the events described by the temporal clause and the main clause. (64c) is ungrammatical because the main clause is imperfective, describing an event in progress, i.e., the consecutive interpretation of the two events is impossible. The sentence type represented by the main clause of (64a) describes a telic situation with an inherent endpoint. The Hungarian equivalent of cut denotes a change affecting the object, and the Hungarian equivalent of the particle up expresses its total affectedness, which makes the continuation of the cutting process impossible. The endpoint of the situation, i.e., the attainment of the total affectedness of the cake, however, is outside the viewpoint of the given sentence. It is precisely this sentence type that is analyzed as telic imperfective by Smith (1991). If the postverbal particle is omitted, the resulting sentence will lack the feature [þtelic]. (65) means almost the same as the main clause of (64a) – except that it does not imply that Eve was in the process of cutting up the whole cake; it allows the interpretation that she just cut into it, or cut one piece from it: (65)

E´va e´ppen ’va´gta a ’torta´t. Eve just cut the cake ‘Eve was just cutting the cake.’

The comparison of (64a) and (65) suggests that what the presence of a resultative particle adds to the sentence is the feature [þtelic]. The relative position of the particle, on the other hand, affects the interpretation of viewpoint aspect. Sentences involving a movement verb and a terminative particle display the same word order variants associated with the same types of aspectual interpretation as sentences involving a change-of-state verb and a resultative particle.


39

For example: (66)

a. Amikor e´szrevettem, Jańos e´ppen tolta ki a biciklit when noticed-I John just pushed out the bicycle (az utca´ra). the street-to ‘When I noticed him, John was just pushing out the bicycle into the street.’ b. Amikor e´szrevettem, Jańos e´ppen ki -tolta a biciklit when noticed-I John just out pushed the bicycle (az utca´ra). the street-to ‘When I noticed him, John had just pushed out the bicycle into the street.’ c. Amikor e´szrevettem, Jańos e´ppen tolta a biciklit. when noticed-I John just pushed the bicycle ‘When I noticed him, John was just pushing the bicycle.’

The ‘V particle’ order in (66a) denotes an action being in progress at the time referred to by the temporal clause, whereas the ‘particle V’ order in (66b) denotes an action completed by the time referred to by the temporal clause. Although both (66a) and (66c) describe situations from an imperfective point of view, (66a) remains a delimited telic situation with an inherent end point. Thus in (66a,b) alike, the verbal particle appears to be the carrier of the feature [þtelic]. Its position, on the other hand, interacts with the viewpoint aspect. The imperfective interpretation of the ‘V particle’ order in (64a) and (66a), and the perfective interpretation of the ‘particle V’ order in (64b) and (66b) raises the possibility of a different analysis, as well. Let us tentatively assume that the resultative or terminative verbal particle is a perfectivity marker; but it functions as such only when it occupies the specifier position of an AspP projection dominating VP, i.e., when it surfaces preverbally (see E´. Kiss 1994). Under this analysis, a [þperfective] Asp head requires the presence of a resultative or terminative particle in Spec,AspP, whereas a [perfective] Asp head allows no other particle but a locative one there. The execution of this approach, however, would run into difficulties. Given that there are also perfective sentences with no particle – e.g. sentences expressing creation or coming into being, like (67a), or sentences describing processes finished in the past, like (67b), the presence of a resultative or terminative particle in Spec,AspP cannot be a condition of perfective aspect. (67)

a. E´va su¨to¨tt egy torta´t. Eve baked a cake b. E´va tegnap fo´´zo¨tt, mosott, e´s ne´zte a te´ve´t. Eve yesterday cooked, washed, and watched the TV-ACC ‘Yesterday, Eve cooked, washed clothes, and watched TV.’

40


The events in (67a,b) are represented in their entirety, with their initial and end points included, hence these predicates are perfective, despite the emptiness of Spec,AspP. In sentences containing a resultative or a terminative particle, e.g. in (64b) and (66b), on the other hand, an empty Spec,AspP cannot but yield an imperfective reading. There is also a further problem. An apparently terminative preverbal particle does not always elicit a perfective interpretation. Compare: (68)

a. A Duna bele -folyik a Fekete-tengerbe. the Danube into flows the Black See-into ‘The Danube flows into the Black See.’ cf. b. A tinta a tollbo´l bele -folyt a zsebembe. the ink the pen-from into flowed the pocket-my-into ‘The ink from the pen flowed into my pocket.’ (69) a. A bokor a´gai ki -hajlanak az utca´ra. the bush’s branches out lean the street-onto ‘The branches of the bush lean out into the street.’ cf. b. Pe´ter ki -hajolt az auto´ ablakań. Peter out leaned the car’s window-on ‘Peter leaned out of the window of the car.’ In the (b) sentences, the terminative particle denotes the resultant position of the moving theme; i.e., these sentences describe telic situations (accomplishments) looked at from a perfective viewpoint, as expected. In the (a) sentences, on the other hand, the very same verbal particle combined with the very same verb functions as a directional adverb rather than a terminative predicate. Although the movement has a goal in these sentences, as well, the attainment of the goal does not complete the process. In (68a), it is the particular semantics and pragmatics of the given predicate– argument complex that block the terminative reading of the particle, and elicit its directonal interpretation. The verb hajlik ‘lean’ in (69), on the other hand, is ambiguous; it has a primary ‘movement’ reading, and a secondary ‘position’ interpretation. The non-human subject of (69a) triggers the latter reading, whereas the human subject of (69b) triggers the former interpretation. Under the ‘movement’ reading, the particle functions as a terminative secondary predicate, whereas under the ‘position’ interpretation, it functions as a directional adverb. What these examples suggest is that not even a preverbal resultative or terminative particle expresses a specific, invariant viewpoint aspect. Verbal particles have an (often decisive) role in aspect marking by means of their contribution to the lexical meaning of the predicate. The type of aspect that is determined compositionally by the lexical meanings of sentence constituents is situation aspect. As is also clear from the above discussion, the correlation attested in (64a,b) and (66a,b) between the preverbal position of the terminative/resultative particle and perfective viewpoint, on the one hand, and between the postverbal position of the terminative/resultative particle and imperfective viewpoint, on the other hand, is an epiphenomenon rather than a direct, biunique relation. Its nature will be clarified in

41


chapter 6, where it is claimed that the [perfective] variant of the Asp head determining the viewpoint aspect of the Hungarian sentence does not tolerate a delimiting (resultative or terminative) constituent in its local environment. The conclusion that the verbal particle primarily marks situation aspect – by means of its lexical meaning, and it affects viewpoint aspect only indirectly – is also confirmed by the behavior of locative particles. Whereas resultative and terminative particles express telicity, locative particles, appearing in a type of stative sentences, those expressing existence or spatial configuration in a given location, are associated with atelicity. Just as sentences containing a resultative or terminative particle are usually perfective, but are also compatible with the imperfective viewpoint, sentences containing a locative particle are usually imperfective, but are also compatible with the perfective viewpoint – that is, they cannot be viewpoint aspect markers, either. Consider: (70)

Pisko´ta´t kavartam, meg-su¨to¨ttem, o¨t percig PRT -baked-I-it five minutes-for sponge.cake-ACC mixed-I hagytam a su¨to´´ben, majd kette´ -va´gtam. left-I-it the oven-in then two.into- cut-I-it ‘I mixed a sponge-cake, I baked it, I left it in the oven for five minutes, then I cut it into two.’

bent inside

The complex sentence in (70) describes a series of consecutive situations, which are represented in their entirety, from a perfective viewpoint. The predicate (a pisko´ta´t) bent hagytam a su¨to´´ben ‘I left (the sponge cake) in the oven’, involving a locative particle, denotes a state, and as a stage-level stative predicate combined with a time adverbial denoting a time span, it allows a perfective reading. I conclude that the verbal particle, whether resultative, terminative, or locative, plays a role in determining the situation aspect, i.e., basically, the [þ/telic] nature, of sentences. Resultative and terminative particles have a telicizing effect, whereas locative particles occur in a type of atelic sentence. They are not direct aspectual operators; they assume an aspectual function by means of their lexical meaning (e.g., in the case of terminatives and resultatives, by means of their delimiting role). Particle position may affect the interpretation of viewpoint aspect (in the case of accomplishment predicates, the imperfective reading is elicited by the ‘V particle’ order). Viewpoint aspect will be argued in chapter 6 to be determined by an Asp head, which imposes certain cooccurrence restrictions on its environment, barring resultative and terminative elements from the domain of a [perfective] Asp.

4.

4.1.

T H E S Y N TA X O F T H E V E R B A L PA RT I C L E

Against the Spec,AspP position of the particle

The discussion of the aspectual role of the verbal particle has led to the conclusion that the verbal particle is a secondary predicate, which plays a role in determining the


42

situation aspect of the sentence by adding a resultative, terminative, or locative component to the situation. This conclusion undermines the analysis of the verbal particle as a viewpoint aspect operator moved from the VP into the specifier of an AspP projection, as proposed e.g. by Pinõn (1995), E´. Kiss (2002a), Alberti (2004), and den Dikken (2004). A further problem of the Spec,AspP approach is that it is also forced to treat the bare nominal complement, an alternative filler of the preverbal slot, on a par with the verbal particle – even though the aspectual operator role of bare nominals is even more questionable. Consider the Hungarian sentence structure assumed in this framework (emerging from the work of Brody (1995), Olsvay (2000a,b), E´. Kiss (2002a), etc.):

TopP

(71) Spec

Top’ Top

FocP

Spec

F’ F

AspP Spec

Asp’ Asp

VP V

XP*

The filling of Spec,AspP triggers V movement to Asp, and the filling of Spec,F(ocus)P goes together with further V movement from Asp to F. The main stress, indicating the left edge of the predicate phrase, falls on the first major post-topic category. In this structure, a stressed preverbal constituent is either an aspectual operator in Spec,AspP (with the V raised to Asp), or a focus expressing exhaustive identification, sitting in Spec,FP (with the V raised to F).14 A bare nominal complement, e.g. that in (72b), clearly does not express any exhaustive identification. According to the exhaustivity test of Szabolcsi (1981), if the (b) sentence of the following pair of examples is a logical consequence of the (a) sentence, the preverbal constituents are not exhaustive foci. (72)

a. Bara´tokat e´s rokonokat va´runk vacsora´ra. friends-ACC and relatives-ACC expect-we dinner-for ‘We are expecting friends and relatives for dinner.’


43

b. Bara´tokat va´runk vacsora´ra. friends-ACC expect-we dinner-for ‘We are expecting friends for dinner.’ (72a) and (72b) can be simultaneously true, hence their preverbal bare nominals cannot sit in Spec,FP, the position of exhaustive focus. If the Hungarian sentence has the structure in (71), then the bare nominals must occupy Spec,AspP; structure (71) leaves no other position for them. Examining the contribution of the preverbal nominal to aspectual interpretation, we find that if the verb has a durative atelic meaning, the ‘bare nominal, V’ complex is also durative atelic, and its preferred reading is imperfective: (73)

a. E´va (ke´t o´ra hosszat) verset olvasott Eve two hour long poem-ACC read ‘Eve was reading poem(s) (for two hours).’ b. Miha´ly csomagot cipel. Michael parcel-ACC carries ‘Michael is carrying a parcel.’

If, on the other hand, the verb in the ‘bare nominal, verb’ complex is [-durative], the preferred reading of the predicate is perfective: (74)

a. E´vańak (ma´jusra) gyereke szu¨letik. is.born Eve-DAT May-by child ‘To Eve, a child will be born by May.’ b. Bence levelet kapott. Bence letter-ACC got ‘Bence got a letter.’

In these sentences, the replacement of the bare nominal with an indefinite noun phrase does not change the aspectual interpretation, e.g.: a. E´va (fe´l o´ra hosszat) olvasott egy verset. Eve half hour long read a poem ‘Eve was reading a poem for half an hour.’ b. Miha´ly cipel egy csomagot. Michael carries a parcel ‘Michael is carrying a parcel.’ (76) a. E´vańak (ma´jusra) szu¨letik egy gyereke. child Eve-DAT May-by is.born a ‘To Eve, a child is born by May.’ b. Bence kapott egy levelet. Bence got a letter (75)


44

As these predicates have the same aspectual interpretation as the corresponding ‘bare nominal, V’ complexes, the bare nominal cannot be an aspect-marker. According to Komlo´sy (1994), a bare nominal like those in (72)–(76) is a predicative element; it is predicated of the internal argument incorporated into the verb. E.g. (74d) can be paraphrased as follows: (77)

Bence received x; x is of the kind ‘letter’

If the bare nominal complement and the verbal particle are indeed alternative fillers of the same syntactic slot, then their shared position must be simply a predicative position, because the predicative feature seems to be the only feature that they share. Both are predicates predicated of the theme argument – with the difference that, whereas the verbal particle has an explicit subject, the subject of a bare nominal is an implicit internal argument. 4.2.

Is the particle a head or a phrase?

Before we attempt to identify this predicative position in Hungarian sentence structure, we have to decide if it is a head or a phrase position, and if it is basegenerated or derived. These questions have been raised in the literature several times (see e.g. Surańyi (2002a), Olsvay (2000b; 2004), Koopman and Szabolcsi (2000)); several relevant arguments have been weighed – even if no conclusive answer has been arrived at. What underlies the head versus phrase question is whether or not the verbal particle or bare nominal complement forms a complex predicate with the verb, as is assumed e.g. in Ackerman (1984), Farkas and Sadock (1989), Brody (1990), Pinõn (1995), E´. Kiss (1998a), or Szendro´í (2003). The complex predicate analysis is supported by the fact that the particle and the V (and often also the bare nominal and the V) form a lexical unit, the meaning of which is sometimes non-compositional (or only partially compositional) – e.g. be-ru´g [lit. in-kick] ‘get drunk’, be-fejez [lit. in-head] ‘finish’, bakot lo´´ [lit. buck-ACC shoot] ‘make a mistake’, csu¨to¨rto¨ko¨t mond [lit. Thursday-ACC say] ‘fail’. This argument, however, is non-conclusive; after all, idioms are allowed by Universal Grammar to have simultaneously a non-compositional meaning and a phrasal structure. The usual syntactic tests of constituent structure argue against the complex predicate analysis of the ‘particle/bare nominal, V’ string, and, in general, against the head analysis of the particle and the bare nominal. Namely, i. The particle can serve as a short answer to a yes-no question. E.g.: (78)

a. ‘‘Meg-

etted az ebe´det?’’ ‘‘Meg.’’ ate-you the lunch PRT ‘Did you eat the lunch?’ ‘Yes.’ b. ‘‘Be-ru´gta´l?’’ ‘‘Be.’’ in kicked-you in ‘Did you get drunk?’ ‘Yes.’ PRT


45

Cross-linguistically, only phrasal constituents can function as elliptical sentences. In Hungarian, too, a verb complement or an adjunct can constitute an elliptical sentence; a subconstituent of them, on the other hand, cannot: a. Magyar mondattani jelense´gekro´´l ´ırod a disszertaćio´dat? Hungarian syntactic phenomena-on write-you the dissertation-your ‘Is it about Hungarian syntactic phenomena that you write your dissertation?’ b. *Nem, francia. no French c. Nem, francia´-k-ro´l. no French-PL-about ‘No, about French ones.’ (80) a. Egy fiaira bu¨szke apa mondta ezt? a his.sons.of proud father said that ‘Was it a father proud of his sons who said that?’ b. *Nem, lańyaira. /*Nem, lańyaira bu¨szke. no his.daughters.of no his.daughters.of proud ‘No, of his daughters.’ / ‘No, proud of his daughters.’ c. Nem, egy lańyaira bu¨szke apa. no a his.daughters.of proud father ‘No, a father proud of his daughters.’

(79)

In (79), the answer francia´kro´l ‘about French-PL-ACC’ stands for the whole noun phrase, as is shown by the case ending, which can only be attached to the right edge of the noun phrase complementing the verb ´ır ‘write’. The missing part of the noun phrase, mondattani jelense´g ‘syntactic phenomenon’, is represented by an empty category. Apparently, a bare verb, or a ‘particle, V’ string can also serve as an elliptic answer: (81)

(82)

‘‘Hı´vtad Pe´tert korcsolya´zni?’’ called-you Peter-ACC to.skate ‘Have you asked Peter to come skating?’ ‘‘Meg-bı´zol a munkata´rsaidban?’’ PRT trust-you the co-workers-your-in ‘Do you trust your colleagues?’

‘‘Hı´vtam.’’ called-I ‘I did.’ ‘‘Meg-bı´zom.’’ PRT trust-I ‘I do.’

In the analysis to be proposed, the answers in (81–82) involve a phrase identified as a PredP, with the V moved to Pred, and with the remnant VP deleted. ii. An indefinite noun phrase can fulfil the same function in preverbal position as a bare nominal. It need not be understood as an exhaustive focus – as is shown by the


46 following exhaustivity test: (83)

a. Pe´ter Peter ‘Peter b. Pe´ter Peter ‘Peter

egy versen a poem-on is working on a egy versen a poem-on is working on a

e´s egy novellań and a short.story-on poem and a short story.’ dolgozik. works poem.’

dolgozik. works

(83b) is a logical consequence of (83a) – hence the preverbal indefinite noun phrase of (83a,b) is not a focus. It must be an alternative filler of the position of the bare nominal, therefore, that position cannot be a head position. iii. A particle can move non-locally, which is typical of phrasal movement: (84)

Fel kell, hogy hı´vjam Marit ma up needs that call-I Mary-ACC this ‘It is necessary that I call up Mary tonight.’15

este. evening

iv. If the particle is generated as part of a complex predicate, the constructions in which it is separated from the V must involve excorporation, which has a theoretically dubious status. Here are two constructions in which the particle has left the alleged complex predicate: (85)

Jańos John is also ‘John

el akart menni a tegnapi elo´áda´sra, e´s el off wanted to.go the yesterday’s performance-to and off ment. went-he wanted to go to yesterday’s performance, and go he did.’

The emphatic is always follows the first stressed major constituent of the second conjunct in a coordinated structure. It can follow an argument or an adjunct, but it cannot follow a subconstituent of them: (86)

Jańos az elso´´ sorban akart u¨lni, e´s az elso´´ sorban is u¨lt. John the first row-in wanted to.sit and the first row-in also sat-he ‘John wanted to sit in the first row, and in the first row he sat.’ b. *Jańos az elso´´ sorban akart u¨lni, e´s az elso´´ is sorban u¨lt. c. Jańos ele´g korań akart el -indulni, e´s ele´g korań is John fairly early wanted PRT to.leave and fairly early also indult el. left.he PRT ‘John wanted to leave fairly early, and fairly early he left.’ d. *Jańos ele´g korań akart elindulni, e´s ele´g is korań indult el. a.


47

The emphatic is can follow the verb, but it cannot follow the ’particle, V’ complex: (87)

a. Jańos menni szańde´kozott az elo´áda´sra, e´s ment is John to.go wanted the performance-to and went-he also ‘John wanted to go to the performance, and go he did.’ b. *Jańos szańde´kozott el-menni az elo ´áda´sra, e´s el-ment is. John wanted off to.go the performance-to and off went also

(86a,c) may involve phrasal movement into the specifier of an is projection, and (87a) may involve head (V) adjunction to is. One or the other of these two operations has applied to the verbal particle in (87), as well. Whether it has been subject to phrasal movement, as suggested above, or to head movement, it cannot be part of a complex verb. v. As is well-known, a particle can undergo focussing and contrastive topicalization, both of which represent phrasal movement:16 (88)

a. A lifttel LE -ment Jańos, nem FEL. the elevator-with down went John, not up ‘It was down that John went with the elevator, not up.’ fel pedig PE´TER. b. Le JAŃOS ment a lifttel, down John went the elevator-with up on.the.other.hand Peter ‘Down, it was John who went with the elevator, up, on the other hand, it was Peter.’

I conclude on the basis of the evidence listed in (i)–(v) that the particle is a phrase (an AdvP) consisting of a mere head. The facts in (i) and (iv) do not support the structure assigned to English particle verbs by Larson (1988b) and Winkler (1996), either. In their structure, the particle is a phrasal modifier of the V head:

(89)

VP V’ AdvP

V

However, if a modifier attached to a nominal is not a possible elliptical sentence, as was argued under (i) above, and cannot be followed by an emphatic is, as was shown under (iv), it is not clear why a modifier attached to the V should behave differently. 4.3.

The PredP hypothesis

If the particle phrase is not adjoined to V, it must be sitting in a specifier position. Then the question is if it occupies Spec,VP, or the specifier of a functional projection.

48


The following coordination facts from E´. Kiss (1998a) (i.e., the ungrammaticality of the (b) examples) would get a natural explanation if the verbal particle or the bare nominal complement occupied Spec,VP – at least under the assumption that nonmaximal projections cannot be coordinated: a. Jańos [VP MEG [V’ ette a po¨rko¨ltet]] e´s [VP MEG [V’ itta a bort]] John up ate the stew and up drank the wine ‘John ate (up) the stew and drank (up) the wine.’ ! b. *Jańos [VP MEG [V’ ette a po¨rko¨ltet] e´s [V’ itta a bort]] (91) a. Jańos [VP KEZET [V’ fogott Pe´terrel]] e´s [VP KEZET [V’ ra´zott hand-ACC shook John hand-ACC clasped Peter-with and Ivańnal]] Ivan-with ‘John clasped hands with Peter and shook hands with Ivan.’ ! b. *Jańos [VP KEZET [V’ fogott Pe´terrel] e´s [V’ ra´zott Ivańnal]]

(90)

However, there is a powerful argument against the Spec,VP-position of particles and bare nominals. The preverbal position of the particle or bare nominal is a landing site for long movement; hence it cannot be a base-generated position; it must be in the specifier of a functional head attracting the particle. The particle or bare nominal can undergo long movement (across a non-finite or even a finite clause) if the matrix verb is a modal or aspectual auxiliary or semi-auxiliary, not denoting a separate event.17 (92)

a. Be fogok/akarok/szeretne´k /szoktam menni az in will-I/want-I/would.like-I/used-I to.go the ‘I will/want/would like/used to go in to the show.’ b. Be szabad menned. in may go-INF-2SG ‘You may go in.’ c. Jańos be kell, hogy fejezze a munka´t. finish the job John PRT needs that ‘John needs to finish the job.’ d. De´lre ebe´det kell, hogy fo´´zzek. cook-I noon-by lunch-PRT needs that ‘By noon, I need to cook lunch.’

elo´áda´sra. show-to

The particle plays the same role whether it immediately precedes its lexical verb or a matrix auxiliary, therefore, it seems reasonable to assume that it occupies the same position, the specifier of a functional projection, in both cases. Following Koster’s (1994) analysis of Dutch, and Csirmaz’s (2004) analysis of Hungarian, I assume that this projection is a PredP.


49

However, if we assign to (90b)–(91b) the structures represented in (93), we lose the explanation of their ungrammaticality. (93)

po¨rko¨ltet] a. *Jańos [PredP MEG [VP ette a PRT ate the stew John b. *Jańos [PredP KEZET [VP fogott Pe´terrel] clasped Peter-with John hand-ACC

e´s [VP itta a bort]] and drank the wine e´s [VP ra´zott Ivańnal]] and shook Ivan-with

A possible way of resolving the contradiction is to assume V-to-Pred movement, as shown in (94). V-to-Pred movement presumably also takes place in sentences involving no verbal particle. PredP is subsumed by AspP, TenseP, and TopP projections. Csirmaz argues in chapter 6 that AspP has an EPP feature, hence it is to be filled by the closest possible filler: by the particle/bare nominal in Spec,PredP, or in case Spec,PredP is empty, by the V. Tense agrees with the nominative subject without attracting it.

(94)

TopP Spec Jánosi

Top’ Top

TenseP Spec

Tense’

Tense

AspP Spec

Asp’

Asp

PredP Spec kezet j

Pred’ Pred fogott k

VP

V tk

DP NP DP t i t j Péterrel


50

Assuming (94), the ungrammatical coordination structures in (90b)-(91b)/(93) could not be derived in any way. If V-to-Pred movement follows coordination, then it violates the coordinate structure constraint in (93). If V-to-Pred movement precedes VP coordination, the resulting V-initial constituent, a Pred’, is not a possible target of coordination. The remnant VP, on the other hand, is predicted to undergo coordination, correctly: (95)

a kerı´te´sre]]]] Jańos [PredP fel [Pred’ ugrott [VP [VP a fo¨ldro´´l John up jumped the ground-from the fence-on e´s [VP a kerı´te´sro´´l a teto´´re]] and the fence-from the roof-on ‘John jumped up from the ground to the fence, and from the fence to the roof.’

As regards ellipsis, it must be a PF rule (the ellipted material must be visible at LF); hence it might be the case that the constraints it is subject to are different from those operative in syntax. Perhaps the string to be ellipted must form a constituent, but not necessarily a maximal one. Then the short answer e.g. in (78a,b), consisting of a mere particle, could be derived by Pred-deletion. The elliptical sentences in (81) and (82) could be the result of VP-deletion following V-to-Pred movement. The proposed structure may have a source similar to the structure assigned to particle verbs by Larson (1988a,b) and Winkler (1996), in which the internal argument (NP2), functioning as the subject of the secondary predicate (XP), c-commands its predicate, as required by the predication theory of Williams (1980) and Rothstein (1985):

(96)

vP NP1

v’ v

VP NP2

V’ V

XP

It is possible that Hungarian sentences containing a complex predicate also derive from a hierarchy like (96), which is obliterated by V movement to Pred. As argued in chapter 9, V movement in Hungarian causes the collapse of the projections deprived of their head, and their reordering according to the phonological weight of the constituents. Koster (1994) proposed the following theory about the motivation of particle movement into Spec,PredP. (Similar ideas were put forth independently by Alberti (1997), as well.) The complement of a verb can be licensed in two ways: it can be an


51

argument or part of the predicate. Complements which are of a predicative nature can only be licensed in the latter way; that is, they must check their [þpred] feature in Spec,PredP. As regards Hungarian, I assume that the finite verb of the sentence (or in verbless sentences, the nominal predicate) also has a [þpred] feature in need of checking, which is checked by V-to-Pred (or in lack of a V, by N/Adj-to-Pred) movement. For example: (97)

a. Jańos John ‘John b. Jańos John ‘John

[PredP olvastai [VP ti az u´jsa´got]] read the newspaper was reading the newspaper. [PredP tana´ri [NP ti ]] teacher is a teacher.’

Csirmaz argues in chapter 6 that the highest element in PredP, i.e., the particle/ bare nominal landing in Spec,PredP, or the V in sentences with no particle, is moved on string-vacuously into the specifier or head position of AspP. The ’V, Particle’ order of telic imperfective sentences, illustrated in (65) and (66a), and in (98) below, is also the result of V movement to Asp. The [þimperfective] Asp head of such sentences does not accept a delimiting terminative or resultative particle in Spec,AspP, hence the closest potential filler of Asp is the V, moving across the particle: (98)

Jańos e´ppen [AspP csuktai [PredP be ti [VP ti az ajto´t]] John just closed in the door ‘John was just closing the door.’

The ‘V Particle’ order attested in sentences containing a focus or a negative particle (or both) is also assumed to be the result of V movement across the particle. These constructions will be analyzed in detail in chapter 9. Briefly, the V in nonneutral sentences (including also imperatives) raises to the head position of a Non-Neutral Phrase (as proposed by Olsvay (2000b)). In the case of negation, the Non-Neutral Phrase merges with a Neg head projecting a NegP, whereas in the case of focussing, the Non-Neutral Phrase (or the NegP derived from it) merges with a Foc(us) head projecting a FocP. The resulting constructions represent various realizations of a phase with the V in its head, and with the post-verbal domain handed over to Spell-Out. At the syntax-phonology interface, the projections having lost their head (VP, vP, and PredP) are flattened, and their constituents are rearranged according to phonological weight, beginning with the lightest constituent, and ending with the heaviest one. Observe the position of the particle in imperative, negative,


52 and focus constructions: (99)

a. [NNPSza´molja´toki [AspP o¨sszej ti [PredP tj ti [VP ti a szavazatokat!]]]] PRT the votes count ‘Count the votes!’ b. [NegP Nem [NNP sza´molta´ki [AspP o¨sszej ti [PredP tj ti [VP ti a szavazanot counted-they PRT the votes tokat]]]]] ‘They did not count the votes.’ c. [FocP csak KETTEN pro [NNP sza´molta´ki [AspP o¨sszej ti [PredP tj ti only two counted PRT [VP ti a szavazatokat]]]]] the votes ‘Only two persons counted the votes.’ 5.

S U M M A RY

This chapter has argued that the verbal particle, a seemingly idiosyncratic category of the Hungarian sentence, can be given a systematic analysis on the basis of which both its functional and its syntactic properties can be predicted. As regards its function, it has been claimed that it plays a role in determining the situation aspect of sentences. The three major types of verbal particles mark three major types of events. Resultative particles mark telic sentences denoting a delimited change of state, terminative particles mark telic sentences denoting a delimited change of location, whereas locative particles mark sentences denoting existence or spatial configuration in a given location. Contrary to wide-spread assumptions, the verbal particle is not a perfectivizer; it only plays an indirect role in determining viewpoint aspect. The syntactic analysis of the verbal particle has led to the conclusion that it is a phrasal category (an AdvP consisting of a mere head), moving first to the specifier of a PredP projection, and then to the specifier of AspP encoding viewpoint aspect. (The latter move might be blocked by a constraint to be discussed in chapter 6.) The particle (as well as the bare nominal complement) moves to Spec,PredP in order to check its [þpred] feature. Owing to checking requirements, the V also moves to Pred. In nonneutral sentences the V raises across the particle into the head of a Non-Neutral Phrase to be merged with a F(ocus) or a Neg head. In the resulting phase, the constituents of the post-head domain are rearranged according to phonological weight. NOTES 1

2

Precedents for this claim can be found in Kiefer (1992), and Kiefer and Ladańyi (2000), who argue that certain types of particles serve to denote particular types of Aktionsart, a notion related to situation aspect. It would also deserve some consideration why resultative adjectives are often case-marked, e.g.: szo ´´ke´-re fest ‘blond-onto dye’, boldog-ga´ tesz ‘happy-to make’, a´tla´tszo´-va´ va´lik ‘transparent-to become’. If only noun phrases can be case-marked, then these phrases may contain an empty noun phrase. The case marking seems to suggest that the resultative phrase is theta-marked by the verb – as proposed by Larson (1988b).

S Y N TA X O F V E R B A L PA RT I C L E 3 4

5

53

Non-compositional idioms like be-ru´g ‘get drunk’ lit. ‘in kick’ represent exceptions to this claim. Although without its verb, a given particle usually gives no information on the quality of the result state it denotes, its choice is not completely idiosyncratic. Be ‘in’, fel ‘up’, and the most unmarked meg denote different kinds of complete affectedness; however, the examination of the semantic differences of the various verbal particles is beyond the scope of this paper. For information on this issue, see Perrot (1966), Szili (2001) etc. The symbol ’ marks stress. Stress is indicated so as to ensure a neutral prosody. Some of the sentences in (6) can be grammatical if they contain a focus, e.g.: (i) ’JAŃOS tala´lta a gyur ˝ ut. ˝ John found the ring ‘It was John who found the ring.’

6

This sentence is a ‘creation’ sentence, discussed in section 2.3 and in chapters 4–5. The theme of a creation predicate is represented by a non-specific indefinite NP – unless the sentence contains a focus, which renders the theme presupposed. In Hungarian, generic plurals have a definite article – see (i); therefore, bare plurals are not ambiguous between a non-specific and a specific (i.e., generic) reading. (i) A go´lya´k Afrika´ban telelnek. the storks Africa-in spend.winter ‘Storks spend the winter in Africa. ’

7

According to En¸c (1991), a universally quantified expression is specific. E´. Kiss (1993) also argues that a quantified expression counts as specific if it denotes a set whose elements are known. In any case, a verbal particle is compatible with a quantified theme argument, e.g.: (i) Jańos meg-hı´vott valakit/ mindenkit. PRT invited somebody/everybody John ‘John invited somebody/everybody.’

8

E´. Kiss (2005b) examines apparent counterexamples to the claim that unergative verbs cannot be associated with a resultative particle, e.g.: (i) Pe´ter meg-reggelizett. Peter PRT breakfasted ‘Peter had breakfast.’ (ii) E´va ki-takarı´tott/el-mosogatott. Eve out cleaned/up washed ‘Eve did the cleaning/did the washing up.’

9

10

(i) is argued to have a theme incorporated into the verb, whereas the predicates in (ii), expressing institutionalized activities, are claimed to have an implicit theme. Actually, these adverbs look directional rather than locational, because in addition to an obsolete stem, they also contain the obsolete -a´/-e´ lative case suffix even if sometimes the case ending has already worn off, as in the case of fel ‘up’. In fact, however, this suffix merely serves to convey the ‘terminus’ theta-role assigned to the locative element by the verb. That is, just as in the case of piros-ra fest ‘red-to paint’, piros ‘red’ denotes the end state of the object, and -ra ‘to’ marks that this end state is in a resultative relation to the verb, in the case of az asztal-ra tesz ‘the table-on put’, or ra´-tesz ‘onto put’, az asztal ‘the room’, and the obsolete stem of ra´ ‘on’ denote the end location of the moving individual, and the -ra ‘to’ suffix of asztal, and the -a´ ‘to’ suffix implicit in ra´ express that this end location is in a terminative relation to the verb. In fact, the versions in (i) and (ii) are also grammatical: ¨ NYVEKET KO tett le az asztalra. (i) Istvań Stephen books-ACC put down the table-on ‘It was books that Stephen put on the table.’


54

(ii) ISTVAŃ tett le ko¨nyveket az asztalra. ‘It was Stephen who put down books on the table.’ A sentence with a verbal particle and a bare nominal theme can be acceptable if the verbal particle is postverbal; however, the resulting sentence is not inherently delimited: ¨ NYVEKET KO tett le az asztalra. (iii) Az elado´ 10 percig the shop-assistant 10 minute-for books-ACC put down the table-on ‘For 10 minutes, it was books that the shop assistant put down on the table.’

11

Such postverbal verbal particles are not resultative or terminative secondary predicates but directional adverbs. Particles like el ‘off’, ki ‘out of’ often denote the end location of centrifugal motion. In that case, the source of the motion can also be expressed, but the particle invariably doubles the terminus argument – see (i), unless it is replaced by it, as in (ii): ha´zbo´l (az utca´rai). (i) E´va kii -szaladt a Eve out ran the house-from (the street-into) ‘Eve (has) run out of the house (into the street).’ cf. (ii) E´va az utca´ra szaladt (a ha´zbo´l).

12

A way of neutralizing the non-specificity requirement on the theme is to focus something in the sentence, as a result of which the theme argument becomes presupposed. In this case, it can be represented by a [þspecific] noun phrase. Thus, (41a) is acceptable in the following context: (i) Jańos John

13

KE´SZI´TETTE PREPARED

mindegyik each

repu¨lo´´ge´pmodellt, air-plaine modell,

nem not

VETTE BOUGHT

ó´ket. them

If the locative particle and its associate are focussed, the examples in (59) become acceptable, for example: ¨B ˜ TT. hever kutya a KU¨SZO ELO (i) OTT there lies dog the threshold before ‘It is in front of the threshold that a dog is lying.’

Chapter 9 will solve the mystery why there is a grammaticality difference between example (58a) and example (i). Essentially, the particle and its associate in (i) are not predicated of the theme subject; they are focussed, which means that they are predicated of the open sentence represented by the rest of the sentence. That is, (i) expresses the following predication relation: (ii) [ahol where 14

15

16 17

kutya dog

hever] lies

[az it

ott, there

a the

ku¨szo¨b threshold

elo ´´tt before

van] is

In fact, there is also a landing site for wide-scope universal quantifiers in the preverbal section of the predicate phrase; this is, however, irrelevant for the discussion. Szendro ´í (2003) assumes that the particle, generated under a complex predicate, undergoes focus movement to Spec,FP in such cases. For arguments that focussing cannot take place via head movement, see Horvath (2005). The construction illustrated in (92) has been analyzed in many different ways in the literature. Farkas and Sadock (1989) analyze it in the autolexical framework, which allows different moduls of grammar to assign partially noncongruent representations to the same linguistic entity. In their analysis, the verbal particle e.g. in (92a) forms a unit with the infinitive selecting it in lexicosemantic structure, and with the compound verb consisting of the auxiliary and the infinitive in morphological structure. Koopman and Szabolcsi (2000) derive (92a) by the cyclic movement of the infinitive phrase combined with VP-extraction, which leaves the verbal particle stranded. The studies in E´. Kiss and van Riemsdijk (2004) assume various versions of cyclic, phrasal particle movement in constructions like (92a) (see e.g. Alberti’s, Brody’s, and Olsvay’s papers). E´. Kiss (2004b), on the other hand,


55

assumes the reanalysis of the VP layers into a single verb phrase, projecting a single PredP. According to Szendro ´í (2003; 2004), particle movement in (92a) is movement for stress. (Semi-)auxiliaries cannot occupy the most heavily stressed leftmost position of a predicate phrase; hence they must be preceded by a focus, a negative particle or by a verbal particle/bare nominal borrowed from their infinitival complement. Den Dikken (2004) argues that particles undergoing long movement, and those undergoing short movement have different landing sites–hence this argument against placing particles in Spec, VP is invalid.

´ R ESZES BOLDIZSA

V E RB A L PA RT I C L E S T E L I C I Z I N G S TAT I V E PS Y C H V E R B S

1.

INTRODUCTION

This chapter aims to analyze two groups of Hungarian psychological verbs and their verbal particles. After discussing some classificatory issues in section 2, I conclude that these particles play a telicising role because the situation type of the base verb they are attached to is different from that of the resulting complex verb. In section 3, I attempt to show that these verbs pose a problem for current theories of the syntax-semantics interface, as they have rather peculiar mapping properties. Section 4 provides a semantic analysis in an aspectual framework which explains the notion of mental change in terms of scale structures. The final section deals with the semantic representation of the causative forms of psychological verbs in Hungarian. 2.

CLASSIFYING PSYCHOLOGICAL VERBS IN HUNGARIAN

We can make a distinction between simple and complex verbs in Hungarian (see Chapter 2). Among the complex transitive verbs expressing psychological changes there are verbs containing the particle meg, and among these, we can distinguish between Experiencer-subject and Experiencer-object verbs. The first subgroup includes such verbs as meg-szeret ‘come to love’ or meg-e´rt ‘come to understand’, ‘grasp’, while the second subgroup ones like meg-ijeszt ‘frighten’ or meg-zavar ‘disturb’. I discuss only Experiencer-subject verbs in this paper. The verbs belonging to this subgroup can be sorted into three types depending on their meaning. (1) Hungarian Experiencer-subject verbs containing the particle meg: a. verbs expressing changes in emotional states: meg-szeret ‘come to love’, meggy u˝ lo¨l ‘come to hate’, meg-irigyel ‘become envious of’, meg-sajna´l ‘take pity on’ etc. b. verbs expressing changes in perceptual states: meg-la´t ‘spot’, meg-hall ‘come to hear’, meg-pillant ‘glimpse’ etc. c. verbs expressing changes in cognitive states: meg-e´rt ‘understand’, meg-tud ‘come to know’ etc. 57 ´ Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 57–73. Katalin E. 2008 Springer.

58

´ R ESZES BOLDIZSA

In considering the semantic role of the particle with respect to the base verb, it may be helpful to choose the Vendlerian verb classes as a starting point. Vendler (1967) classified English verbs (and verb phrases) into aspectual categories on the basis of their duration, culmination and temporal structure: they can denote states, activities, accomplishments or achievements (see Chapter 6). States have no internal structure, hence their time interval is divisible. Activities are ongoing durative events without a necessary endpoint. Accomplishments are durative events with an obligatory endpoint (culmination), while achievements are non-durative, being identical with their own instantaneous culmination. The typical achievements (e.g. reach the top) are preceded by a preparatory phase. Several authors introduce a fifth verb class whose members lack such preparatory phases (e.g. sneeze, flash etc.). Smith (1991) refers to them as semelfactive verbs, while others call them punctuals. Smith (1991) argues for a two-component theory of aspect, introducing the notions of situation aspect and viewpoint aspect (see Chapter 6). Situation aspect encodes the differences between the Vendlerian verb classes based on the temporal structure of the situations, while viewpoint aspect presents different perspectives on situations. As Smith shows, (a)telicity, a distinction which turns on whether an event has a natural endpoint or outcome, is a feature used in classifying types of situation aspect. Activities and semelfactives are atelic events, while accomplishments and achievements are telic. On the other hand, (im)perfectivity – a distinction which has to do with viewing a situation either in its totality or as a partial one – belongs to viewpoint aspect. I set aside cognitive change-of-state verbs because in my opinion they require a special analysis (though certain general observations apply to them, too), and wish to discuss only subclasses (1a) and (1b) in detail. In order to analyze their meaning, first I would like to clarify the role of the particle meg in an informal way. Members of (1a), (1b) and (1c) are often categorized as inchoative verbs which express a kind of situation aspect, since their particles are supposed to carry a special meaning related to the beginning of an event. However, I agree with Kiefer (2000a) that this is an incorrect approach, because unlike other complex verbs the verbs belonging to these classes do not express the beginning of an emotional, perceptual or cognitive state, but rather a process or an (instantaneous) change which results in a state.1 Thus, it cannot be said that they introduce any situation type. So what is the role of meg in psych verbs, if it does not introduce a particular situation type the same way as other particles (or meg in its other uses) do which slightly modify but otherwise preserve the meaning of the base verb? It cannot telicize the base verb in the sense of creating a telic state from an atelic one, as the situation type of states cannot be made telic. Still, I think there is another straightforward sense of ‘telicization’ in which these particles telicize the stative verbs, because the result of their application is a new telic verb, with an aspectual type different from that of the base verb. Filip (2000) argues that though verbal prefixes in Russian make the predicate they apply to a telic one, they cannot be markers of perfectivity because they behave like derivational morphemes, while verbal aspect in Slavic languages is a

T E L I C I Z I N G S TAT I V E P S Y C H V E R B S

59

grammatical (inflectional) category. She analyses them as so-called ‘eventuality description modifiers’ which affect the situation type of the predicate they modify, being mappings from the sets of eventualities of any type into sets of events ((PROCESS [ STATE [ EVENT) ! EVENT). In terms of Filip’s approach, I take the Hungarian verbal particles contained in psych verbs to be eventuality description modifiers, mappings of the kind STATE ! EVENT (the description EVENT covers accomplishments and achievements), and suppose that they are not markers of perfectivity, which is a type of viewpoint aspect. At this point we may ask what the aspectual classification of the verbs belonging to (1a) in terms of situation aspect is: should we consider them accomplishments, or achievements with a preparatory phase? On one approach, members of (1a) would qualify as accomplishments containing a durative phase and a culmination point in their lexical event structure. Another possibility may be that this structure does not include any duration: on this approach, they would be categorized as a kind of achievements with a presupposed preparatory phase. My proposal is that Hungarian psych verbs in (1a) can be classified as accomplishment verbs on the basis of the following contrast: Pe´ter de´lutań 4 o´ra´ra meg-tala´lta a kulcsot. PRT-find the key-ACC Peter-NOM afternoon 4-by ‘Peter found the key by 4 o’ clock in the afternoon.’ b. Pe´ter az o ´´sz ve´ge´re meg-kedvelte az iskola´t. the school-ACC. Peter-NOM the autumn end-by PRT-like ‘Peter came to like the school by the end of the autumn.’ (3) a. Pe´ter de´lutań 4-kor meg-tala´lta a kulcsot. PRT-find the key-ACC Peter-NOM afternoon 4-at ‘Peter found the key at 4 o’ clock in the afternoon.’ b. #Pe´ter de´lutań 4-kor meg-kedvelte az iskola´t. PRT-like the school-ACC. Peter-NOM afternoon 4-at ‘Peter came to like the school at 4’ o clock in the afternoon.’

(2)

a.

Meg-tala´l ‘find’ is a particleþverb combination that expresses an achievement and presupposes a preparatory phase of searching in most contexts (when it cooccurs with the time adverbial 4 o´ra´ra ‘by four o’clock’ in (2a) it certainly has this presupposition). In general, achievements can be anchored to a moment of time, independently of whether they presuppose a preparatory phase or not. (3a) shows that this is valid for the complex verb meg-tala´l. However, meg-kedvel cannot be an achievement verb because adding an adverb expressing a point of time (a ‘momentary adverbial’) as in (3b), results in semantic anomaly. (Smith (1991, p. 55) points out that in English an inceptive interpretation arises when certain accomplishments are combined with momentary adverbials, but this is not the case in the examples above). As all the verbs belonging to (1a) (in contrast to those in (1b)) behave like megkedvel with respect to modification, I conclude that (1a) should be categorized as a

60

´ R ESZES BOLDIZSA

subclass of accomplishments (while those in (1b) are obviously achievements, and those in (1c) may be either accomplishments or achievements). To sum up, I propose that we should consider the meaning of the particle meg applied to the base verbs of (1a) an ‘eventuality description modifier’ (EDM). This EDM changes the situation aspect type, resulting in an accomplishment by complementing the original state with a preceding durative phase. The subjects of the sentences containing the verbs in (1a) have the thematic role Experiencer, so they belong to the group of non-agentive accomplishment verbs. The base verbs of many accomplishments express the activity/process leading to a result (e.g., eszik ‘eat’ is the activity leading to the total consumption of some food expressed by meg-eszik valamit ‘eat up something’), while those in (1a) do not. I would like to point out two differences between accomplishments that contain a base verb expressing an activity leading to a result state and those in (1a), whose base verbs express the result state itself: (A) accomplishments with activity-expressing base verbs can generally be modified with the adverbs partially/halfway, but the members of (1a) cannot. (4)

a. Pe´ter re´szben/fe´lig megette az alma´t. the-apple-ACC Peter-NOM partly/halfway PRT-eat ‘Peter ate the apple partly/halfway.’ b. #Pe´ter re´szben/fe´lig meg-kedvelte Marit. Mary-ACC. Peter-NOM partly/halfway PRT-like ‘Peter came to like Mary partly/halfway.’

(4b) cannot be interpreted as meaning that Peter’s emotional state of liking at the end of some process was only half of a normal state of liking, because a stative eventuality does not allow such degrees. (B) While in general it is the object of transitive accomplishment verbs that ‘measures out’ the temporal course of the event (see Tenny (1994)), the verbs of (1a) are different because their subject is the measuring-out argument. We may think about the relevant change of state as the intensification of an attitude’s emotional component. Based on (A) and (B), I propose that complex Hungarian accomplishment and achievement verbs can be categorized according to the thematic role of the subject and the meaning of the base verb (i.e., whether it expresses a result state). Subject thematic role: þ/ Agent (þ/AG) Base verb meaning: þ/ result state (þ/RS) We can distinguish four groups, which might be helpful in considering their aspectual structure:


61

Groups (A) and (B) include verbs whose base verbs do not express result states. (A) RS, þAG : meg-eszik valamit ‘eat up something’, ki-takarı´t valamit ‘clean up something’ etc. (B) RS, AG: le-esik ‘fall down’, fel-emelkedik ‘rise up’, megsza´rad ‘dry out’ etc. Groups (C) and (D) include verbs whose base verbs express result states. (C) þRS, AG: (C1): transitive/ intransitive psych verbs: meg-kedvel valakit (come to like someone), meg-haragszik valakire (get angry with someone) etc. (C2) achievement verbs: fel-forr (boil up) etc. (D) þRS, þAG: inchoative accomplishment verbs: fel-a´ll (stand up), le-u¨l (sit down) etc. Before presenting a detailed semantic analysis, I discuss their so-called mapping properties to draw some conclusions about psych verbs in general. 3.

PSYCH VERBS IN MAPPING THEORIES

There are several current theories about the syntactic realization of the arguments of predicate types in natural language. These so-called mapping theories belong to the interface between syntax and lexical semantics, devised to provide different accounts of the regular patterns in the syntactic expression of arguments. These regularities are usually formulated as ‘linking rules’ that associate certain semantic roles with grammatical functions. Since the role of an argument depends on the meaning of the predicate selecting it, the latter is an important factor in determining the structure of the clause. Some scholars (e.g., Dowty (1991)) claim that it is the arguments’ protothematic properties that count, others are of the opinion that it is the aspectual value of the predicate (Tenny (1994)), while others again (e.g., Croft (1991)) consider causation as the decisive factor. The authors analyzing the mapping properties of psych verbs draw on Croft’s generalization (Croft (1986)) that partly resulted from the investigation of two main classes of English psych verbs, the fear and the frighten classes. What is surprising about these classes is that their members come in pairs that are near synonyms, but their arguments are mapped in quite different ways onto syntactic positions, as the following examples illustrate: Experiencer-subject fear-class (5) a. John fears dogs. c. John likes Mary. e. John regards Mary as pompous.

Stimulus-subject frighten-class b. Dogs frighten John. d. Mary pleases John. f. Mary strikes John as pompous.

´ R ESZES BOLDIZSA

62

Croft’s generalization amounts to this: Experiencer-subject psych verbs are always stative, while Stimulus-subject verbs can be either stative or inchoative. The Hungarian transitive psych verbs considered in this chapter constitute a counterexample to this generalization, as they express specific changes in the mental state of the Experiencer argument which is their subject, so it seems that these verbs have rather peculiar mapping properties. In what follows, I will present the difficulties with psych verbs by sketching one of the current mapping theories. Tenny (1994) proposed a theory of argument structure based on the aspectual properties of the verb. The central principle of her approach is the Aspectual Interface Hypothesis (Tenny (1994, pp. 115–116)): Aspectual Interface Hypothesis: The universal principles of mapping between thematic structure and syntactic argument structure are governed by aspectual properties. Constraints on the aspectual properties associated with direct internal arguments, indirect internal arguments, and external arguments in syntactic structure constrain the kinds of event participants that can occupy these positions. Only the aspectual part of thematic structure is visible to the universal linking principles. According to Tenny, two aspectual properties may affect the realization of arguments, namely, ‘measuring out’ and ‘delimitedness’. The latter corresponds to the property of ‘telicizing’ mentioned earlier. The following examples illustrate these notions: (6) (7)

The gardener ripened the fruit. The ball cracked the glass.

(8)

John ate the apple.

(9) (10)

Mary climbed the ladder. John hiked to the top of the hill.

Sentences (6)–(10) show that it is the entity that moves or undergoes some change of state that measures out the event (in (9) it is the length of the ladder traversed), while the goal of the movement delimits the motion event. Tenny postulates two constraints to ensure that the argument measuring out the event is expressed as a direct internal argument and the one delimiting the event as either a direct or indirect internal argument: Measuring-Out Constraint on Direct Internal Arguments: a. The direct internal argument of a simple verb is constrained so that it undergoes no necessary internal motion or change, unless it is motion or change which ‘measures out the event’ over time (where ‘measuring


63

out’ entails that the direct argument plays a particular role in delimiting the event). b. Direct internal arguments are the only overt arguments which can ‘measure out the event’. c. There can be no more than one measuring out for any event described by a verb. The Terminus Constraint on Indirect Internal Arguments: a. An indirect internal argument can only participate in aspectual structure by providing a Terminus for the event described by the verb. The Terminus causes the event to be delimited. b. If the event has a Terminus, it also has a Path, either implicit or overt. c. An event as described by a verb can only have one terminus. Tenny formulates the following constraint on external arguments: The Non-Measuring Constraint on External Arguments (NMC): An external argument cannot participate in measuring out or delimiting the event described by a verb. An external argument cannot be a Measure, a Path or a Terminus. As I noted, though the meanings of such pairs as fear vs. frighten are closely related, the Experiencer of fear is syntactically realized as the subject while the Experiencer is a grammatical object in the sentences containing frighten. This seems to contradict the Non-measuring-out Constraint which does not allow external arguments to measure out the event. However, as Tenny points out, Experiencersubject psych verbs in English do not refute the NMC because they can be interpreted only as having a static meaning, so their Theme cannot be a participant undergoing any change of state. Tenny’s aspectual mapping theory thus makes the following predictions with respect to the two classes of psych verbs: a. When the internal argument is a Theme, the sentence can only be interpreted as having a static meaning. (fear-class) b. When the direct internal argument is an Experiencer and the subject is a Theme, it is the Experiencer that measures out the event, so the sentence can be interpreted as expressing some change of state. (frighten-class) However, since Tenny accepts Croft’s generalization, we could not explain the characteristics of Hungarian psych verbs within the framework of her theory. Moreover, a more general problem arises: why can only internal arguments measure out and delimit events? In other words, is it necessary that external arguments do not have the measuring-out property? Tenny answers this question affirmatively, stressing that the external argument does not belong to the so-called event nucleus which is the aspectual structure forming part of a more general event structure. Though a verb may exactly determine

´ R ESZES BOLDIZSA

64

the change which its direct internal argument has to undergo, it does not determine what should happen to the external argument during the event. According to Tenny, the external argument’s possible changes, movements or activities are not determined by the verb at all, and these processes cannot be considered a scalar change in a single property (Tenny (1994, p. 87)). For example, if John bakes a cake, the cake undergoes a determinate change as it gets gradually baked until the end of the process. By contrast, while baking the cake, John may perform different activities in various orders. I think Tenny’s theory is inadequate in this respect. As Dowty (1991, pp. 570– 571) pointed out, it is not clear whether Tenny’s constraints are empirical generalizations or theoretical postulates. If they are contingent generalizations, they can be challenged by empirical data. However, if they are not, one should give a proper analysis of the apparent counterexamples in line with the postulates. I think this would be impractical, considering the verbs Dowty refers to. He mentions several transitive verbs in English (such as enter, exit, pierce, reach, leave and cross) whose external arguments are apparently assigned the measuring-out role: (11)

At the turtle race, the winning turtle crossed the finish line in 42 seconds.

(12)

The crowd exited the auditorium in 21 minutes.

(13–15) show that the sentences containing these verbs pass the measuring-out tests proposed by Tenny. (13)

John entered the icy water a little further.

(14) (15)

The turtle crossed the finish line slowly. a. John entered the water completely. entails b. John was completely in the water.

Being transitive, these verbs cannot be unaccusatives (by the definition of ‘unaccusative’), and it would be implausible to argue that they have two internal arguments in their lexical representation. I agree with Dowty that this would indeed be arguing in a circle to save the original postulates, since we don’t have any evidence suggesting that these verbs actually have two internal arguments. As we can see, there are some verbs that do not fit Tenny’s NMC. In this respect, the Hungarian verbs in (1a) are similar to Dowty’s examples, as they have an external argument which measures out the event. The English verbs which Dowty mentions and the Hungarian psych verbs show the limits of the aspect-based mapping theories, and also that one cannot entirely rely on Croft’s generalization in analyzing psych verbs. 4.

A S E M A N T I C A N A LY S I S

My aim in this section is to provide a compositional semantic analysis for the two verb groups (1a) and (1b), assuming that the meaning of the complex verb is jointly


65

determined by the meaning of the particle and that of the base verb. As the base verb denotes a state which the subject argument has to reach by the end of the mental event, we may reasonably assume that the meaning of the particle contributes the dynamics of the change of state. Several theories have already been proposed to deal with aspectual phenomena in different languages. Among the formal approaches, Krifka’s mereological theory (Krifka 1992) based on the part-whole relationship especially stands out. Krifka intends to explain the role of object arguments in the first place by using algebraic structures to model the relevant properties of arguments and verb meanings. He assumes that objects and events form algebraic semi-lattices with respect to the so-called sum operation. His theory captures the traditional opposition between atelic and telic situations by introducing the notions of cumulative and quantized reference types, which can be given an exact algebraic representation. In general, we can characterize the relevant predicates of physical objects or events in these terms. Nominal phrases with cumulative reference type (i.e., mass nouns like wine) denote both the parts and the sum of two quantities of matter. By contrast, NPs with quantized reference cannot be truly predicated of the parts of the entities belonging to their extensions: the proper parts of a glass of wine cannot be called glasses of wine. We can make the same distinction among event predicates, too: run has cumulative reference, so it is atelic, while run three miles has quantized reference and is telic. In the case of VPs with a quantized Theme argument (e.g., drink a glass of wine) the parts of the event and those of the Theme can be brought into correspondence with each other (within algebraic semantics, this part-by-part correspondence is expressed by a special function known as a homomorphism). The VP is quantized (telic) in such cases because the Theme argument is itself quantized, so for instance, we can consider the time by which the Theme argument, e.g., a glass of wine is totally consumed as the endpoint of drinking a glass of wine. An important alternative to the mereological theories of aspect is the scalebased analysis (Hay et al 1999). A scale is a set of points ordered along a given dimension. This dimension can be length, volume, temperature, brightness, etc. This approach does not require a part-by-part correspondence (homomorphism) between arguments and events, because it assumes that it is not the Theme that undergoes the change but some of its properties (volume, area, etc.). This change can be characterized with the help of a scale. The aspectual classification of a sentence depends on whether the degree of change – the so-called difference value – is quantized (delimited) or not. I can see two advantages of this analysis as compared to the mereological aspectual theories. First, besides explaining the changes in object arguments, it can provide an account of the changes in subject arguments, too (e.g., in the arguments of certain psych verbs). Second, it accounts for sentences like (16): (16) Peter’s cheeks reddened in ten seconds.

66

´ R ESZES BOLDIZSA

Sentence (16) poses a problem for Krifka’s mereological theory: the parts of Peter’s cheeks cannot be made to correspond to the parts of the reddening event because usually all or most of the parts are reddening simultaneously. However, within the scale-based approach this change can be explained by representing the intensity of the relevant property (redness) on a scale. Ramchand (1997) points out that there are other types of Themes (or Patients) besides those of verbs of creation or consumption. She designates these Patient arguments using the labels Patþ/ and PatLOC. Objects of type Patþ/ are the arguments of the so-called degree achievement verbs like the transitive lengthen, widen, and the intransitive cool. (17) shows that the verbs belonging to this class are aspectually ambiguous: (17) The soup cooled (in an hour/for an hour). Hay, Kennedy and Levin explain this by assuming that the hearer infers the difference value by drawing a conversational implicature, because the degree of change is not given explicitly in the meaning of the sentence. The telic reading provides the most informative interpretation of the sentence because a certain delimited degree is available in most contexts, namely the room temperature. By contrast, the durative time adverbial cancels this implicature, so the sentence may also receive an atelic reading. However, if we add an adverb denoting an explicit degree (e.g., completely), the aspectual ambiguity disappears and only the telic reading remains. In my opinion, the scalar analysis of Hay, Kennedy and Levin may provide a general theoretical framework for the analysis of the verbal particles in Hungarian. It seems that many Hungarian verbal particles signal that the change reached a scalar limit, that is, they can delimit the scalar difference value which is relevant for the given change. As this value is often contextually given, this fact should be reflected in the meaning representation of the particles, which means that their meaning should include the scale structure on which the contextually provided value can be reached. For example, the particle ki (out) in (18) presumably refers to the fact that the color of Peter’s cheeks reached a certain (conventionally given) degree. (18)

Pe´ter arca kipirult. Peter-NOM cheeks PRT-redden. ‘Peter’s cheeks reddened.’

I turn to presenting the formal details of the scalar analysis. According to Kennedy and Levin (2002), the meaning of an intransitive verb expressing a gradable change can be characterized by using the INCREASE relation: (19)

V1¼‚x‚d‚e.INCREASE(GV(x))(d)(e)


67

INCREASE is a three-place relation between the Theme argument x characterized by the gradable property (GV) associated with the verb, the degree-of-change d and the change-of-state event e. It expresses that the degree of the gradable property of the Theme increased by degree d during the change. dA scale is defined as an ordering on a set of values. In analyzing psych verbs, let us assume that in their case the scalar structure (S) is the partially ordered set of the intensity of certain selected psychological states represented by points p1, . . . ,pn. The degrees (d) are the intervals on the scale. We can define the set of positive degrees (POS (S)) corresponding to the change of intensity from a lesser to a greater degree as (20) shows. (20)

POS(S) ¼ {d S j $p12d 8p22 S [p2 < p1 ! p22d]}

The instances of the scheme d (where i is a positive real number) denote the positive degrees. The addition of degrees is defined in (21): (21)

d þ d ¼ d< 0, i þ j>

(22) gives the difference value: (22) jjINCREASE(j(x))(d)(e)jj ¼ 1 iff j(x) (SPO(e)) þ d ¼ j(x) (EPO(e)) Here SPO is the starting point of the ‘run-time’ (temporal interval) of event e involving a change of state, while EPO is the endpoint of the run-time. (23) gives the truth conditions of (22) in detail: jjINCREASE(j(x))(d)(e)jj¼1 $ the degree to which x has property j at the beginning of the event increased with degree d results in the degree to which x has property j at the end of the event e. (23) and (25) show the application of the INCREASE relation:

(22)

(23) John lengthened the rope. (23)

a. (Paraphrase): John caused the length of the rope to increase by some amount.

The non-causative part of the meaning of (24) can be represented by (25): (24)

$e,d[ INCREASE(long (rope))(d)(e)]

The difference value d in (25) is existentially bound by default. Hay, Kennedy and Levin distinguish four ways of specifying the difference value: (A) The difference value is explicitly expressed in the sentence

´ R ESZES BOLDIZSA

68

(A1) by quantized measure phrases: (26) The workers are widening the road 5 meters. (27) The lake cooled 4 degrees in four days. (A2) by non-quantized measure phrases: a bit, a quantity, a part (28) The soup is cooling a bit. (A3) by scalar adverbs: among them, the maximizing ones (completely, totally, halfway) have a quantizing effect, as they indicate that the change reached a certain limit on the scale. By contrast, minimizing adverbs express that the change cannot go beyond a given limit (slightly, partially, somewhat). (B) The difference value can be lexically inferred in the case of verbs derived from closed-scale adjectives, since it is given in the semantic characterization of the adjective. The adjectives empty, straight and dry are closed-scale ones, because their associated scales have a maximal value, while long, wide and short and the corresponding derived verbs are open-scale ones. (C) The difference value can be inferred from the context: (29)

The tailor lengthened Peter’s pants.

In (29), it is the context of the utterance that provides the difference value, which is based on the average length of Peter’s pants. (D)

In the ‘Elsewhere case’ when neither a measure phrase nor the scalar properties of the predicate nor the context provides the difference value, Hay, Kennedy, and Levin assume that it is existentially bound in the formula, so we get a nonquantized, atelic predicate. A relevant example is the verb lengthen in (24), which is atelic in its default interpretation.

According to my proposal, in case of the verbs in (1a), the particle meg creates accomplishment verbs from stative ones, while it creates achievements from stative base verbs in the case of group (1b) I think it is plausible to suppose that emotional states come about as a result of an event involving a mental change of state whose endpoint is the emotional state denoted by the base verb. The events expressed by verbs like meg-szeret ‘come to love’ or meg-sajna´l ‘come to take pity on’ involve the intensification in the emotional component of a sympathetic attitude, while those expressed by e.g., meg-gy u˝ lo¨l ‘come to hate’ or meg-uta´l ‘come to loathe’ involve the intensification of an antipathetic attitude. Though I will use the scale-based analysis to represent the meaning of these sentences, I see no obstacle in complementing it with some definitions widely used in


69

the mereological approach. Following Krifka (1992), I suppose that the domains of eventualities and objects form separate semi-lattices structured by an idempotent, commutative and associative sum operation (t). The extensions of the predicates ST, T and E contain states, time intervals and events respectively. The following definitions are needed for the analysis: (30)

8x,y[xvy $ xty¼y] (part relation (partial order))

(31) (32)

8x,y[x@y $ xvy : x¼y] (proper part (strict partial order)) 8x,P[ATOM(x, P) $ P(x) : $y[y@x P(y)] (x is a P-atom)

That is, x is a P-atom with respect to predicate P if and only if it has property P but has no proper parts with property P. (33)

8P[ATOM(P) ! 8x [P(x) ! $y[y v x ATOM(y, P)]]] (P has atomic reference)

Predicate P has atomic reference if and only if each entity x belonging to its extension has an atomic part. (34)

ATOM(T) 8t[Ta(t) $ ATOM(t, T)]

The set of time intervals has atomic reference: the set of atoms (Ta) includes the time points. (35)

8e, t[EPO(e)¼t $ Ta(t) t v(e) 8t’ [t’ v(e) ! t’ t ]]

The function EPO assigns the endpoint of their runtime to eventualities. is a linear order (temporal precedence/identity) on the set of intervals. Following Pinõń (2001a), we can define the following relations on the unified domain of events and states: (36)

Complete temporal precedence (): a strict partial order

(37)

Immediate temporal precedence: v v’¼def v v’ :$ v’’[v v’’ v’’ v’] Final proper part: v @fin v’¼def v @ v’ : $ v’’ [v’’@ v’ v v’’]

(38)

The proper part of an eventuality is a final proper part of it if the proper part starts later but ends at the same time as the eventuality. According to Pinõń (2001a), accomplishment and achievement verbs both express changes in certain states, so he refers to them with the general term ‘Change-of-State predicates’. He defines the notion of change of state as

´ R ESZES BOLDIZSA

70 follows: (39)

Change-of-State(e, x, s, P) ¼ def Theme(e, x) e s P(s) Theme(s, x) 8e’ [e’@ e : (e’ @fin e) ! : $s’ [e’ s’ P(s’) Theme(s’, x)]]

A change of state is a four-place relation between events, ordinary objects, states, and types of states. An object x comes to be in a state s of type P by virtue of e just in case x is the Theme of e, e immediately precedes s, s is of type P, x is the Theme of s, and no proper part e’ of e that is not also a final proper part of e is immediately followed by a state s’ of type P of which x is the Theme. The verbs in (1a) and (1b) denote a mental change-of-state (MCS) which is a variant of the change-of-state relation. I define a stative base verb as a five-place relation between Experiencers, Themes/Stimuli, states, times (endpoints) and degrees. (40) MCS(e, x, y, s, P,d ) ¼def Exp(e, x) (Theme(e, y) | Stimulus(e, y)) P(y)(x)(s)(EPO(e)) (d) 8e’[(e’@e : (e’@fin e)) ! 8s’8t : [e’ s’ P(y)(x)(s’)(t)(d)]] Definition (40) is intended to cover psych verbs whose object argument can be either Theme ((1a)) or Stimulus ((1b)). We have to introduce a further distinction concerning the relation between psychological states and the verbal predicates (the base verbs of the verbs in (1a)) that can be applied to them. I suppose that the mental state has its characteristic property (e.g., liking) right from the beginning of the change (which distinguishes it from other psychological states that the Experiencer may undergo during the relevant time interval), so the so-called mental change-of-state is not really a change from one state to another but rather a change in the degree of intensity of the relevant state which is assumed to be present all along. Still, for some pragmatic reasons we cannot apply the corresponding verbal predicate properly until this degree has reached a certain value. This is presumably due to a more general restriction on the felicitous use of natural language predicates, which could adequately be spelled out in a detailed analysis of vagueness. Let us assume a closed scale with bounds 0 and 1, and let us make a distinction between the characterizing predicate and the properly applicable verbal predicate. The latter can be properly applied only if the intensity of the state reached a certain (relatively high) degree, let us say 0.9 on the scale (of course, it is an idealization because this limit is vague.) I assume that in the initial situation the characterized state has a minimal non-zero degree so that the characteristic property can be predicated of it, though I will not introduce it into the formal representation for the sake of simplicity. The degrees between 0.9 and 1 are ones where the intensity may increase further, as the following

71


sentence shows: (41)

Jańos egyre inka´bb meg-kedvelte John-NOM more and more PRT-like ‘John came to like Mary more and more.’

Marit. Mary-ACC.

This sentence is not about how the state of liking originally arose: it presupposes that John already likes Mary and asserts that the degree of his emotion increased further. To represent the meaning of the particle meg contained by the verbs in (1a), a modified version of the INCREASE relation is needed. The re-defined INCREASE is a three-place relation between states, degrees and events: (42)

jjINCREASE (s) (d ) (e)jj¼1 iff the intensity of state s increases by degree d during the event e.

Second, we have to build the scalar limit value into the meaning of the particle. (43) shows the meaning of the particle meg applicable to the verbs of (1a). (43)

[meg1 Part] ) ‚P,y,x,e$s,t0[P(y)(x)(s)(t0) (d ) INCREASE(s)(d )(e) MCS(e,x,y,s,P,d ) 0.9 j]

where d is a degree by which the intensity of a state s has to increase in order to reach or exceeda certain limit (0.9) from whichthe verbalpredicate P can beproperlyapplied to s. Figure 1 illustrates three possible mental change-of-state events, each starting at t0 and ending within the range between degrees 0.9 and 1 at t1, t2 and t3 respectively:

t1 t2 t3 d

d

d

MCS events The denotation of the transitive verb kedvel ‘like’ is a five-place relation: (44)

[kedvelV] ) ‚y,x,s,t,d[Liking(s) Exp(s,x) Theme(s,y) Hold(s,t,d)]

where the relation Hold expresses that a state of intensity d holds at time t. Thus we can compositionally derive the meaning of the verb meg-kedvel ‘come to like’ from the meanings of the particle and the base verb. (45)

[[meg1 Part][kedvel V] V] ) ‚y,x,e$s, t0[Liking(y) (x) (s) (t0) (d)

´ R ESZES BOLDIZSA

72

INCREASE (s) (d)(e) MCS (e, x,y, s, ‚y,x,s,t,d[Liking(s) Exp(s,x) Theme(s,y) Hold (s,t,d)], d ) 0.9 j] The verb meg-kedvel ‘come to like’ is a three-place relation between Experiencers, Themes and events, denoting events that lead to a certain degree of a result state s by an increase in the intensity of this state. I take the particle meg as attached to the base verbs of those in (1b) to denote atomic mental changes of state (punctual eventualities), so it has a meaning which is different from that of meg1. Moreover, the definition of MCS has to be modified, because the notion of intensity cannot be applied to perceptual mental states, so the degree argument is unnecessary. (46–48) show how the meaning of the verb meg-la´t ‘spot’ can be compositionally derived from the base verb la´t ‘see’: (46) (47) (48)

[meg2 Part] ) ‚P,y,x,e$s[ATOM(e,E) MCS(e,y,x,s,P)] [la´t V] ) ‚y,x,s[Seeing (s) Exp (s, x) Stim(s, y)] [[meg2 Part] [la´t V]V] ) (y, x, e $s [ATOM(e,E) MCS(e, x,y,s, ‚y,x,s [Seeing(s) Exp(s,x) Stim(s,y)])]

Due to their atomic, punctual event structure, the verbs in (1b) cannot co-occur either with unbounded (durative) or with bounded (time-span) temporal adverbials. In their case, the role of the particle is restricted to the introduction of the atomic change of state. 5.

T H E R O L E O F P S Y C H O L O G I C A L C A U S AT I O N

In this section I will briefly outline a problem needing further research. We have already seen that within the scale-based approach the Causer role is not considered directly relevant to the scalar semantics of change-of-state verbs. However, in Hungarian the causative -tAt suffix can be applied to psych verbs expressing changes in emotional states, so we should take into account a meaning component of mental causation as well. Komlo´sy (2000) distinguishes between two basic types of causative verbs: in the case of the verbs belonging to the so-called ‘causing’ type, causation amounts to some kind of physical manipulation: (49)

(49)

a. Mari tisztı´tja a ruha´t. Mary-NOM cleans the clothes-ACC ‘Mary is cleaning the clothes.’ b. Mari forgatja a kereket. Mary-NOM turns the wheel-ACC ‘Mary is turning the wheel.’

The other causative verb group (the so-called ‘factitives’) includes verbs whose base verbal predicate expresses the action of an Agent, while the newly introduced Agent (or non-agentive Causer) forces the original Agent to perform the action.


(50)

Mari levelet ´ırat Mary-NOM letter-ACC write-CAUS ‘Mary has Peter write a letter.’

73

Pe´terrel. Peter-with.

As Komlo´sy notes, the meaning of meg-kedveltet is not ‘to force someone to like someone’, but ‘to make it happen that someone likes someone’. Consider (51): (51)

Anna meg-kedveltette Ann-NOM PRT-like-CAUS ‘Ann made Peter like John.’

Jańost John-ACC

Pe´terrel. Peter-with.

This means that psychological verbs belong to the causing type, but the relevant states come about as a result of some ‘psychological manipulation’ instead of physical intervention. At this point, it may seem unclear whether a causal relation of this kind can be adequately represented in the formalism with the CAUSE relation as (e.g.) in Dowty (1979) (or with its modified version applying to events), or whether we have to introduce another special relation to account for fact causation, because it may seem that the Causer arguments of psych verbs can be states, events and even facts, which contrast them with cases of purely physical manipulation. (52)

Az (a teńy) hogy Mari beteg, meg-ijesztette Jańost. The (fact) that Mary-NOM ill PRT-frighten-CAUS John-ACC. ‘The fact that Mary is ill frightened John.’

At first sight, sentence (52) seems to justify the assumption that the Causer arguments of the mental CAUSE relation may be facts besides events and states. (I should add that the relation CAUSE has only propositional arguments in Dowty’s original definition.) It might be thought that this would necessitate the reinterpretation of thematic roles (in order to introduce the category of facts as Causers), but this would contradict the highly plausible claim that facts are causally inert, not being able to exercise any causal effect directly. At most, a fact may affect a person’s behavior only after he grasped it and the subsequent mental process influences his other mental states. This can be adduced as an important argument if we are to insist that psychological causation should be uniformly represented as a relation between eventualities in all cases. NOTE 1

Kiefer (2000a: p. 291) argues that the function of the particle meg of psych verbs cannot be that of introducing situation aspect. According to him, in the case of meg-szeret and similar psych verbs, the particle meg cannot introduce inchoative or resultative situation aspect, because it does not meet one of his criteria for introducing that aspect type: the meaning component introduced by the particle does not simply modify but instead essentially transforms the core meaning of the static base verb.

Ń Õ CHRISTOPHER PIN

D E F I N I T EN E S S E F F E C T V E R B S

1.

DIAGNOSING DEFINITENESS EFFECT VERBS

Hungarian has a sizable class of verbs that exhibit a so-called definiteness effect (DE). The verbs belonging to this class (DE-verbs) display certain properties that clearly distinguish them from related verbs with similar lexical content. Descriptively, DEverbs are verbs of change which express that something comes into existence or becomes available in a particular fashion.1 The aim of any account of DE-verbs in Hungarian should be to make this sort of characterization more precise, because not every verb of change which expresses that something comes into existence or becomes available in a particular fashion is a DE-verb. In this section, I will discuss three general prohibitions that DE-verbs exhibit before turning to my account of them in section 2. 1.1.

Prohibition against definites

Perhaps the most striking feature of DE-verbs is that they do not permit their internal argument to be realized as a definite noun phrase just in case the clause that they head lacks a focused constituent2 and the aspectual value of the clause is an accomplishment or an achievement:3 (1)

a.

b. (2)

a.

b. (3)

a.

b. (4)

a.

Sa´ra ´ırtde egy levelt. Sarah wrote a letter.ACC ‘Sarah wrote a letter.’ a levelt. #Sa´ra ´ırtde Sarah wrote the letter.ACC Rebeka festettde egy kepet. Rebecca painted a picture.ACC ‘Rebecca painted a picture.’ kepet. #Rebeka festettede a Rebecca painted the picture.ACC Dańiel tala´ltde egy toja´st. Daniel found an egg.ACC ‘Daniel found an egg.’ tojast #Dańiel tala´ltade a Daniel found the egg.ACC E´rkezettde egy vende´g. arrived a guest ‘A guest arrived.’ 75

Katalin E´. Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 75–90. 2008 Springer.


76 b. #E´rkezettde arrived

a the

vendeg. guest

The natural way of correcting the (b)-sentences in (1)–(4) would be to employ a corresponding verb with a verbal particle (which is typically meg ‘PRT’, akin to a perfective verbal prefix; see chapter 2): (5)

a. Sa´ra meg-ı´rta a levelet. Sarah PRT-wrote the letter.ACC ‘Sarah wrote the letter.’ b. Rebeka meg-festette a ke´pet. Rebecca PRT-painted the picture.ACC ‘Rebecca painted the picture.’ c. Dańiel meg-tala´lta a toja´st. the egg.ACC Daniel PRT-found ‘Daniel found the egg.’ d. Meg-e´rkezett a vende´g. PRT-arrived the guest ‘The guest arrived.’

If the clause in question has a focused constituent or if it is understood as an activity, then (what superficially appears to be) a DE-verb can appear with a definite noun phrase that realizes its internal argument, as seen in (6a) and (6b), respectively, for festette ‘painted’:4 (6)

a. REBEKA festette a ke´pet (e´s nem SA´RA) REBECCA painted the picture.ACC and not SARAH ‘It was Rebecca who painted the picture (and not Sarah).’ b. Rebeka ege´sz de´lutań festette a ke´pet. Rebecca whole afternoon painted the picture.ACC ‘Rebecca painted the picture the whole afternoon.’

In the case of achievements (e.g., (3) and (4)), an activity reading is excluded for independent reasons, which leaves focusing as an option: (7)

toja´st. DAŃIEL tala´lta a DANIEL found the egg.ACC ‘It was Daniel who found the egg.’ b. #Dańiel ege´sz de´lutań tala´lta a Daniel whole afternoon found the a.

toja´st.’ egg.ACC

The most straightforward assessment of these data is that the verb forms ´ırta ‘wrote’, festette ‘painted’, tala´lta ‘found’, and e´rkezett ‘arrived’ in (5)–(7) are not instances of the DE-verbs ´ırde ‘write’, festettede ‘paint’, tala´ltade ‘find’, and

DEFINITENESS EFFECT VERBS

77

e´rkezikde ‘arrive’, respectively, but rather of corresponding non-DE-verbs (as already suggested by the lack of the superscript ‘de’). Although these corresponding non-DEverbs would be morphologically indistinguishable from the DE-verbs, they would differ semantically, lacking a DE-meaning. If correct, a scenario emerges in which the meaning of a DE-verb and a counterpart with a verbal particle are both based on or derived from the meaning of a corresponding non-DE-verb, as schematized in (8a) and (8b) for fest ‘paint’. Arguably, the activity reading of fest ‘paint’ (festact; see (6b)) is also derived, as is suggested in (8c). In this scenario, there is no direct connection between a DE-verb, a counterpart with a verbal particle, and its activity counterpart – they are related only via the corresponding non-DE-verb. (8)

a. fest b. fest c. fest

(‘paint’) (‘paint’) (‘paint’)

,! ,! ,!

festde meg-fest festact

(‘paint’) (‘PRT-paint’) (‘paint’)

The alternative would be to say that there is a single DE-verb (e.g. festde ‘paint’) whose DE-meaning ‘disappears’ or is ‘neutralized’ in certain contexts (e.g. in megfest ‘PRT-paint’ and in (6)), but not only does this seem less promising, it has also never been worked out. Strictly speaking, the prohibition against definites with DE-verbs does not rule out specific indefinites, though one has to be clear about which sense of ‘specific indefinite’ is at issue. If the speaker simply has a particular referent in mind (which is the epistemic construal), then DE-verbs are compatible with specific indefinites in this sense, as illustrated for hozde ‘bring’ in (9c) (see also Bende-Farkas 2001, p. 57): (9)

Sa´ra hozottde egy lańyt a bulira. sarah brought a girl.ACC the party.to ‘Sarah brought a girl to the party.’ lańyt a bulira. b. #Sa´ra hoztade a Sarah brought the girl.ACC the party.to a bulira, akit c. Sa´ra hozottde egy (bizonyos) lańyt Sarah brought a certain girl.ACC the party.to who.ACC re´go´ta ismerek. for-a-long-time know.I ‘Sarah brough a certain girl to the party who I’ve known for a long time.’ a.

Less straightforward is the question of whether DE-verbs are compatible with specific indefinites in En¸c ’s (1991) sense of a contextually familiar set of individuals from which a novel individual is chosen. As Bende-Farkas (2001, p. 56) points out, a discourse such as the following is incoherent, which suggests that the answer is no: (10)

a. Ke´t dia´k elte´vedt. two student got-lost ‘Two students got lost.’


78 b. #Jańos John

tala´ltde egy found a

lańyt. girl.ACC

The girl that John finds in (10b) cannot (intentionally) be one of the students that gets lost in (10a). At the same time, however, overt partitives seem to be much more acceptable: (11)

az elte´vedt dia´kok ko¨zu¨l. a. ?Jańos tala´ltde egyet students from-among John found one.ACC the lost ‘John found one of the lost students.’ a meghı´vott vende´gek ko¨zu¨l. b. ?Sa´ra hozottde egyet guests from-among Sarah brought one.ACC the invited ‘Sarah brough one of the invited guest.’

An interfering factor in (10) is that the meaning of tala´lde ‘find’ (in contrast to that of meg-tala´l ‘PRT-find’) excludes the possibility of a previous search, and yet the coherence of the discourse in (10) would require at least a modest search by John for the lost students. Note that if one is willing to construe John’s finding of a girl in (10b) as purely accidental and as unrelated to the fact that two students get lost in (10a), it is possible that the girl John finds in (10b) just happens to be one of the lost students. Such an accidental finding is apparently easier to construe in (11a), where there is less pressure to accommodate a preceding discourse (precisely because there is no preceding discourse). In view of these considerations, probably the safest strategy for the time being is not to build into the meaning of DE-verbs en bloc the prohibition that the novel referent they introduce may belong to a contextually familiar set of individuals, since this would run the risk of ruling out even accidental inclusion in such a set. 1.2.

Prohibition against strong quantifiers

DE-verbs do not allow their internal argument to be realized as a (strongly) quantificational noun phrase: a. *Rebeka minden ke´pet festettde. Rebecca every picture.ACC painted b. Rebeka minden ke´pet meg-festett. Rebecca every picture.ACC painted ‘Rebecca painted every picture.’ (13) a. Sa´ra minden alma´t evettde. Sarah every apple.ACC ate b. Sa´ra minden alma´t meg-evett. Sarah every apple.ACC PRT-ate ‘Sarah ate every apple.’ (14) a. *E´rkezettde minden vende´g. arrived every guest (12)


b. Meg-e´rkezett minden PRT-arrived every

79

vende´g. guest

Other strong quantifiers such as a legto¨bb ‘most’ and mindegyik ‘each’ are also excluded, but these are morphosyntactically definite, because they trigger the definite object conjugation:5 (15)

a. *Sa´ra Sarah b. Sa´ra Sarah ‘Sarah

mindegyik alma´t ettede. each apple.ACC ate mindegyik alma´t meg-ette. each apple.ACC PRT-ate ate each apple.’

Observe that DE-verbs are compatible with indefinite weak quantifiers such as numerals and sok ‘many, much’:6 (16)

a. Rebeka festettde ha´rom ke´pet. Rebecca painted three picture.ACC ‘Rebecca painted three pictures.’ b. Dańiel tala´ltde sok toja´st. Daniel found many egg.ACC ‘Daniel found many eggs.’

An interesting exception to the prohibition against strong quantifiers is the determiner mindenfe´le ‘every kind of’, which is acceptable with DE-verbs: (17)

a. Rebeka mindenfe´le ke´pet festettde. (cf. (12)) Rebecca every-kind-of picture.ACC painted ‘Rebecca painted every kind of picture.’ b. Sa´ra mindenfe´le alma´t evettde. (cf. (13a)) Sarah every-kind-of apple.ACC ate ‘Sarah ate every kind of apple.’ vende´g. (cf. (14a)) c. E´rkezettde mindenfe´le arrived every-kind-of guest ‘Every kind of guest arrived.’

These examples suggest that it is not universal quantification per se that is responsible for the unacceptability in the (a)-sentences of (13)–(15) but rather the sorts of entities that are universally quantified over. 1.3.

Prohibition against wide-scope indefinites

DE-verbs do not allow an indefinite noun phrase that is linked to their internal argument to take scope over negation (cf. (1)–(4)): (18)

a. *Sa´ra Sarah

egy nem ´ırtde not wrote a

levelet. letter.ACC

80


b. *Rebeka nem festettde egy ke´pet. Rebecca not painted a picture.ACC c. *Dańiel nem tala´ltde egy toja´st. Daniel not found an egg.ACC d. *Nem e´rkezettde egy vende´g. not arrived a guest The natural way of repairing these sentences would be to drop the indefinite article: (19)

a. Sa´ra nem ´ırt levelet. Sarah not wrote letter.ACC ‘Sarah didn’t write a letter.’ b. Rebeka nem festett ke´pet. Rebecca not painted picture.ACC ‘Rebecca didn’t paint a picture.’ c. Dańiel nem tala´lt toja´st. Daniel not found egg.ACC ‘Daniel didn’t find an egg.’ d. Nem e´rkezett vendeg. not arrived guest ‘A guest didn’t arrive.’

However, as suggested by the lack of the superscript ‘de’, the verbs in (19) are not DE-verbs, because the positive versions of these sentences involve semantically and syntactically incorporated complements which are semantically unspecified for number (where ‘unspecified for number’ means one or more): (20)

a. Sa´ra levelet ´ırt. Sarah letter.ACC wrote ‘Sarah wrote letters (lit.: letter-wrote).’ b. Rebeka ke´pet festett. Rebecca picture.ACC painted ‘Rebecca painted pictures (lit.: picture-painted).’ c. Dańiel toja´st tala´lt. Daniel egg.ACC found ‘Daniel found eggs (lit. egg-found).’ d. Vende´g e´rkezett. guest arrived ‘Guests arrived (lit. guest-arrived).’

Another way of repairing the sentences in (18) is to append the element sem ‘neither’ to the indefinite: (21)

egy levelet a. Sa´ra nem ´ırtde Sarah not wrote a letter.ACC ‘Sarah didn’t write a single letter.’

sem. neither


81

b. Rebeka nem festettde egy ke´pet sem. Rebecca not painted a picture.ACC neither ‘Rebecca didn’t paint a single picture.’ sem. c. Dańiel nem tala´ltde egy toja´st Daniel not found an egg.ACC neither ‘Daniel didn’t find a single egg.’ d. Nem e´rkezettde egy vende´g sem. not arrived a guest neither ‘A single guest didn’t arrive.’ Observe that indefinites headed by numerals pattern more generally like those headed by egy ‘a’ but with the difference that in this case the option of deleting the numeral is not available as a repair strategy (cf. (19)): (22)

a. *Rebeka Rebecca b. Rebeka Rebecca ‘Rebecca

nem not nem not didn’t

(cf. (16a)) festettde ha´rom ke´pet. painted three picture.ACC festettde ha´rom ke´pet sem. painted there picture.ACC neither paint three pictures at all.’

The difficulty in (18) and (22a) seems to be that the indefinites cannot remain in the scope of negation for some reason7 and yet they cannot take wide scope over negation either, because the DE-verbs prohibit this. In (19), (21), and (22b), the indefinites have been altered so that they can remain in the scope of negation, which renders the sentences acceptable. In this connection, notice that the indefinites in (18) and (22a) must take wide scope over negation with the corresponding verbs with a verbal particle from (5):8 a. Sa´ra nem ´ırt meg egy levelet. (cf. (18a)) letter.ACC Sarah not wrote PRT a ‘There was a letter that Sarah didn’t write.’ b. Rebeka nem festett meg egy ke´pet. (cf. (18b)) picture.ACC Rebecca not painted PRT a ‘There was a picture that Rebecca didn’t paint.’ c. Dańiel nem tala´lt meg egy toja´st. (cf. (18c)) Daniel not found PRT an egg.ACC ‘There was an egg that Daniel didn’t find.’ d. Nem e´rkezett meg egy vende´g. (cf. (18d)) guest not arrived PRT a ‘There was a guest that didn’t arrive.’ (24) Rebeka nem festett meg ha´rom ke´pet. (cf. (22a)) picture.ACC Rebecca not painted PRT three ‘There were three pictures that Rebecca didn’t paint.’

(23)

These data support the view that the reason why the indefinites in (18) and (22a) must take wide scope over negation is independent of the phenomenon of DE-verbs.


82

The role played by DE-verbs in this matter is that they do not permit the indefinites to take scope over negation, in contrast to the corresponding verbs with a verbal particle, which do. 2.

A N A LY Z I N G T H E D E F I N I T E N E S S E F F E C T

As already hinted at in section 1.1, a leading idea of the analysis to follow is that verbs which participate in the definiteness effect are systematically polysemous in that their definiteness effect meaning is simply one of two or more meanings that they may have (e.g., recall (8)). The convention adopted was that if V is a verb that participates in the definiteness effect, then Vde is the syntactic representation of V on its definiteness effect meaning. I speak of ‘systematicity’ because the polysemy in question is not an idiosyncratic property of certain individual verbs but is rather always a more general property of a class of verbs. Moreover, the polysemy is syntactically conditioned in that the definiteness effect meaning is available only if certain syntactic conditions are satisfied. In this section, I will first touch upon the syntactic side of the analysis before turning to its semantic side. There is by now a sizeable literature on the definiteness effect in Hungarian, beginning with Wacha (1978, chapter 4) and continuing with Szabolcsi (1986), Harlig (1989, chapter 5), Szabolcsi (1992, chapter 4), Bende-Farkas (1995), Ka´lmań (1995), E´. Kiss (1995), Maleczki (1995), Alberti (1998), Bende-Farkas (2001), Maleczki (2001), Ka´lmań and Varasdi (2005), and Maleczki (2005) (and this list is not necessarily exhaustive). While it would take a separate work to discuss the various proposals (and such a discussion would be complicated by the fact that certain authors continue to revise their earlier proposals), it is only recently that formally explicit analyses of the definiteness effect have begun to appear (most notably, Bende-Farkas 2001 and Ka´lmań and Varasdi 2005). The present approach is probably closest in spirit to Szabolcsi’s (1986) original treatment, which was informally cast in Dowty’s (1979) framework, though it aims to disencumber her notion of EXIST of real existence and to substitute for it the dynamic semantic notion of the introduction of a novel discourse referent.9 2.1.

Two syntactic conditions

Basically, there are two syntactic conditions on DE-verbs. The first is that the clause that a DE-verb appears in be neutral, i.e., lack a focused constituent in preverbal position; and the second is that the direct internal argument10 of a DE-verb be realized by a DP, i.e., by a noun phrase that cannot be syntactically incorporated. On the polysemy hypothesis, the first condition amounts to saying that if a verb has a DE-meaning, then it cannot have that meaning in a non-neutral clause (which, of course, still allows it to appear with a non-DE-meaning in a non-neutral clause). If the model of Hungarian syntax that one adopts requires the verb to occupy different positions in neutral and non-neutral clauses, then then this first condition can be stated as a positional constraint. For example, in E´. Kiss model of chapter 9 the verb moves to Pred in neutral clauses and but moves to an even higher functional

83


projection in non-neutral clauses, thus the first condition in her approach would be tantamount to saying that DE-verbs have to occupy Pred . However, in a model that attributed the same position to the verb in both neutral and non-neutral clauses, explicit reference would have to be made to the lack of a focused constituent. Adopting E´. Kiss’s model for concreteness, the first syntactic condition on DE-verbs could be formulated as the following constraint: (25)

Vde: [Pred0 [Pred Vde i ] [VP [V ti] . . . ]]

This constraint requires that a DE-verb appear in Pred , having raised there from V. As pointed out above, if the verb raises no further than Pred in E´. Kiss’s approach, then the clause in question must be neutral, i.e., there cannot be a focussed constituent in the specifier of a higher projection, hence this information does not have to be encoded separately in (25). For an example of this constraint in action, consider the relevant part of the structure assigned to the sentence in (1a): (26)

Sa´raj [PredP [Pred0 [Pred ‘Sarah wrote a letter.’

´ırtde i ]

[VP [V ti]

egy

levelet tj]]]

As mentioned at the outset, the second syntactic condition is that the direct internal argument of a DE-verb be realized by a DP. An explicit statement of this condition would require reference to the link between argument structure and syntactic subcategorization, which I do not explicitly discuss here, thus the following formulation will have to suffice for present purposes: (27)

The direct internal argument of a DE-verb is syntactically realized as a DP.

Assuming that DPs cannot be syntactically incorporated, this constraint rules out the possibility that the direct internal argument of a DE-verb is syntactically incorporated. In Hungarian, a syntactically incorporated constituent appears unfocused in preverbal position without a determiner (recall (20)). In this sense, incorporated constituents are like verbal particles. Although it is still a matter of debate whether incorporated constituents (and verbal particles, for that matter) are syntactically heads (e.g. N s) or phrases (e.g. NPs), they are clearly not DPs. In sum, the effect of the condition in (27) is that DE-verbs do not syntactically incorporate their direct internal argument. Nevertheless, from the present perspective, there is a curious mismatch between syntax and semantics in the case of DE-verbs, because although DE-verbs do not syntactically incorporate their direct internal argument, they do semantically incorporate it, which is the topic that I turn to next. 2.2.

A dynamic semantic analysis

There is both uniformity and diversity in the semantics of DE-verbs. The uniformity is seen in the idea that every DE-verb introduces a novel discourse marker

84


corresponding to its internal argument. The diversity is revealed in the idea that although every DE-verb specifies an end condition that comes to hold of its internal argument, the exact value of this end condition varies across subclasses of DE-verbs, hence there is no single end condition that does duty for all DE-verbs. 2.2.1. The framework The background framework presupposed here is an event semantics with a provision for the handling of discourse referents. Although event semantic frameworks (e.g., that of Krifka 1992) usually do not offer a means of dealing with discourse referents, it is feasible to extend them with a treatment of discourse referents familiar from dynamic semantic frameworks. The leading idea behind the treatment of discourse referents in a dynamic semantic framework (e.g., see Heim 1982, Krifka 1993, Chierchia 1995, Muskens 1996) is to analyze a (declarative) sentence as operating on or updating information states in such a way that the information about the active discourse markers thus far is inherited from the ‘input information state’ and any changes that the meaning of the sentence makes to the inventory of discourse markers is encoded in the ‘output information state’. For present purposes, an information state can be modeled as a set of discourse referent assignment functions (or more simply, discourse referent assignments), the idea being that such functions can adequately encode the necessary information about discourse referents. In this setting, sentences can be analyzed as functions from (input) information states to (output) information states, and in what follows, such functions will be represented by expressions of the form ‚I‚g[ . . . ], where I is a variable for information states (a set of discourse referent assignments) and g is a variable for discourse referent assignments. Expressions of this type are also known as updates. Furthermore, in order to account for semantic composition in this framework, natural language predicates are analyzed as having both an argument for discourse referent assignments and an argument for information states. In addition, verbs are assumed to have an event argument,11 as is usual in an event semantics. More technically, the models assumed here include a universe of discourse A (a, b, c, . . . ), which contains as pairwise disjoint subsets a set O of ordinary individuals (x, y, z, . . . ), and a set E of events (e, e0 , e00 , . . . ). The models also include a set D of discourse referents (1, 2, 3, . . . ), which can be identified with the set of positive integers, and a set G of discourse referent assignments (g, ”, g0 , . . . ), which is the set of (total) functions from D to A. If d is a discourse referent, then g ¼ ”[d!a] means that g is identical to ” except that g(d) ¼ a. The more compact statement g ¼ ”[d] means that there is an a such that g ¼ ”[d!a]. In this case, the sortal status of a is determined by the predicate in question. Finally, in addition to an interpretation function, the models also include the usual (static) assignments to variables, but these do not figure explicitly in the semantic representations. 2.2.2. Uniformity in novelty As mentioned above, the unity in the semantics of DE-verbs stems from the claim that they all introduce a novel discourse referent. This idea is implemented as in (28), where an intransitive DE-verb Vde,1 is analyzed as a


85

four-place relation between discourse referent assignments g, information states I, events e, and predicates P such that a discourse referent 1 is introduced so that the corresponding relation Vde holds between e and g(1) and P holds between g, I, and g(1), as in (28a), and the analysis of a transitive DE-verb is parallel but with an extra argument x for the verb’s external argument, as in (28b).12 Evidently, this analysis takes seriously the idea that a DE-verb has the force of a (dynamic) existential quantifier built into its meaning. (28)

Vde,1 a. ‚P‚e‚I‚g[9v[I(v) g¼v[1]] V(e,g(1)) P(g,I,g(1)) I(g)], if Vde,1 is intransitive b. ‚P‚x‚e‚I‚g[9v[I(v) g¼v[1]] V(e,x,g(1)) P(g,I,g(1)) I(g)], if Vde,1 is transitive

For an extended application of the present approach, let us work through the analysis of the following text: (29)

Sa´ra ´ırtde,5 egy levelet. Feladta5. letter.ACC she-mailed-it5 Sarah wrote5 a ‘Sarah wrote a letter. She mailed it.’

The semantic derivation of the (dynamic) event predicate corresponding to the first sentence in (29) is detailed in (30).13 (30)

a. Sa´ra (Sarah) sarah b. ´ırde,5 (‘write5-’) ‚P‚x‚e‚I‚g[$v[I(v) g¼v[5]] write(e,x,g(5)) P(g,I,g(5)) I(g)] c. egy levelet (‘a letter.ACC’) ‚y‚I0 ‚g0 [letter(y) I 0 (g0 )] de,5 5 egy levelet (‘write a letter.ACC’) d. ´ır ‚P‚x‚e‚I‚g[$v[I(v) g¼v[5]] write(e,x,g(5)) P(g,I,g(5)) I(g)](‚y‚I0 ‚g0 [letter(y) I0 (g0 )]) ¼ ‚x‚e‚I‚g[$v[I(v) g¼v[5]] write(e,x,g(5)) letter(g(5)) I(g)] e. Sa´ra ´ırde,5 egy levelet (‘Sarah write5 a letter.ACC’) ‚x‚e‚I‚g[$v[I(v) g¼v[5]] write(e,x,g(5)) letter(g,I,g(5)) I(g)](sarah)¼ ‚e‚I‚g[$v[I(v) g¼v[5]] write(e,sarah,g(5)) letter(g(5)) I(g)]

As seen in (30a), proper names are treated simply as constants. The analysis of ´ırde,5 ‘write5-’ in (30b) is an instance of the pattern in (28b). Since the internal argument of


86

´ırde,5 ‘write5-’ is a (dynamic) predicate of ordinary objects and not an individualdenoting argument, we are forced to analyze egy levelet ‘a letter.ACC’ in (30c) as a (dynamic) predicate of ordinary objects, but this causes no difficulty, because it is feasible to analyze indefinite DPs such as egy levelet ‘a letter.ACC’ in this way. In (30d) and (30e), respectively, the result of functionally applying the meaning of ´ırde,5 ‘write5-’ first to the meaning of egy levelet ‘a letter.ACC’ and then to the meaning of Sa´ra ‘Sarah’ is given. In order to transform the (dynamic) event predicate in (30e) into an update (i.e., function from information states to information states) that is the logical type of sentence meanings, the event variable has to be existentially bound. This can be accomplished by an assertion operator !3 (‘!’ is also mnemonic for Hungarian a´llı´t ‘claim, assert’) that introduces a novel discourse referent for the event argument, as in (31), where E is a variable for relations between discourse referent assignments, information states, and events: (31)

!3

‚E‚I‚g[$v[I(v) g¼v[3]] E(g,I,g(3)) I(g)]

Applying !3 to the (dynamic) event predicate in (30e), we obtain the following: (32)

!3 Sa´ra ´ırde,5 egy levelet (‘Sarah write5 a letter.ACC’) ‚E‚I‚g[$v[I(v) g¼v[3]] E(g,I,g(3)) I(g)](‚e0 ‚I0 ‚g0 [$v0 [I0 (v0 ) g0 ¼v0 [5]] write(e,sarah,g0 (5)) letter(g0 (5)) I(g)]) ¼ ‚I‚g[$v[I(v) g¼v[3]] $v0 [I(v0 ) g¼v0 [5]] write(g(3),sarah,g(5)) letter(g(5)) I(g)] def = S-i

This formula denotes a function from (input) information states I to (output) information states (‚g[ . . . ]) such that the discourse referent assignments g update I with new assignments to the discourse referents 3 and 5 so that Sarah writes g(5) in an event g(3), g(5) is a letter, and I holds of g. In other words, I is updated to contain only those assignments g such that Sarah writes a letter g(5) in an event g(3). The second sentence of (29) has an implicit subject pronoun that is anaphorically dependent on Sa´ra ‘Sarah’ and an implicit object pronoun that is anaphorically dependent on the indefinite description introduced by ´ırde,5 ‘write5-’, which is instantiated as egy levelet ‘a letter.ACC’. The (dynamic) event predicate corresponding to this sentence is given in (33a), and the result of applying !3 to it is shown in (33b). (33)

a. Feladja5 (‘she mail it5’) ‚e‚I‚g[mail(e,sarah,g(5)) I(g)] b. !7 Feladja5 (‘she mail it5’) ‚E‚I‚g[$v[I(v) g¼v[7]] E(g,I,g(7)) I(g)](‚e0 ‚I0 ‚g0 [mail(e0 ,sarah,g0 (5)) I0 (g0 )]) ¼ ‚I‚g[$v[I(v) g¼v[7]] mail(g(7),sarah,g(5)); I(g)] def = S-ii


87

Observe that the meaning of the verb feladja5 ‘she mail- it5’ does not introduce a novel discourse referent for its internal argument—rather, it simply makes use of one (namely, 5) that has already been introduced. Of course, in combination with the assertion operator the sentence does introduce a novel discourse referent for the event argument of the verb. The meaning of the sentence, then, both ‘tests’ the input information state I to verify that the value that g assigns to 5 is mailed by Sarah and updates I with the information that g assigns to 7 an event in which Sarah mails g(5). In order to semantically conjoin the two sentences of (29) to yield a text, we require a notion of dynamic conjunction that applies to two updates to yield a new update. Technically, dynamic conjunction is implemented by applying the update denoted by the first sentence to the information state that results when the update denoted by the second sentence is applied to the input information state. The syntactic marker of dynamic conjunction is E´S (Hungarian e´s means ‘and’), as defined in (34), where S and T are variables for updates. (34)

E´S

‚S‚T‚I‚g[S(g,‚v[T(v,I)])]

The result of applying E´S to the two sentences in (29) (in their order of occurrence) is shown as follows: (35)

!3 Sa´ra ´ırde,5 egy levelet E´S !7 Feladja5 (‘Sarah write5 a letter.ACC and she mail it5) ‚S‚T‚I‚g[S(g,‚v[T (v,I)])](S-i)(S-ii) ¼ ‚I‚g[S-i(g,‚v[S-ii(v,I)])]¼ ‚I‚g[$v[I(v) g¼v[3]] $v0 [I(v0 ) g¼v0 [5]] write(g(3),sarah,g(5)) letter(g(5)) $v00 [I(v00 ) g¼v00 [7]] mail(g(7),sarah,g(5)) I(g)]

The resulting formula for the text denotes a function from (input) information states I to (output) information states (‚g[ . . . ]) such that I is updated with assignments g that assign to 5 a letter that Sarah writes and mails, to 3 an event in which she writes g(5), and to 7 an event in which she mails g(5). This extended application of the present dynamic semantic approach to the text in (29) has shown how the meaning of a DE-verb introduces a novel discourse referent for its direct internal argument and can bind further occurrences of this referent beyond the syntactic scope of its clause. 2.2.3. Accounting for the prohibitions The analysis presented in the previous section can account for the three prohibitions of DE-verbs that were discussed in section 1. The first prohibition, against definites (see section 1.2), follows from the fact that DE-verbs introduce a novel discourse referent for their direct internal argument.

88


Although a full demonstration of how this prohibition is derived would require a more detailed treatment of definites than I can provide here, the idea (following Heim 1982) is that definites are familiar in the sense that they presuppose that their discourse referent is already assigned to a salient individual satisfying the descriptive content in question in the input information state.14 To see how this idea is implemented in the present framework, let us consider the semantic derivation of the unacceptable sentence in (1b). Since ´ırde ‘write’, like DE-verbs in general, takes a (dynamic) predicate argument for their internal argument, the definite DP a9 levelet ‘the9 letter.ACC’ should be analyzed as a predicate and not as a individual term: (36)

a9 levelet (‘the9 letter.ACC’) ‚x‚I‚g[8v[I(v) ! x¼v(9) letter(v(9))] I(g)]

In this analysis, the meaning of a9 levelet ‘the9 letter.ACC’ denotes a relation between updates and ordinary individuals x such that every assignment v in the input information state I assigns the discourse referent 9 to x, which is a letter. Observe that no new information is added to I, which is meant to capture the idea that definites are familiar. Given the analysis of a9 levelet ‘the9 letter.ACC’ in (36), it is sufficient to derive the verb phrase of (1b) (cf. (30d)) to see what goes wrong: (37)

#ı´rjade,9 a9 levelet (‘write9 the9 letter.ACC’) ‚P‚x‚e‚I‚g[$v[I(v) g¼v[9]] write(e,x,g(9)) P(g,I,g(9)) I(g)](‚y‚I0 ‚g0 [8v0 [I0 (v0 ) ! y¼v0 (9) letter(v0 (9)) I0 (g0 )]) ¼ ‚x‚e‚I‚g[$v[I(v) g¼v[9]] write(e,x,g(9)) 8v0 [I(v0 ) ! g(9)¼v0 (9) letter(v0 (9))] I(g)]

The problem is that whereas the meaning of ´ırjade,9 ‘write9-’ introduces the novel discourse referent 9 for its internal argument, the meaning of a9 levelet ‘the9 letter.ACC’ presupposes that this referent is familiar. Technically, a conflict arises because the meaning of the verb updates the input information state with a novel assignment of 9 to a letter, and yet the meaning of the definite presupposes that all of the assignments in the input information state already assign 9 to a letter. More generally, it is clear that this kind of conflict will always arise between DE-verbs and definite DPs for their direct internal argument. The prohibition against strong quantifiers (see section 1.2) follows from the fact that DE-verbs take a predicate and not an individual argument as their direct internal argument. The problem is that strong generalized quantifiers such as minden vende´g ‘every guest’ (see (14)) and mindegyik alma´t ‘each apple.ACC’ (see (15a)) cannot be analyzed as predicates of individuals, which would be necessary in order for DEverbs to apply to them.15 The apparent exception to this prohibition is the determiner mindenfe´le ‘every kind of’ in (17), presumably because this determiner is second-order, quantifying


89

over sets of individuals. Since DE-verbs provide a predicate argument that may be quantified over, it is expected that they should be compatible with mindenfe´le ‘every kind of’. For an idea of what is at issue, consider the following static semantic analysis of the sentence in (17c): (38)

E´rkezikde mindenfe´le vende´g (‘arrive every-kind-of guest’) 8P[P guest P 6¼ Ø ! $e$x[arrivede(e,x) P(x)]]

Informally, this formula states that for every kind of guest there is an event in which such a guest arrives. Although the recasting of this static analysis in a dynamic framework would require us to address more precisely the interaction of DE-verbs with DPs headed by mindenfe´le ‘every kind of’, the basic point is that DE-verbs allow in principle for their predicate argument to be quantified over. Finally, the third prohibition, against wide-scope indefinites (see section 1.3), is also due to the fact that DE-verbs take a predicate argument as their internal argument. If the indefinite DP representing the direct internal argument of a DE-verb is a predicate and not a quantifier, then it does not bear scope and hence cannot move out of the scope of negation, as seen in (18), which contrasts with (23), in which the object DPs are quantifiers and therefore can move out of the scope of negation. However, the problem in (18) is that the indefinite DPs cannot remain in the scope of negation either,16 unlike in (19), where the indefinites are nouns or NPs (but not DPs), thus a conflict arises. A way to resolve this conflict is to append the element sem neither to the indefinites, thereby making them negative polarity items and allowing them to remain in the scope of negation, as in (21). In conclusion, the idea that DE-verbs take a predicate argument and introduce a novel discourse referent corresponding to their direct internal argument, as implemented in the dynamic semantic analysis in the previous section, is successful in accounting for the three prohibitions that DE-verbs exhibit and which are discussed in section 1. NOTES 1

2

3

4

Note that this characterization excludes the classical example of a DE-verb, namely, van ‘there is’, because it is not a verb of change, and in this respect it is not consonant with Szabolcsi’s (1986) original classification. However, although I will focus on DE-verbs that are verbs of change, one can think of the class of DE-verbs as being divided into two nonoverlapping subclasses, a small closed one containing van ‘there is’ (and perhaps akad ‘occur, turn up’: Akadt egy proble´ma ‘turned-up a problem’) and a large open one containing verbs of change. By ‘focused constituent’ is meant an identificational focus in preverbal position – see chapter 9. A clause without a focused constituent is also called ‘neutral’. Since Hungarian has both an indefinite and a definite object conjugation (which are not glossed here), the exact form of the verb depends on the (in)definiteness of the direct object noun phrase, the indefinite conjugation being the ‘elsewhere’ choice. A focused constituent is marked by SMALL CAPS.

90 5

6

7

8 9

10

11 12

13

14 15

16


The situation with a legto¨bb ‘most’ is less straightforward, because a noun phrase headed by a legto¨bb ‘most’ can only appear in clauses with a focused constituent, but DE-verbs are excluded from such clauses to begin with, if my reasoning above is correct. Keve´s ‘few, little’ could also be listed here, but a noun phrase headed by keve´s ‘few, little’ obligatorily appears in the preverbal focus position, and so it is not immediately apparent whether such clauses should count as neutral or not. Arguably, however, such clauses are neutral, because it is the inherent negative meaning of keve´s ‘few, little’ ( nem sok ‘not many, not much’) that attracts it to the focus position. According to Anna Szabolcsi (pers. comm.), indefinites headed by egy are so-called positive polarity items, which seems descriptively correct. Presumably, this would also generally be true of indefinites headed by numerals. One effect of sem ‘neither’ would then be to create a negative polarity item out of a positive polarity item. A verbal particle normally appears after its host verb in negative sentences. Early versions of the present approach were presented at the Sixth Symposium on Logic and Language in Budapest on 30 Aug. 1998 and at the Fifth International Conference on the Structure of Hungarian in Budapest on 21 May 2001. See van Geenhoven (1998) for a Semantic account of incorporation in the same vein, though she does not discuss Hungarian, where as I point out below, DE-verbs do not syntactically incorporate their direct internal argument. In case of more than one internal argument, the ‘direct internal argument’ is the theme/patient argument. The term ‘event’ is used in its broad sense, as covering processes and states as well. By convention, a superscript indicates the discourse referent introduced and a subscript, the intended antecedent discourse referent. In what follows, tense will be ignored and the question of how proper names should be treated in a dynamic framework will be set aside (see Muskens 1996). Strictly speaking, Heim does not attribute a uniqueness presupposition to definites. McNally (1998) analyzes existential there as taking a predicate argument in order to account for the parallel restriction in English. Recall footnote 7 for the suggestion that such DPs are positive polarity items and that the addition of sem neither makes them negative polarity items.

´ N Õ CHRISTOPHER PIN

WEAK AND STRONG ACCOMPLISHMENTS

1.

DISTINGUISHING ACCOMPLISHMENTS

Traditional tests for distinguishing accomplishments from activities in English (e.g. see Dowty 1979, chapter 2.2.3) include compatibility with temporal in-phrases and the availability of two interpretations when modified by almost, as seen by the contrasts in (2)–(6), where paint a picture and write a paper are accomplishment verb phrases and paint pictures and write papers, activity verb phrases. Following Rapp and von Stechow (1999), the two readings of almost will be called the ‘counterfactual’ and the ‘scalar’ interpretation, respectively. (1) a. Rebecca painted a picture. b. Rebecca painted pictures. (2) a. Rebecca painted a picture in an hour. b. #Rebecca painted pictures in an hour. (3) a. Daniel wrote a paper. b. Daniel wrote papers. (4) a. Daniel wrote a paper in three days. b. #Daniel wrote papers in three days. (5) a. Rebecca almost painted a picture. counterfactual: Rebecca did not begin painting a picture scalar: Rebecca did not finish painting a picture b. Rebecca almost painted pictures. counterfactual: Rebecca did not begin painting pictures (no scalar interpretation) (6) a. Daniel almost wrote a paper. counterfactual: Daniel did not begin writing a paper scalar: Daniel did not finish writing a paper b. Daniel almost wrote papers. counterfactual: Daniel did not begin writing papers (no scalar interpretation) Despite initial appearances, I want to argue that the two aforementioned criteria do not diagnose a single class of accomplishments. In particular, the criterion of compatibility with in-phrases picks out a larger class of accomplishments than the availablity of two readings when modified by almost. I will refer to as ‘weak accomplishments’ those accomplishments that are compatible with in-phrases but which 91 Katalin E´. Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 91–106. 2008 Springer.


92

exhibit only a counterfactual interpretation when modified by almost and call ‘strong accomplishments’ those accomplishments that are compatible with in-phrases and have both a counterfactual and a scalar interpretation in combination with almost. This classification is summarized as follows: (7) a. weak accomplishments are compatible with in-adverbials but have only the counterfactual interpretation with almost b. strong accomplishments are compatible with in-adverbials and have both a counterfactual and a scalar interpretation with almost c. accomplishments are compatible with in-adverbials (i.e., every accomplishment is either weak or strong) In a way, weak accomplishments are less robust accomplishments than strong accomplishments are because they share with activities the property that only a counterfactual interpretation is available when modified by almost – in this sense, they are closer to activities than strong accomplishments are. The main empirical argument for the distinction between weak and strong accomplishments comes from languages that distinguish them more sharply than English does.1 One such language is Hungarian, which has a class of verbs that may head verb phrases which (by means of aspectual composition) express weak but not strong accomplishments. To see this, let us consider the behavior of the verbs fest ‘paint’ and ´ır ‘write’ with respect to the two diagnostics mentioned in (7): (8)

(9)

a.

b.

(10)

(11)

a.

b.

Rebeka festett egy ke´pet. Rebecca painted a picture.ACC ‘Rebecca painted a picture.’ Rebeka egy o´ra alatt festett egy ke´pet. (cf. (2)) Rebecca an hour under painted a picture.ACC ‘Rebecca painted a picture in an hour.’ Rebeka majdnem festett egy ke´pet. (cf. (5)) Rebecca almost painted a picture.ACC ‘Rebecca almost painted a picture.’ counterfacutal: Rebecca did not begin painting a picture (no scalar interpretation) Dańiel ´ırt egy dolgozatot. Daniel wrote a paper.ACC ‘Daniel wrote a paper.’ Dańiel ha´rom nap alatt ´ırt ety dolgozatot. (cf. (4)) Daniel three day under wrote a paper.ACC ‘Daniel wrote a paper in three days.’ Dańiel majdnem ´ırt egy dologozatot. (cf. (6)) Daniel almost wrote a paper.ACC ‘Daniel almost wrote a paper.’ counterfactual: Daniel did not begin writing a paper (no scalar interpretation)


93

Since fest egy ke´pet ‘paint a picture.ACC’ and ´ır egy dolgozatot ‘write a paper.ACC’ pattern according to (7a), they are weak accomplishments. Strong accomplishments in Hungarian are overwhelmingly based on verbs with verbal particles.2 The strong accomplishments corresponding to fest ‘paint’ and ´ır ‘write’ are meg-fest ‘PRT-paint’ and meg-ı´r ‘PRT-write’ (both with the verbal particle meg). In contrast to fest egy ke´pet ‘paint a picture.ACC’ and ´ır egy dolgozatot ‘write a paper.ACC’, meg-fest egy ke´pet ‘PRT-paint a picture.ACC’ and meg-ı´r egy dolgozatot ‘PRT-write a paper.ACC’ pattern according to (7b) and hence are strong accomplishments:3 (12)

(13)

a.

b.

(14)

(15)

a.

b.

Rebeka meg-festett egy ke´pet. picture.ACC Rebecca PRT-painted a ‘Rebecca painted a picture’ Rebeka egy o´ra alatt meg-festett egy ke´pet. (cf. (2)) Rebecca an hour under PRT-painted a picture.ACC ‘Rebecca painted a picture in an hour’ Rebeka majdnem meg-festett egy ke´pet. (cf. (5)) PRT-painted a picture.ACC Rebecca almost ‘Rebecca almost painted a picture’. counterfactual: Rebecca did not begin painting a picture scalar: Rebecca did not finish painting a picture Dańiel meg-ı´rt egy dolgozatot. paper.ACC Daniel PRT-wrote a ‘Daniel wrote a paper.’ Dańiel ha´rom nap alatt meg-ı´rt egy dolgozatot. (cf. (4)) Daniel three day under PRT-wrote a paper.ACC ‘Daniel wrote a paper in three days’ Dańiel majdnem meg-ı´rt egy dolgozatot. (cf. (6)) PRT-wrote a paper.ACC Daniel almost ‘Daniel wrote a paper in three days’ counterfactual: Daniel did not begin writing a paper scalar: Daniel did not finish writing a paper

Another respect in which the morphologically simple verbs fest ‘paint’ and ´ır ‘write’ differ from meg-fest ‘PRT-paint’ and meg-ı´r ‘PRT-write’ is that the former are compatible with bare plural object noun phrases, whereas the latter are not: Rebeka ke´peket festett. Rebecca pictures.ACC painted ‘Rebecca painted pictures.’ b. #Rebeka meg-festett ke´peket. Rebecca PRT-painted pictures.ACC (17) a. Dańiel dolgozatokat ´ırt. wrote Daniel papers.ACC ‘Daniel wrote papers.’ (16)

a.


94 b.

Dańiel Daniel

meg-ı´rt PRT-wrote

dolgozatokat. papers.ACC

It would be attractive to relate this contrast to the idea that meg-fest ‘PRT-paint’ and meg-ı´r ‘PRT-write’ are strong accomplishments, whereas fest ‘paint’ and ´ır ‘write’ are not, and this is indeed what I aim to do.

2.

A N A LY Z I N G A C C O M P L I S H M E N T S

The intuition behind the distinction between weak and strong accomplishments that I want to pursue is that strong accomplishments encode a notion of finishing that weak accomplishments do not encode. It is this notion of finishing that accounts both for the (optional) presupposition of strong accomplishments that there is an earlier event of the same type that affects the object in question (as witnessed by the two interpretations with majdnem ‘almost’; see (13b) and (15b)) and for the observation that strong accomplishments are incompatible with bare plural noun phrases (see (16b) and (17b)). Since weak accomplishments do not exhibit these two properties, the conclusion is that they do not encode this notion of finishing. In this section, I will explicate the notion of finishing that is needed and illustrate its application in the analysis of strong accomplishments.4 2.1.

The framework

The background framework assumed here is an event semantics with a provision for the handling of presuppositions. Although event semantic frameworks (e.g., that of Krifka 1992) usually do not offer a means of analyzing presuppositions, it is feasible to extend them with a treatment of presuppositions familiar from dynamic semantic frameworks. The leading idea behind the treatment of presuppositions in a dynamic semantic framework (e.g., see Krifka 1993, Chierchia 1995) is to analyze (declarative) sentences as operating on or updating contexts in such a way that the presuppositional content of a sentence must be satisfied in the ‘input context’ and the assertive content of the sentence, in the ‘output context’. For present purposes, a context can be modeled as simply a set of possible worlds (also called a context set), the idea being that the context represents the information that is still under consideration. In this setting, sentences can be analyzed as functions from (input) context sets to (output) context sets, and in what follows, such functions will be represented by expressions of the form ‚C‚w[ . . . ], where C is a context set variable and w is a world variable. Expressions of this type are known as updates. Furthermore, in order to account for semantic composition in this framework, natural language predicates should be analyzed as having both a world and a context set argument. In addition, verbs are assumed to have an event argument,5 as is usual in an event semantics.


2.2.

95

Weak accomplishments

To get a feel for how the mechanics of the present framework function, let us consider how the sentence in (8) is derived. (In what follows, tense will be ignored.) The verb of creation fest ‘paint’ is analyzed as a five-place relation between worlds w, context sets C, events e, and two ordinary individuals x and y (agents and things painted, respectively) such that x paints y in e of w in C, as in (18a).6 Both egy ke´pet ‘a picture.ACC’ and Rebeka are basically treated as generalized quantifiers but with the difference that they apply to four-place relations R between worlds, context sets, events, and ordinary individuals and yield a three-place relation between worlds, context sets, and events, as in (18b)–(18c). (18) a. fest (‘paint’) ‚y‚x‚e‚C‚w[C(w) paint(w,e,x,y)] def = PAINT b. egy ke´pet (‘a picture.ACC’) ‚R‚e‚C‚w[C(w) $x[R(w,C,e,x) picture(w,x)]] def = a-picture c. Rebeka ‚R‚e‚C‚w[C(w) R(w,C,e,rebecca)] def = REBECCA In order for fest ‘paint’ to be able to combine with egy ke´pet ‘a picture.ACC’, the verb’s type must be raised so as to take a generalized quantifier argument, and the result of this type-raising is shown in (19), where fest" ‘paint’ indicates the type-raised variant of fest for generalized quantifier objects and ] is a variable for generalized quantifiers.7 (19) fest" (‘paint’) ‚R‚]‚x‚e‚C‚w[](w,C,e, ‚y‚e 0 ‚C 0 ‚w 0 [R(w 0,C 0 ,e 0 ,x,y)])](paint) ¼ ‚]‚x‚e‚C‚w[](w,C,e,‚y‚e 0 ‚C 0 ‚w 0 [paint(w 0,C 0 ,e 0 ,x,y)])] ¼ ‚]‚x‚e‚C‚w[](w,C,e, ‚y‚e 0 ‚C 0 ‚w 0 [C 0 (w 0 ) paint(w 0,e 0 ,x,y)])] The result of applying fest" to egy ke´pet is given in (20), which yields the information that x paints a picture in e of w in C. (20) fest" egy ke´pet (‘paint a picture.ACC’) ‚]‚x‚e‚C‚w[](w,C,e,‚y‚e 0 ‚C 0 ‚w 0 [C 0 (w 0 ) paint(w 0,e 0 ,x,y)])](a-picture) ¼ ‚x‚e‚C‚w[C(w) $y[paint(w,e,x,y) picture(w,y)]] Since the generalized quantifier representing Rebeka can unproblematically apply to this relation, no type-raising is necessary, and the result of the application is the


96

following event predicate, which relative to a context set C and a world w denotes the set of events in which Rebecca paints a picture in w of C: (21) Rebeka fest" egy ke´pet (‘Rebecca paint a picture.ACC’) ‚x‚e‚C‚ w[C(w) $y[paint(w,e,x,y) picture(w,y)]](REBECCA) ¼ ‚e‚C‚w[C(w) $y[paint(w,e,rebecca,y) picture(w,y)]] The final step is to existentially bind the event argument, which may be achieved by means of a declarative operator, as defined in (22), where E is a variable for event predicates (i.e., relations between worlds, context sets, and events). (22)

‚E‚C‚w[C(w) $e[E(w,C,e)]]

DECL

The result of applying the declarative operator to the event predicate in (21) is the update shown in (23), which adds to the context set the information that there is an event in which Rebecca paints a picture. (23)

Rebeka fest" egy ke´pet ( (8)) ‚E‚C‚w[C(w) $e[E(w,C,e)]] (‚e‚C‚w[C(w) $y[paint(w,e,rebecca,y) picture(w,y)]]) ¼ ‚C‚w[C(w) $e$y[paint(w,e,rebecca,y) picture(w,y)]]

DECL

As an aside, in order to retrieve the corresponding static proposition (construed as a set of possible worlds) from an update, a special ‘down’ operator (+) is employed that applies the update to the set of possible worlds W (also called the world set), as defined in (24), where U is a variable for updates. (24) +

‚U‚w[U(w,W)]

Applied to the update in (23), this operator yields the following prepositional formula, which denotes the set of worlds in which Rebecca paints a picture: (25) +

Rebeka fest" egy ke´pet ( (8)) ‚U‚w[U(w,W)](‚C‚w[C(w) $e$y[paint(w,e,rebecca,y) picture(w,y)]]) ¼ ‚w[W(w) $e$y[paint(w,e,rebecca,y) picture(w,y)]]

DECL

By replacing the (variable) context set with the (fixed) world set, the ‘down’ operator effectively trivializes the contribution of context, precisely because it makes every context the universal context. By aspectual composition (following Krifka 1992), the event predicate in (21) is quantized with respect to its event argument, which is a property that accomplishments


97

have and one that is presumably needed for compatibility with temporal alatt-phrases (in-phrases). In Krifka’s framework, a verb with an incremental theme argument that combines with an object noun phrase whose nominal head is quantized results in a verb phrase that is quantized with respect to its event argument. Adapting the notion of quantized reference to three-place relations P between worlds, context sets, and individuals (ordinary individuals or events), we can say that P is quantized (QUA) just in case the definition in (26) is satisfied, where a and b stand for individuals. The event predicate in (21) is quantized in this sense.8 def

(26) QUA = ‚P[8w8a8b[(P(w,W,a) P(w,W,b)) ! :b @w a]] Without going into the semantics of temporal alatt-phrases (in-phrases), the following analysis of egy o´ra alatt (‘in an hour’) suffices for present purposes: (27) egy o´ra alatt (‘in an hour’) ‚E‚e‚C‚w[C(w) E(w,C,e) an-hour(w,e) QUA(E)] Applied to the (quantized) event predicate in (21), this function restricts the context set to those worlds and events in which Rebecca paints a picture in an hour: (28) Rebeka egy o´ra alatt fest" egy ke´pet ( (9a)) ‚e‚C‚w[C(w) $y[paint(w,e,rebecca,y) picture(w,y)] an-hour(w,e) QUA(‚e‚C‚w[C(w) $y[paint(w,e,rebecca,y) picture(w,y)]])] Returning to the update in (23), observe that there is no presupposition involved in this analysis of the sentence in (8), which seems to be correct, given that only a counterfactual interpretation is available with majdnem ‘almost’ (see (9b)). However, due to the lack of a presupposition, there is no straightforward way for majdnem ‘almost’ to modify the reading represented in (23) to yield a scalar interpretation. Although a detailed discussion of the lexical semantics of majdnem ‘almost’ is beyond the scope of this chapter,9 we can reasonably think of it as a modifier of updates with the following content: (29) majdnem (‘almost’) ‚U‚C‚w[C(w) :U(w,C) $w 0 [C(w 0 ) U(w 0,C) close(w 0,w)]] def = almost In prose, the effect of majdnem ‘almost’ is to negate the update for the output worlds w but to presuppose that it holds of a world w 0 ‘close’ to w in the context set C. Applied to the update in (23), the analysis of majdnem ‘almost’ restricts the context set to those worlds in which Rebecca does not paint a picture but she almost


98 does so: (30)

Rebeka majdnem fest" egy ke´pet ( (9b)) ‚C‚w[almost(w,C,‚C 0 0 ‚w 0 0 [C 0 0 (w 0 0 ) $e$y[paint(w 0,e,rebecca,y) picture(w 0,y)]])] ¼ ‚C‚w[C(w) :$e$y[paint(w,e,rebecca,y) picture(w,y)] $w 0 [C(w 0 ) $e$y[paint(w 0,e,rebecca,y) picture(w 0,y)] close(w 0,w)]]

DECL

From the present perspective, the fact that the sentence in (9b) lacks a scalar reading is tantamount to saying that it does not encode the necessary notion of finishing – in other words, it is a weak accomplishment and not a strong accomplishment. For the analysis of strong accomplishments, a notion of finishing needs to be introduced, which is what I turn to next. 2.3.

Finishing and strong accomplishments

As an initial guide to finishing, let us consider how the verb finish is understood. The meaning of a sentence such as that in (31a) seems to presuppose that Rebecca or someone else began painting the picture in question. This is supported by the observation that the corresponding negative sentence also implies that Rebecca or someone else began painting the picture: (31) a. Rebecca finished painting a/the picture. b. Rebecca didn’t finish painting a/the picture. presupposition: Rebecca or someone else began painting the picture in question More generally, finish is used to assert that a final part of an event of a particular type takes place and to presuppose that the initial part of such an event takes place earlier. Since the present goal is not so much to analyze the meaning of the verb finish per se but rather to appeal to a slightly revised notion of finishing, the strategy will be to make the presupposition dependent on whether the final part of the event in question is also a proper part of that event (in which case there is a presupposition regarding the initial part) or whether the final part is an improper part of that event (in which case there is no need for a presupposition regarding the initial part). In other words, the notion of finishing that we are after should have the informal paraphrase ‘either an event of a particular type takes place, or a final part of an event of a particular type takes place and it is presupposed that the initial part of the event takes place earlier’. Arguably, the meaning of finish only allows for the second case (hence there is necessarily a presupposition), whereas the revised notion of finishing also allows ‘a final part’ to be construed as an improper part (in which case the presupposition falls away – this is the first case), which is what we want here. Furthermore, the meaning of finish allows the agent of the presupposed event to differ from the agent of the asserted event, whereas the revised notion of finishing


99

that we aim for requires them to be identical. In sum, then, the revised notion of finishing sought here is both broader and narrower than the meaning of finish. Another important component of the meaning of the verb finish is that if an event of a particular type is finished, then an event of that type cannot be continued. To illustrate, if Rebecca finished painting a picture is true, then no continuation can be part of a larger event in which Rebecca paints a picture, for otherwise she would have merely stopped painting a picture but not yet have finished painting one. Indeed, this condition on the lack of a continuation can account for why finish but not stop is incompatible with activities (Dowty 1979, p. 57): (32) a. Rebecca finished painting a picture. b. #Rebecca finished painting pictures. (acceptable only if a fixed number of pictures is understood) (33) a. Rebecca stopped painting a picture. b. Rebecca stopped painting pictures. The problem with (32b) is that there could be a continuation which is part of a larger event in which Rebecca paints pictures, hence the condition that a continuation with respect to that event type not be possible is not satisfied.10 Since it seems reasonable to treat the condition on the lack of a possible continuation as a presupposition of finish (e.g., it cannot be negated), this condition also figures as a presupposition in the notion of finishing that will be defined next. Now that the vital ingredients of finishing have been introduced (again, these are: an assertion of the whole event or of a final part of the event, a presupposition of the initial part of the event in the case of the latter, and a presupposition of a lack of a continuation of the event type in both cases), the next step is to define a relation that captures this notion of finishing. Strictly speaking, we need to define two relations, one for the case of the whole event, and the other for the case of a final part of the event. The first relation is finish-1, a six-place relation between worlds w, context sets C, events e, ordinary individuals (agents) x, generalized quantifiers ], and relations T (representing transitive verbs), as defined in (34).11 (34)

def

finish-1 = ‚T‚]‚x‚e‚C‚w[](w,C,e,T(x)) 8w 0 8e 0 8x 0 [W(w 0 ) ](w 0,W,e 0 ,T(x 0 )) ! :$e00 [e 0 w 0 e 0 ](w 0,W,e 0 w 0 e 0 ,T(x 0 ))]]]

In this definition, ‘’ stands for temporal precedence and ‘ ’ for (mereological) sum. In prose, the first line asserts that an event of the type determined by T and ] takes place, whereas the second and third lines express the presupposition that any event of this type lacks a continuation. Observe that the world set W is used in the formulation of the second presupposition, because the lack of a continuation should not depend on a particular context set C. The second relation is finish-2, which is more complex than finish-1 because it involves the additional presupposition that a proper part of the patient is affected in


100

an earlier event of the type determined by T and ]: (35)

def

finish-2 = ‚T‚}5D‚x‚e‚C‚w[

$y 0 [C(w) ](w,C,e,‚y‚e 0 ‚C 0 ‚w 0 [T(w 0,C 0 ,e 0 ,x,y –w 0 y 0 )]) 8w 0 [C(w 0 ) ! $e 0 [](w 0,C,e 0 , ‚y‚e 0 ‚C 0 ‚w00 [T(w00,C 0 ,e00 ,x,y 0 ) y 0 @w 0 y]) e 0 w 0 e]]] 0 0 8w 8e 8x 0 [W(w 0 ) ](w 0,W,e 0 ,T (x 0 )) ! :$e00 [e 0 w 0 e00 ](w 0,W,e 0 w 0 e00 ,T(x 0 ))]]] In this definition, ‘–’ stands for mereological difference and ‘@’ for proper part. Note that the mereological difference of two individuals x and y (x–wy, for a world w) is that part of x which remains when y is subtracted from it. Informally, the second through sixth lines of this formula state that there is an individual y 0 such that the difference of y (the patient) and y 0 participates in e, whose type is determined by T and ] (this is the assertion), and that y 0 participates in an earlier event e 0 of the same type (this is the first presupposition). The seventh and eight lines express the second presupposition, familiar from (34), that an event of this type lacks a continuation. Let us consider how the relations finish-1 or finish-2 figure in the analysis of the sentence in (12). One of these first applies to a relation T (representing a transitive verb), then to a generalized quantifier ] (representing the object noun phrase), and finally to an ordinary individual x (representing the external argument of the verb). If we think of finish-1 and finish-2 as alternative interpretations of the verbal particle meg (meg1 vs. meg2), as shown in (36a), then the two representations of the morphologically complex verb meg-fest ‘PRT-paint’ are derived by applying finish-1 and finish-2 to

PAINT

(from (18a)), as seen in (36b):

(36) a. meg1 (‘PRT’) finish-1 meg2 (‘PRT’) finish-2 b. meg1-fest (‘PRT-paint’) ‚]‚x‚e‚C‚w[finish-1(w,C,e,x,],PAINT)] meg2-fest (‘PRT-paint’) ‚]‚x‚e‚C‚w[finish-2(w,C,e,x,],PAINT)] Next comes the verb phrase, which is derived by applying one of the formulas in (36b) to the generalized quantifier representing the object noun phrase egy ke´pet ‘a picture.ACC’ (from (18b)): (37) a. meg1-fest egy ke´pet (‘PRT-paint a picture.ACC’) ‚x‚e‚C‚w[finish-1(w,C,e,x,a-picture,PAINT)]


101

b. meg2-fest egy ke´pet (‘PRT-paint a picture.ACC’) ‚x‚e‚C‚w[finish-2(w,C,e,x,a-picture,PAINT)] These formulas in turn serve as possible inputs to the generalized quantifier representing the subject noun phrase Rebeka (from (18c)): (38) a. Rebeka meg1-fest egy ke´pet (‘Rebecca PRT-paint a picture.ACC’) ‚e‚C‚w[REBECCA(w,C,e, ‚y‚e 0 ‚C 0 ‚w 0 [finish-1(w 0,C 0 ,e 0 ,y,a-picture,PAINT)])] b. Rebeka meg2-fest egy ke´pet (‘Rebecca PRT-paint a picture.ACC’) ‚e‚C‚w[REBECCA(w,C,e, ‚y‚e 0 ‚C 0 ‚w 0 [finish-2(w 0,C 0 ,e 0 ,y,a-picture,PAINT)])] Finally, the result of applying the declarative operator (from (22)) to these two event predicates are the (now unpacked) updates shown in (39) and (40), respectively. Rebeka meg1-fest egy ke´pet ( one reading of (12)) ‚C‚w[C(w) $e$y[paint(w,e,rebecca,y) picture(w,y)] 8w 0 8e 0 8x 0 [W(w 0 ) $y[paint(w 0,e 0 ,x 0 ,y) picture(w,y)] ! :$e00 [e 0 w 0 e00 $y[paint(w 0,e 0 w 0 e00 ,rebecca,y) picture(w 0,y)]]]] (40) DECL Rebeka meg2-fest egy ke´pet ( another reading of (12)) ‚C‚w[$y 0 [$e$y[paint(w,e,rebecca,y –w y 0 ) picture(w,y)] 8w 0 [C(w 0 ) ! $e 0 [paint(w 0,e 0 ,rebecca,y 0 ) y 0 @w 0 y picture(w 0,y) e 0 w 0 e]]] 8w 0 8e 0 8x 0 [W(w 0 ) $y[paint(w 0,e 0 ,x 0 ,y) picture(w,y)] ! :$e00 [e 0 w 0 e00 $y[paint(w 0,e 0 w 0 e00 ,rebecca,y) picture(w 0,y)]]]]

(39)

DECL

The update in (39) asserts that Rebecca paints a picture y (in its entirety) and presupposes that an event of this type cannot be continued. In contrast, the update in (40) asserts that there is a y 0 such that Rebecca paints the mereological difference between a picture y and y0 and presupposes that she paints y 0 earlier, where y 0 is a proper part of a picture y, and also presupposes that an event of the type determined by T and ] cannot be continued.12 Strictly speaking, nothing in this update forces the y in the assertion and the y in the (first) presupposition to be identical, though both have to be pictures. How ever, since y0 has to be a proper part of both, this constrains the referents of the two instances of y to be at least overlapping, e.g., two properly overlapping pictures. No harm is caused by this slight looseness, because there is no prohibition against the possibility that y 0 is a proper part of another picture as well as long as Rebecca ends up painting one of them.

102


Comparing the update in (40) to the one in (39), it is not so difficult to see how modification by majdnem ‘almost’ (see (29)) would yield a scalar interpretation in the case of the former but a counterfactual interpretation in the case of the latter, which would correspond to the two readings of (13b). In this respect, the update in (39) is similar to the one in (23) for the weak accomplishment in (8) (cf. also (30)). Since in both of the latter cases the existence of the whole event is asserted, the negation contained in majdnem ‘almost’ would serve to negate the existence of the whole event, with the presupposition that the whole event takes place in a ‘close’ world in the context set. Of course, the two updates in (23) and (39) are still distinguished by the presupposition of the latter that the event type cannot be continued, but this has no bearing on the role of majdnem ‘almost’. In contrast, in the case of (40), the effect of majdnem ‘almost’ would be to negate the assertion that there is an event in which Rebecca paints the remainder of the picture, but it would preserve the presupposition that she paints a proper part of it in an earlier event. This presupposition would be preserved (and the second one as well) because of the presupposition of majdnem ‘almost’ that there is an event of this type in a ‘close’ world in the input context set. If this presupposition of majdnem ‘almost’ is satisfied, then the presuppositions of the update in (40) would also be satisfied, which would apply to every world in the context set – what would get negated is only that Rebecca paints the remainder of the picture. In sum, the content of majdnem ‘almost’ modifies the assertive content of an update, and where there is a presupposition of an earlier event, as on the construal of strong accomplishments with finish-2, we predict a scalar reading, but where there is not such a presupposition, as in the case of weak accomplishments and the other construal of strong accomplishments (with finish-1), we predict only a counterfactual reading. 2.4.

Negation

It is well-known that negation is normally a ‘hole’ with respect to presuppositions, i.e., the presuppositions of a sentence are preserved when the sentence is negated, as was illustrated for finish in (31b). My analysis of weak and strong accomplishments predicts that they will behave differently under negation. In particular, since weak accomplishments are not presuppositional, this should be confirmed by negation, which should simply negate the existence of an event of the type in question, whereas since strong accomplishments have one or two presuppositions, these presuppositions should be preserved under negation. Consider the following contrasts in this light:13 (41)

a. Rebeka nem festett egy ke´pet sem. (cf. (8)) Rebecca not painted a picture.ACC none ‘Rebecca didn’t paint a picture.’ unambiguous: Rebecca didn’t begin painting a picture


103

b. Rebeka nem festett meg egy ke´pet sem. (cf. (12)) pictures.ACC none Rebecca not painted PRT a ‘Rebecca didn’t paint a picture.’ ambiguous: (i) Rebecca didn’t begin painting a picture; (ii) Rebecca didn’t finish painting a picture (42) a. Dańiel nem ´ırt egy dolgozatot sem. (cf. (10)) Dańiel not wrote a paper.ACC none ‘Daniel didn’t write a paper.’ unambiguous: Daniel didn’t begin writing a paper b. Dańiel nem ´ırt meg egy dolgozatot sem. (cf. (14)) paper.ACC none Daniel not wrote PRT a ‘Daniel didn’t write a paper.’ ambiguous: (i) Daniel didn’t begin writing a paper; (ii) Daniel didn’t finish writing a paper The sentences in (41a) and (42a) are negations of weak accomplishments and the effect of negation is simply to deny the existence of an event of corresponding type – such sentences have only one reading. In contrast, the sentences in (41b) and (42b) are negations of strong accomplishments and we find that they have two readings, depending on whether or not an initial part of the event in question is presupposed. The two readings of the sentences in (41b) and (42b) depend on whether the presupposition of an earlier event is absent or present (which in turn depends on which version of meg ‘PRT’ is chosen), but it is easy to see that negation preserves the second presupposition regarding a lack of a continuation of the event type. More precisely, just as strong accomplishments in positive sentences are not compatible with bare plural object noun phrases (cf. (16b) and (17b)), nor are the corresponding negations of strong accomplishments compatible with such noun phrases, as shown in (43b) and (44b). Rebeka nem festett ke´peket. (cf. (16a)) Rebecca not painted pictures.ACC ‘Rebecca didn’t paint pictures.’ b. #Rebeka nem festett meg ke´peket. (cf. (16b)) Rebecca not painted PRT pictures.ACC (44) a. Dańiel nem ´ırt dolgozatokat. (cf. (17a)) Daniel not wrote papers.ACC ‘Daniel didn’t write papers.’ b. #Dańiel nem ´ırt meg dolgozatokat. (cf. (17a)) Daniel not wrote PRT papers.ACC

(43)

a.

Naturally, the simple verbs in (43a) and (43b), which express weak accomplishments and hence lack the presupposition that the event type cannot be continued, are perfectly compatible with bare plural object noun phrases.

104


In the present framework (see also Krifka 1993, pp. 282–283), negation can be analyzed as a modifier of updates with the assertion that the update does not hold of the output worlds w and the presupposition that there is at least one world w 0 in the input context set that the update holds of, as shown in (45). Insofar as this presupposition of negation is satisfied by at least one world in the input context set, the presuppositions of the update are also satisfied in the input context set, which accounts for the fact that negation preserves any presuppositions of the sentence that it negates.14 (45) nem (‘not’)

‚U‚C‚w[C(w) :U(w,C) $w 0 [U(w 0,C)]]

The analysis of the negated weak accomplishment in (41a) (ignoring the possible contribution of the negative polarity marker sem ‘none’) is the result of applying nem ‘not’ to the update in (23): (46) nem DECL Rebeka fest" egy ke´pet ( (41a)) ‚C‚w[C(w) :$e$y[paint(w,e,rebecca,y) picture(w,y)] $w 0 [C(w 0 ) $e 0 $y 0 [paint(w 0,e 0 ,rebecca,y 0 ) picture(w 0,y 0 )]]] In this case, since the weak accomplishment does not have any presuppositions of its own, the effect of negation is simply to update the context set to contain those worlds in which Rebecca does not paint a picture. The analysis of the first reading of the negated strong accomplishment in (42b) is given in (47), which is the result of applying nem ‘not’ to (the packed form of) the update in (39) (cf. (38a)). (47) nem DECL Rebeka meg1-fest egy ke´pet ( first reading of (42b)) ‚C‚w[C(w) :$e[REBECCA(w,C,e, ‚y‚e 0 ‚C 0 ‚w 0 [finish-1(w 0,C 0 ,e 0 ,y,a-picture,PAINT)])] $w 0 $e00 [REBECCA(w 0,C,e00 , ‚y‚e 0 ‚C 0 ‚w 0 [finish-1(w 0,C 0 ,e 0 ,y,a-picture,PAINT)])]] Like the update in (46), this one negates in the output context set the existence of the whole event in which Rebecca paints a picture, but in contrast to the former, it presupposes that this event type cannot be continued. This presupposition survives negation because of the presupposition of negation that the update holds in at least one world in the input context set. Finally, the analysis of the second reading of the negated strong accomplishment in (42b) is shown in (48), which is the result of applying nem ‘not’ to (the packed form of) the update in (40) (cf. (38b)). (48) nem DECL Rebeka meg2-fest egy ke´pet ( second reading of (42b))


105

‚C‚w[C(w) :$e[REBECCA(w,C,e, ‚y‚e 0 ‚C 0 ‚w 0 [finish-2(w 0,C 0 ,e 0 ,y,a-picture,PAINT)])] $w 0 $e00 [REBECCA(w 0,C,e00 , ‚y‚e 0 ‚C 0 ‚w 0 [finish-2(w 0,C 0 ,e 0 ,y,a-picture,PAINT)])]] Due to the relation finish-2, this update carries the additional presupposition that Rebecca paints a proper part of the picture in an earlier event – what is negated is that she paints the remainder of the picture in the output context set. 2.5.

Conclusion

Although both strong and weak accomplishments are quantized, as suggested by their compatibility with temporal alatt-phrases (in-phrases), strong accomplishments, in contrast to weak accomplishments, are analyzed with the help of one of relations finish-1 and finish-2. Accordingly, the sentence in (12) (as analyzed in (39) and (40)) is a strong accomplishment, whereas the one in (8) (as analyzed in (23)) is a weak accomplishment. Thanks to the relations finish-1 and finish-2, strong accomplishments have one or two presuppositions that weak accomplishments lack. The first presupposition, contributed by finish-2, is very palpable and states that the initial part of the event in question has taken place earlier with the same agent, which leads to the scalar reading with majdnem ‘almost’ in (13b) and (15b) and the second reading of (41b) and (42b). The second presupposition, contributed by both finish-1 and finish-2, is less obvious at the phrasal level, but it cashes out at the verb level. More specifically, this presupposition constrains the kind of object noun phrases that verbs expressing strong accomplishments can combine with, as a consequence of the condition that events of the type in question cannot be continued, as seen in (16b) and (17b). A goal in further work will be to integrate more explicitly the distinction between weak and strong accomplishments into a theory of aspectual composition. NOTES

1

2

3

4

Earlier versions of this chapter were presented at the Eight Symposium on Logic and Language in Debrecen, Hungary on 27 Aug. 2004 and at the University of Leipzig on 29 June 2005, and I wish to thank those audiences for their comments and questions. Arguably, the counterfactual reading is the more prominent one in (5a) and (6a), which may suggest that the ‘default’ tendency is to interpret (1a) and (3a) as weak accomplishments. Verbal particles in Hungarian are often called ‘preverbs’ or (less accurately) ‘verbal prefixes’. For background on verbal particles, see chapter 2 and E´. Kiss (2002a, chapter 3.6). I employ a hyphen to separate a verbal particle from its host verb, although they would be written together according to Hungarian orthography Out of the blue, it would be more natural to use definite object noun phrases in (12)–(15), but I use indefinite noun phrases in order to maintain as close as possible the parallel with the unprefixed verbs, which resist definite objects (see chapter 4). Ka´lmań (1995, pp. 226–227) presents a similar intuition about the difference between verbs such as ´ır ‘write’ and meg-ı´r ‘PRT-write’, but his focus is on the so-called definiteness effect exhibited by ´ır ‘write’ and comparable simple verbs, which I discuss in chapter 4. For a recent more general discussion

106

5 6

7

8

9

10

11

12

13

14


of the various functions that verbal particles in Hungarian have, see chapter 2 and E´. Kiss (2004a). In particular, note that the verbal particle meg has more than one meaning and that another one is analyzed in chapter 3. The term ‘event’ is used in the broad sense, as covering processes and states as well. If I am right in saying that definiteness effect verbs take a predicate argument as their internal argument (as I argue in chapter 4), then for consistency fest ‘paint’ in (18a) (cf. (8)) – which is really festde – should also receive such an analysis. However, since the analysis of definiteness effect verbs is not the main issue in this chapter, I put aside the definiteness effect here and assume for convenience that such verbs take an individual argument as their internal argument, which also has the benefit of keeping the comparison between fest ‘paint’ and meg-fest ‘PRT-paint’ as close as possible. Ultimately, though, the meaning of festde is not identical to the meaning of fest ‘paint’, though it is based on it. This is not the only way to combine a verb with a generalized quantifier object (e.g., movement at Logical Form would be an alternative), but nothing crucially depends on this choice here. This definition of quantized reference makes use of the world set W, because the question of whether P is quantized should not depend on a particular context set C. See Rapp and von Stechow (1999) for a treatment of German fast ‘almost’ and Morzycki (2001) for a recent analysis of almost. This is because paint pictures is an activity, where the corresponding relation is cumulative with respect to its event argument (see Krifka 1992). It would be straightforward to modify the definition of finish-1 for application to relations representing intransitive verbs, but I focus on the case of transitive verbs here. Engelberg (2002, p. 396) claims that the German verb aufessen ‘eat up’ is ambiguous between having a presupposition regarding the existence of an earlier event and not having this presupposition. In present terms, aufessen ‘eat up’ is a strong accomplishment, similar to Hungarian meg-eszik ‘PRT-eat’ in this respect. The negative polarity marker sem ‘none’ is needed in (41)–(42) in order for the indefinite noun phrases to take narrow scope with respect to negation. Without sem ‘none’, noun phrases headed by egy ‘a’ are so-called positive polarity items and must take wide scope with respect to negation (but this is disallowed with the simple verbs – see chapter 4.) This strategy was also adopted in the analysis of majdnem ‘almost’ in (29), as discussed above.

ANIKO CSIRMAZ

PA RT I C LE S A N D A T W O C O M P O N E N T T H E O RY O F A S P E C T

This chapter argues that Hungarian is best described by adopting a two component theory of aspect, which distinguishes situation and viewpoint aspect. Situation aspect concerns properties of the event description, and viewpoint aspect establishes aspectual properties external to the event description. It is argued that Hungarian particles, similarly to non-particle constituents, play a role in both types of aspect: their presence can affect situation aspect properties, and their position can vary according to the viewpoint specification. The effect on either aspectual component is a consequence of the denotation of the particle or the similar non-particle constituent. It is also shown that situation and viewpoint aspect characterizes all event descriptions, not only those which show an overt difference between different aspect markings. The proposed account assumes a compositional account of situation aspect, and derives the distribution of particles by appealing to the semantic properties of particles and viewpoint aspect. The chapter is organized as follows. Section 1 outlines the two component theory of aspect. Section 2 compares situation and viewpoint aspect and argues that they need to be distinguished in Hungarian. Section 3 proposes that viewpoint aspect is marked even in those environments where this aspect marking is consistently absent. Section 4 considers some possibilities for encoding viewpoint aspect, and proposes a hybrid approach to derive the attested particle patterns. On a more general note, section 5 extends the conclusions of section 3 and argues that no special category of neutral aspect (Smith 1991) needs to be assumed for languages which fail to overtly mark viewpoint aspect. 1.

T W O C O M P O N E N T T H E O RY O F A S P E C T

A number of authors argue against a homogeneous view of aspect (Smith 1991, Klein 1994, Olsen 1997, Bertinetto 2001, Borik 2002, among others). Even though the terminology used varies widely, they all agree that it is necessary to distinguish properties that are ‘internal’ to the eventuality from those that are imposed on the eventuality from ‘outside’.1 To describe these two notions of aspect, I will use the terms situation aspect and viewpoint aspect, respectively. The remainder of this section contains a brief discussion of the two types of aspect and specific assumptions about them.

107 Katalin E´. Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 107–128. 2008 Springer.

108

ANIKO CSIRMAZ

1.1.

Situation aspect

Situation aspect encodes the Vendlerian distinctions of states, activities, achievements and accomplishments. The features [durative], [telic] and [dynamic] encode these classes (Smith 1991), as shown in the following table. Table 1 Situation aspect classes [durative]

[dynamic]

[telic]

þ þ þ

þ þ þ

þ þ

States Activities Achievements Accomplishments

The feature [durative] distinguishes instantaneous and durative event descriptions2 and [dynamic] differentiates states from dynamic event descriptions. The feature [telic] singles out event descriptions with an inherent endpoint (Smith 1991), a characterization which is qualified below. The Vendlerian categories are exemplified in the following examples. (1)

(2)

(3)

(4)

State Jańos aludt J-NOM slept ‘Jańos slept’ Activity Jańos futott ran J-NOM ‘Jańos ran’ Achievement Jańos meg nyerte J-NOM perf won ‘Jańos won the race’ Accomplishment Jańos ´ırt egy J-NOM wrote one ‘Jańos wrote a letter’

a the

versenyt race-ACC

levelet letter-ACC

Since this discussion focuses on the contrast between telicity and viewpoint aspect, the features [dynamic] and [durative] will be ignored in the remainder of this section. The feature [telic], as noted above, can be characterized, as in Smith 1991, as identifying event descriptions with an inherent endpoint. For purposes of this discussion, I assume a definition of telicity in terms of the subinterval property (Dowty 1979).3


109

Atelic event descriptions such as activities have the subinterval property: if Jańos runs from 1:00 to 3:00, then he also runs during all subintervals of this time interval.4 In contrast, telic event descriptions lack the subinterval property. If Jańos wrote a letter during the time interval stretching from 1:00 to 3:00, then he did not write a letter during every single subinterval of this interval, since he finished the letter at three o’clock. The telicity of the event description is not necessarily determined by the lexical properties of the verb (Smith 1991, Verkuyl 1993). E´. Kiss 1987, Kiefer (1992, 1994), and references cited therein note that particles can affect the aspectual properties of an event description in Hungarian. Accordingly, a particle can telicize an inherently atelic event description, as the following examples show. (5)

a. Jańos J-NOM ‘Jańos b. Jańos J-NOM ‘Jańos

futott ran ran’ ki futott (az utca´ra) out ran the street-onto ran out to the street’

(atelic)

(telic)

The telicity of the event description can be shown by adverb modification. Foradverbs (as well as its Hungarian a´t and -ig-adverb counterparts) can only measure the event time of atelic event descriptions. In-adverbs, and their Hungarian alatt counterparts modify telic event descriptions. (6)

a. Jańos ke´t e´s fe´l o´rań a´t futott J-NOM two and half hour-on across ran ‘Jańos ran for two hours and a half’ (atelic) b. Jańos ke´t e´s fe´l perc alatt ki futott J-NOM two and half minute under out ran az utca´ra the street-onto ‘Jańos ran out to the street in two minutes and a half’ (telic)

As the previous examples show, the telicity of event descriptions is compositionally determined, based on the interpretation of the constituents that are part of the event description. A given particle does not necessarily telicize a lexically atelic predicate. The same particle telicizes an inherently atelic predicate in (7a) and (8a), and appears in an atelic event description in (7b) and (8b), respectively. (7)

a. Jańos ki futott az utca´ra the street-onto J-NOM out ran ‘Jańos ran out to the street’ (telic) b. Az a´g ki hajolt az u´t the branch-NOM out leaned the road ‘The branch leaned above the road’ (atelic)

fo¨le´ above

110 (8)

ANIKO CSIRMAZ

a. Jańos fel e´pı´tett egy ha´zat one house-ACC J-NOM up built ‘Jańos built a house’ (telic) b. Jańos fel olvasott J-NOM up read ‘Jańos read aloud’ (atelic)5

In sum, instead of the lexical entry of individual predicates or particles, it is the interpretation of the constituents that determines the telicity of the resulting complex predicate. As noted by E´. Kiss (chapter 2), the particles which telicize an inherently atelic event description have a resultative or terminative interpretation. The particle and its optional associate az utca´ra (‘the street-onto’) are goals in (7a), and the particle fel (‘up’) denotes a result in (8a). The effect of particles on telicity is not surprising, since goals and resultatives generally yield a telic event description, as shown in the following examples. The parallelism between telicizing particles and their lexical counterparts extends even further; in absence of particles, resultatives and goals appear in the immediately preverbal position which is otherwise occupied by particles (see section 4 for accounts of movement to the preverbal position). (9)

(10)

Jańos (ke´t perc alatt) az utca´ra futott J-NOM two minute under the street-onto ran ‘Jańos ran out to the street’ Jańos (fe´l o´ra alatt) tiszta´ra mosta J-NOM half hour under clean-onto washed a ruha´t the dress-ACC ‘Jańos washed the dress clean’

As in the case of particles, it is a constituent interpreted as a goal or resultative which telicizes the event description. Accordingly, whenever the preverbal sublative constituent is not interpreted as a goal no telic event description arises. In (11) the sublative a falra ‘onto the wall’ expresses location, and thus appears in an atelic event description. (11)

Az a´g a falra the branch-NOM the wall-onto ‘The branch leaned onto the wall’

hajlott leaned

It was shown that telicity, a situation aspect property of event descriptions, is compositionally determined. In this respect, particles fail to show a surprising behavior, since they contribute to the situation aspect property of the complex event description just as lexical resultatives and goals do.


1.2.

111

Viewpoint aspect

As described above, the term situation aspect refers to properties of the event description proper. Viewpoint aspect, which may be either perfective or imperfective, expresses properties that are external to the event description.6 If the event description is imperfective, only a part of the event is relevant or asserted. With perfective aspect, the complete (maximal) event is considered. The perfective–imperfective distinction is illustrated below for Hungarian. Formally, the position of the particle correlates with the viewpoint specification of the event description: the particle immediately precedes the verb if the event description is perfective, and follows it whenever the event description is imperfective (Kiefer 1992, 1994, E´. Kiss 2002a).7 In terms of interpretation, the two viewpoint aspect categories can be distinguished by the ordering with respect to a when-clause. While the perfective matrix event is interpreted as entirely following the event described in the when-clause (12), the imperfective event both precedes and follows the event in the when-clause (13). (12)

(13)

Jańos J-NOM ‘Jańos Jańos J-NOM ‘Jańos

haza ment (amikor meg la´tta home went when perf saw went home (when he saw Mari)’ ment haza (amikor meg la´tta went home when perf saw was going home (when he saw Mari)’

Marit) M-ACC Marit) M-ACC

I assume a characterization of viewpoint aspect based on Reichenbach 1947 and Klein 1994. Perfective and imperfective aspect are defined as different orderings of two time intervals: the event time and the reference time. The event time is the runtime of the event, and the reference time is a time interval distinct from the event time. The role of reference time is viewed somewhat differently by a number of authors. Reichenbach 1947 and Hornstein 1990, for instance, distinguish complex perfective tenses from simple tenses. In complex perfective tenses the event time is not ordered with respect to the utterance time, but the reference time that is distinct from the event time. Demirdache and Uribe-Etxebarria 2000, 2004, in contrast, assume that the relation between reference time and event time encodes perfect and imperfective aspect. In perfect event descriptions, the reference time follows the event time, and in imperfectives, the reference time is contained within the event time. It can be shown that apart from perfect and imperfective, a perfective aspect must also be distinguished. Perfective aspect, which can be described as the reference time containing the event time, is overtly marked in Mandarin (e.g. Smith 1991) and Bole (Gimba 2000), among others. In order to accommodate perfective aspect as well as imperfective and perfect, I follow Klein 1994, Olsen 1997, Iatridou et al. 2001 and von Fintel and Iatridou 2002 in assuming that these aspectual categories are encoded by two distinct heads. A perfect head Perf encodes perfect aspect. A viewpoint aspectual head Asp encodes perfective and

112

ANIKO CSIRMAZ

imperfective aspect. Accordingly, it can impose two orderings: either the reference time contains the event time (perfective) or the reference time is contained within the event time (imperfective). In addition, I assume that the initial and final endpoints of the event time cannot be contained in the reference time of an imperfective event description, as suggested in different terms by Smith 1991.8 This ordering of the event time and the reference time is similar to the description of perfective and imperfective aspect in Smith 1991. Smith 1991 characterizes perfective event descriptions as representing the whole event or situation, and imperfectives as spanning an interval which is internal to the situation. The relative ordering of the perfective and imperfective event description and the situation in Smith 1991 is the same as assumed above for reference time and event time. Thus the proposed formalization captures the convergence of a number of different approaches. The different interpretations of perfective and imperfective event descriptions in (12) and (13) can be described straightforwardly by the present approach. The time established by the when-clause is ordered with respect to the reference time rather than the event time. In all of these examples, the reference time follows the time of the whenclause.9 If the event description in the matrix clause is perfective, then the event follows the time of the when-clause, since the reference time includes the event time. With imperfective viewpoint specification, the matrix event is interpreted as ongoing at the time of the when-clause. Again, this interpretation follows from the reference time being (properly) included in the event time of an imperfective event description. I assume that perfective or imperfective situation aspect is encoded by an appropriate aspectual head in Asp (Demirdache and Uribe-Etxebarria 2000, Iatridou et al. 2001). Asp takes vP as its complement and in turn, AspP serves as the complement of T. The reference time (i) is merged as an argument of Asp, in Spec,AspP. (14)

TP T

AspP

i Asp

vP

It was argued above that situation aspect encompasses properties of the event description (telicity, durativity and dynamicity). Viewpoint aspect, in contrast, reflects the ordering of the event time and the reference time. This account mirrors the intuition that situation aspect describes properties ‘internal’ to the event, and viewpoint aspect those ‘outside’ of it. The following section presents further evidence for the separation of the aspect components in Hungarian. 2.

THE INDEPENDENCE OF THE TWO ASPECT SYSTEMS

As suggested above, situation and viewpoint aspect are independent components of the aspectual system. This section offers some arguments from Hungarian for the separation of the two components.


113

The framing effect, described by Jespersen 1931, is shown by imperfective, but not by any perfective event description. The framing effect is the requirement that the event be anchored to some specific time, which can be provided by the present tense (the utterance time) or, with past tense marking, by another past event. In the absence of an overt or contextually salient event, the past tense imperfective event description is ungrammatical. (15)

a. Jańos megy fel a le´pcso´ń NOM goes up the stair-on J‘Jańos is going up the stairs’ (present progressive) b. Jańos ment fel a le´pcso´ń *(amikor csengettek) rang-3pl J-NOM went up the stair-on when ‘Jańos was going up the stairs *(when the bell rang)’ (past progressive)

The difference between the two aspectual components is also shown by the ordering of events, which is sensitive to the viewpoint rather than situation aspect specification. As the examples in section 1.2 show, the ordering with respect to a whenclause depends on the perfectivity of the event description. Perfective event descriptions follow the time of the when-clause, while imperfective ones are already in progress at that time. (16)

a. Amikor csengettek, Jańos le ment a le´pcso´ń J-NOM down went the stair-on when rang-3PL ‘When the bell rang, Jańos went down the stairs’ (perfective; consecutive) b. Amikor csengettek, Jańos ment le a le´pcso´ń J-NOM went down the stair-on when rang-3PL ‘When the bell rang, Jańos was going down the stairs’ (imperfective; in progress)

The interpretation of present tense forms is also sensitive to the viewpoint rather than the situation aspect properties of event descriptions. The present imperfective form of dynamic predicates, like states, has an ongoing interpretation. A present perfective, however, form can only be interpreted as referring to a future or habitual event. (17)

a. Jańos le megy a le´pcso´ń J-NOM down goes the stair-on ‘Jańos goes down the stairs’ (perfective; future or habitual) b. Jańos megy le a le´pcso´ń J-NOM goes down the stair-on ‘Jańos is going down the stairs’ (imperfective; in progress)

114

ANIKO CSIRMAZ

Finally, the subinterval property, which holds both in the situation and viewpoint aspect domain, also provides an argument for the separation of these components. As noted in section 1, telicity, a situation aspect property, can be defined in terms of the subinterval property. If a telic event description holds for a time interval, then it doesn’t hold for all subintervals of that interval. In contrast, an atelic predicate (18a) holds for the event time as well as all subintervals of the event time. The subinterval property can also hold in the viewpoint aspect domain. Imperfective event descriptions have the subinterval property: if the imperfective event description, as in (18b), is true of a time interval, then it is also true of all subintervals of that interval. Perfective event descriptions, in contrast, lack the subinterval property. If the perfective event description in (18c) is true for the interval between 1:00 and 2:00, for example, then it does not hold for all subintervals of that interval.10 (18)

a. Jańos gyalogolt J-NOM walked ‘Jańos walked’ / ‘Jańos was walking’ (atelic perfective or imperfective) b. Jańos ment haza J-NOM went home ‘Jańos was going home’ (telic imperfective) c. Jańos haza ment J-NOM home went ‘Jańos went home’ (telic perfective)

The situation and the viewpoint aspect components can be modified by durative adverbs. The durative for-adverb modifies predicates of times which have the subinterval property, while an in-adverb modifies those which lack that property. In (19a), the situation aspect of the event description lacks the subinterval property, hence the event time can be modified by an in-adverb. Imperfective viewpoint aspect, in contrast, has the subinterval property, hence modification by a for-adverb is grammatical. The simultaneous modification by the two durative adverbs expresses that for ten minutes, Jańos was engaged in an activity which would result in his reaching the top of the mountain in two hours (cf. also de Swart 1998, 2000). No similar ambiguity can be observed in Hungarian with respect to the correlates of the English durative adverbs. In the discussion I treat Hungarian -ig adverbs as equivalent to English for-adverbs and requiring the subinterval property, and Hungarian alatt adverbs as behaving identically to in-adverbs. The behavior of durative in-adverbs in Hungarian is distinct from their English counterparts, since alatt adverbs, unlike in-adverbs, can only modify perfective telic event descriptions (19b). English in-adverbs can modify any telic event description, independently of its viewpoint aspect specification. The Hungarian equivalent, an alatt adverb, only modifies telic event descriptions. As shown in (19c,d), an imperfective telic event description can be modified by an -ig adverb, the equivalent of a for-adverb.


115

Modification by both an -ig and an alatt adverb, parallel to the English (19a), is ruled out (19e) (E´. Kiss (chapter 1)). (19)

a. ?For ten minutes, Jańos was going up the mountain in two hours b. Jańos ke´t o´ra alatt fel ment a hegyre J-NOM two hour under up went the mountain-onto ‘Jańos went up the mountain in two hours’ (perfective telic) c. Jańos ment fel a hegyre J-NOM went up the mountain-onto (amikor do¨ro¨gni kezdett) when rumble-INF began ‘Jańos was going up the mountain when it started to rumble’ (imperfective telic) d. Jańos tı´z percig ment fel a hegyre J-NOM ten minute-until went up the mountain-onto ‘Jańos was going up the mountain for ten minutes’ (imperfective telic) e. *Jańos tı´z percig ment fel ke´t o´ra alatt J-NOM ten minute-until went up two hour under a hegyre the mountain-onto ‘For ten minutes, Jańos was going up the mountain in two hours’ (imperfective telic)

The different behavior of Hungarian alatt adverbs and English in-adverbs can be captured as the Hungarian adverb permits no subinterval property either in the situation or the viewpoint aspect component. In English, in-adverbs are grammatical whenever the situation aspect lacks the subinterval property, indepently of the properties of the viewpoint aspect component. In contrast, Hungarian alatt adverbs require lack of the subinterval property in both aspectual components. Concerning -ig and for-adverbs, the English adverbs and their Hungarian counterparts behave alike. These adverbs can modify an event description if it has the subinterval property, either in the situation or the viewpoint aspect component. Since an event description in Hungarian cannot be modified by an alatt and an -ig adverb at the same time, there is no direct argument available for the conflicting subinterval properties of situation and viewpoint aspect. In spite of the lack of direct evidence for the two aspectual components in Hungarian, it can be shown that both aspectual components exist. An atelic event description, such as (20), allows modification by an -ig adverb in the absence of correlates of imperfective event descriptions (framing effect, obligatory simultaneous interpretation with a when-clause and obligatory ongoing interpretation). Atelic event descriptions can be thus modified by an -ig adverb even

116

ANIKO CSIRMAZ

when they are perfective, since situation aspect specification has the subinterval property. (20)

Jańos tı´z percig gyalogolt J-NOM ten minute-until walked ‘Jańos walked for ten minutes’

In contrast, whenever telic event descriptions appear with -ig adverbs (19d), they show properties of imperfectives. Since these properties, generally associated with imperfectives, necessarily arise, I conclude that the time associated with viewpoint aspect can be modified by -ig adverbs, even if the event description in question is telic.11 In sum, several arguments support the distinction between situation and viewpoint aspect in Hungarian. It was shown that even though the subinterval property can be interpreted in both aspectual domains, the specifications are independent for situation and viewpoint aspect. In addition, a number of properties are determined by the viewpoint specification of the event description rather than situation aspect, including the framing effect, event ordering and the interpretation of present tense forms. 3.

VIEWPOINT ASPECT IN HUNGARIAN

As argued above, situation and viewpoint aspect needs to be distinguished in Hungarian. These two aspectual components cannot be collapsed, since the subinterval property in the situation aspect domain does not have the same effects as that in the viewpoint aspect domain. Whenever the event description is imperfective, having the subinterval property in the viewpoint aspect component, it displays properties associated with imperfectives in general (the framing effect, obligatory simultaneous interpretation and ordering with when-clauses). The previous examples illustrating the effects of different viewpoint aspect specifications involved particle verbs with a telicizing particle. These event descriptions show an overt difference between perfective and imperfective forms: the particle is immediately preverbal in the former, but appears postverbally in the latter. This section proposes that perfective or imperfective aspect is specified for all event descriptions in Hungarian, including those which fail to show an overt difference between the two readings. In section 5, it is suggested that the generality of viewpoint aspect specification is, in fact, universal and is also present in languages which uniformly lack overt viewpoint aspect specification. As noted by Kiefer 1994, the Hungarian event descriptions which lack particles may be interpreted as imperfectives, and the same holds for verbs modified by nontelicizing particles. All of the following event descriptions have a perfective or imperfective interpretation, which results in a variable ordering with respect to when-clauses. The available readings can be disambiguated by the adverbs e´ppen (‘just then’) and ro¨gto¨n (‘immediately’), which appear with imperfective and perfective event descriptions, respectively.


117

a. Amikor csengettek, Jańos (e´ppen) telefona´lt J-NOM just then phoned when rang-3PL ‘When the bell rang, Jańos was talking on the phone’ (in progress; imperfective) b. Amikor csengettek, Jańos (ro¨gto¨n) telefona´lt J-NOM immediately phoned when rang-3PL ‘When the bell rang, Jańos called immediately’ (consecutive; perfective) (22) a. Amikor csengettek, Jańos (e´ppen) fel mosott J-NOM just then up washed when rang-3PL ‘When the bell rang, Jańos was washing the floor’ (in progress; imperfective) b. Amikor csengettek, Jańos (ro¨gto¨n) fel mosott J-NOM immediately up washed when rang-3PL ‘When the bell rang, Jańos washed the floor immediately’ (consecutive; perfective)

(21)

The event descriptions lacking telicizing particles also show a variability in the interpretation of present tense forms. The interpretations of imperfective event descriptions, with an ongoing interpretation, and that of perfective event descriptions, with a futurate or habitual interpretation, are both possible. (23)

(24)

Jańos J-NOM ‘Jańos Jańos J-NOM ‘Jańos

telefona´l phones is speaking on the phone’ / ‘Jańos often calls’ fel mos up washes is washing the floor’ / ‘Jańos often washes the floor’

The variable interpretation of these event descriptions is most straightforwardly accounted for by assuming that they can have either a perfective or an imperfective viewpoint specification. I assume that similarly to the event descriptions containing a telicizing particle, the event descriptions in question can also be either perfective or imperfective, depending on the aspectual head merged in Asp. The interpretation of the event description simply varies according to the viewpoint specification of the event description.12 It is attractive to adopt a parallel treatment of predicates with and those lacking a telicizing particle, where event descriptions can appear with either a perfective or an imperfective Asp head. The difference between event descriptions with a telicizing particle (which show word order variation according to the viewpoint aspect specification) and those without such a particle can be captured as the consequence of the

118

ANIKO CSIRMAZ

denotation of the particle. Section 4 presents an account which derives the behavior of the two types of event description. 4.

PA RT I C L E S A N D A S P E C T

I suggested above that it is desirable to assume that all event descriptions are specified as either perfective or imperfective in Hungarian. This section offers an account of the interaction of particles and the generally available viewpoint aspect, including the alternating or constant position of the former.13 Before delving into the interaction of viewpoint aspect and particles, let us consider again how particles affect situation aspect. It was argued in section 1 that particles can telicize the event description only when they denote an endpoint, namely a result or goal. In this respect, particles behave like other constituents; all of these elements only affect situation aspect properties when denoting an endpoint. Particles also behave like non-particle constituents with similar meaning in their interaction with viewpoint aspect. The alternation of telicizing particles in perfective and imperfective event descriptions is parallel to non-particle resultatives and goals. If a neutral event description is perfective, both particles and non-particles precede the verb. With imperfective viewpoint aspect, both follow the event, as shown in (25) and (26), respectively.14 a. Jańos le ment a le´pcso´ń J-NOM down went the stair-on ‘Jańos went down the stairs’ (perfective) b. Jańos ment le a le´pcso´ń J-NOM went down the stair-on ‘Jańos was going down the stairs’ (imperfective) (26) a. Jańos a teto´´re ment J-NOM the roof-onto went ‘Jańos went to the roof’ (perfective) b. Jańos ment a teto´´re J-NOM went the roof-onto ‘Jańos was going to the roof’ (imperfective)

(25)

Bare plurals and particles which do not denote an endpoint fail to show an alternation related to viewpoint aspect. Both perfective and imperfective event descriptions appear with these constituents in preverbal position. (27)

a. Jańos fel mosott J-NOM up washed ‘Jańos washed the floor’ / ‘Jańos was washing the floor’ (perfective / imperfective)


119

b. Jańos ablakot mosott J-NOM window-acc washed ‘Jańos cleaned a window / windows’ / ‘Jańos was cleaning a window / windows’ (perfective / imperfective) Three accounts of the position of particles (and related constituents) are discussed below. It is argued that the account which fares best appeals to the effect of constituents on situation aspect and to the aspectual properties of viewpoint aspect. It is also shown that particles do not show an unexpected behavior. This conclusion is in line with the point made above for situation aspect, where I noted that particles and non-particle constituents with similar semantic properties have the same effect on aspectual properties. As noted, particles appear in a preverbal position in perfective event descriptions and in certain cases, in imperfective event descriptions as well. Building on Farkas and Sadock 1989, Koopman and Szabolcsi 2000, E´. Kiss (chapter 2) and others, I assume that particles are phrases rather than heads and are merged within the vP. As discussed in E´. Kiss (chapter 2), a number of accounts have been proposed for the identity of the preverbal landing site of particles and for the motivation of movement. In this section I consider two families of these proposals, those of particle movement targeting Spec,AspP (Pinõń 1995, E´. Kiss 2002, Alberti 2004) and Spec,PredP (Koster 1994, E´. Kiss (chapter 2)). I propose that a hybrid account, which adopts assumptions from both of these approaches, accounts most successfully for the attested patterns. 4.1.

Particles in Spec,AspP

A number of accounts propose that preverbal particles in neutral sentences move to Spec,AspP (Pinõń 1995, E´. Kiss 2002a, Alberti 2004). The Spec,AspP position as the landing site reflects the intuition that particles affect the aspectual properties of the event description. E´. Kiss 2002a, building on Pinõń 1995, Kiefer 1994 and others, proposes that prefixes in preverbal positions act as aspectualizers or aspectual operators, perfectivizing the event description. The variation in the position of particles is straightforwardly described: since the particle is a perfectivizer, it cannot move to the specifier position of an imperfective Asp head (28)

ment a le´pcso´ń ] a. Jańos [AspP le down went the stair-on J-NOM ‘Jańos went down the stairs’ (perfective)

This account, also discussed in E´. Kiss (chapter 2), raises a number of questions. Of these, I mention two below. If particle movement to the preverbal position targets Spec,AspP in neutral clauses and the same account is invoked for bare noun objects, then presumably bare nouns are identical to perfectivizing particles in their interpretational effects.

120

ANIKO CSIRMAZ

This is, however, not the case. Bare objects neither telicize the event description, nor do they enforce a perfective aspect of the event description. (29)

Jańos kenyeret evett J-NOM bread-ACC ate ‘Jańos ate bread’ / ‘Jańos was eating bread’ (perfective / imperfective)

The same holds for those particles which do not denote an endpoint. They precede the verb in perfective and imperfective event descriptions alike. (30)

Jańos fel mosott J-NOM up washed ‘Jańos washed the floor’ / ‘Jańos was washing the floor’ (perfective / imperfective)

If particles and bare noun objects are to be treated the same way, then the aspectual properties of these elements and their effect on the event description cannot be responsible for preverbal movement. In addition, particles and other constituents can affect the aspectual properties of the event description even if they are not in the preverbal (arguably Spec,AspP) position. Situation aspect is determined compositionally, as noted in section 1. Thus a particle denoting a goal or result can telicize the event description irrespectively of the position where it appears. The considerations above address the motivation of particle movement. Particles are not necessarily aspectualizers, and can affect aspectual properties even when they are not in the preverbal position. Particle movement, then, cannot be driven by the need of ‘aspectualizers’ to appear in Spec,AspP. It is still possible, however, that particles move to a Spec,AspP position; this hypothesis will be adopted in section 4.3. Before proceeding, however, let us consider the arguments and motivation for a different position, Spec,PredP as the landing site. 4.2.

Particles in Spec,PredP Based on Koster 1994 and Csirmaz 2004, E´. Kiss (chapters 2, 9) argues that preverbal particles (as well as other elements with particle-like distribution) are located in the specifier of PredP. PredP is merged on top of the verb phrase and restricts the range of constituents which can appear in its specifier position. Following Koster 1994, E´. Kiss (chapters 2, 9) assumes that only predicative elements can appear in Spec,PredP. With this treatment of particles, E´. Kiss avoids a commitment to the aspectualizer nature of particles. In addition, the movement of predicative elements to Spec,PredP allows a uniform treatment of particles, bare noun objects and other constituents which immediately precede the verb in neutral clauses. The proposed structure is shown below.


(31)

Jańos [PredP le ment a down went the J-NOM ‘Jańos went down the stairs

121

le´pcso´ń ] stair-on

Concerning the motivation of movement to Spec,PredP, E´. Kiss (chapter 2) adopts the proposal of Koster 1994 for Dutch particles. According to Koster 1994, constituents theta-marked by the verb can be licensed in two ways: (a) they can be arguments or (b) they may be part of the predicate. In the latter case, they suggest, the constituent must move to a position within PredP to be licensed. Implementing this account for Hungarian, E´. Kiss (chapter 2) notes that Hungarian verbs are predicative, and argues that particles and bare nouns are predicative as well. In accordance with the predicative licensing requirement, verbs must always move to the Pred head. In addition, whenever present, particles and bare nouns need to move to Spec,PredP in order to be licensed.15 Given the predicative nature of particles and bare nouns, their preverbal distribution is derived. Under this approach it remains unaccounted for, however, why particle position can interact with viewpoint aspect. Recall that telicizing particles, which denote an endpoint, only precede the verb in perfective event descriptions. If the viewpoint aspect is imperfective, they follow the verb. Since the PredP projection does not, by assumption, house viewpoint aspectual operators or have an aspectual interpretation, the correlation is unexpected. The following section bridges the gap between the position of particles and their interaction with viewpoint aspect. 4.3.

A hybrid account

The issues raised above can be accounted for if the two previous accounts are merged. The motivation for particle movement is adopted from E´. Kiss (chapter 2), and the landing site of particle movement is identified as Spec,AspP. This hybrid account avoids the problems raised, as shown below. In addition, it also allows the Hungarian aspect system to be assimilated to those independently argued for (Reichenbach 1947, Klein 1994, among others). Before turning to particle position and viewpoint aspect, let us consider how situation aspect is determined in the proposed hybrid framework. As noted in section 1, situation aspect is determined compositionally. For concreteness, I assume that situation aspect properties are determined within the vP. The compositionality of situation aspect amounts to allowing the vP-internal constituents to affect situation aspect properties. For instance, a particle or other constituent denoting an endpoint, merged within vP, will appear in a telic event description. Irrespective of the inherent properties of the verb, the constituents denoting an endpoint result in telicity. Viewpoint aspect, as noted above, is crucially distinct from situation aspect, since it is determined by the Asp head. I assume that an aspectual head, namely perfective or imperfective is merged in Asp, unambiguously determining the viewpoint aspect properties of the event description.

122

ANIKO CSIRMAZ

It is the AspP projection where particles move when they appear in preverbal position. I assume that the AspP dominates PredP, as shown below. (32)

[AspP particlei Aspperf [PredP ti [vþPred] [vP . . . ]]]

With respect to the trigger of movement to Spec,AspP, I assume that the Asp head has an EPP feature which requires the closest constituent to move to an AspP-internal position. I also assume that the EPP feature is category-neutral; that is, it can be checked by any overt constituent as long as economy conditions are satisfied (e.g. the moved constituent is closest to Asp). As in the previous example, whenever a particle is present, it is the particle which appears in Spec,AspperfP, being the constituent closest to Asp.16 In the absence of a particle or other constituent in Spec,PredP, it is the Pred head that moves to Asp, checking the EPP feature of Asp. (33)

[AspP [[vþPred]iþAsp] [PredP ti [vP . . . ]]]

The constituent which moves to AspP is determined by the constituents within PredP. If the specifier of PredP is filled, then the specifier moves to check the EPP feature. If the specifier is not filled, then the complex Pred head raises. The proposed account, where the EPP feature of Asp can be satisfied by either an XP or a head is at odds with the proposal of Alexiadou and Anagnostopoulou 1998. They argue that the EPP feature of T can be checked by a head or a phrase, subject to parametric variation. In null subject languages EPP is checked by a head, hence no subject raising to Spec,TP is triggered by that feature. In contrast, in non-null subject languages the EPP feature of T is checked by an XP, hence the movement of the subject or that of some other XP to Spec,TP. Unlike Alexiadou and Anagnostopoulou 1998, I suggest that either an XP or a head can move to check the EPP feature of Asp in Hungarian. The head, Pred moves only if the movement of the phrase in Spec,PredP is ruled out by the incompatibility of imperfective aspect and constituents denoting an endpoint in AspP. Within the framework of Alexiadou and Anagnostopoulou 1998, this amounts to the claim that languages can also opt for moving either a phrase or a head to check an EPP feature, an optionality resolved by economy or other, well-formedness conditions17. If all particles move to Spec,AspP, the behavior of particles with perfective and imperfective aspect can be accounted for by appealing to aspectual properties. Recall from the previous discussion that only telicizing particles, which denote the endpoint of an event, are excluded from the specifier of imperfective AspP. If the event description is perfective, all particles can appear in Spec,AspP. (34)

a. Jańos ment haza (imperfective) J-NOM went home ‘Jańos was going home’ b. Jańos haza ment (perfective) J-NOM home went ‘Jańos went home’


(35)

123

a. Jańos fel olvasott (imperfective / perfective) J-NOM up read ‘Jańos was reading aloud’ / ‘Jańos read aloud’

The ungrammaticality of an endpoint-denoting particle in the specifier of an imperfective Asp follows from the definition of imperfectives proposed in section 1. A perfective event description was defined as one where the reference time contains the event time. For the reference time of imperfective event descriptions, it was assumed that it is properly contained in the event time. An additional constraint on imperfectives, also noted in section 1, is that the endpoints of the event time cannot be contained in the reference time. No such restriction holds for perfective event descriptions, where the reference time – contrary to imperfectives – must contain the complete event time. I suggest that if a particle denoting the endpoint of the event appears in Spec,AspP, then it delimits the reference time, just as it delimits the event time of telic event descriptions. If the reference time is delimited, then it cannot serve as an argument of imperfective aspect. This leads to the ungrammaticality of imperfective viewpoint aspect with a telicizing particle in Spec,AspP. Locative particles and bare noun objects are not delimiters, and hence do not delimit the event or reference time. These constituents are free to move to Spec,AspP in perfective and imperfective event descriptions alike to check the EPP feature of Asp. Contrary to the ‘aspectualizer’ account sketched in section 4.1, in this system it is not the aspectualizer nature of particles which drives their movement. Instead, particles interact with viewpoint aspect, determined in the Asp head, as described above. The particles which denote an endpoint move to Spec,AspP in perfective event descriptions, but fail to move (and so stay in PredP) if the viewpoint aspect is imperfective. In contrast, the particles that do not describe an endpoint move to Spec,PredP in perfective and imperfective event descriptions alike. Two aspects of the account are worth emphasizing. First, the trigger of particle movement is independent from aspectual properties of either the particle or the event description itself. Based on E´. Kiss (chapter 2), it was assumed that it is the proximity of Spec,PredP constituents that forces particle movement to check the EPP feature of Asp. Second, the landing site of particles and other preverbal elements is Spec,AspP. By tying together particles and viewpoint aspect, it is derived that preverbal particles can interact with reference time. Telicizing particles, which delimit the reference time when they appear in that position, are excluded from the specifier of an imperfective AspP since they delimit the reference time. It is also worth noting that this account does not appeal to particles specifically. It is expected, then, that the conclusions also hold of non-particle telicizing constituents. It was shown in (26), repeated below, that this is indeed the case. Non-particle resultatives and goals pattern with particles; they are preverbal in perfective event descriptions but follow the verb if the event description is imperfective.

124 (36)

ANIKO CSIRMAZ

a. Jańos a teto´´re ment J-NOM the roof-onto went ‘Jańos went to the roof’ b. Jańos ment a teto´´re J-NOM went the roof-onto ‘Jańos was going to the roof’ 5.

NEUTRAL ASPECT

In the preceding sections I suggested that all Hungarian event descriptions are specified for either perfective or imperfective aspect, including those which show no overt difference between these two forms. This proposal can be extended to languages which consistently fail to show an overt viewpoint aspect difference, either by the lack of overt aspect marking restricted to certain tenses or universally in the given language. I propose that if an event description has both perfective and imperfectivelike interpretation, this arises from the ambiguity in viewpoint aspect specification. This proposal is at odds with the neutral aspect account of Smith 1991 (p77 ff). Smith 1991 considers two accounts of event descriptions with aspectually variable interpretation: an account in terms of neutral aspect, and one where the event descriptions lack viewpoint specification. Smith 1991 argues that these event descriptions have a third, neutral viewpoint aspect marking. Smith cites two arguments to support this conclusion. First, the event descriptions give rise to both perfective and imperfective like interpretation with respect to event ordering. This excludes an account where the viewpoint aspect is reduced to unambiguously perfective or imperfective specification of the event description. In addition, Smith 1991 shows that a number of readings which would be expected if the event descriptions lacked viewpoint specifications do not arise. The missing reading illustrated and discussed in detail is the one which is available for overt imperfective marking of achievements. In the terms of Smith 1991, the overtly marked past imperfective marking in French yields an interpretation which focuses on the preliminary interval of an instantaneous achievement. The imperfective event description marks the preparatory stage leading up to the entering in the store, but does not assert that the culmination took place. (37)

Il entrait dans un magasin a store he enter-PAST.IMPERF in ‘He was entering a store’ (Smith 1991:197)

No similar interpretation is available for achievements in the future, which, according to Smith 1991, is only compatible with the neutral viewpoint aspect. Future tense, as in gagnera (‘will win’) can only refer to the achievement itself, indicating that imperfective aspect is not available for future event descriptions.

125


(38)

# Le cheval gagnera le course mais il ne gagnera pas the horse will win the race but he ne will win not ‘The horse will be winning the race but he will not win’ (Smith 1991:80)

Smith 1991 assumes that if no viewpoint aspect was specified, then the preparatory stage interpretation of achievements would be available. She concludes that it is neutral viewpoint aspect which allows a perfective- and imperfective-like interpretation, but disallows readings such as the preparatory interpretation of achievements. I propose that the conclusion of Smith 1991 does not necessarily hold. It is possible to account for the data by assuming that the event descriptions in question are ambiguous; they can either be specified as perfective or imperfective. Specifically, I assume that the viewpoint aspectual operator present in the structure is either perfective (Perf) or imperfective (Imperf). The ambiguous viewpoint aspect specification accounts for the fact that the event descriptions can be interpreted as either perfective or imperfective. With respect to the disallowed readings, I suggest that their availability is more generally restricted. The preparatory stage reading, for instance, is not necessarily available with overt imperfective marking either. As noted by Smith 1991, overt imperfective marking on an achievement can trigger a preparatory stage interpretation, which refers to the stage preceding the culmination. The preparatory reading is illustrated in (37) and (39). Similarly, the overt perfective marking a stative predicate such as have can trigger an aspect shift to an inceptive interpretation. In the Mandarin (40), the overt perfective marking le triggers an inceptive interpretation. (39) (40)

John was reaching the top (when the rain started to pour) Zhangsan you le yi da bi qian Z has le one big amount money ‘Zhangsan got a lot of money’ (Lin 2004a,b)

While overt perfective and imperfective marking can trigger the readings mentioned above, they do not necessarily do so. As shown below, the overt imperfective marking in Mandarin fails to yield the preparatory stage interpretation for achievements. Similarly, the overt perfective marking in Italian does not allow states to have a coerced inceptive interpretation. (41)

*zhege laoren zai chuan this old.man at bed de shihou, ta de jiaren GEN gen time he shen pang body side ‘This old man was dying in the around him’

shang zheng right.then dou zou family all TOP

zai want zai sit

si die ta de at he

bed with his family sitting

GEN

126 (42)

ANIKO CSIRMAZ

Gianni si e` fidato del suo manager in occasione Gianni si is trusted of his manager in time di quel difficile compito of that difficult situation ‘Gianni trusted / *came to trust his manager in that difficult situation’

The availability of coerced readings, including the preparatory stage interpretation of achievements, can vary even if the viewpoint aspect is overtly marked. In sum, the absence of coerced readings is not an argument against the presence of perfective or imperfective aspect marking. I assume that the possibility of coercion, where a viewpoint aspect marking allows a reading distinct from the readings defined in section 1, needs to be defined independently for a given language and viewpoint aspect.18 An economy consideration also supports the proposed account, which allows for either perfective or imperfective specification of all event descriptions. If the variable interpretation of ambiguous event descriptions follows from their ambiguous viewpoint aspect specification, then it is not necessary to assume the third, independent category of neutral aspect.19

6.

S U M M A RY

This chapter discussed a two-component theory of aspect. It was argued that the distinction between situation and viewpoint aspect is desirable on both general grounds and based on considerations specific to Hungarian. I showed that Hungarian particles interact with both kinds of aspect: the presence or absence of some particles can affect situation aspect, and the position of telicizing particles interacts with viewpoint aspect. It was argued that even though overt distinctions between perfective and imperfective aspect are limited, it is desirable to assume perfective or imperfective specification for all event descriptions in Hungarian. As for the specific implementation of aspect, I suggested that the two types of aspect are encoded differently. Situation aspect is calculated compositionally, while viewpoint aspect is encoded by an aspectual operator heading the designated AspP projection. In order to account for the interaction of particles and viewpoint aspect, I suggested that particles can raise to Spec,AspP from a lower Spec,PredP position. A particle in Spec,AspP can interact with viewpoint aspect by delimiting the reference time just as it delimits the event time. Since the reference time of imperfective event descriptions cannot include the endpoint of the event, no telicizing constituent can appear in the specifier of an imperfective AspP. The account proposed is not specific to particles, since it relies on the position within PredP for triggering movement and the denotation of the moved constituent for the interaction with viewpoint aspect. It applies, instead, to all types of constituents in AspP, without a restriction on their categorical or semantic properties.


127

NOTES

1

2

3

4

5

6

7

8

For comments and discussions about the research described in this chapter, I wish to thank Sabine Iatridou, David Pesetsky, Shigeru Miyagawa, the audience of Console XII and the Cornell Symposium on Language Universals, as well as the participants of the Hungarian research project (Katalin E´. Kiss, Christopher Pinõń, Boldizsa´r Eszes, Bala´zs Surańyi), Csaba Olsvay and Annama´ria Bene. For the Mandarin, French and Italian data, I would like to thank Feng-fan Hsieh, Jimmy Lin, Valentine Hacquard and Giorgio Magri. Following Bach 1986, I use the term eventuality as encompassing both states and (dynamic) events. The properties discussed hold not of eventualities themselves, but of event descriptions. The same event can be described in different ways, with the descriptions having different aspectual properties. The event of John walking to the store, for instance, can be described as the atelic John walked or as the telic John walked to the store. Instantaneous event descriptions hold during an atomic time interval. The event time of an instantaneous event description is treated as atomic in the linguistic system. It is not atomic, however, in the actual world, since time is dense. The classifications in terms of telicity or the subinterval property are largely identical, except for the characterization of semelfactives. As discussed in Csirmaz (chapter 10), semelfactives lack not only the subinterval property but also an inherent endpoint, since they are not delimited. According to the definition of Smith 1991 then, they are atelic, but given telicity defined in terms of the subinterval property, they count as telic. Semelfactives can be described as non-delimited and telic (since they lack the subinterval property), encoding the fact that semelfactives behave like telic rather than atelic event descriptions, for example for durative adverbial modification. The part relation is not strictly interpreted, but allows for the granularity effect. Dowty 1979 notes (and Hinrichs 1985 discusses extensively) the granularity effect which holds for activity predicates and iterated events. With granular predicates of times, the predicate does not hold for all subintervals, but only for those that are large enough to be the minimal intervals for the given predicate. For instance, if John walked is true for an interval i, then it does not hold for all subintervals of i. The time when he lifts his left heel off the ground does not qualify as walking. In contrast, states hold for all subintervals of the maximal time that the stative predicate is true of. In spite of the granularity effect, atelic events (as well as states) have the subinterval properly. The particle predicate fel olvas is not necessarily atelic. An overt object is incremental, and is mapped to the event just as in the case of the particle-less predicate olvas ‘read’. A quantized (indefinite singular or any definite) object yields a telic event description with both predicates. Following Olsen 1997 and Smith 1991, I assume that progressives are not distrinct from imperfectives. Rather, progressives are a subtype of imperfectives, which require the event description to be dynamic. The alternation does not affect all particles (for a detailed discussion on the particles which show such an alternation, as well as restrictions on imperfectives, see Kiefer 1992, 1994 as well as references cited there). For the present purposes, the particles with alternating position can be described as those with a transparent, compositional endpoint or resultative interpretation. I adopt the formulation of perfective and imperfective aspect from Iatridou et al. 2001. Both perfective and imperfective aspect take two arguments, a predicate with a time argument and a time interval (syntactically, these are realized of the complement and specifier of Asp, where the viewpoint aspect head is merged). The event description is treated as a predicate which takes the event time (i’) as an argument. The reference time argument of viewpoint aspect is ordered with respect to the event time, as noted above. The modality of imperfective and progressive event descriptions is ignored here, since it is not directly relevant to the discussion. (i) [[perfective]]¼‚P<st>. ‚i. $i¢. [i¢ i & P(i¢)] (ii) [[imperfective]]¼‚P<st>. ‚i. $i¢. [i i¢ & P(i¢)]

9

This ordering is not universal. On the variable ordering of the time of when-clauses and the reference time, see Moens and Steedman 1988.

128 10

11 12

13

14

15

16

17

18

19

ANIKO CSIRMAZ

I assume, following Dowty 1979, that the subinterval property is a property of predicates of times (predicates with a time interval argument). A predicate of times has the subinterval property iff whenever the predicate is true for an interval i, it is also true for all subintervals of i. Within the situation aspect domain, the relevant predicate of times is the event description applying to the event time. In the situation aspect domain, it is the predicate which applies to the reference time. Modification of the reference time is discussed in more detail in Csirmaz (chapter 10). The two viewpoint aspect specifications are not freely available. No imperfective marking is possible, for instance, with instantaneous event descriptions (Kiefer 1994). In this section I focus on telicizing particles and other constituents which are preverbal in perfective, and possibly imperfective, event descriptions. I leave the description and account of a number of constructions, including the ‘existential’ construction and those containing postverbal bare objects, and the pre- and postverbal indefinite count objects (Kiefer 1994) for further research. While the contrast with non-particle goals is less salient for a number of speakers, it is nevertheless present. The preferred interpretation with respect to a when-clause, for instance, is consecutive for (26a) and simultaneous for (26b). The correlation between viewpoint aspect readings and word order is obscured by focused resultatives and goals, since focused particles and non-particle goals as well as resultatives can precede the verb in imperfectives (see also fn (18)). Particles and bare nouns are in complementary distribution; see E´. Kiss (chapter 2) for an overview of accounts and discussion. Since the verb in Pred cannot move in these cases, it must be ensured that it is the particle in Spec,PredP which counts as closest to the Asp head. This can be ensured if closeness is calculated in terms of branching nodes or c-command and linearity (but not if a proximity metric such as that of Pesetsky and Torrego 2001 is used). As argued by E´. Kiss (chapter 2), the verb can move to a projection higher than AspP, namely NonNeutP. I assume that verb movement is generally possible in Hungarian, even if the verb did not move to Asp. If verb movement (more specifically, Pred movement, following v-to-Pred raising) is triggered by a higher head, then Pred can move to that head via Asp. This account presupposes that no locality constraints ban the movement of Pred to a higher head (e.g. there are no phase boundaries intervening between the two positions). Coercion processes can be constrained, for instance, by restricting the distribution of coercion operators as in de Swart 1998, 2000. Alternatively, the viewpoint specification can be unconstrained (contrary to Smith 1991), with the ordering of the event time and the reference time being only required to intersect. Both accounts are compatible with the data presented by Kratzer 2004. Kratzer 2004 shows that an imperfective reading is available for finite event descriptions in German, but not for nonfinite event descriptions which appear under a root modal. The lack of imperfective readings is due to the size of the complements of modals, which cannot contain an AspP projection. AspP is present in finite clauses, where Asp can be either perfective or imperfective (as argued above) or it can be an unspecified ordering between event time and reference time (as noted in this footnote). In either case, it is derived that telic event descriptions do not need to culminate in finite clauses.


F R O M T H E G R A M M ATI C A L I Z AT I O N O F V I E W P O I N T A SP E C T TO T H E G R A M MAT IC A L I Z ATI O N O F S I T U AT I O N A S P E C T 1.

INTRODUCTION

As illustrated in chapter (1), in present-day Hungarian basically all accomplishment and achievement predicates denoting a delimited change of state or change of location have a verbal particle. In Standard Hungarian, there is still a handful of verbs left which are ambiguous with respect to situation aspect; they can be interpreted both as atelic process predicates and as telic accomplishment or achievement predicates. In spoken Hungarian, however, these verbs have also developed a clearly telic variant associated with a verbal particle. For example: (1)

a. Standard Hungarian: A kurato´rium e´rte´keli a pa´lya´zatokat. the council evaluates the applications b. Spoken Hungarian: A kurato´rium ki-e´rte´keli a pa´lya´zatokat. the Council out evaluates the applications (2) a. Standard Hungrian: Az orvos ferto´´tlenı´tette a sebet. the doctor sterilized the wound b. Spoken Hungarian: Az orvos le -ferto´´tlenı´tette a sebet. the doctor down sterilized the wound (3) a. Standard Hungarian: Elleno´´rzo¨m a megadott telefonsza´mot. check-I the given phone-number ‘I check the given phone number.’ b. Spoken Hungarian: Le -elleno´´rzo¨m a megadott telefonsza´mot. down check-I the given phone-number In Spoken Hungarian, Latin prefixed verbs are also supplied with a Hungarian particle, for example: le-degrada´l ‘down-degrade’, a´t-transzforma´l ‘throughtransform’, el-deforma´l ‘off-deform’, ki-disszida´l ‘out-defect’. Compare: (4)

a. Standard Hungarian: Az auto´ ajtaja deforma´lo´dott az u¨tko¨ze´skor. the car’s door deformed the crash-at ‘The door of the car deformed at the crash.’ 129

´ Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 129–157. Katalin E. 2008 Springer.


130

b. ‘Spoken Hungarian: Az auto´ ajtaja el -deforma´lo´dott az u¨tko¨ze´skor. the car’s door off deformed the crash-at This kind of non-standard usage of the verbal particle is also creeping into literary Hungarian. Observe, for example, the following citation from a 2004 ´ . Na´dasdy: poem by A (5)

‘‘Minek what-for ‘What is

a the love

szeretet, love for

ha if if

nem hunyok szemet, ha gonoszul not shut-I eye-ACC if wickedly I don’t shut my eyes, if wickedly ki -a´bra´zolom, kinek mije van?’’ out represent-I who-to what is I represent (out) who has what?’ ´ da´m Na´dasdy: ‘‘Az a´bra´zola´s’’ ‘The Representation’ A (E´let e´s Irodalom July 23, 2004) In Standard Hungarian, codified e.g. in Hungarian dictionaries, a´bra´zol ‘represent’ has no verbal particle associated with it, nevertheless, the intended telicity of the predicate forces the poet to supplement it with a particle. Standard Hungarian, which is based primarily on the classical literature of the 19th century and the first half of the 20th century, is more conservative than Spoken Hungarian. The slight discrepancy between Standard Hungarian and Spoken Hungarian with respect to telicity marking suggests that the grammaticalization of telicity by means of a resultative or terminative particle has been a historical process which has basically been completed in the spoken language, and is nearing to its completion in the more conservative standard language. The gradual grammaticalization of situation aspect marking, i.e., the extension of telicity marking to all sentences describing a delimited change of state or a delimited change of location, has spanned about 800 years in the history of Hungarian. The first surviving coherent Hungarian text, Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral speech and prayer’, written between 1192–95, contains only a single instance of the verbal particle. In the Old Hungarian language of those times, aspect was marked by a rich system of verbal inflection, including both a viewpoint aspect morpheme, and a tense morpheme. The emergence of verbal particles marking telicity, i.e., the grammaticalization of situation aspect, was soon followed by the simplification of verbal inflection and the disappearance of viewpoint aspect marking. This chapter will raise the question if the near simultaneity of these two processes was a mere coincidence. Although it will not challenge the claim of Smith (1991), Csirmaz (chapter 6), and others that situation aspect and viewpoint aspect represent

G R A M M AT I C A L I Z AT I O N O F S I T U AT I O N A S P E C T

131

two separate subsystems of aspect, it will argue that the overt morphological encoding of both systems would be redundant to a large extent, because viewpoint aspect correlates with situation aspect in a predictable way in the majority of cases, and the language also provides clues to detect the occasional lack of correlation. The chapter is organized as follows. Section 2 describes the aspectual system of Old Hungarian, analyzing its complex inflection as a combination of viewpoint aspect marking and tense marking. Section 3 provides a morphosyntactic representation of the tense and aspect system of Old Hungarian, employing the framework of Demirdache and Uribe-Etxebarria (2004). Section 4 briefly surveys the stages of the emergence of telicity marking, and of the gradual disappearance of viewpoint aspect marking. Section 5 argues that the overt systematic morphological encoding of both types of aspect would be largely redundant because a given situation type defines a default value of viewpoint, and the non-default cases can also be easily identified with the help of various morphological and syntactic means. 2.

THE ENCODING OF TENSE AND VIEWPOINT ASPECT IN OLD HUNGARIAN

Whereas present-day Hungarian only has two tenses, an unmarked present tense, and a past tense marked by -t/tt, Old Hungarian had a variety of verb forms denoting past events. A verb stem like mond ‘say’ occurred in the following temporal–aspectual versions: mond ‘say.PRES.3SG’, mond-a, mondo-tt, mond vala, mondott vala. The analysis of the contexts in which these verb forms occur will lead us to the conclusion that they had the following temporal–aspectual values: (6) mond mondott monda mond vala mondott vala

– – – – –

simple present present perfect simple past past imperfective past perfect

The simple present was used to describe states and processes going on at the time of the utterance. Observe the following examples, quoted in present-day spelling: (7)

La´tja´tuk feleim szu¨mtu¨khel, mik vogymuk?1 see-you fellows-my eyes-your-with what are-we ‘Do you see, my fellows, with your eyes what we are?’ (Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral Speech and Prayer’, (1192–95, in Benko´´ (1980:47))

The -t/tt suffix, which is identical with the past tense suffix (as well as the past participle suffix) of present-day Hungarian, has been claimed in traditional

132


literature to represent a past tense suffix used in personal communication. Whereas in medieval Hungarian literature, the general tense of story telling is the simple past marked by -a/e, there is often a switch to the -t-marked tense when the story contains a direct quotation. This happens in the following section of the Vienna Codex (1450): (8)

Orpha megapola´ o ´´ napa´t s meg -fordola. Orpha kissed her mother-in-law and around-turn-PAST-3SG ‘Orpha kissed her mother-in-law, and turned around. Rut eggyeso´´le o´´ napa´val. Kinek monda´ Noe´mi: Ruth joined her mother-in-law whom said Naomi Ruth joined her mother-in-law. Naomi told her: Ime te rokonod meg -fordolt.2 look your relative around-turn-PERF-3SG Look, your relative has turned around.’ (Vienna Codex (1450:2), quoted by E. Abaffy (1992:154))

In the following quotation from a repentant psalm, the speaker, talking to the Lord in first person singular, uses only -t/tt-marked verbs: (9)

Jaj nekem, mert haragossa´ eń ellenem to¨tt-em eń woe to.me because angry me against make-PERF-1SG my megva´lto´mat . . . Igaz utat elhagytam, e´s az ja´ratlan savior true way-ACC leave-PERF-1SG and the untrodden utakon messze forgott-am.3 ways-on far move-PERF-1SG ‘Woe be to me because I have made my savior angry against me! I have left the true way, and I have moved far on untrodden ways.’ (Festetich Codex (1493:396), cited by E. Abaffy (1992:155))

As illustrated by these examples, the sentences containing a verb supplied with a -t/tt suffix ‘‘present a state of affairs with characteristics due to the prior situation’’ – which is the description of present perfect tense by Smith (1991:147). In such sentences, the speaker focusses on the resultant state of a past event, which is still in effect at the time of the utterance. In other words, a -t/tt marked verb form describes a past event seen from a present viewpoint. The aspect of these verbs, establishing a relation between a past event and a present reference time, is a type of perfective, and the tense, establishing a relation between a present reference time and a present utterance time, is the present. The illusion that the -t/tt-marked tense is a past tense used in personal communication arises because present perfect involves a reference time including the utterance time, which is typical of personal communication, e.g. of story-telling by an affected party. Biblical or historical story-telling, i.e., story-telling by a


133

non-participating party, rarely uses a reference time including the utterance time; such stories rarely have a result state extending to the time of the story-telling. The -a/e-marked past tense appearing in some of the sentences quoted above, called ‘historical past’, represents the Old Hungarian simple past. It was used to describe past events in the case of which the event time and the reference time are identical; in other words, the past is not seen as related to the present. This is the tense of story-telling by a non-participant: (10)

Sza´ll-a ala´ poklokra. descend-PAST.3SG down hells-into ‘He descended into hell. Harmadnapon halottaibo´l felta´mad-a. third-day-on dead-his-from resurrect-PAST.3SG The third day He arose again from the dead. Felmeń-e a mennyekbe . . . ascend-PAST.3SG the heavens-into He ascended into heaven’ (Apostle’s Creed)

Crucially, this is the tense used in personal communication, as well, if the sentence contains a time adverbial referring to a past point of time or a past period. In such sentences, the time adverbial specifies the reference time: (11)

Enyingen ´ır -a´ -m ez levelet peńteken4 Enying-at write -PAST -1SG this letter Friday-on ‘I wrote this letter at Enying on Friday’ (a letter from 1529, quoted from Magyar Nyelv (37:277) by E. Abaffy (1992:153))

The verb form mond vala represents the past imperfective. It was used, on the one hand, in cases when the reference time, i.e., the viewpoint, is internal to the time span of a past event. This is the context triggering the use of past imperfective e.g. in the following quotation from a fable of Sza´z fabula ‘Hundred fables’ by Ga´spa´r Heltai (1566): (12)

Melle´je gyu´´lveń az mezo´í egerek, ott ja´tszadoznak near-him gathering the field mice, there play-IMPERF.3PL vala ko¨rnyu¨le, e´s a kergete´sbe az egyik rea´ja the one on.him be-PAST around.him and the chase-in szo¨kelle´k az oroszlańra.5 hopped the lion-on ‘Having gathered around him, the field mice were playing around him, and during the chase one of them hopped on the lion.’

134


(Ga´spa´r Heltai: 16. fabula, in Heinrich (1897:45)) The reference time is specified by the event time of the predicate rea´ja szo¨kelle´k ‘hopped on him’. The playing event denoted by ja´tszadoznak vala ‘were playing’, taking place in the past (prior to the utterance time), is going on before, during, and after the reference time. Another function of the past imperfective was to mark past habitual events. Observe, for example, the following quotation from the legend of Saint Margaret. The sentences which describe Saint Margaret’s usual activities all contain a verb in the past imperfective. On the other hand, when the author, Lea Ra´skai, tells a unique event in the life of Saint Margaret, she switches to the simple (historical) past: (13)

Ez nemesse´ges szent szu´´z magyari kira´lnak nemes leańya this noble saint maid Hungarian king’s noble daughter hetet tart val-a, week-ACC keep-IMPERF.3SG be-PAST az kohnyań fo´´z val-a az sororoknak, the kitchen-in cook-IMPERF.3SG be-PAST the sisters-DAT fazekat mos val-a, pot-ACC wash-IMPERF.3SG be-PAST ta´lakat mos val-a, bowls-ACC wash-IMPERF.3SG be-PAST, az halakat megfaragja val-a carve-IMPERF.3SG be-PAST the fish-ACC Aze´rt mikoron szent Margit asszony egy napon szoka´sa therefore when saint Margaret lady one day her.habit szereńt az mosdo´vı´znek mosle´ka´t akarna´ according the washing.water’s leavings-ACC want-COND.3SG kivinni az refekto´riumbo´l, de nem vi-het-e´ el to.take.out the refectory-from but not take-POT-PAST.3SG off az vı´znek sokasa´gaé´rt, hı´v-a hozza´ egy sorort6 the water’s mass-for call-PAST.3SG to.it a sister ‘This noble saint maid, the Hungarian king’s noble daughter, would have turns on duty, she would cook in the kitchen for the sisters, she would wash pots, she would wash bowls, she would carve the fish . . . When one day lady Saint Margaret wanted to take out the dirty leavings of the water for washing, but could not take it because of the mass of the water, she called a sister to it’ . . . (Szent Margit e´lete [Saint Margaret’s life] (1510/1990:11))

The past imperfective could also be combined with an achievement verb marked by a verbal particle, in which case it expressed that the event did not culminate; it stopped after a preparatory phase. E.g.


(14)

135

Es oz gyimilcsnek u´l keseroü¨ vola vize, hugy so bitter was water-its that and that fruit-DAT turkokat mige´ szokosztja vol-a. burst-IMPERF.3SG be-PAST throat-their-ACC PRT ‘And that fruit had such a bitter juice that it was bursting their throats.’ Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral Speech and Prayer’ 1192–95, in Benko´´ (1980:47))

This sentence relates the story of Adam and Eve in Paradise, therefore we know that the apple which ‘‘was bursting their throats’’ (mige´ szokosztja vola) did not actually burst their throats, i.e., it did not kill them, merely it was about to kill them. Let us assume that habitual activities, e.g. those illustrated in (13), form an extended event. Then it is true for all three types of contexts in which the imperfective past occurs that the reference time (i.e., the viewpoint) is internal to the event time, and the utterance time follows the reference time. The verb form mondott vala represents the past perfect tense. It was used to describe past situations from a reference point which followed the event time but preceded the utterence time. For example: (15)

a. E´s megemle´kez-e´-k Pe´ter az ige´ro´´l, kit the word-about that and commemorate-PAST-3SG Peter mondo-tt val-a nekik7 say-PERF.3SG be-PAST them ‘And Peter commemorated the word that he had told them.’ (Mu¨nchen codex (1466:103), quoted by Bańhidi (1941:24)) b. Meglel-t-em eń juhomat, ki elvesze-tt val-a8 find-PERF-1SG my sheep that astray.go-PERF.3SG be -PAST ‘I have found my sheep that had gone astray.’ (Mu¨nchen codex (1466:146), quoted by Bańhidi (1941:24))

In both sentences, the time of the event described in the first clause represents the reference time for the verb of the second clause. The verb of the second clause denotes an event that took place prior to the reference time, hence its aspect is perfective. The utterance time follows the reference time, therefore the tense of this verb is past.

3.

T H E S Y N TA C T I C R E P R E S E N TAT I O N O F O L D

HUNGARIAN TENSE AND VIEWPOINT ASPECT

The current standard representations of complex tense – aspect systems (e.g. Hornstein 1993) go back to the theory of Reichenbach (1947), which derives the different English tenses (incorporating also aspect) from the varying relations of speech time, reference time, and event time. From among the neo-Reichenbachian


136

theories, I have chosen the approach of Demirdache and Uribe-Etxebarria (2004) as the framework of representing Old Hungarian tenses because of the clear and plausible syntactic claims that it makes. In the framework of Demirdache and Uribe-Etxebarria (2004), times are related by Tense and Aspect. Tense relates the time of the utterance (UT) to a reference time (RT). Aspect (in our terminology, viewpoint aspect), on the other hand, relates the reference time to the time at which the event/state denoted by the VP occurs or holds (ET). T(ense) is a spatiotemporal ordering predicate with the meaning of ‘after’ (past), ‘before’ (future) or ‘within’ (present). Likewise, Asp(ect) is a spatiotemporal ordering predicate with the meaning of ‘after’ (perfect), ‘before’ (prospective), or ‘within’ (progressive). Syntactically, Tense and Aspect are heads relating two time-denoting arguments projected in syntax. The external argument of T is the utterance time, its internal argument is the reference time specified by AspP. The external argument of Asp is the reference time, whereas its internal argument is the time of the event denoted by the VP (see Demirdache and Uribe-Etxebarria 2004, 143–145). In the case of simple tenses, Event Time and Reference Time are identical. In this framework, simple present can be represented as follows:

(16)

Simple Present TP UT

T’ T within

AspP RT i

Asp’ Asp within

VP

ET i

VP……

Utterance Time is internal to the coindexed Reference Time/Event Time; i.e., the Event Time (and the Reference Time which is non-distinct from it) extend over the present.


137

The present perfect, marked in Old Hungarian by -t/tt, corresponds to the following configuration:

(17) Present Perfect TP UT

T’

T within

AspP RT

Asp’ Asp after

VP ET

VP

The Utterance Time is internal to the Reference Time (i.e., the viewpoint), and the Reference Time is ordered after the Event Time. In other words, a past event is looked at from a present viewpoint. This is how the simple past can be represented in the framework of Demirdache and Uribe-Etxebarria:

(18)

Simple Past TP UT

T’ T after

AspP Asp’

RT i Asp within

VP ET i

VP


138

The Utterance Time is ordered after the Event Time, which is coindexed with the Reference Time. Observe the syntactic representation of the English past continuous or the Old Hungarian past imperfective:

(19) Past Imperfective/Progressive TP UT

T’ T after

AspP

RT

Asp’

Asp within

VP

ET

VP

The Utterance Time is ordered after the Reference Time, which is internal to the Event Time. This is how past perfect is represented in the framework of Demirdache and Uribe-Etxebarria:

(20) Past Perfect TP UT

T’ T after

AspP RT

Asp’ Asp after

VP ET

VP


139

The Utterance Time is ordered after the Reference Time, and the Reference Time is ordered after the Event Time. According to the proposed analysis, Old Hungarian (the Hungarian language between the 10–16th centuries) possessed a verbal inflection system which included both tense and aspect marking. Aspect had two values: imperfective (with the Reference Time within the Event Time), marked by a phonologically null suffix, and perfective (with the Reference Time after the Event Time), marked by a -t/tt suffix. (The perfective tenses of Old Hungarian represented the perfect subtype.) Tense also had two values: present (with the Utterance Time within the Reference Time), marked by a phonologically null suffix, and past (with the Utterance Time after the Reference Time), marked by -a/e. In the case of the simple tenses (simple present and simple past), the Reference Time coincides with the Event Time; so I assume no separate AspP projection. As there is only a tense marker for the verb to pick up, no auxiliary is needed. Observe the morphological make-up of simple tenses:

(21) a. Simple present (mond):

b. Simple past (monda):

TP T within 0

TP T after -a

VP V mond

VP V mond

In the case of the complex tenses also marking aspect, the verb combines with the aspect marker. The tense morpheme, when phonologically non-null, appears on an auxiliary homophonous with the copula. (The constraint that requires the insertion of lexical support for the phonologically salient tense suffix is still at work also in present-day Hungarian. As Bartos (2000:712, 726) formulates it: the verb stem cannot be combined with two analytical suffixes.) That is:

(22) a. Present perfect (mondott): TP T 0

AspP

Asp -tt

VP V mond


140

b.

Past imperfective (mond vala):

c.

Past perfect (mondott vala):

TP T (val)-a

TP AspP

Asp 0

T (val)-a VP V mond

AspP

Asp -tt

VP V mond

In present-day Hungarian, the inflected verb is subject to the Mirror Principle of Baker (1985), i.e., the suffixes representing the heads of the morphosyntactic projections extending the V appear attached to its right-hand side in the opposite order. The universally valid Mirror Principle was presumably in effect in Old Hungarian, as well. As Bartos (2000) shows, however, the reverse morpheme order attested on the Hungarian verbal and nominal stems cannot be the result of successive cyclic head movement to the left, because the occasional specifiers of the projections extending the head remain on its left-hand side in surface structure. Therefore, Bartos assumes, instead of head movement, an operation called ‘morphosyntactic merger’, which he defines as follows:

(23)

Morphosyntactic Merger

a. A [þaffix] category X can morphosyntactically merge with a potential stem Y under structural adjacency. This yields a word domain {x; y} (where x realizes X, and y realizes Y) at the level of morphology, and a head-chain <X,Y> in syntax. b. Structural adjacency X and Y are structurally adjacent iff (i) X c-commands Y, and (ii) there is no Z, such that Z is of the same projection level as X, and X c-commands Z, and Z c-commands Y.

The structures in (22) are built cyclically, so once e.g. Asp is merged with the VP, Morphosyntactic Merger creates a word domain {V,Asp}, located under V. The morphemes of the word are linearized in accordance with the Mirror Principle. Notice that Morphosyntactic Merger and the Mirror Principle yield the right order in the case of the complex tenses only if vala (i.e., the copula þ past tense) is also treated as one of the suffixes of the lexical verb, i.e., if the

141


aspect-marked verb plus vala string is analyzed as a single word domain. There is also some independent evidence supporting this assumption. In yes-no questions the interrogative clitic is attached to the finite V, hence it can be used to test where the right edge of the V is. Present-day Hungarian has preserved a single relic of the complex verb forms of Old Hungarian: the past conditional, e.g. mondo-tt vol-na say-PERF.3SG be-COND ‘would have said’. In the case of past conditional predicates, the interrogative clitic is attached to the whole verbal complex, instead of the V bearing the -t suffix: (24)

a. Mondtam volna say-PERF-1SG be-COND ‘Would I have told you b. ??Mondtam-e volna

-e nektek a hı´rt, ha nem -Q you-DAT the news if not the news if it is not true?’ nektek a hı´rt, ha nem igaz?

igaz? true

In the Old Hungarian simple tenses, the agreement morpheme(s) appear on the V plus Tense complex. In the Subjective Conjugation, the verb only bears a subject agreement suffix. In the Objective Conjugation, there is also an object agreement morpheme, often fused with the subject agreement morpheme, marking the presence of a 3rd person definite object: (25)

a. Subjective conjugation mond-e´ -k say -PAST-1SG mond-a´ -l say -PAST-2SG mond-a -0 say -PAST-3SG b. Objective conjugation mond-a´ -m say -PAST-OBJ.1SG ‘I said it’ mond-a´ -d say -PAST-OBJ.2SG mond-a´ -0 say -PAST-OBJ.3SG

mond-a´ -nk say -PAST-1PL mond-a´ -tok say -PAST-2PL mond-a´ -nak say -PAST-3PL mond-a´ -nk say -PAST-OBJ.1PL ‘we said it’ mond-a´ -tok say -PAST-OBJ.2PL mond-a´ -k say -PAST-OBJ.3PL

Interestingly, in complex tenses, the morphemes representing AgrO and AgrS appear between the aspect marker and the tense marker, on the aspect-marked V: (26)

a. Subjective Conjugation mond-t -am val-a say -PERF-1SG be-PAST mond-t -a´l val-a say -PERF-2SG be-PAST

mond-t -unk val-a say -PERF-1PL be-PAST mond-t -atok val-a say -PERF-2PL be-PAST


142

mondo-tt -0 val-a mond-t -ak val-a say -PERF-3PL be-PAST say -PERF-3SG be-PAST b. Objective Conjugation mond-t -a -m val-a mond-t -uk val-a say -PERF-OBJ.1PL be-PAST say -PERF-OBJ-1SG be-PAST ‘I had said it’ ‘we had said it’ mond-t -a -d val-a mond-t -a´ -tok val-a say -PERF-OBJ-2PL be-PAST say -PERF-OBJ-2SG be-PAST mond-t -a -0 val-a mond-t -a´ -k val-a say -PERF-OBJ-3PL be-PAST say -PERF-OBJ-3SG be-PAST In view of the Mirror Principle, the morpheme order in (26) means that the agreement projections intervene between AspP and TP, as follows:

(27)

TP T

AgrSP

AgrS

AgrOP AgrO

AspP

Asp

VP V

The problem is that in the case of simple tenses, involving no AspP projection, the agreement morphemes appear outside the tense morpheme – see (25), i.e., AgrOP and AgrSP presumably subsume TP. To resolve this contradiction, I tentatively assume that the agreement projections AgrOP and AgrSP always dominate the first projection extending the verb with an analytic suffix.

4.

T H E E M E R G E N C E O F S I T U AT I O N A S P E C T M A R K I N G

Although the complex tenses of Old Hungarian have survived up till now in the easternmost Csańgo´ and Sze´kely dialects of Hungarian, they gradually disappeared from standard Hungarian in the Middle Hungarian period (between 1526–1772). Their loss had been preceded by a period of neutralization of the functional differences of the various non-present forms. A few surviving texts also display signs of the loosening of


143

the formal system: in the Middle Hungarian period we sporadically find past perfect forms like mondo-tt vol-t, in which not only the verb stem bears the -t perfectivity marker but also the temporal auxiliary bears it instead of the -a past tense suffix. In literary prose, the various complex verb forms survived well into the 19th century, but they were only used for variety’s sake, or for an archaic or elevated effect. Present-day Hungarian has a single non-present tense, marked by -t, the Old Hungarian perfectivity marker reinterpreted as a past tense suffix. Parallel with the loss of its complex tense-aspect system, Hungarian has developed systematic means of marking telicity, the key feature of situation aspect. As was discussed in Chapter 2, there are three major types of telic predicates in Hungarian: (i) those expressing a delimited change of state, (ii) those describing a delimited change of location, and (iii) those expressing creation/coming into being. In the case of type (i) predicates, telicity is marked by a resultative verbal particle selected by the verb, e.g.: (28)

Jańos lenyı´rta /lenyı´rja a John down cut /down cuts the ‘John cut down/will cut down the grass.’

fu¨vet. grass

In the case of telic predicates of type (ii), telicity is expressed by a lexically selected terminative verbal particle, which often doubles a postverbal terminative noun phrase or PP: (29)

Jańos be -ment/ be -megy (a munkahelye´re). John in went/in goes his workplace-to ‘John went/will go to his workplace.’

If the verbal particles are omitted, these sentences denote atelic processes: (30)

a. Jańos nyı´rta /nyı´rja a fu¨vet. John cut /cuts the grass ‘John was/is cutting the grass.’ b. Jańos ment /megy a munkahelye´re. John went /goes his workplace-to ‘John was/is going to his workplace.’

In type (iii) telic sentences, telicity is a consequence of the fact that a verb expressing creation/coming into being is combined with a non-specific indefinite theme argument (see chapter 4 of this book). Compare: (31)

a.

Su¨to¨ttem su¨temeńyeket. baked-I cookies ‘I have baked cookies.’


144 b.

Su¨to¨ttem a su¨temeńyeket (amikor felhı´vta´l). baked-I the cookies when phoned-you.me ‘I was baking the cookies (when you phoned me).’

(31a), containing an indefinite object, is telic, whereas (31b), with a definite object, is atelic. The means of marking telicity, i.e., the category ‘verbal particle’ combinable with any process verb, on the one hand, and the definite/indefinite articles, on the other hand, are still missing in 12th-13th century Hungarian texts. In Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral speech and prayer’, the first surviving longer coherent Hungarian prose, written in 1192–95 (see Benko´´ 1980:47), none of the numerous change-of-state and change-of-location predicates has a verbal particle yet. The telicity of sentences expressing a bounded change is sometimes indicated by the perfective viewpoint aspect: (32)

es odutta vola neki paradicsumut and give-PERF-OBJ.3SG be-PAST him Paradise-ACC ‘And had given him Paradise for a house.’

hazoa´.9 house-for

When the -a/e-marked simple past is used, the boundedness of the sentence can only be inferred from the context: (33)

a. es veteve´ ´´t u ez munka´s vila´g bele´10 and throw-PAST.3SG him this laborious world into ‘and threw him into this laborious world’ b. kit u´r ez nopun ez homis vila´g timnu¨ce beleu¨l mente11 whom Lord this day-on this false world’s prison from save-past.3SG ‘whom the Lord saved from the prison of this false world on this day’

In present-day Hungarian, we would say, instead of odutta ‘gave’, veteve´ ‘threw’, and mente ‘saved’, the particle verbs oda-adta ‘thereto-gave’, be-vetette ‘in-threw’, and ki-mentette ‘out-saved’. Sentences expressing a bounded change of state also lack a particle in this 12th century text: (34)

a. es bulcsa´ssa mend u´´ and forgive-IMPER-OBJ.3SG all his ‘and he should forgive all his sins’ b. hugy ougya mend u´´ that solve-IMPER-OBJ.3SG all his ‘that he should resolve all his sins’ c. Es szobodochcha u´´t u¨rdu¨ng and save-IMPER-OBJ.3SG him devil’s

bu´ńe´t12 sin bu´ńe´t13 sin ildetuïtu´´l14 harassments-from


145

‘and he should save him from the devil’s harassments’ Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral Speech and Prayer’ contains a single verbal particle in an interesting function: (35)

Es oz gyimilcsnek u´l keseroü¨ vola vize, hugy turkokat so bitter was water-its that throat-their-ACC and that fruit-DAT mige´ szokosztja vol-a.15 PRT burst-IMPERF.OBJ.3SG be-PAST ‘And that fruit had such bitter juice that it was bursting their throats.’

In this example, also quoted under (14), the combination of past imperfective and a telicizing particle serves to denote the preparatory phase of an achievement situation which fails to culminate. In such sentences expressing a blocked achievement, both the imperfective inflection and the verbal particle play a role; the latter denotes the inherent boundedness of the situation, and the former, the incompleteness of it. Whereas in Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral Speech and Prayer’ from 1192–95, the use of the verbal particle is restricted to a very special context, a ´ magyar Ma´ria-siralom ‘Old Hungarian Lament of Mary’ hundred years later, in O (Benko´´ (1980:52)), delimited motion is already expressed by a particle plus V complex: (36)

Buábeleu¨l ki -nyuhhad16 grief-her-from out pull-IMPER-2SG ‘Pull her out from her grief’

In Saint Margaret’s legend from 1510, delimited changes of state are also marked by a verbal particle: (37)

a. az halakat megfaragja vala17 the fishes-ACC PRT carve-IMPERF.OBJ.3SG be-past ‘she would carve the fish’ b. az o´´ kezeńek bo´´re meg- hasadoz vala18 the her hand’s skin PRT crack.IMPERF.3SG be-PAST ‘the skin of her hands would crack’ Here is an example of a delimited change of location:

(38)

el - megyen vala az kapitulumha´zba19 off go-IMPERF.3SG be-PAST the capitulum-house-to ‘she would go to the capitulum house’ (Szent Margit e´lete [Saint Margaret’s life] (1510/1990:12))

146


Examples (37)–(38) are imperfective sentences denoting past habits. Their verbal particles serve to denote the boundedness of the individual actions, whereas the imperfective suffix expresses habituality. In the case of sentences in the historical past, the verbal particle denotes the boundedness of the situations. As the fables of Ga´bor Pesti from 1536, abounding in perfective verb forms, indicate, the particle also accompanies perfective verbs: (39)

Az mezo´´t, hogy ott o¨tte´l, mind el there eat-PERF-2SG all PRT the field-ACC that -pusztı´tottad nekem.20 destroy-PERF-OBJ.2SG me ‘By having eaten there, you have destroyed all the field for me.’ (Ga´bor Pesti, Az farkasro´l e´s ba´rańyro´l ‘On the wolf and the sheep’, (1536/1950:79b))

An interesting ‘‘snapshot’’ of the change taking place in the syntax of late Old Hungarian is provided by the so-called Mu¨ncheni emle´k ‘Mu¨nchen relic’, a Hungarian section in a codex written in the early 16th century presumably by a German monk interested in languages (for details, see Haader (2004)). The relic contains, among others, two versions of the prayers Pater noster and Ave Maria in Hungarian. The scribe first transscribed phonetically what he heard from a Hungarian person reciting the two prayers, and then he copied the prayers from a written Hungarian source. Whereas the written prayers, representing an earlier stage of Old Hungarian, contain practically no verbal particles, in the oral versions, representing the actual usage of the early 16th century, the verbs bocsa´t ‘forgive’ and szabadı´t ‘liberate’ already occur with the verbal particle meg. Since the 14th century, more and more accomplishment and achievement verbs have come to be lexicalized with a verbal particle, until we have reached the present stage when practically all verbs denoting a delimited change of state or location take a particle (at least in the spoken language). In the course of this development, simple particleless verbs denoting a change of state or change of location have assumed an unambiguously atelic reading. Verbs of creation and coming into being assume a telic interpretation if their theme argument, whose creation or coming into being is asserted, is a non-presupposed, non-specific indefinite noun phrase. In the following pair of examples representing present-day Hungarian, the (a) sentence is unambiguously atelic, and the (b) sentence is unambiguously telic: (40) a. Jo¨ttek vende´gek. came guests ‘Guests have come/came.’ b. (E´ppen) Jo¨ttek a vende´gek (amikor telefona´lta´l). (just) came the guests (when phoned-you) ‘The guests were coming (just when you phoned).’


147

The system of determiners that makes this type of aspectual distinction possible emerged in the late Old Hungarian period. The definite article, derived from the demonstrative pronoun, appeared in the 14th century; and the indefinite article, derived from the numeral one, appeared in the 15th century. In the Mu¨nchen relic from the early 16th century, the transcription of the oral Pater noster already says szabadı´ts meg minket a gonosztu´l ‘save PRT us the evil-from’, whereas in the more conservative written version gonosztu´l ‘evil-from’ has no article yet. The slow extension of the articles to their various present-day functions reached a stage resembling the usage of present-day Hungarian in the 16th century (cf. I. Gallasy (1991, 1992)). Thus at the beginning of the period of Middle Hungarian, both means of marking telicity, the verbal particle and the system of definite/indefinite determiners, were ready in the language. The emergence of the system of telicity marking was soon followed by the attrition of the system of complex tenses. Middle Hungarian was the period of the gradual loss of the functional differences between the complex verb forms marking both tense and viewpoint aspect. By the 19th century, the present perfect (e.g. mondo-tt) had been reinterpreted as a past tense, and the past tense forms, i.e., the simple past monda, the past imperfective mond vala, and eventually even the past perfect mondott vala came to be used as stylistic variants of the regular past tense marked by -t. That is, Hungarian, representing a language type with a grammaticalized viewpoint aspect in the period of Old Hungarian, went through a typological shift in the period of Middle Hungarian, as a result of which Modern Hungarian has become a language with a grammaticalized situation aspect. The syntactic structure of the Modern Hungarian sentence involves a PredP projection according to the evidence presented in Chapter 2. PredP dominates VP, with the V raised to Pred. Spec,PredP is filled alternatively by a verbal particle or a bare nominal complement. The PredP projection actually must have existed in the Old Hungarian period, as well. Although Old Hungarian – especially its earlier stages – used the verbal particle only sporadically, the use of non-referential, predicative bare nominals was wide-spread. They appeared preverbally, presumably in the same Spec, PredP where they appear in present-day Hungarian. For example: (41)

a. isa por es homu vogymuk21 indeed dust and ash are-we ‘indeed we are dust and ashes’ (Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral speech an prayer’ 1192–1195, in Benko´´ (1980:47)) b. Es oz gyimilcsnek u´l keseroü¨ vola vize22 and that fruit-DAT so bitter was water-its ‘And the juice of that fruit was so bitter ’ (Halotti besze´d e´s ko¨nyo¨rge´s ‘Funeral Speech and Prayer’ 1192–95, in Benko´´ (1980:47))


148

According to the evidence of these examples, Spec,PredP, the position harboring the verbal particle in present-day Hungarian, was already available when the verbal particle became widely used. When Spec,PredP came to be associated with telicity marking, the projection assumed a kind of aspectual function. (Although Csirmaz argues in chapter 6 for separate PredP and AspP projections in present-day Hungarian, there is string-vacuous movement from Spec,PredP to Spec,AspP, i.e., the two projections are as good as conflated.) This paved the way for the reinterpretation of the original AspP projection as a TP, with the -t/tt perfectivity marker reinterpreted as a past tense suffix, and for the disappearence of the original TP projection. A similar process must have taken place in various Slavic languages (except Bulgarian, which has both developed particle-like verbal prefixes, and has preserved its complex tenses (cf. Bertinetto 2001)). German may very well be a language undergoing this kind of type-shift at present. On the one hand, more and more German telic verbs are associated with a telicizing particle. On the other hand, the functional differences between the perfect and the imperfective past tenses have been neutralized, and speakers use the different non-present forms as stylistic variants in most dialects. The fact that the appearance of situation aspect marking was followed by the disappearance of viewpoint aspect marking both in Hungarian and in several other languages suggests that these two processes are related. Perhaps the overt marking of (a)telicity makes the systematic morphological marking of (im)perfectivity redundant; perhaps viewpoint aspect can be inferred from situation aspect in the unmarked case. This is the question that section 5 will address.

5.

CAN VIEWPOINT ASPECT BE INFERRED FROM S I T U AT I O N A S P E C T ?

It is a basic tenet of the two-component aspect theory of Smith (1991) that situationmarking and viewpoint-marking represent two independent subsystems of aspect. This is the view that Csirmaz holds concerning Hungarian in chapter 6, as well. She claims that a viewpoint aspect head with the value perfective or imperfective is present in every Hungarian sentence. Although it is phonologically unrealized, its presence is indicated by its interaction with time adverbials, and by its effect upon interpretation. The question is what clues enable Hungarian speakers to identify its value. The fact that the disappearance of the morphological marking of viewpoint aspect went on parallel with the grammaticalization of situation aspect marking suggests that the basic clues of identifying viewpoint must be provided by situation aspect. This section will argue that there is a fairly close correlation between situation aspect and viewpoint aspect; viewpoint aspect is predictable on the basis of situation aspect in the majority of cases. Hungarian has also developed certain lexical, morphological and syntactic means to help the identification of viewpoint in cases when the default correlation between situation type and viewpoint does not hold.


5.1.

149

The unmarked cases

The close correlation between telicity and perfectivity, and the somewhat looser correlation between atelicity and imperfectivity is common knowledge; several studies have undertaken the task of trying to disentangle these notions, and to separate tests of (a)telicity from those of (im)perfectivity (see e.g. Depraetere (1995) and Bertinetto (2001)). Telic predicates, describing inherently delimited events, i.e., achievements, and accomplishments with their culmination point attained, presuppose the perfective viewpoint. The defining property of telic events is their natural final point. The perfective, on the other hand, is defined by Smith (1991:103) as the viewpoint that includes the initial and final endpoints of a situation. Thus a telic sentence, e.g. (42), describing an event consisting of a change-of-state process and a resultant state, is also perfective, representing the event with its initial and end points included: (42)

Erzsi meg -ette a vacsora´t Elisa up ate the supper ‘Elisa ate up the supper.’

If the marker of telicity (the verbal particle referring to the resultant state of the theme) is omitted, the situation is not a telic accomplishment any longer; it is of the ‘process’ type, and its viewpoint is imperfective: (43)

Erzsi ette a vacsora´t Elisa ate the supper ‘Elisa was eating the supper.’

As opposed to accomplishments and achievements, states are imperfective in the unmarked case, as a typical state situation (see Smith 1991: 37) does not include the initial point and the endpoint of the given state. According to the definition of Smith (1991:111), ‘‘imperfective viewpoints present part of a situation, with no information about its endpoints’’. For example: (44)

Jańos tudja a matematika´t John knows the mathematics ‘John knows mathematics.’

It is only in the case of processes that the speaker can include or exclude the initial point and the endpoint of the situation in his/her viewpoint at will. In the case of processes involving a specific theme argument, like those in (43) and (45a,b), the fact that no verbal particle is spelled out indicates that the situation is unclosed, hence imperfective. (45)

a. Jańos ´ırja a dolgozatot. John writes the paper ‘John is writing the paper.’

150


b. A su¨temeńy su¨l. the cake bakes ‘The cake is baking.’ In the case of unergatives predicates, like those in (46), the possibility of a verbal particle is excluded; it is the bare verb that can be interpreted either perfectively or imperfectively.23 The present tense form makes the imperfective reading likelier (46a), whereas the past tense form favors the closed, perfective interpretation (46b) – although an appropriate time adverbial can also elicit the perfective reading of the present tense form (46c), and the imperfective reading of the past tense form (46d): (46)

a. Jańos teniszezik. John plays.tennis ‘John is playing tennis.’ b. Jańos teniszezett. John played.tennis ‘John played tennis.’ c. Jańos de´lutań teniszezik, aztań moziba megy. John afternoon plays.tennis then cinema-to goes ‘In the afternoon, John plays tennis, then he goes to the cinema.’ d. Jańos e´ppen teniszezett, amikor kerestem. John just played.tennis when searched-I-him ‘John was playing tennis when I looked for him.’

That is, in the case of unergative process predicates, tense provides a clue of viewpoint aspect interpretation; however, it can be overwritten by time adverbials, or by the context, e.g. that of a series of consecutive events. In sum, the major types of situations correspond to the following viewpoints in the unmarked cases: (47)

accomplishment ! perfective achievement ! perfective state ! imperfective process: unaccusative verbs ! imperfective unergative verbs ! imperfective or perfective 5.2.

The marked cases

5.2.1. Situation type shift In addition to the correspondences under (47), the literature also knows of telic predicates represented from an imperfective viewpoint, and states represented from a perfective viewpoint. I will claim that some of these cases can also be reduced to the correlations in (47), because their predicate has undergone a situation type shift – expressed by the regular morphological means of situation aspect marking – prior to viewpoint aspect interpretation. That is, certain


151

apparently perfective states are not states any more; they have been telicized beforehand. Similarly, there are also apparently telic imperfective sentences in the case of which the lexically determined telicity of the predicate is changed to atelicity before the imperfective viewpoint is imposed on it. Consider a few common cases of situation type shift: i. Situation type shift from state to accomplishment/achievement. Most stative predicates, e.g. van ‘be’, tud ‘know’, szeret ‘love, like’, la´t ‘see’, tud ‘know’ can be transformed by means of a resultative verbal particle into telic change-of-state predicates, for example: meg-van, meg-szeret, meg-la´t, meg-tud. As is demonstrated in detail in chapter 3, these predicates represent the state that they denote as a resultant state, attained as a consequence of a change. Whereas the stative predicates in the (a) sentences of (48)–(51) are imperfective, their telic counterparts in the (b) sentences are naturally perfective. Compare: a. Volt peńz az e´pı´tkeze´shez. was money the building.project-for ‘There was money for the building project.’ b. Ke´t he´ten belu¨l meg-volt a peńz az e´pı´tkeze´shez.24 two week within PRT was the money the building.project-for ‘The money was obtained for the building project within two weeks.’ (49) a. Jańos hala´laíg szerette E´va´t. John death-his-until loved Eve ‘John loved Eve until his death.’ b. Jańos ro¨vid ido´´ alatt meg-szerette E´va´t. PRT loved Eve John short time in ‘John came to love Eve in a short time.’ (50) a. Ga´bor la´tja a va´rat az ablaka´bo´l. Gabriel sees the castle the window-his-from ‘Gabriel sees the castle from his window.’ b. Ga´bor hirtelen meg-la´tta a bara´tja´t az ablaka´bo´l. saw his friend the window-his-from Gabriel suddenly PRT ‘Gabriel suddenly caught sight of his friend from his window.’ (51) a. Sańdor tudja az eredmeńyt. Alexander knows the result ‘Alexander knows the result.’ b. Sańdor tı´z perc alatt meg -tudja az eredmeńyt. PRT knows the result Alexander ten minute in ‘Alexander will learn the result in 10 minutes.’

(48)

ii. Situation type shift from accomplishment to process Accomplishment predicates are complex predicates consisting of a process component and a resulting state component. Imposing an imperfective viewpoint on them


152

means the deletion of the resulting state component marked by the verbal particle. The resulting predicate is non-distinct from an imperfective process predicate. Compare:25 (52)

a. Istvań Stephen ‘Stephen b. Istvań Stephen ‘Stephen

fe´l o´ra alatt el -olvasta az u´jsa´got. PRT read the newspaper half hour in read the newspaper in half an hour.’ fe´l o´ra hosszat olvasta az u´jsa´got. half hour long read the newspaper was reading the newspaper for half an hour.’

In the case of achievements, the initial point of the event is simultaneous with the attainment of the resultant state, hence the omission of the second subevent does not leave an imperfective process predicate. Compare: (53)

a helyes va´laszt. found-I the correct answer ‘I found the correct answer.’ b. *Tala´ltam a helyes va´laszt (amikor az o´ra ve´gete´rt). found-I the correct answer (when the class ended) ‘*I was finding the correct answer (when the class ended).’

a. Meg-tala´ltam PRT

iii. Situation type shift from accomplishment/achievement to state The core predicate in the following sentences is a telic accomplishment; however, quantification over it creates an open, atelic situation. (54a) involves a universal quantifier, whereas (54b) involves a habitual operator manifested in the combination of present tense and a durative time adverbial. The viewpoint aspect of the sentences is, naturally, imperfective: (54) a. E´va fe´rje hala´la o´ta minden nap ki-megy a temeto´´be. Eve her.husband’s death since every day out goes the cemetery-to ‘Since her husband’s death, Eve has been going to the cemetery every day.’ b. Jańos ke´t hete haza-kise´ri Marit iskola utań. John two week-for home accompanies Mary school after ‘For two weeks, John has been seeing Mary home after school.’ 5.2.2. Syntactically encoded situation type–viewpoint mismatches a. Incomplete accomplishments as telic imperfectives If the verbal particle, denoting the resulting state of an accomplishment, also contributes to the lexical meaning of the predicate, it cannot simply be cut off in order to derive an imperfective process. Nevertheless, such accomplishments can also be represented from the imperfective viewpoint, with the


153

endpoint of the event outside the speaker’s perspective – by syntactic means. As was shown in chapter 6, an imperfective Asp head does not tolerate a delimiting (resultative or terminative) element in its specifier. Hence in the case of a PredP involving a telic particle verb, with the particle occupying Spec,PredP, the EPP feature of the imperfective Asp head can be satisfied in the most economic way by V movement from Pred to Asp, instead of particle movement from Spec,PredP to Spec,AspP. The V-to-Asp movement crosses the particle in Spec,PredP, yielding a ‘V particle’ order, which is indicative of the imperfective viewpoint. Compare the following sentences. The perfective value of the Asp head of (55a) and (56a) is tested by a completive time adverbial, whereas the imperfective value of the Asp head of (55b) and (56b) is ensured by a durative time adverbial. (55)

a. Ge´za tı´z perc alatt [AspP feli [PredP ti ment [VP tj ti a toronyba]]] Geza ten minute in up went the tower-to ‘John went up to the tower in ten minutes.’ b. Ge´za e´ppen [AspP menti [PredP felj ti [VP ti tj a toronyba]]] Geza just went up the tower-to amikor meg-la´ttam when PRT saw-I-him ‘John was just going up to the tower when I caught sight of him.’

(56) a. Rozi fe´l o´ra alatt [AspP o¨sszei [PredP ti ´ırta[VP tj ti a neveket]]] Rosy half hour in together wrote the names ‘Rosy wrote down the names in half an hour.’ b. Rozi ege´sz de´lutań [AspP ´ırtai [PredP o¨sszej ti [VP ti tj a neveket]]] Rosy whole afternoon wrote together the names ‘Rosy was writing down the names the whole afternoon.’ The imperfective viewpoint aspect of such telic constructions can also be corroborated by lexical means: the postverbal particle can be supplemented with the morpheme -fele ‘-wards’, which apparently serves to indicate that the given element is a goal to be aimed at rather than a terminus attained. That is: (57)

a. Ge´za e´ppen ment felfele a toronyba, amikor megla´ttam. ‘Geza was just going upwards in the tower when I caught sight of him.’ b. Rozi ege´sz de´lutań ´ırta o¨sszefele a neveket. ‘Rosy was writing downwards the names the whole afternoon.’

In the case of achievements, the imperfective viewpoint is not possible. Since the process and the resultant state are inseparable, simultaneous subevents of achievements, the perspective cannot focus on the process alone, excluding the


154

resultant state: hegycsućsot]]] (58) a. *Jańos[AspP e´rtei PredP elj ti [VP ti tj a PRT the mountaintop John reached amikor ki-to¨rt a vihar. when out broke the storm ‘John was reaching the mountaintop when the storm broke out.’ b. *Fleming e´ppen [AspP tala´ltai [PredP felj ti [VP ti tj a penicillint]]] PRT the penicillin Fleming just invented amikor . . . when . . . ‘Fleming was just inventing penicillin when . . . ’ The following predicates do contain an independent, albeit short, process component separable from the resultant state, so they are accomplishments rather than achievements; that is why they are compatible with an imperfective Asp head: (59)

a. Jańos [AspP csuktai [PredP be(fele)j ti [VP ti tj az ajto´t]]] John closed in(wards) the door ‘John was closing the door, when . . . ’ b. Miklo´s e´ppen [AspP kapcsoltai [PredP be(fele)j ti [VP ti tj a Nicholas just turned in(wards) the amikor . . . when . . . ‘Nicholas was just turning on the TV when . . . ’

amikor . . . when . . . te´ve´t]]] TV

ii. Imperfective adverbial participle constructions derived from telic predicates The following constructions assert a state which is the resultant state of a previous accomplishment or achievement.26 (60)

a. A the ‘The b. Az the ‘The c. A the ‘The

pad be van festve. bench in is paint-ADV bench is painted/is in a painted state.’ ajto´ be van csukva. door in is close-ADV door is closed/is in a closed state.’ leve´l meg volt ´ırva. was write-ADV letter PRT letter was written/was in a written state.’

I assume that this construction is a raising construction, involving a copula functioning as a raising verb and an adverbial small clause, that is, e.g. (60a) is derived from the following structure: (61)

van [AdvP a pad be-festve]

Although the predicate of the small clause is derived from an accomplishment or achievement verb, this verb is buried under an AdvP. The head of AdvP is the


155

suffix -va/ve realized on the verbal head, which derives a participle phrase functioning as a stative adverbial from the PredP: (62)

van [AdvP

-ve [PredP1

bei

festj [VP1 ti tj

a pad]]]

A pad is first raised into the subject position of the adverbial small clause (as in (61)), and then into the subject position of the matrix predicate. The verbal particle is moved into the matrix Spec,PredP – see (63a), unless the matrix verb is preceded by a focus or a negative particle – see (63b): (63)

a. A padi [AspP bej [PredP2 tj vank [VP2 tk [AdvP ti tj festve]]]] b. A padi CSUŃYAŃ van [AdvP ti be -festve] PRT painted the bench uglily is ‘The bench is painted IN AN UGLY WAY.’

(As has been made clear by Szendro´í (2003), van, a light verb, cannot bear phrasal stress, therefore it cannot occupy the leftmost position of a predicate phrase, the locus of main stress, but must be preceded by a verbal particle, a focus, or a negative particle.) The adverbial small clause in (60)–(63) obviously has no AspP projection. Only the stative matrix predicate projects an AspP, which is understood as imperfective, in accordance with the correlations in (47). The verbal particle, subordinated to the AdvP denoting a state resulting from a former – possibly much earlier – telic event, does not have a delimiting role; it serves to express the complete affectedness of the theme. Compare: (64)

a. A pad the bench ‘The bench b. A pad the bench ‘The bench

festve van painted is is painted/has paint on it.’ be van festve. in is painted is (completely) painted.’

Not having the feature terminative or resultative, the particle can freely move into the specifier of the imperfective Asp head. In Hungarian, the copula has not only an imperfective variant (van), but also a perfective one (lesz). This seems to be a relic of the earlier viewpoint-aspect marking period of the language. If the matrix predicate of e.g. (60c) is replaced by the past form of lesz, then, naturally, the viewpoint aspect of the sentence is perfective. (The present tense form of the perfective copula, lesz, is ambiguous; it not only means perfective present, but also functions as the future tense form of the imperfective copula.) (65)

A leve´l ke´t perc alatt meg lett ´ırva. PRT be.PERF-PAST write-ADV the letter two minute in ‘The letter has been/was written in two minutes.’

156


The difference between the viewpoint aspects of (60c) and (65) confirms that a resultative or terminative verbal particle generated in a VP subordinated to an AdvP projection is aspectually inert in the matrix cycle, where it is raised only for phonological reasons. That is, sentences like (60c) involve no real mismatch between a telic situation and an imperfective viewpoint; telicity and imperfectvity are associated with different cycles of the syntactic structure.

5.

CONCLUSION

This chapter has argued that in the period of Middle Hungarian, the Hungarian language has gone through a typological shift: it has developed a system of marking telicity, and parallel with that process, it has lost its complex tense-aspect system marking viewpoint aspect. It has been demonstrated that the lack of overt viewpoint aspect marking in present-day Hungarian does not result in the loss of any aspectual information. In the unmarked case, predicates marked as telic have the perfective viewpoint, whereas stative atelic predicates have the imperfective viewpoint.27 There is a single situation type, the process, which has no default viewpoint belonging to it. More precisely, unergatives, which cannot be telicized, and transitive and unaccusative verbs behave differently in this respect. As for unergatives, a present tense form is understood as imperfective, and a past tense form is understood as perfective in the unmarked case. In the case of transitive and unaccusative predicates, which have the ability to take a telicizing verbal particle, the imperfective reading is the unmarked choice in both tenses. Seemingly perfective states and seemingly imperfective telic situations may arise as a consequence of a situation type shift. For example, perfective predicates like meg-szeret valakit ‘come to like someone’, meg-tud valamit ‘come to know something’ involve an atelic verb turned telic by means of a resultative particle, thus their perfectivity is a consequence of the telicity they have assumed. The situation type shift from an accomplishment to a process, yielding an imperfective viewpoint, often consists in the deletion of the particle marking the resultant state component of the accomplishment. In another type of imperfective telic constructions, quantification over an accomplishment or achievement results in an open, atelic situation requiring an imperfective interpretation. Imperfective – telic mismatches are sometimes encoded by syntactic means. For example, an accomplishment or achievement verb can be transformed into a stative adverbial by the addition of an adverbial participle suffix. If the resulting participle phrase is subordinated to a stative raising verb, the sentence has an imperfective viewpont. Although the verbal particle may eventually end up in the Spec,AspP position of the matrix raising predicate, it has no delimiting (resultative or terminative) function, therefore it is compatible with an imperfective Asp head. Imperfective accomplishments, in the case of which the perspective extends only over (a part of) the process subsection of the event, are marked by a ‘V, particle’ order instead of the regular ‘particle, V’ order – because the imperfective


157

Asp head does not tolerate a delimiting (terminative or resultative) element in its specifier.

NOTES

2

3

4 5

6

7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

24

The spelling of quotations has been modernized; the original spelling is provided in footnotes. Here is the original spelling of (7):Latiatuc feleym zumtuchel mic vogumuc. Orpha megapola o: napat m_eg‘fordola Rvt e_ gge`so: le` o: napaual Kin.c moda Noemi Ime te rokonod meg‘fordolt. IAy neekem merth haragossaa een ellenem tewttem, een meg waaltoomath . . . Igaz wtath . . . el hattam, ees az yaarathlan wtakon mezze . . . forgottam. Eny¨ngen y¨ram ez lewelet penteken. Melle˛ie gyu¨lueń a` mezoï egerec ott ia´tzadoznac vala ko¨rnyu´´le. E´s a` kergette´sbe az egyic reaía szeckeleć az oroszlańra. Ez nemesseges zent zvz magery kyralnak nemes leyanyä .hetet tart vala az cohnyän fevz vala az sororoknak .fazekat mos vala .talakat mos vala .az halakat meg farayä vala . . . Es meg e`mle`kezec Pe´ter az ig_ero: l kit mondot vala ne`ki ic. meg‘ leltem e`n iuhomat ki e`l ve`zetuala. es odutta vola neki paradisumut hazoá´ es veteve wt ez muncas vilagbele. kit vr ez nopun ez homus vilag timnucebelevl mente. es bulscassa mend w bunet. hug ovga mend w bunet. Es zoboducha wt urdung ildetuitvl. Es oz gimilsnec vvl keseruv uola vize. hug turchucat mige zocoztia vola. buabeleul kyniuhhad. az halakat meg farayä vala. az ev kezenek bevre meg hasadoz vala. el megyen vala az capitulum hazba. Az mezewt hogy¨ ot ewttel my¨nd el puzty¨tottad nekem. y´sa pur es chomuv uogmuc. Es oz gimilsnec vvl keseruv uola vize. Recall from chapter 2 that the verbal particle is a secondary predicate predicated of the theme argument; hence it is not licensed in the case of verbs taking no theme. The particle verb meg-van ‘PRT-is’ is, in fact, ambiguous. It can also be used statively, to denote the existence of a specific subject, e.g.: $

1

(i) Jańosnak meg-van az Anna Karenina. John-DAT PRT is the Anna Karenina ‘John has Anna Karenina.’ 25

26

27

A similar proposal was put forward by Pinõn (1995). In the derivation of the progressive aspect proposed by him, the first step is performed by a PR operator which creates process predicates from event predicates by separating the process stage of events. The construction in question resembles the passive, yet it is more restricted than e.g. the English passive. Its aspect is fixed (but see the discussion of (65)), and only verbs denoting a delimited change-of-state can occur in it – see e.g. Alberti (1997). A similar point is made by Bohnmeyer and Swift (in press).

ANIKO CSIRMAZ

A C C U S AT I V E C A S E A N D A S P E C T

This chapter explores the effect of non-theta marked accusative constituents on aspect. A significant body of research (Tenny 1994, Wechsler and Lee 1996, Svenonius 2002, Kratzer 2004, among others) argues that properties of objects or other accusative case marked constituents can affect the aspectual properties of events. I show that Hungarian non-theta marked accusative constituents do interact with aspectual properties, albeit in different ways. From among these constituents, reflexive pronouns and body parts are required by a constraint on the independently established aspectual structure of events (as argued by Rappaport Hovav and Levin 2001 for English). Other accusative constituents, however, directly affect the properties of events. The latter type of non-theta marked constituents are delimiters and thus yield telic, delimited event descriptions. Analyzing accusative non-theta marked constituents as delimiters also permits an account of the aspectual restrictions on the events modified by these adjuncts. In addition to the aspectual characterization, the chapter also provides a discussion of other properties of non-theta marked accusative constituents. The properties addressed include the argument or adjunct status, case marking of these elements, as well as their interaction with the left periphery of Hungarian clauses. The chapter is organized as follows. The first section discusses non-theta marked reflexives and body parts. A brief general description of structurally case marked adjuncts follows. The remaining sections focus on accusative situation delimiters: temporal and spatial measure phrases, multiplicatives and predicate modifying adjectives. 1.

R E S U LTAT I V E S

AND REFLEXIVES

Within non-theta marked accusative constituents, reflexive pronouns and body parts must be distinguished from other accusative constituents. The latter include temporal and spatial phrases, as well as multiplicatives and accusative adjectives. While other accusative constituents affect the aspectual properties of events, accusative reflexives and body parts do not influence the aspectual properties directly. Rather, their presence is enforced by a restriction on the aspectual structure of the event (as argued for reflexives and other non-theta marked accusative constituents in Engish by Rappaport Hovav and Levin 2001). Non-theta marked reflexives and other non-argument objects in English have been discussed extensively in the generative literature (Simpson 1983, Levin and Rappaport Hovav 1995, Rappaport Hovav and Levin 2001, among others). These ‘fake’, or non-theta marked reflexives obligatorily appear as objects of the predicate, 159 Katalin E´. Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 159–200. 2008 Springer.

160

ANIKO CSIRMAZ

but lack a thematic role. With certain predicates, discussed below, the appearance of these fake reflexives is obligatory in the presence of a resultative constituent. (1) Juli rekedtre kiaba´lta J-NOM hoarse-onto shouted ‘Juli shouted herself hoarse’

maga´t self-ACC

Similarly to English, Hungarian fake reflexives only appear in the presence of resultatives. If no resultative phrase is present, then these reflexives are ungrammatical. (2) Juli kiaba´lt J-NOM shouted ‘Juli shouted’

(* maga´t)1 herself-ACC

I argue that the distribution of these constituents is constrained by an event structure restriction, which requires the presence of an argument for each subevent. In this account, I adopt the proposal of Rappaport Hovav and Levin 2001, who make the same claim for English reflexives and other non-theta marked objects in English In the remainder of this section I show that an account based on a syntactic constraint, the Direct object restriction (Levin and Rappaport Hovav 1995), is compelling at first sight. Then I present arguments for accounting for fake reflexives by adopting a constraint on event structure instead of a syntactic restriction. 1.1.

Syntactic account

Building on Simpson 1983, Levin and Rappaport Hovav 1995 propose the Direct Object Restriction (DOR) to account for the distribution of fake reflexives in English. They note that resultatives appear to be predicated of direct objects, which can be stated as a requirement of the former. (3)

Direct Object Restriction (DOR) Resultatives are predicated of an NP in object position, whether the object is an argument of the verb or not

The DOR is operative prior to movement, and is satisfied by the only overt argument of passives and unaccusatives as well as the direct object of transitive predicates. The three structures are illustrated below for Hungarian. (4)

a. Juli darabokra to¨rte a va´za´t J-NOM pieces-onto broke the vase-ACC ‘Juli broke the vase into pieces’ b. A va´za darabokra van to¨rve2 broken the vase-NOM pieces-onto is ‘The vase is broken into pieces’

(transitive)

(passive)

161


c. A va´za darabokra to¨rt the vase-NOM pieces-onto broke ‘The vase broke into pieces’

(unaccusative)

In contrast, no argument of an unergative predicate is present in the object position. A resultative is ungrammatical, as predicted by the DOR. It is possible, however, for a nonargument object to appear with unergatives and satisfy the restriction. In fact, as noted above, the non-argument object is required for unergatives modified by a resultative. (5) Juli betegre nevette *(maga´t) J-NOM sick-onto laughed self-ACC ‘Juli laughed herself sick’ In addition to reflexives, inalienable body parts can also appear as non-argument objects to satisfy the DOR. (6) Juli vo¨ro¨sre sı´rta *( a szeme´t) J-NOM red-onto cried the eye-POSS,ACC ‘Juli cried her eyes red’ As noted by E´. Kiss (chapter 2), resultative particles3, which denote a result state in some abstract sense, have the same effect as lexical resultatives. As predicted, resultative particles can occur with transitive, passive and unaccusative predicates (7). With unergative predicates, a fake reflexive or a non-argument body part must also be present (8). a. Juli o¨ssze to¨rte a va´za´t J-NOM together broke the vase-ACC ‘Juli broke the vase (to pieces)’ b. A va´za o¨ssze van to¨rve the vase-NOM together is broken ‘The vase is broken (to pieces)’ c. A va´za o¨ssze to¨rt the vase-NOM together broke ‘The vase broke (to pieces)’ (8) a. Juli el sı´rta *(maga´t) self-ACC J-NOM away cried ‘Juli started crying’ b. Juli el nevette *(maga´t) J-NOM away laughed self-ACC ‘Juli started laughing’ c. Juli ki sı´rta *( a szeme´t) the eye-POSS,ACC J-NOM out cried ‘Juli cried her eyes out’

(7)

162

ANIKO CSIRMAZ

Apart from resultatives, goals also denote an end state. If the event culminates, then the goal is reached. Let us consider whether goals, which also induce the DOR, require fake reflexives or body parts similarly to resultatives. Levin and Rappaport Hovav 1995 discuss intransitive motion verbs. They argue that motion verbs in some languages, including English, German and Dutch, show variability, behaving as either unaccusative or unergative predicates. Directed motion verbs are unaccusative, while manner of motion verbs are unergative. Thus whenever a motion verb is modified by a goal – and is therefore a directed motion predicate – it is expected to be grammatical without a nonargument object. Manner of motion verbs, modified by a resultative rather than a goal, require a fake reflexive or body part. The dual classification of motion verbs also holds for Hungarian. The ungrammaticality of fake reflexives with directed motion verbs in (9a) is consistent with the unaccusativity of directed motion verbs. Verbs expressing manner of motion, which are not modified by a goal adjunct, show a different behavior. If fut (‘run’) is modified by the resultative particle ki – meaning ‘to a sufficient extent’ rather than a spatial goal – then the fake reflexive is obligatory, as predicted by the unergative status of the predicate (9b). (9)

a. Juli a ce´lig J-NOM the goal-until ‘Juli ran to the goal’ b. Juli ki futotta *( J-NOM out ran ‘Juli ran herself tired’

futotta ran

(* maga´t) self-ACC

maga´t) self-ACC

Similarly to lexical goals, particles which denote a goal (‘terminative’ particles in the terminology of E´. Kiss (chapter 2)4) appear with motion verbs and do not require a non-theta marked object. (10)

a. Juli haza ment (* maga´t) self-ACC J-NOM home went ‘Juli went home’ b. Juli le futott (a partra) (* maga´t) the bank-onto self-ACC J-NOM down ran ‘Juli ran to the riverbank’

While the DOR and the variable status of motion verbs appears to account for the distribution of fake reflexive objects in Hungarian, a revision is in order. The next section discusses counterarguments to the DOR. It also introduces an account of DOR-effects that appeals to the correlation of subevents and arguments (Rappaport Hovav and Levin 2001).


1.2.

163

The correlation between subevents and arguments

As noted above, non-argument objects (fake reflexives and body parts) are required in certain Hungarian resultative constructions. While DOR, a syntactic restriction, appears to account for the data, it can be shown that an account based on the aspectual properties of events – in line with Rappaport Hovav and Levin 2001 – is superior. The DOR cannot derive the distribution of non-theta marked reflexives and body parts, since resultatives are not always predicated of the object. As shown by Rappaport Hovav and Levin 2001 for English, some resultatives and goals describe the end state of the subject rather than an overt object. The same is true for Hungarian. Eszes (chapter 3), adopting the description of Kiefer 2000, argues that a stative predicate, prefixed with a perfective particle meg, describes the end state of the subject. The resulting complex predicate, which includes the prefix, is resultative: the state described by the lexical verb szeret (‘like’) starts at the end of the event denoted by the complex predicate. Furthermore, the resulting state holds for the subject rather than the object, in violation of the DOR. (11)

a. Jańos meg szerette Marit Mari-ACC J-NOM perf liked ‘Jańos came to like Mari’ b. Jańos meg tudta az eredmeńyt J-NOM perf knew the result-ACC ‘Jańos came to know the result’

Lexical goal adjuncts can also denote the final state of the subject rather than the object.5 (12)

a. A hajo´sok a Horn-fokig ko¨vette´k the mariners-NOM the Horn cape-until followed a ra´dioútası´ta´sokat the radio orders-ACC ‘The mariners followed the radio orders to Cape Horn’ b. Juli cigańykerekeket hańyt ege´szen az ajtoíg did straight the door-until J-NOM cartwheels-ACC ‘Juli did cartwheels straight to the door’

In (12a), it is the mariners that arrive at Cape Horn. Radio signals do not need to be anywhere close to Cape Horn – they may be transmitted from a control room in the other hemisphere. Similarly, the object in (12b) does not denote an entity – and consequently it cannot denote the entity that is in the final state described by the resultative or goal.

164

ANIKO CSIRMAZ

As the preceding examples illustrate, the DOR does not hold universally, and so cannot be then regarded as the source of the object requirement of resultative constructions. Rappaport Hovav and Levin 2001 argue that the variable behavior of motion verbs – where they either pattern as unaccusatives or unergatives – also suggests that an account other than the DOR should be adopted. As noted above, motion verbs show variable behavior: if they denote a directed motion, then they are unaccusative. In absence of a directed interpretation, when the motion verbs describe a manner of motion, they behave as unergative predicates (Levin and Rappaport Hovav 1995). In order to account for the required non-theta marked object and the variable behavior of motion verbs, Rappaport Hovav and Levin 2001 (following Grimshaw and Vikner 1993 and van Hout 1996) introduce a correspondence between subevents and arguments. (13)

Argument per subevent condition There must be at least one argument XP in the syntax per subevent in the event structure6

Given the Argument per subevent condition (ASC), it is the number of subevents rather than the reference of the resultative that enforces the presence of objects. Let us consider how the effects of the DOR are derived, by adopting the arguments of Rappaport Hovav and Levin 2001 for Hungarian. (14)

Juli a´lomba sı´rta maga´t J-NOM sleep-into cried self-ACC ‘Juli cried herself to sleep’

(14) contains two subevents: (a) Juli cried, and (b) as the consequence of crying, she fell asleep. The two subevents are, to some extent, independent, which can be shown by adverbial modification. Following Rappaport Hovav and Levin 2001, if the resultative a´lomba sı´r (‘cry to sleep’) is modified by the adverb quickly, then the adverb can only be understood as modifying the second result subevent. As the following paraphrase shows, while crying can take a long time, falling asleep happens quickly. In addition, the verb cannot be modified by quickly in absence of the resultative. (15)

(16)

Juli gyorsan a´lomba sı´rta maga´t J-NOM quickly sleep-into cried self-ACC ‘Juli quickly cried herself to sleep’ Juli o´ra´kig sı´rdoga´lt, de aztań gyorsan a´lomba but then quickly sleep-into J-NOM hours-until cried sı´rta maga´t cried self-ACC ‘Juli was crying for hours, but then quickly cried herself to sleep’


(17)

165

*Juli gyorsan sı´rt J-NOM quickly cried ‘Juli cried quickly’

Similarly, for other examples where a non-theta marked object appears, it can be shown that there are two independent subevents. Given the ASC, the existence of two subevents requires the presence of an additional ‘argument’. In contrast, whenever an intransitive predicate denotes a single subevent, no reflexive appears. This accounts for the lack of fake reflexives with unaccusative verbs. (18) A va´za darabokra the vase-NOM pieces-onto ‘The vase broke into pieces’

to¨rt broke

(* maga´t) self-ACC

Rappaport Hovav and Levin 2001 also argue that the variable behavior of motion verbs can be described by appealing to their aspectual properties. If the motion is directed towards a goal, then the goal is attained as a direct result of that motion. The subevents are temporally dependent, and do not count as distinct subevents for the ASC, which can be satisfied by the unique argument. If the motion verb occurs with a resultative and describes a manner of motion, then the result subevent is temporally distinct from the process subevent. The ASC requires two arguments, resulting in a non-theta marked reflexive or body part. While the variable behavior of motion verbs can be described by appealing to the ASC, the unaccusative – unergative distinction can be shown to exist as well. In other words, directed motion verbs are unaccusative and manner of motion verbs are unergative, as noted earlier. The different argument structure of these predicates is shown by auxiliary selection facts. In Germanic and Romance languages, directed motion verbs, like other unaccusative predicates, occur with the auxiliary be; manner of motion verbs, as well as other unergative verbs, with have (Rosen 1984, Zaenen 1993, Levin and Rappaport Hovav 1995, Hoekstra 1988, Chierchia 2004).7 (19)

a. Hij he ‘He b. Hij he ‘He

heeft / *is gelopen has is run ran’ is / ?heeft naar huis gelopen is has to home run ran home’ (Zaenen 1993)8

The auxiliary selection facts show that motion verbs are unaccusative and unergative, independently from the interpretation of the subevents. Since there is a distinction in terms of argument structure, the behavior of motion verbs fails to provide an argument for the ASC over the DOR, contrary to Rappaport Hovav and Levin 2001.

166

ANIKO CSIRMAZ

The inadequacy of the DOR is shown, however, by examples such as (12). Thus while the ASC should be adopted over the DOR, the former does not account for all the data discussed above. In sum, the distribution of non-theta marked reflexives and body parts in Hungarian can be described by adopting the treatment of Rappaport Hovav and Levin 2001 proposed for English. The presence of these constituents is required by the Argument per Subevent Condition, a constraint which requires at least one argument for each subevent in the event structure. 1.3.

Arguments and case checking

The preceding discussion shows that the presence of non-theta marked reflexives and body parts is related to the ASC, a constraint on arguments and event structure. This section explores some properties of these constituents, specifically, the argument status and the nature of case marking of fake reflexives and body parts. Let us first consider whether constituents required by the ASC are arguments or adjuncts. The argument or adjunct status of a constituent can be ascertained by whether it is an island for extraction. For the associates of particles9, U¨ro¨gdi 2003 argues that they are arguments, allowing extraction. In absence of a particle, non-theta marked goals and locatives are adjuncts and form an island for movement. (20)

Kinek ´ırtad ra´ a verset who-DAT wrote-2SG onto-3SG the poem-ACC a t sı´rko¨ve´re? the gravestone-onto ‘On whose gravestone did you write the poem?’ b. *Kinek ´ırtad meg a verset wrote-2SG perf the poem-ACC who-DAT a t sı´rko¨ve´re? the gravestone-onto ‘For whose gravestone did you write the poem?’ (U¨ro¨gdi 2003)

a.

As (21) shows, extraction out of body parts is marked. (21)

# Kinek sı´rta vo¨ro¨sre [t a the who-DAT cried red-onto ‘Whose eye did Juli cry red?’

szeme´t] Juli? eye-poss-ACC J-NOM

The ungrammaticality of extraction, however, does not necessarily indicate that the body part is an adjunct. The possessor of the body part is lexically determined to be coreferential with the subject; (21) is then pragmatically marked, on par with (22).

167


(22)

# Kinek vesztette el [t a away the who-DAT lost ‘Whose head did Juli lose?’

feje´t] head-POSS-ACC

Juli? J-NOM

As the notation in the examples above also indicates, the nature of the ungrammaticality in (20b) and (21) is different. While (20b) is an instance of island violation, (21) is pragmatically marked. I take this difference to indicate that non-theta marked body parts, and presumably also reflexives, are arguments. If these constituents are arguments, then it may be expected that their case is licensed similarly to that of other object arguments as well. I argue below that accusative reflexives and body parts, as well as theta-marked objects, are structurally case marked.10 Theta marked objects do not show an entirely parallel behavior with non-theta marked accusative reflexives and body parts. The former show case alternation; they appear with nominative rather than accusative case marking in impersonal passive constructions (23). The derived subject in an impersonal passive must, however, be affected (Alberti 1998). This affectedness requirement rules out a number of theta marked arguments, such as the theme in (23b), as derived subjects. (23)

a.

A kapu be van festve the gate-NOM in is painted ‘The gate is / has been painted’ b. *Mari meg van la´tva M-NOM perf is seen ‘Mari has been seen’ (E´. Kiss 2002a)

Due to the affectedness requirement, non-theta marked accusative constituents, including accusative multiplicatives, also fail to appear as derived subjects in impersonal passives: (24)

a. Jańos oda u¨to¨tt egyet one-ACC J-NOM there hit ‘Jańos hit (it)’ b. *oda van egy u¨tve there is one-NOM hit ‘It is hit (once)’

The lack of case alternation with accusative adjuncts thus cannot be interpreted as an argument against structural case marking. Definiteness agreement, however, provides an argument for the structural case marking of Hungarian non-theta marked accusative constituents. All accusative constituents trigger definiteness agreement. Given an overt object, the finite verb agrees with the object in definiteness (Bartos 1999, Den Dikken 2004, among others). Definite objects trigger definite agreement (25a). If the object is

168

ANIKO CSIRMAZ

indefinite or the predicate is intransitive, the verb shows indefinite agreement (25b–d). (25)

a. Jańos olvasta a ko¨nyvet J-NOM read-DEF-PAST.3SG the book-ACC ‘Jańos read the book’ b. Jańos olvasott ( egy ko¨nyvet) book-ACC J-NOM read-INDEF-PAST,3SG a ‘Jańos read (a book)’ c. Jańos ke´t ko¨nyvet olvasott J-NOM two book-ACC read-INDEF-PAST,3SG ‘Jańos read two books’ d. Jańos aludt J-NOM slept ‘Jańos slept’

Let us assume that definiteness agreement is tied to accusative case checking, similarly to phi-feature agreement and nominative case.11 Definiteness agreement involves the object checking a definiteness feature on v. v, in turn, checks an abstract case feature of the object. Under this view, definiteness agreement with an accusative adjunct provides an argument for case licensing by v, and hence for structural case marking. Non-subcategorized reflexives trigger definiteness agreement with the verb. Reflexives and body parts are thus parallel to objects in both the position they are merged and case marking: (26)

a. Jańos rekedtre *kiaba´lt/ J-NOM hoarse-onto shouted-INDEF-PAST,3SG/ kiaba´lta maga´t shouted-DEF-PAST,3SG self-ACC ‘Jańos shouted himself hoarse’ b. Juli vo¨ro¨sre *sı´rt / sı´rta J-NOM red-onto cried-INDEF-PAST,3SG / cried-DEF-PAST,3SG a szeme´t the eye-POSS-ACC ‘Juli cried her eyes red’

As the data show, Hungarian fake reflexives and body parts, similarly to their English counterparts, are structurally case marked objects. As argued above, their presence is required by the ASC. The distribution and properties of reflexives and body parts significantly differ from those of other non-theta marked accusative constituents in Hungarian. In the following sections I show that other accusative constituents, in spite of forming a heterogeneous set, are all situation delimiters. As delimiters, they impose a


169

non-delimitedness restriction on the aspectual properties of the predicates that they modify. These accusative constituents sharply contrast with reflexives and body parts, whose presence is required by the properties of the event predicate. The following section offers a brief discussion of accusative adjuncts as situation delimiters. The following sections discuss various types of delimiters: in section 3, I examine accusative temporal and measure phrases, while sections 4 and 5 address accusative multiplicatives and adjectives, respectively.

2.

A C C U S AT I V E A D J U N C T S A R E S I T U AT I O N D E L I M I T E R S

As noted above, fake reflexives and body parts are not the only non-theta marked constituents which can appear with accusative case in Hungarian. Other accusative constituents without theta roles include temporal and measure phrases as well as multiplicatives and adjectives. Before delving into the details of Hungarian adjuncts in more detail, let us briefly discuss the range adjuncts which can appear with structural case marking. Structurally case marked adjuncts can appear in a variety of typologically unrelated languages, including Korean, Chinese, Russian, Polish and Greek. These adjuncts all (a) measure the duration of the event, (b) measure a distance covered during the event, or (c) count occurrences of the event. Wechsler and Lee 1996 argue that these adjuncts are all situation delimiters, which serve as extensive measures of the event.12 I suggest that non-theta marked accusative constituents can be situation delimiters, but are not necessarily so in all languages. I show below that while Korean accusative constituents are extensive measures of the event, they do not yield a delimited event description. Non-theta marked accusative constituents in Hungarian are situation delimiters. The term situation delimiter refers to a constituent which establishes the endpoint of an event and yields a complex, telic or delimited event description. The generalization can be stated as in (27). (27)

Non-theta marked accusative constituents in Hungarian are situation delimiters

(27) does not extend to accusative reflexives and body parts. For these constituents, it was argued above that their presence is determined by the Argument per Subevent Condition (Rappaport Hovav and Levin 2001). In other words, they do not affect the aspectual properties of the event themselves. The range of events which can be modified by delimiters is restricted to nondelimited events in Hungarian.13 This follows from the uniqueness requirement on delimiters, proposed by Goldberg 1991 and Tenny 1994. They argue that an event can have only one delimiter, stated as the Single delimiting constraint.

170 (28)

ANIKO CSIRMAZ

Single delimiting constraint The event described by a verb may only have one measuring-out argument and be delimited only once. (Tenny 1994)

Hungarian conforms to the constraint above. Since non-theta marked accusatives delimit the event, they must modify non-delimited eventualities. In the ungrammatical examples of (29), the event descriptions which are modified by the non-theta marked accusatives are telic. The goal or the terminative particle el (‘away’) yields a telicized (and delimited) event description. The non-theta marked accusatives are italicized in the examples. (29)

a. * Jańos el futott fe´l o´ra´t ( a boltba) half hour-ACC the store-into J-NOM away ran ‘John ran into the store for half an hour’ b. * Jańos el futott sza´z me´tert ( a boltba) hundred meter-ACC the store-into J-NOM away ran ‘Jańos ran a hundred meters to the store’ c. * Jańos el futott egyet (a boltba) one-ACC the store-into J-NOM away ran ‘Jańos ran to the store’ d. * Jańos nagyot el futott (a boltba) the store-into J-NOM large-ACC away ran ‘Jańos ran hard to the store’

In absence of a terminative particle or goal, the predicates which are modified by the accusative constituents are non-delimited – and as such, are amenable to modification by the accusative modifiers: (30)

a. Jańos J-NOM ‘Jańos b. Jańos J-NOM ‘Jańos c. Jańos J-NOM ‘Jańos d. Jańos J-NOM ‘Jańos

futott fe´l o´ra´t ran half hour-ACC ran for half an hour’ futott sza´z me´tert ran hundred meter-ACC ran a hundred meters’ futott egyet ran one-ACC ran’ nagyot futott large-ACC ran ran hard’

It should be noted that multiplicatives are not delimiters by virtue of their denotation. In addition to accusative multiplicatives, Hungarian also has a non-structurally case


171

marked multiplicative, which appears with the suffix -szer.14 These suffixed multiplicatives are not situation delimiters and do not fall under the generalization in (28). Since suffixal multiplicatives are not delimiters, they are not subject to the Single delimiting constraint. They can freely modify any event, including telic event descriptions: (31)

Jańos egyszer el futott a away ran the J-NOM once ‘Jańos ran to the store once’

boltba store-into

As noted above, the non-theta marked accusative (or more generally, structurally case marked) constituents are not universally situation delimiters (contrary to Wechsler and Lee 1996). In Korean, an accusative multiplicative can modify an achievement, for instance: (32)

Cheli-ka Mary-lul panci-lul two pen-ul M-ACC ring-ACC twice-ACC Ch-NOM hay-ss-ta do-PAST-DECL ‘Cheli presented Mary with a ring twice’ (Wechsler and Lee 1996)

senmwul-ul gift-ACC

I propose that the difference between Hungarian and Korean is due to the event structure restriction on adjuncts with structural case. In Hungarian, all non-theta marked accusatives are situation delimiters, as in (27) and repeated below. (33)

Non-theta marked accusative constituents in Hungarian are situation delimiters

Korean, in contrast, can structurally case mark all extensive measure adjuncts of events, including non-delimiting multiplicatives. The remaining sections of this chapter discuss Hungarian non-theta marked accusative constituents in more detail. It is argued that these situation delimiters are not homogeneous. They are either measure phrases (temporal and spatial measure phrases) or introduce and modify a result (multiplicatives and adjectives) and accordingly show diverging behavior. All of these constituents are, however, situation delimiters and thus modify only non-delimited event descriptions. 3.

TEMPORAL AND MEASURE PHRASES

Structurally case marked temporal and measure phrases can measure temporal duration or spatial distance:15

172 (34)

ANIKO CSIRMAZ

a. Juli ke´t o´ra´t J-NOM two hour-ACC ‘Juli slept for two hours’ b. Juli ke´t me´tert J-NOM two meter-ACC ‘Juli ran two meters’

aludt slept futott ran

As noted above, accusative temporal and measure phrases are situation delimiters, and as such, modify only non-delimited event descriptions. a. Juli ke´t o´ra´t futott J-NOM two hour-ACC ran ‘Juli ran for two hours’ b. Jańos ke´t o´ra´t el borozgatott J-NOM two hour-ACC away drank.wine ‘Jańos drank wine for two hours’ c. Juli ke´t napot volt beteg two day-ACC was sick J-NOM ‘Juli was sick for two days’ d. Az auto´ ke´t me´tert gurult two meter-ACC rolled the car-NOM ‘The car rolled two meters’ e. * Juli ke´t o´ra´t el futott a boltba two hour-ACC away ran the store-into J-NOM *‘Juli ran into the shop for two hours’ (36) a. Juli sza´z me´tert futott J-NOM hundred meter-ACC ran ‘Juli ran a hundered meters’ b. * Juli sza´z me´tert futott a boltba the shop-into J-NOM hundred meter-ACC ran ‘Juli ran a hundred meteres to the shop’ (35)

The restriction to modifying atelic event descriptions thus follows from the generalization in (27) and the uniqueness constraint in (28). Other properties follow from the category and semantics of the modifiers. 3.1.

Temporal and measure phrases are adjuncts

In contrast with fake reflexives and body parts, accusative temporal and measure phrases impose no restrictions on the argument structure of the predicate.16 The phrases can modify transitive and unaccusative events as well as unergative ones: (37)

a. Juli ke´t o´ra´t hallgatta a J-NOM two hour-ACC listened the ‘Juli listened to music for two hours’

zene´t music-ACC

173


b. Juli ke´t me´tert two meter-ACC J-NOM ‘Juli pushed the car two meters’17

tolta pushed

az the

auto´t car-ACC

These phrases, unlike fake reflexives, are not merged as direct objects of the verb. I suggest that they are not even arguments, but rather adjuncts of the predicate. While no extraction facts can argue for adjuncthood – since no constituent is available for extraction – more abstract considerations support this status.18 Baker 1988 (building on Marantz 1984) argues that semantic selection by a predicate can only apply to arguments, but not adjuncts. A constituent whose interpretation varies according to the verb is thus an argument of the predicate. Based on semantic selection facts, as well as those based on extraction, Baker 1988 argues that benefactives, certain instrumentals and locatives are arguments of the verb and not adjuncts. English locatives, for instance, can be interpreted as either goals or locations, depending on the specific verb. (38)

a. Carmel went in the room b. Carmel sat in the room (Baker 1988)

The Chichewa benefactives discussed by Baker also have a range of different interpretations, including the beneficiary of the event or a participant adversely affected by the event. The agent can also fulfill some responsibility for the ‘beneficient’, but not for his direct benefit, supporting the argument status of these constituents. The interpretation of Hungarian temporal and measure phrases does not vary according to the verb choice. They have a constant meaning; in all cases, they measure temporal or spatial duration, respectively. The interpretation of the modifiers thus supplies some support for their adjunct status. Another argument in favor of the adjunct status is provided by the interpretation of these constituents. Temporal and spatial measure phrases are interpreted as measuring out the event, by establishing a homomorphic mapping between the event and the interval measured by the modifier. If temporal and measure phrases were arguments, then they would be required to be direct objects by Tenny 1994’s Measuring Out constraint: (39)

Measuring Out constraint (ii) Direct internal arguments are the only overt arguments which can ‘measure out’ the event (Tenny 1994)

The phrases are, however, not necessarily direct internal arguments, as the transitive predicates in (37) show. Consequently, they cannot be arguments, since they ‘measure out’ the event. The constant interpretation of temporal and measure phrases and their compatibility with transitive predicates suggests that they are adjuncts rather than arguments.

174

ANIKO CSIRMAZ

Let us consider the distribution of the accusative temporal and measure phrases in more detail. Accusative temporal and durative adjuncts are first merged inside the vP, along with arguments and certain adjuncts. As the vP deletion data (E´. Kiss 2002a)19 show, the accusative modifiers can be elided along with other material within the vP. (40)

a. Jańos ke´t kilome´tert futott, Pe´ter viszont nem P-NOM but not J-NOM two kilometer-ACC ran [vP futott ke´t kilome´tert] ran two kilometer-ACC ‘Jańos ran two kilometers, but Peter did not’ b. Jańos ke´t o´ra´t futott, Pe´ter viszont nem P-NOM but not J-NOM two hour-ACC ran [vP futott ke´t o´ra´t] ran two hour-ACC ‘Jańos ran for two hours, but Peter did not’

The accusative modifiers are not confined to a postverbal slot. They can also move to a preverbal position, which precedes particles and other constituents in Spec,PredP.20 I assume that the modifiers below are located in the specifier position of some functional projection above PredP.21 Accusative temporal modifiers share this preverbal slot with temporal modifiers such as ke´t o´rań a´t (‘for two hours’), which also measure the time of non-delimited events. (41)

/ ke´t o´rań a´t a. Jańos [PredP ke´t o´ra´t J-NOM two hour-ACC / two hour-on across olvasott ]] [PredP u´jsa´got newspaper-ACC read ‘Jańos read newspapers for two hours’ a´t b. Jańos [PredP ke´t o´ra´t / ke´t o´rań / two hour-on across J-NOM two hour-ACC besze´lgetett ]] [PredP el away talked ‘Jańos chatted for two hours’

The functional projection where preverbal accusative temporal and measure phrases appear is located below the durative Asp projection of Cinque 1999. Cinque argues that durative adverbs are located above the progressive aspect head and one of the projections which house celerative, repetative or frequentative aspect: (42)

Aspdurative > Aspprospective > AsprepetitiveII (Cinque 1999)

AspcelerativeII >

AspfrequentativeII >

Given this ordering, if accusative temporal modifiers can move to the specifier of Aspdurative, then they are expected to precede the adverbs located in the specifiers of

175


the lower functional projections. The predicted order is not possible with accusative temporal modifiers; accusative durative phrases cannot precede the predicted modifiers in Hungarian.22 (43)

a. *Jańos ke´t o´ra´t gyorsanCelerativeII futott ran J-NOM two hour-ACC fast ‘Jańos ran fast for two hours’ b. *Jańos fe´l o´ra´t majdnemProspective repu¨lt flew J-NOM half hour-ACC almost ‘Jańos almost flew for half an hour’

Temporal modifiers with a´t can precede the adverbs, as predicted by Cinque 1999: (43)

a. Jańos ke´t o´rań a´t gyorsanCelerativeII futott ran J-NOM two hour-on across fast ‘Jańos ran fast for two hours’ b. Jańos fe´l o´rań a´t majdnemProspective repu¨lt flew J-NOM half hour-on across almost ‘Jańos almost flew for half an hour’

While a´t-adverbs can appear in the predicted specifier position of AspdurativeP, this option is not available for accusative temporal adverbs. I assume that accusative temporal and measure modifiers can move to the specifier of a functional projection which is located below the prospective, celerative, frequentative and repetitive Asp projections of Cinque 1999. The distinct positions of durative modifiers are also shown by the time intervals they can measure. As argued in Csirmaz (chapter 10), accusative temporal phrases can only measure the event time and not the reference time. The predicate applying to the reference time has the subinterval property, and is thus amenable to modification by a´t temporal phrases: (44)

Ke´t e´s fe´l o´rań a´t csak Jańos two and half hour-on across only J-NOM ‘For two hours and a half, only Jańos arrived’

e´rkezett arrived

meg perf

Accusative temporal phrases, in contrast, can only modify the event time, and not the reference time. Being restricted to event time modification, accusative temporal modifiers cannot modify the reference time of an instantaneous event description: (45)

*Ke´t e´s fe´l o´ra´t csak Jańos e´rkezett meg two and half hour-ACC only J-NOM arrived perf ‘For two hours and a half, only Jańos arrived’

176

ANIKO CSIRMAZ

In Csirmaz (chapter 10), I argue that the locality of modification and case checking is responsible for this restriction on accusative modifiers. Accusative modifiers must be merged within the vP. Given the locality requirement on adverbial modification, a modifier merged vP-internally can only measure the local event time. As argued above, temporal and measure phrases are adjuncts, which are merged within vP and can move to a functional projection above PredP. In the following subsection I argue that these adjuncts are structurally case marked and consider how case marking is licensed. 3.2.

Case marking

Temporal and measure modifiers are NPs, a categorization which is supported by the presence of case marking as well as the availability of definite phrases as modifiers (E´. Kiss 2004a).23 (46)

(47)

a. Jańos ke´t o´ra´t futott J-NOM two hour-ACC ran-INDEF-PAST.3SG ‘Jańos ran for two hours’ b. Jańos ve´gig futott ke´t o´ra´t J-NOM to end ran-INDEF-PAST.3SG two hour-ACC ‘Jańos ran two hours to the end’ c. Jańos ve´gig futotta a ke´t o´ra´t J-NOM to.end ran-DEF-PAST.3SG the two hour-ACC ‘Jańos ran the (allotted) two hours to end’24 a. Jańos ke´t kilome´tert futott J-NOM two kilometer-ACC ran-INDEF-PAST.3SG ‘Jańos ran two kilometers’ b. Jańos le futott ke´t kilome´tert two kilometer-ACC J-NOM down ran ‘Jańos ran two kilometers’ c. Jańos le futotta a ke´t kilome´tert J-NOM down ran-DEF-PAST.3SG the two kilometer-ACC ‘Jańos ran the (allotted) two kilometers’

Definite modifiers also support the structural case hypothesis. As noted in section 1, case checking by v and definiteness agreement by Hungarian finite verbs go hand in hand. The preceding examples show that definite temporal and measure phrases are case marked by and trigger definiteness agreement in the verb. The case marking of temporal and measure phrases cannot be treated as identical to direct object and fake reflexive case marking, discussed in section 1. The latter only appear with intransitives, and as shown in the preceding section, temporal and measure modifiers can appear with direct objects as well. I assume that the checking of nominative and accusative case proceeds in Hungarian as in English and other nominative – accusative languages. The


177

nominative case feature of a nominal is checked by finite T and the accusative case feature, by the v head. Since the accusative case of both the direct object and the modifier is checked by v, the v head must be able to check more than one case feature. Ura 2000 and Bobaljik and Branigan 2006 argue that it is possible for a given functional head – T or v – to check multiple case features. This possibility is a part of UG, but is a marked option, attested in only a few languages. The same head checking multiple case features accounts for multiple nominative constructions in Japanese (Ura 2000) as well as ergative patterns and dative subjects in causative constructions (Bobaljik and Branigan 2006). I suggest that that multiple case feature checking is also an option for v in Hungarian, and for other languages with accusative adjuncts.25 I assume that feature checking is subject to the same locality restrictions with single or multiple case checking.26 Accusative constituents must be merged within vP to allow case checking by the v head. The checking of accusative case on objects and on adjuncts thus does not need to be separated; the two are identical for the present purposes. As shown above, temporal and measure modifiers are structurally case marked adjuncts. Given restrictions on case feature checking, they must be merged within the vP, which restricts temporal modifiers to modifying the event time. The modifiers are not restricted to postverbal position; they can move to a low preverbal functional projection which is located above PredP but below the Aspdurative projection of Cinque 1999. In addition, temporal and measure phrases only modify atelic event descriptions, which follows from their status as situation delimiters. 4.

M U LT I P L I C AT I V E S

The last group of constituents to be discussed includes accusative multiplicatives and adjectives, which share some similarities with, but are also significantly different from accusative temporal and measure modifiers. I suggest that both accusative multiplicatives (which are formally numerals with accusative case marking) and accusative adjectives introduce and modify a result, and thus delimit the event they modify. The aspectual properties of multiplicative and adjectival constructions are argued to follow from this account. The syntactic properties of multiplicatives and adjectives, however, show some variability according to the predicate modified.27 While this section discusses accusative multiplicatives, section 5 addresses accusative adjectives in more detail, showing that the two groups show largely parallel behavior. 4.1.

Multiplicatives and aspectual properties

In section 2, I proposed that non-theta marked accusative constituents in Hungarian are situation delimiters. In this section, I show that the situation delimiter status of accusative constituents (also Kiefer 1994, K. E´. Kiss, p.c.) accounts for most

178

ANIKO CSIRMAZ

aspectual restrictions that they impose on the event description they modify. In addition, this account allows a uniform treatment of the types of event descriptions which are modified by accusative multiplicatives. Since accusative multiplicatives are delimiters, they yield complex telic event descriptions. Kiefer 1992 and Pinõń 2001b observe that these event descriptions cannot be modified by alatt modifiers (48), in spite of being telic. (48)

a. *Juli ke´t o´ra alatt J-NOM two hour under ‘Juli ran in two hours’ b. *Juli ke´t o´ra alatt J-NOM two hour under ‘Juli rested in two hours’

futott ran pihent rested

egyet one-ACC egyet one-ACC

The telicity of the resulting event description can still be shown by modification by other durative modifiers which modify telic event descriptions (K. E´. Kiss, p.c.).28 (49)

Juli [ke´t o´rań belu¨l] futott egyet one-ACC J-NOM two hour-in inside ran ‘Juli ran (some) in two hours’

Other tests, including the conjunction test (Verkuyl 1993, Kiefer 1992) also show that the event description is telic. If two temporal phrases are conjoined, then telic event descriptions, modified by the conjoined phrases, only allow the interpretation where distinct eventualities occur during each of the temporal intervals (50a–c). In absence of the accusative multiplicative, the event description is atelic. Similarly to other atelic event descriptions, it can be interpreted as a single eventuality which holds during both intervals (50d). (50)

a. Juli de´lelo´´tt is e´s de´lutań is ´ırt J-NOM morning too and afternoon too wrote egy levelet one letter-ACC ‘Juli wrote a letter in the morning and one in the afternoon’ (two eventualities) *‘Juli wrote a letter in the morning and the afternoon’ (single eventuality) b. Juli de´lelo´´tt is e´s de´lutań is futott egyet one-ACC J-NOM morning too and afternoon too ran ‘Juli ran in the morning and she also ran in the afternoon’ (two eventualities) *‘Juli ran in the morning and in the afternoon, too’ (single eventuality)

179


c. Juli de´lelo´´tt is e´s de´lutań is pihent J-NOM morning too and afternoon too rested ‘Juli rested in the morning and she also rested in the afternoon’ (two eventualities) *‘Juli rested in the morning and in the afternoon too’ (single eventuality) d. Juli de´lelo´´tt is e´s de´lutań is pihent J-NOM morning too and afternoon too rested ‘Juli rested in the morning and in the afternoon too’ (can be a single eventuality)

egyet one-ACC

Events modified by accusative multiplicatives behave like other telic event descriptions, and in absence of accusative multiplicatives, the event descriptions are atelic. Since accusative multiplicatives are situation delimiters, the Single delimiting constraint in (28) demands that they modify only non-delimited events.29 Kiefer 1994 and Pinõń 2001b note that multiplicatives can occur with activities, which is consistent with the delimiter status of multiplicatives. (51) Jańos se´ta´lt egyet J-NOM walked one-ACC ‘Jańos walked’ Accusative multiplicatives can also appear with semelfactives (52a) (Kiefer 1992, Pinõń 2001), but not with achievements (52b). The modification of semelfactives by multiplicatives is consistent with the view of delimitedness in Tenny 1994. Tenny 1994 assumes that delimited events have an endpoint distinct from the process part of the event. Achievements, but not semelfactives, yield an end state which is distinct from the state or process which precedes the endpoint.30 (52)

a. Jańos J-NOM ‘Jańos b. *Jańos J-NOM ‘Jańos

bo´lintott nodded nodded’ e´rkezett arrived arrived’

egyet one-ACC egyet one-ACC

Delimitedness crucially differs from the Vendlerian notion of telicity; it is possible for a telic event description to be non-delimited if it lacks a distinct endpoint. Semelfactives are a case in point, since they are instantaneous – and therefore telic – and lack a result state. Accusative multiplicatives also fail to appear with temporal and measure phrases, as shown below. Since all three modifier types are situation delimiters – being non-theta marked accusative constituents – only one of them can be present in the structure. Multiple occurrences of delimiters are excluded by the Single delimiting constraint of Tenny 1994.

180 (53)

ANIKO CSIRMAZ

a.

Juli futott egyet one-ACC J-NOM ran ‘Juli ran’ b.* Juli ke´t o´ra´t futott egyet one-ACC J-NOM two hour-ACC ran ‘Juli ran for two hours’ c.* Juli ke´t kilome´tert futott egyet one-ACC J-NOM two kilometer-ACC ran ‘Juli ran two kilometers’

The delimiter status of multiplicatives also excludes modification by terminative particles.31 Terminative particles which appear with semelfactives are also delimiters, specifying an endpoint of the event. Since events can be delimited only once, the cooccurrence of these particles and multiplicatives is excluded. (54)

(55)

a. *Jańos fel ugrott egyet J-NOM up jumped one-ACC ‘Jańos jumped up’ b. Jańos fel ugrott J-NOM up jumped ‘Jańos jumped up’ a. *A feńy fel villant flashed the light-NOM up ‘The light flashed once’ b. A feńy fel villant flashed the light-NOM up ‘The light flashed once’ c. A feńy villant egyet the light-NOM flashed one-ACC ‘The light flashed once’

egyet one-ACC

Before turning to non-aspectual properties of accusative multiplicatives, let us consider the type of events which they can modify in more detail. As shown above, accusative multiplicatives can modify not only activities, but also semelfactive events. Pinõń 2001b offers a definition of egyet which encodes that it can modify either activities or semelfactives. The definition states that an event e of type R has a runtime which is a proper part of some time interval t. In addition, there is no event e 0 of type R within t such that e 0 is larger than e or distinct from e. (56)

egyet ! ‚R ‚x ‚e [R(e,x) (e) t 8e 0 [R(e 0 , x) (e 0 ) t :(e 0 e) ! e 0 ¼ e] ($e 0 $e 0 0 $y [R(e 0 ,y 0 ) R(e 0 0 ,y 0 ) e 0 e 0 0 ] $e 0 $e 0 0 $y [R(e 0 ,y) R(e 0 ,y) e 0 quickly)

The manner adverb can also precede the multiplicative, where it takes wide scope.

184

ANIKO CSIRMAZ

(63) Jańos [AspCelerativeP gyorsan ke´tszer [PredP klikkelt]] quickly twice clicked J-NOM ‘Jańos clicked quickly twice’ (quickly > twice) An accusative multiplicative, unlike its suffixed counterpart, cannot precede and take scope over the manner adverb.38 (64)


ketto´´t gyorsan klikkelt two-ACC quickly clicked clicked quickly twice’ (*twice > quickly) gyorsan ketto´´t klikkelt quickly two-ACC clicked clicked quickly twice’ (quickly > twice)

The relative scope and position of manner adverbs and multiplicatives supports the proposed description, where accusative multiplicatives are in Spec,PredP, while suffixed multiplicatives can move to positions higher than PredP. Accusative multiplicatives may move to Spec,PredP, a focus or contrastive topic position in the preverbal domain, and are excluded from other positions. This restriction is shown by the obligatory narrow scope that multiplicatives take with respect to preverbal quantifiers, which precede PredP. The linear order of the preverbal suffixed multiplicative and the quantifier corresponds to their scope relations, as in (65a). An accusative multiplicative which precedes the quantifier, as in (65b), is a contrastive topic, and as such takes scope under the quantifier.39 (65)

a. Ke´tszer mindenki ko¨hintett twice everyone-NOM coughed ‘Twice everyone coughed’ (twice > everyone) b. Ketto´´t mindenki ko¨hintett two-ACC everyone-NOM coughed ‘Everyone coughed twice’ (*twice > everyone)

In order to account for the pattern above, I assume that accusative multiplicative phrases can move to Spec,PredP and suffixed multiplicatives can move to projections above PredP (both modifiers can also appear as contrastive topics). I suggest that a non-contrastive multiplicative in the preverbal quantifier or topic domain, as in (65a), is first merged in that position. I assume that accusative multiplicatives can only be merged within the vP. Suffixed multiplicatives, in contrast, can be merged as adjuncts to topic positions and precede quantifiers in the preverbal field. The distribution of accusative multiplicatives resembles that of temporal measure phrases. As argued in Csirmaz (chapter 10), accusative temporal phrases can modify the event time, but not the reference time. With both multiplicatives and temporal phrases, modification is restricted to the vP and PredP. I assume that in


185

both cases, this restriction arises from the locality of case checking and modification. Before concluding this section, I briefly address a restriction on accusative multiplicatives which does not follow from the previous assumptions. Multiplicatives cannot appear with particles, independently of the interpretation of the latter. The restriction holds even if the particle does not delimit the event, as in the following examples. The for-modification of the event descriptions below shows that the event descriptions in question are non-delimited, atelic. (66)

a. *Jańos J-NOM ‘Jańos a’. Jańos J-NOM ‘Jańos b. *Jańos J-NOM ‘Jańos b’. Jańos J-NOM ‘Jańos c. *Jańos J-NOM ‘Jańos c’. Jańos J-NOM ‘Jańos

el me´la´zott egyet away mused one-ACC mused’ ke´t o´rań a´t el me´la´zott two hour-on across away mused mused for two hours’ fel olvasott egyet up read one-ACC read aloud’ fe´l o´rań a´t fel olvasott (a ko¨nyvbo´´l) half hour-on across up read the book-from read aloud for half an hour (from a book)’ fel mosott egyet up washed one-ACC mopped the floor’ tı´z percen a´t fel mosott ten minute-on across up washed washed the floor for then minutes’

To account for the data above, I follow E´. Kiss 2004, who argues that multiplicatives are obligatorily non-specific.40 Particles, whether delimiting or non-delimiting, require a specific direct object. The specificity requirement of particles extends to direct objects independently of their thematic role. It also carries over to multiplicatives merged in direct object position, which modify activities.41 The nonspecificity constraint on multiplicatives is at odds with the specificity requirement imposed by particles. I argued in this section that accusative multiplicatives are situation delimiters. A number of aspectual properties, of the predicates the multiplicatives modify as well as those of the resulting predicate, follow from this property. In the next section I show that these conclusions also carry over to accusative adjectives. 5.

A C C U S AT I V E A D J E C T I V E S

In addition to the non-theta marked accusative constituent discussed above, certain adjectives can also appear with accusative case marking.

186 (67)

ANIKO CSIRMAZ

a.

b.

c.

Juli nagyot / hatalmasat / o´ria´sit large-ACC / enormous-ACC / gigantic-ACC J-NOM tu¨sszentett sneezed ‘Juli gave a large / enormous / gigantic sneeze’ Juli kicsit u¨to¨tt J-NOM small-ACC hit ‘Juli hit someone slightly’ Juli jo´t futott J-NOM good-ACC ran ‘Juli had a good run’

The discussion in this section is restricted to nagyot (large-ACC), but the conclusions carry over to other adjectives. I show that accusative adjectives largely behave like accusative multiplicatives and argue that the parallel behavior of these accusative constituents follows from the fact that both introduce and modify a result, thus accusative adjectives, like multiplicatives, are situation delimiters. They impose the same restrictions on aspectual properties of the events that they modify. I also show that adjectives have a wider distribution than multiplicatives, since they can also modify the degree argument of a predicate. Finally, I address and argue against an account of multiplicatives and adjectives in terms of cognate object ellipsis.

5.1.

Adjectives and aspectual properties

As noted above, accusative adjectives show the same restrictions on aspectual properties as multiplicatives. While adjectives can modify activities and semelfactives, delimited events and states are ruled out. (68)

a. Jańos nagyot se´ta´lt J-NOM large-ACC walked ‘Jańos walked a large distance’ b. Jańos nagyot bo´lintott J-NOM large-ACC nodded ‘Jańos nodded (emphatically)’ c. Jańos nagyot kia´ltott J-NOM large-ACC shouted ‘Jańos gave a loud cry’ d. *Jańos nagyot nyert J-NOM large-ACC won ‘Jańos won’ e. *Jańos nagyot fel e´rt J-NOM large-ACC up reached ‘Jańos reached the top’

a csućsra the top-onto


187

f. *Jańos nagyot va´rt a mega´llo´ban J-NOM large-ACC waited the stop-in ‘Jańos waited at the stop (for a long time)’ I assume that accusative adjectives, similarly to multiplicatives, introduce and modify a result, yielding the restrictions observed above. The result may be a sound (as in (68c)), or a more abstract result of the action. Massam 1990 also suggested a result account for English cognate objects, where the cognate object describes the result of the process described by the verb. (69)

John laughed a menacing laugh

I assume that the same result-based account can be invoked for English cognate objects as well as Hungarian accusative adjectives and multiplicatives. 5.2.

Restrictions on argument structure

Accusative adjectives, like accusative multiplicatives, can modify unergative predicates, but not transitive ones. (70)

Jańos nagyot se´ta´lt J-NOM large-ACC walked ‘Jańos walked a large distance’ b. Jańos nagyot eńekelt J-NOM large-ACC sang Jańos sang wonderfully’ c. *Jańos nagyot eńekelt egy one J-NOM large-ACC sang ‘Jańos sang a song wonderfully’ d. *Jańos nagyot ko¨nyvet olvasott J-NOM large-ACC book-ACC read ‘Jańos read (from) a book’

a.

dalt song-ACC

With respect to unaccusative predicates, accusative adjectives show some variation. They are similar to multiplicatives in allowing the modification of unaccusative semelfactives. (71)

a. A labda nagyot pattant the ball-NOM large-ACC bounced ‘The ball bounced high’ b. A ko´´ nagyot csobbant the stone-NOM large-ACC splashed ‘The stone made a loud splash’ c. A za´r nagyot kattant the lock-NOM large-ACC clicked ‘The lock clicked loudly’

188

ANIKO CSIRMAZ

Unlike multiplicatives, adjectives can also modify atelic unaccusative predicates with a degree argument.42,43 a. A va´ros nagyot va´ltozott the city-NOM large-ACC changed ‘The city changed a lot’ b. *A va´ros va´ltozott egyet the city-NOM changed one-ACC ‘The city changed’ (73) a. Az a´rfolyam nagyot esett the exchange rate-NOM large-ACC fell ‘The exchange rate fell considerably’ b. *Az a´rfolyam esett egyet one-ACC the exchange rate-NOM fell ‘The exchange rate fell’

(72)

Modification of an unaccusative predicate without a degree argument is, however, ungrammatical. (74)

a. *Nagyot folyt a vı´z large-ACC flowed the water-NOM b. *Nagyot ve´rzett a katona the soldier-NOM large-ACC bled ‘The soldier bled heavily’

As shown above, accusative adjectives are three-way ambiguous with respect to their argument structure status. Adjectives can modify the degree argument of any predicate. In addition, they also show the ambiguity which was argued to characterize multiplicatives. It was noted earlier that accusative adjectives impose the same restrictions on aspectual properties as multiplicatives, since they only modify dynamic non-delimited events.44 The aspectual restrictions also hold with degree argument modification. The accusative adjective delimits the degree argument, and hence the event as well. The Single delimiter constraint allows only non-delimited, atelic event descriptions to be modified. Given that states lack a degree argument, they cannot be modified by degree-modifying accusative adjectives either. 5.3.

Arguments and adjuncts

As shown in the previous subsection, accusative adjectives are three-way ambiguous. First, when modifying activities, the adjectives can pattern with multiplicatives and be merged in the direct object position. Second, if an activity has a degree argument, then that argument is amenable to adjectival modification. In this case the adjective is merged as an adjunct, and allows direct objects in the argument structure. Finally,


189

accusative adjectives are similar to multiplicatives in the modification of semelfactives. Both kinds of modifiers are adjuncts, allowing unaccusative semelfactives to be modified. Accusative adjectives and multiplicatives behave alike with respect to the positions they occupy. Adjectives are first merged within the vP, and can be elided by vP-deletion. (75)

a. Jańos J-NOM [vP

nagyot bo´lintott, Pe´ter viszont nem large-ACC nodded P-NOM but not bo´lintott nagyot] nodded large-ACC ‘Jańos nodded emphatically, but Pe´ter did not’ b. Jańos ketto´´t bo´lintott, Pe´ter viszont nem P-NOM but not J-NOM two-ACC nodded bo´lintott ketto´´t] [vP nodded two-ACC ‘Jańos nodded twice, but Pe´ter did not’

Both accusative adjectives and multiplicatives can move to the preverbal position which they occupy in the first conjunct. For adjectives as well as for multiplicatives, this position is Spec,PredP. As expected, adjectives must follow manner adverbs which are located above PredP. (76)

Bayonne (* nagyot) [AspCelerativeP large-ACC B-NOM [PredP (nagyot) ordı´tott]] large-ACC shouted ‘Bayonne quickly shouted loudly’

gyorsan quickly

Accusative adjectives also resemble multiplicatives in excluding particles or other constituents which appear in Spec,PredP. I assume that the ungrammaticality of particle constructions with adjectives is due to the obligatory non-specific interpretation of the adjectives, as in the case of multiplicatives. (77)

*Jańos nagyot el me´la´zott J-NOM large-ACC away mused ‘Jańos mused (for a long time)’

Accusative adjectives and multiplicatives are thus largely parallel. Given the similarities between these constituents, it has been proposed (E´. Kiss 2004) that accusative adjectives appear when the multiplicative egy (‘one’) is elided. (78)

a. Jańos nagyot futott J-NOM large-ACC ran ‘Jańos ran hard / a long distance’

190

ANIKO CSIRMAZ

b. Jańos J-NOM ‘Jańos c. Jańos J-NOM ‘Jańos

[egy nagyot] futott one large-ACC ran ran hard / a long distance’ egy nagyot futott one large-ACC ran ran hard / a long distance’

Those occurrences of accusative adjectives which modify a degree argument are not covered by this account, since multiplicatives are excluded in these cases. For the remaining occurrences of accusative adjectives, the account ensures a parallel treatment of adjectives and multiplicatives. Let us consider the scope of this account, and see whether numeral ellipsis can apply to arguments. The numerals which modify arguments cannot be elided; the interpretation of arguments varies significantly according to the presence or absence of an overt numeral. A singular count noun without an overt quantifier or determiner (a bare count noun, such as the direct object szendvicset (‘sandwich-ACC’) in (79a)) has a cumulative interpretation. With an overt numeral, as in (79b), the object has a non-cumulative, singular interpretation. Bare arguments modified by an adjective share the neutral interpretation (80). a. Jańos szendvicset evett J-NOM sandwich-ACC ate ‘Jańos was eating a sandwich / sandwiches’ b. Jańos egy szendvicset evett J-NOM one sandwich-ACC ate ‘Jańos was eating a sandwich’ (80) a. Jańos nagy szendvicset evett J-NOM large sandwich-ACC ate ‘Jańos was eating a large sandwich’ b. Jańos nagyot evett J-NOM large-ACC ate ‘Jańos was eating a large one’ b’ Jańos egy nagyot evett J-NOM one large-ACC ate ‘Jańos was eating a large one’

(79)

The interpretation of accusative adjectives does not differ from multiplicatives with an adjective, as in (78). In both cases, the adjective is non-cumulative, since it can delimit the event. The presence or absence of the numeral thus does not affect the properties of accusative adjectives, unlike when it appears with arguments. The ellipsis of the numeral, if it accounts for the parallel behavior of accusative multiplicatives and numerals, does not extend to arguments. In addition, the ellipsis of numerals or multiplicatives is not otherwise attested in Hungarian. If this ellipsis was generally available, then count noun arguments could have a non-cumulative interpretation, contrary to fact.


191

Given the different interpretations of ‘bare’ arguments and accusative adjectives, I believe that a different account is needed. I suggest, rather, that the non-cumulative interpretation and delimiter status of accusative adjectives correlates with their accusative case marking.45 It was argued above that non-theta marked accusative constituents in Hungarian are situation delimiters.46 The similar interpretation of multiplicatives and adjectives is expected then, since both are situation delimiters. In addition, they also share a similar, result interpretation. Thus it is the situation delimiter status and the result interpretation of adjectives rather than the ellipsis of a multiplicative which are responsible for the parallel behavior of adjectives and multiplicatives. In the last section, I briefly explore other approach to multiplicatives and adjectives. The alternative account derives the properties of these accusative constituents by assuming the ellipsis of a cognate object. I argue that while this approach is appealing, it is not without problems either. 5.4.

Cognate objects

The distribution of accusative multiplicatives and adjectives resembles that of cognate objects in English. Masssam 1990 argues that English cognate objects are syntactically direct objects47 and as a consequence, they appear with unergatives, but not with unaccusatives and transitives. Massam 1990 also argues that English cognate objects are objects of result.48 The syntactic position of English cognate objects is the same as that of Hungarian multiplicatives modifying activities, and their interpretation is also similar. On the assumption that the distribution and interpretation of Hungarian cognate objects is similar to their English counterparts, an account in terms of cognate object ellipsis seems viable. This account, however, cannot be maintained. While the properties of cognate objects and the accusative multiplicatives / adjectives are similar, their distribution is different. On the one hand, not all cognate objects can be elided in Hungarian, as shown below. (81)

a. Jańos szo¨rny u˝ hala´lt halt J-NOM horrible death-ACC died ‘Jańos died a horrible death’ b. *Jańos szo¨rny u˝ t halt J-NOM horrible-ACC died ‘Jańos died a horrible death’49

On the other hand, a number of accusative adjectives cannot appear as modifiers of a cognate object. In fact, as noted above, it is not necessarily the case that an overt nominal is possible whenever an accusative adjective can modify the predicate.50 (82)

a. *Jańos nagy ugra´st J-NOM large jump-ACC ‘Jańos jumped high / far’

ugrott jumped

192

ANIKO CSIRMAZ

b. Jańos nagyot ugrott J-NOM large-ACC jumped ‘Jańos jumped high / far’ (83) a. *Jańos se´ta´lt egy se´ta´t walk-ACC J-NOM walked a ‘Jańos took a walk’ b. Jańos se´ta´lt egyet J-NOM walked one-ACC ‘Jańos took a walk’ Furthermore, multiplicatives and adjectives can also modify unaccusative semelfactives, which differs from the unergative predicates appearing with cognate objects. Given these differences between cognate objects and accusative multiplicatives and adjectives, they cannot be collapsed. Whenever a cognate object can alternate with an accusative multiplicative or adjective, I assume that the cognate object is elided, as in (84b). (84)

a. Juli ha´rom le´pe´st le´pett elo´´re step-ACC stepped forward J-NOM three ‘Juli made three steps forward’ b. Juli ha´rmat le´pett elo´´re J-NOM three-ACC stepped forward ‘Juli made three steps forward’

Since the example involves ellipsis, it is not an instance of an accusative multiplicative (or adjective) construction as discussed above. 6.

S U M M A RY

This chapter offered a description of non-theta marked accusative constituents. These constituents were all shown to interact with the aspectual structure of the event, even if a non-accusative counterpart fails to show such effect. I argued that the effect of accusative constituents on aspectual properties is not uniform. On the one hand, the presence of non-theta marked reflexives and body parts is enforced by a restriction on event structure, specifically the Argument per subevent condition (Rappaport Hovav and Levin 2001). The accusative constituents fail to affect the aspectual properties in this case. On the other hand, the remaining accusative constituents – which include accusative temporal and spatial measure phrases, multiplicatives and adjectives – are situation delimiters. Situation delimiters directly affect aspectual properties: they modify only non-delimited events and yield delimited ones. The situation delimiter status is in line with Wechsler and Lee 1996, who argue that only situation delimiter adjuncts can be structurally case marked. In other respects, non-theta marked accusative constituents show a heterogeneous behavior. A survey of these constituents shows that they can be merged as


193

either arguments or adjuncts, depending on the specific modifier or predicate modified. Accusative constituents also vary in the positions they occupy in the preverbal field; an accusative modifier can appear as the specifier of PredP or move to a projection located higher than PredP. In spite of the variation among non-theta marked accusative constituents, they all interact with the aspectual properties of the event.

NOTES

1

2

3

4

I gratefully acknowledge the comments and discussions provided by Sabine Iatridou, David Pesetsky, as well as by Katalin E´. Kiss, Christopher Pinõń, Boldizsa´r Eszes, Bala´zs Surańyi, Csaba Olsvay and Annama´ria Bene. The verb displays definiteness agreement with the object (Bartos 1999, Den Dikken 2004, E´. Kiss 2002). Since object agreement is not directly relevant to the present discussion, it is generally ignored in this section. For the purposes of this discussion, I am treating the impersonal passive construction as passive. It differs from English passives – among others – in allowing passivization of unaccusative as well as transitive predicates (Alberti 1998). I use the term particle to refer to postpositions and adverbs which appear in the immediately preverbal position in neutral sentences. Hungarian particles include the equivalents of Germanic particles, but also a wider range of constituents (for instance, locative adverbs such as ott (‘there’) (see E´. Kiss (chapter 2) for a detailed discussion of particles). A number of constituents, including noun phrases without a determiner or numeral, also show a distribution similar to that of particles. The term verb modifier (E´. Kiss (chapter 2)) includes all of these constituents. The third type of particles, those denoting location, are not discussed in this section. Locative particles do not denote an end state in any sense. As expected, unergative predicates with locative particles do not require a non-argument accusative constituent: (i) ott telefona´l Juli (a sarokban) there phones J-NOM the corner-in ‘Juli is speaking on the phone in the corner’

5

The goal is an adjunct rather than an argument in these examples. A non-theta marked constituent is an argument when it is the associate of a particle (U¨ro¨gdi 2003), but an adjunct otherwise. In (12) no particles are possible which would have the goal as an associate. The non-theta marked goal is an adjunct then and extraction is impossible from the goal: (i) *Kinek ko¨vette´k [t az ajtajaíg] a rendo ´´ro¨k the door-POSS-UNTIL the policemen-NOM who-DAT followed a ra´dioútası´ta´sokat? the radio orders-ACC ‘To whose door did the policemen follow the radio orders?’

6

7

8

The term argument is presumably used as encompassing subjects and objects, and do not require these ‘arguments’ to be thematically marked. Furthermore, Rappaport Hovav and Levin 2001 qualify the ASC and note that the argument-per-subevent condition holds only for temporally independent subevents. If a given event has two subevents that are temporally not independent, then the ASC requires one, not two arguments. I am abstracting away from some crosslinguistic variation as well as variation between the results of different unaccusativity diagnostics. Following Zaenen 1993, Levin and Rappaport Hovav 1995 mark Hij heeft naar huis gelopen (19b) as being marginal (?). As a reviewer points out, Hoekstra describes this structure as fully grammatical,

194

9

10

11 12

13

14 15

ANIKO CSIRMAZ

with the PP interpreted as on his way home. Thus the auxiliary that appears with a resultative interpretation is is, and the auxiliary appearing with a PP denoting direction is heeft. ¨ ro¨gdi 2003 analyzes the associate – particle relation as the one holding between a full XP and a clitic U which doubles the XP. The associate in (20a) is the PP kinek a sı´rko¨ve´re (‘onto whose gravestone’). For a number of languages with non-theta marked accusative constituents, it can be shown that the case marking is, in fact, structural. Direct objects and the non-theta marked accusatives, both structurally case marked, show similar case alternations in some languages. For instance, these Korean accusative constituents alternate between nominative and accusative according to the stativity of the predicate alike (Maling 1989, Wechsler and Lee 1996). Finnish accusatives – objects as well as non-theta marked constituents – all appear with nominative rather than accusative case marking in passives. Also, all of these accusative constituents have partitive case in negative sentences (Vainikka 1989, Kiparsky 2001, among others). In this assumption, I side with Den Dikken 2004, in contrast to Bartos 1999. Extensive measures have the property of additivity. Given two events, the measure of the sum of the events must be the sum of the measure of the separate events. Duration is an extensive measure: if an event e1 is 2 h long and event e2 lasts for an hour, then the duration of e1 e2 will be the sum of the durations, 3 h. Average speed, in contrast, is not an extensive measure. If e1 is an event of driving at 50 km/h for 1 h and e2 is one of driving at 60 km/h for an hour, then the speed of e1 e2 is not 110 km/h, but rather 55 km/h. For the present purposed, delimitedness can be treated as telicity. The definition of delimitedness in Tenny 1994 calls for a distinct endpoint of the event, which coincides with telicity as defined by Smith 1991. The relation of telicity and delimitedness are discussed in more detail in section 4. The allomorphs of the suffix (-szor, -szer or -szo¨r) are determined by vowel harmony. Theta-marked durative and measure arguments, such as those below, are ignored in this discussion. (i) Jańos hu´sz kilome´tert tett meg egy J-NOM twenty kilometer-ACC did perf one ‘Jańos covered twenty kilometers in one day’ (ii) Jańos ke´t e´vet to¨lto¨tt Szudańban Sudan-in J-NOM two year-ACC spent ‘Jańos spent two years in Sudan’

16

17

nap day

alatt under

With the exception of the requirement imposed by measure modifiers, since they require a path argument. In these examples, the verb shows definite agreement as a reflex of the case checking of the definite object. A reviewer points out that if the case of the measure phrase is also checked by v, then the ‘mirror image’ of (37b), with a definite measure phrase and an indefinite the marked object should also show definite agreement. This ‘mirror image’ structure is, however, ungrammatical, either with or without definiteness agreement. (i) *Juli a ke´t me´tert J-NOM the two meter-ACC ‘Juli pushed a car two meters’

tolt(a) pushed(-def)

egy a

auto´t car-ACC

The ungrammaticality is expected (also noted in footnote (23)), since the measure phrase is homomorphic to the event. Like other homomorphic constituents, a definite measure adverb obligatorily appears with a particle such as ve´gig (to the end): (ii)

Juli ve´gig tolta a ke´t me´tert J-NOM to.end pushed the two meter-ACC ‘Juli pushed the two meters to the end’

Even in the presence of a particle, however, a thematic object and an accusative measure phrase are ungrammatical. Again, the presence or absence of definiteness agreement on the verb is

195


immaterial: (iii)

*Juli ve´gig tolt(a) egy J-NOM to.end pushed a ‘Juli pushed a car two meters’

auto´t car-ACC

a the

ke´t two

me´tert meter-ACC

The ungrammaticality of an overt object and an accusative measure phrase is independent of the definiteness of the accusative constituents. Rather, it appears to follow from a ban on multiple accusative constituents in the presence of a particle. When modified by the particle ve´gig (to the end), the causative futtat (make sy run) can appear with either an accusative measure phrase (iv) or an accusative agent (v). (iv) Juli ve´gig futtatta a harminc J-NOM to.end run-CAUS-PAST the thirty ‘Juli had a boy run the entire thirty kilometers’ (v) Juli ve´gig futtatta a fiu´t J-NOM to.end run-CAUS-PAST the boy-ACC ‘Juli had the boy run the entire thirty kilometers’

kilome´tert kilometer-ACC

egy a

a the

kilome´teren kilometer-on

harminc thirty

fiu´val boy-with

If both the causee agent and the measure phrase are accusative, the structure is ungrammatical, independently of the definiteness of the accusative constituents or the definiteness agreement on the verb: (vi) *Juli ve´gig futtatta a fiu´t J-NOM to.end run-CAUS-PAST the boy-ACC ‘Juli had the boy run the entire thirty kilometers’ (vii) *Juli ve´gig futtatta a fiu´t J-NOM to.end run-CAUS-PAST the boy-ACC ‘Juli had the boy run thirty kilometers’ (viii) *Juli ve´gig futtatott egy fiu´t boy-ACC J-NOM to.end run-CAUS-PAST a ‘Juli had a boy run thirty kilometers’

a the

harminc thirty

harminc thirty harminc thirty


kilome´tert kilometer-ACC kilome´tert kilometer-ACC

In absence of a particle, the two accusative constituents are grammatical. (ix) Juli harminc kilome´tert futtatta kilometer-ACC run-CAUS-PAST J-NOM thirty ‘Juli had the boy run thirty kilometers’

18

20

fiu´t boy-ACC

The restriction on the cooccurrence of multiple accusative case and particles remains to be accounted for. It must be noted, however, that this restriction cannot be reduced to the Single Delimiting Constraint (28). The thematic objects in these examples (that of push or the agent of run) do not measure out the event, and are thus permitted to cooccur with a measure phrase. If the temporal or measure phrase is the associate of a verbal modifier, then it behaves as an argument and allows extraction, as argued by U¨ro¨gdi 2003 for other types of adjuncts (K. E´. Kiss, p.c.). (i) Minek aludta a´t Juli a t what-DAT slept-3SG through J-NOM the ‘What did Juli sleep through the large part of?’

19

a the

nagy large

re´sze´t? part-POSS-ACC

The possibility of extraction in (i) fails to provide an argument for the adjunct or argument status of accusative temporal and measure phrases, since all adjuncts permit extraction if they are the associates of a verbal modifier. In contrast with E´. Kiss 2002a, I assume that the deleted projection is vP rather than VP. Here I follow E´. Kiss (chapters 2, 9) in assuming that particles and other elements move to Spec,PredP when they are in preverbal position. The constituents in Spec,PredP are verb modifiers, which include particles and noun phrases without a determiner or numeral as well as certain quantifiers and adverbs.

196 21

ANIKO CSIRMAZ

The data show that modifiers are not located in Spec,PredP when preceding the verb. Similarly to verbal modifiers in Spec,PredP, however, they must follow the verb if a focused constituent or negation is present. I assume that a stress-based account of focus (where nuclear stress is assigned to the leftmost constituent within the projection which houses accusative modifiers) can account for this distribution (see E´. Kiss (chapter 9)) for a detailed discussion of foci). ´ NOS futott (i) *Ke´t o´ra´t / ke´t kilome´tert JA ran two hour-ACC / two kilometer-ACC J-NOM ‘It was Jańos who ran for two hours / two kilometers’ ´ NOS futott ke´t (ii) JA o´ra´t / ke´t kilome´tert ran two hour-ACC / two kilometer-ACC J-NOM ‘It was Jańos who ran for two hours / two kilometers’

22

23

Accusative temporal and measure phrases can linearly precede these adverbs if the accusative modifiers appear as contrastive topics. The modifiers (as well as a number of other constituents) as contrastive topics can exceptionally appear in preverbal position. Contrastive topics are characterized by an intonation pattern of a brief fall followed by a long rise. They are interpreted as implicating the existence of alternatives to the denotation of the topic; an alternative predicate holds for the alternatives to the topic (E´. Kiss 2002). Not all particles can appear with a definite accusative adjunct. (47c) is in contrast with (i) below. The particles in the grammatical examples express completion rather than a goal or resultative. A result or goal particle is ruled out by the uniqueness restriction on delimiters (28). (i) *Jańos el / fel futotta away / up ran J-NOM ‘Jańos ran two kilometers’

a the

ke´t two


It is worth noting that incremental themes, which map to the event homomorphically, pattern with temporal and measure phrases with the distribution of verbal modifiers (E´. Kiss 2004a). Perfective events with indefinite objects appear without modifiers (ii), while perfective events with definite objects appear with a verbal modifier (iii). An event description with no verbal modifier, where the object is definite, is interpreted as atelic (iv) (E´. Kiss (chapter 2). (ii) Jańos ke´t alma´t evett J-NOM two apple-ACC ate ‘Jańos ate two apples’ (iii) Jańos meg ette a ke´t alma´t the two apple-ACC J-NOM perf ate ‘Jańos ate the two apples’ (iv) Jańos ette a ke´t alma´t the two apple-ACC J-NOM ate ‘Jańos was eating the two apples’ 24

A reviewer raises the question of whether both accusative time adverbs in (46a,b) are arguments. Both accusative time adverbs permit extraction, supporting their argument status: (i) [Hańy o´ra´t] futott ve´gig [(a) ha´rombo´l to.end the three-from how.many hour-ACC ran ‘How many hours from (the) three did Juli run to the end?’

25

26

t] J-NOM

Juli?

I assume that multiple case checking is defined for heads in a given language; thus while v can check multiple case features in Hungarian, T does not necessarily do so. Accusative case on adjuncts differs from the case of theta marked objects. The case of adjuncts is not subject to Burzio’s generalization; they can appear with unaccusative as well as with unergative and transitive predicates. The common account of accusative adjuncts and objects can be upheld even in


27

197

spite of this difference. It has been argued (Yip et al 1997, Schu¨tze 1997, Reuland 2000 and Woolford 2003, among others) that the relevant generalization is not connected to the availability of an agent theta role. Rather, the object receives nominative case if there is no nominative subject. The section does not address all numerals and adjectives with accusative case marking. Specifically, those examples where case marking appears as a result of ellipsis of the direct object, as in (ib) and (iib), are excluded (in general, the case marking or postposition appears on an adjective or numeral if the head noun is elided). (i) a. Jańos la´tott egy ke´pet one picture-ACC J-NOM saw ‘Jańos saw a picture’ b. Jańos la´tott egyet one-ACC J-NOM saw ‘Jańos saw one’ (ii) a. Jańos nagy su¨temeńyt evett ate J-NOM large cake-ACC ‘Jańos was eating a large cake’ b. Jańos nagyot evett J-NOM large-ACC ate ‘Jańos was eating a large one’ Generally no overt noun can appear with the multiplicatives and accusative adjectives discussed below; they do not arise as the result of ellipsis. (iii) a. Jańos futott egyet one-ACC J-NOM ran ‘Jańos ran’ b. Jańos nagyot / jo´t J-NOM large-ACC / good-ACC ‘Jańos ran far / an exhausting run’

28

29

futott ran

While temporal modification by alatt adverbs is generally marked, they are not outright ungrammatical for all speakers (K. E´. Kiss, p.c.). Thus event descriptions modified by an accusative multiplicative do not pattern identically with telic event descriptions, but they allow – even if marginally – temporal modification similarly to telic predicates. As noted in section 2, Hungarian has two kinds of multiplicatives. Multiplicatives which appear with the suffix -szer, are not delimiters. It is expected then that they can modify all eventuality types. (i) Jańos egyszer nyert won J-NOM once ‘Jańos won once’ (ii) Juli egyszer ko¨nyvet olvasott book-ACC read J-NOM once ‘Juli once read a book / books’ (iii) Jańos egyszer fel ugrott up jumped J-NOM once ‘Jańos once jumped up’

30

31

Smith 1991 requires telic event descriptions to have a resultant state. Since semelfactives lack such a state, they are atelic according to this classification, which encodes the same distinction as Tenny 1994’s delimitedness. On the ungrammaticality of non-terminative particles, see later in the section. The assumption that accusative multiplicatives are delimiters does not account for all aspectual restrictions imposed by these multiplicatives. States cannot be modified by accusative multiplicatives, in spite of being nondelimited (i). I assume that modification of states is independently excluded.

198

ANIKO CSIRMAZ

(i) a. *Juli o¨ru¨lt egyet J-NOM was.happy one-ACC ‘Juli was happy’ b. *Juli va´rt egyet a J-NOM waited one-ACC the ‘Juli waited at the stop’ 32

mega´llo´ban stop-in

The symbol marks immediate precedence. Pinõń 2001b’s definition also excludes frequentative verbs, such as in (i). I believe, however, that frequentatives do not need to be excluded. (i) A kutya ugra´lt egyet the dog-NOM jumped one-ACC ‘The dog jumped (several jumps)’

33 34

More precisely, by a strictly cumulative predicate. Contrary to Pinõń 2001b, I assume that not only semelfactives can be modified by a multiplicative higher than one. An activity can be modified by any multiplicative if it can be delimited by a conventionalized unit, as in (ii). This parallel behavior of semelfactives and activities also supports a uniform treatment of the two event types. (i) Jańos ketto ´´t bo´lintott J-NOM two-ACC nodded ‘Jańos nodded twice’ (ii) Jańos ketto´´t aludt (a sza´lloda´ban) J-NOM two-ACC slept the hotel-in ‘Jańos slept two nights in the hotel’

35 36

37

38

I explicitly exclude states from the range of constituents modified. The variable position of accusative multiplicatives raises the question of how they relate to universal mapping relations. I suggested that with both semelfactives and activities, the multiplicative introduces and modifies a result of the event. The specific semantic import of multiplicatives is different with the two predicate types. For instance, the multiplicative telicizes acivities, and leaves the (Vendlerian) telicity of semelfactives unaffected. The mapping procedures may be sensitive to this difference, and yield different first merge positions for the multiplicatives. This issue remains to be further explored. Manner adverbs are also merged within vP and moved to a position above PredP (E´. Kiss 2002a assumes that they are adjoined to PredP instead). The accusative multiplicative can linearly precede the adverb if the multiplicative is a contrastive topic, marked in boldface in (i). Even in this case, it takes narrow scope with respect to the adverb: (i)

39

40

Jańos ketto ´´t gyorsan klikkelt J-NOM two-ACC quickly clicked ‘Jańos clicked quickly twice’ (quickly > twice)

This linear order is only possible if the multiplicative is a contrastive topic. Contrastive topic quantifiers are interpreted as taking narrow scope, in violation of the usual correspondence of linear order and scope of preverbal constituents in Hungarian (E´. Kiss 2002a). As noted in the previous footnote, contrastive topics are characterized by rising intonation and implicate the existence of alternatives to the topic. Contrastive topics have narrow scope with respect to other preverbal constituents. An optional contrastive topic intonation also allows a narrow scope reading for the multiplicative in (65a). E´. Kiss 2004a treats verbs modified by accusative multiplicatives (and adjectives) as verbs of creation, which appear with non-specific objects in general. The source of the nonspecificity restriction is the interpretation of the verb in her account. However, verbs of creation other than the predicates under discussion can also appear with a specific (including a definite) object if the creation verb is modified by a particle.

A C C U S AT I V E C A S E A N D A S P E C T (i) Jańos meg / le J-NOM perf / down ‘Jańos wrote a letter’

41

42

egy one

levelet letter-ACC

Since predicates with multiplicatives and adjectives cannot appear with either a specific accusative constituent or a particle, the examples parallel to (i) must be independently excluded. It is worth noting that the ungrammaticality of specific multiplicatives is not due to the obligatory non-referential nature of multiplicatives (and adjectives, discussed in the following section). Definite temporal and measure modifiers are also non-referential, but denote a contextually salient measure or a salient span of time / space. Due to the nonspecificity restriction on multiplicatives and adjectives, they cannot trigger definiteness agreement (a reflex of structural case checking by v). In fact, as noted in section 3, the restriction also extends to adjunct accusative temporal and measure phrases. Transitive predicates with a degree argument cannot be modified by accusative adjectives. The source of this ungrammaticality remains to be explored.

(i) *Jańos J-NOM ‘Jańos (ii) *Jańos J-NOM ‘Jańos 43

´ırt wrote

nagyot emelte a su´lyzo´t large-ACC lifted the weight-ACC lifted the weight high’ nagyot va´ltoztatta a szoba´t large-ACC changed the room-ACC changed the room considerably’

The degree argument can be modified by a measure phrase as well: (i) Az a´rfolyam ke´t pontot the exchange rate-NOM two point-ACC ‘The exchange rate fell two points’

44

45

46 47 48

199

esett fell

The predicate esik (‘fall’) also behaves as a non-delimited event, even though it is delimited in the real world. I assume that esik is a non-delimited event in Hungarian, and is delimited by either a nontheta marked accusative constituent, a terminative particle or a goal. This difference between real world and linguistic properties is reminiscent of the durativity of events. Even instantaneous events have some duration, but they are treated as instantaneous, having an atomic duration in the linguistic system. Accusative adjectives results have a predicted delimited interpretation, as noted earlier for non-theta marked accusative constituents in general. As for why accusative case can appear on adjectives, I assume that the adjectives that modify an understood object, time, distance or intensity of an eventuality can appear with a case feature. Alternatively, it may be argued that in the cases discussed, the adjectives modify some argument (e.g. time) or modifier (e.g. result) of the eventuality. In this approach, the modified constituent has a case feature, and in some cases, may not be overtly expressed. The case marking of the complex constituent thus appears on the overt adjective, as with the elided nouns discussed above. An evaluation of these accounts, as well as a more specific account of the interpretation of these adjectives and the mechanism of case marking is left for future research. With the exception of reflexives and body parts. The verb is also merged with the (projection of the) cognate object in Hale and Keyser 2002. A result object or result state interpretation is more general; it is argued, for instance, for Hebrew cognate objects by Mittwoch 1998.

200 49

ANIKO CSIRMAZ

Szo¨rnyet (‘monster-ACC’), which resembles szo¨rnyut ˝ (‘horrible-ACC’), is etymologically unrelated. The interpretation of szo¨rnyet and szo¨rnyut ˝ is also different: (i) Jańos szo¨rnyet halt J-NOM monster-ACC died ‘Jańos died an accidental death’ While in some environments both an accusative multiplicative or adjective and a cognate object is possible (ii) (K. E´. Kiss, p.c.), this is not necessarily the case (iii): (ii) Jańos ha´rom le´pe´st / ha´rmat le´pett step-ACC three-ACC stepped J-NOM three ‘Jańos took three steps’ (iii) Jańos *ha´rom kopoga´st / ha´rmat kopogott knock-ACC three-ACC knocked J-NOM three ‘Jańos knocked three times’

50

Numerals are acceptable with a number of cognate objects, but not always: (i) Jańos ha´rom le´pe´st le´pett step-ACC stepped J-NOM three ‘Jańos took three steps’ (ii) *Jańos ha´rom ugra´st ugrott three jump-ACC jumped J-NOM ‘Jańos jumped three times’


A P PA R E N T O R R E A L ? O N T H E C O M P L E M E N TA RY D IS T R I B U T I O N O F I D E N T I F I C AT I O N A L F O C U S A N D T H E V E R B A L PA RT I C L E

1.

INTRODUCTION

Chapter 2 of this book has argued that in the Hungarian sentence, the primary locus of the marking of event type (or, in a different terminology, the primary locus of the marking of situation aspect) is the immediately preverbal slot, identified as Spec,PredP. This is the canonical position of resultative and terminative particles marking accomplishments and achievements, and this is the canonical position of locative particles marking predicates of existence or spatial configuration in a given location. Interestingly, the immediately preverbal position is also the position of identificational focus. The question this chapter aims to answer is if the complementary distribution of the focus and the verbal particle in preverbal position is apparent or real, and – whether it proves to be apparent or real – what brings about the parallel properties of particle-initial and focus-initial projections. The chapter first surveys the different analyses of the relative position of the preverbal verbal particle, bare nominal complement, and identificational focus. Section 2 presents the problem to be accounted for, and its earliest accounts, which share the view that the focus and the particle/bare nominal are in complementary distribution. Section 3 discusses Brody’s (1990, 1995) focus theory, which assigns the verbal particle/bare nominal and the focus to separate projections. In Brody’s focus theory, focus movement into the specifier of FocP is combined with V-movement into Foc, and both movements are motivated by feature checking. Section 4 introduces Szendro´í’s (2003) stress-based focus theory, which motivates focus movement by the requirement that the most emphatic constituent occupy the position assigned highest stress by the regular rules of grammar – while maintaining a separate landing site for focus constituents. Section 5 deals with Horvath’s (2000, 2005) focus theory, in which focus movement is phrasal movement to the specifier of an Exhaustive Identification Phrase, not accompanied by V-movement. Section 6 discusses the merits and problems of E´. Kiss’s (2003) complementary distribution theory, which adopts a predicative analysis of focus, placing the verbal particle, the bare nominal, and the focus in the specifier of the same PredP projection. Based on evidence presented in section 3, showing that the PredP projection harboring the verbal particle/bare nominal is dominated by AspP, whereas the predicative projection harboring the focus is external to AspP, section 7 discards the ‘complementary distribution’ element of E´. Kiss’s (2003) theory, and puts forth a new version of focus movement combined with V-movement. The proposal is shown to be exempt from the problems raised by earlier theories of a similar kind. The main 201 ´ Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 201–223. Katalin E. 2008 Springer.


202

problem of former approaches involving V-movement, the free word order of postverbal constituents, is eliminated by invoking Behaghel’s (1932: 6) Law of Growing Constituents in a phase-theoretic framework. 2.

THE PROBLEM AND ITS EARLIEST SOLUTIONS

The verbal particle and the verb form a close lexical, semantic and phonological unit in Hungarian (the particleþV string is even spelled as one word; it is hyphenated here merely for expository purposes). No wonder the fact that a focussed constituent causes the verbal particle to appear postverbally has puzzled linguists for two centuries. Compare: (1)

a. Jańos meg- vette a ko¨nyvet. bought the book John PRT ‘John bought the book.’ b. JAŃOS vette meg a ko¨nyvet. ‘It was John who bought the book.’ c. Jańos A KO¨NYVET vette meg. ‘It was the book that John bought.’

Bare nominal complements also share the variable the behavior of verbal particles in the presence and absence of focus: (2)

a. Jańos ko¨nyvet vett. John book-ACC bought ‘John bought some book(s).’ b. JAŃOS vett ko¨nyvet. ‘It was John who bought some book(s).’ c. Jańos TEGNAP vett ko¨nyvet. ‘It was yesterday that John bought some book(s)’.

A preverbal constituent other than a verbal particle or bare nominal complement, called focus, expresses exhaustive identification, similarly to an English cleft constituent. The problem that Hungarian linguists faced and still face is how to account for the similar syntactic distribution and the different semantic interpretation of the different types of preverbal elements. In the first formal analysis of Hungarian phrase structure by Brassai (1863–65), the immediately preverbal position was identified as a kind of focus position, and the alternative preverbal presence of the verbal particle and the focus was derived from the assumption that they are alternative fillers of the focus slot. The complementary distribution of the focus and the verbal particle (or the complementary distribution of the focus and the verb modifier, a category subsuming verbal particles and bare nominal complements) was also maintained in the first generative analyses of Hungarian sentence structure, for example, in Horvath (1986), and in E´. Kiss (1987, 1994). E´. Kiss (1994) derived the different interpretations of a preverbal verbal particle and a preverbal focussed argument from their different referential properties. She claimed that focussing expresses identification, which involves the exclusion of alternatives only if the focussed constituent denotes a member of a closed set of distinct entities.

A P PA R E N T O R R E A L ?

203

Farkas (1986) was the first to notice that the idea that the verbal particle and the focus are alternative fillers of the same position also has unwelcome syntactic consequences. For example, if both were located in the same Spec,VP slot, we ought to be able to coordinate VPs no matter if their specifiers are filled by a focus or by a verbal particle – which is not the case: (3)

*[[VP JAŃOS [V’ vette meg a ko¨nyvet]] e´s [VP el- [V’ olvasta pro]]] and PRT read it John bought PRT the book ‘It was John who bought the book, and he read it.’

Instead of assuming the complementary distribution of the focus and the verb modifier, Farkas (1986) postulated a syntactic feature [þINV] on the VP of sentences with focus (as well as on the VPs of negative sentences and imperatives, which also display a V Particle order irrespective of whether or not they have a preverbal focus). 3.

B R O D Y ’ S F O C P T H E O RY

Brody (1990; 1995) also discarded the complementary distribution of verb modifiers and foci, arguing that they occupy different positions associated with different interpretations. The verb modifier is base-generated left-adjoined to the verb, whereas the identificational focus sits in the specifier of a functional projection called FocusP (abbreviated as FP or FocP). The focus is immediately followed by the verb as a consequence of V-to-Foc movement across the verbal particle:

(4)

FocP Spec

Foc’ Foc

VP

V’

V+

VM

XP

XP

V

t j Pétert t i JÁNOSi hívtaj fel John called up Peter ’It was John who called up Peter.’

204


Brody (1995) also reformulated his focus theory in the Minimalist framework. In Minimalism, movement is always a ‘‘last resort’’, triggered by the need of a constituent to have its uninterpretable morphological feature checked. The morphological feature involved is claimed to be a [þf] feature, which figurs as the head of the FocP projection, and also appears on the Tense head. To have its [þf] feature checked, the Tense (or, in the case of a strong Tense, the TenseþV complex) must move into Foc, stranding the verbal particle. The [þf] feature of the head of FocP must be checked in a specifier–head configuration by a phrase carrying the feature [þf], attracted into Spec,FocP. A number of recent theories have questioned various aspects of Brody’s approach. Horvath (2000) criticizes the cross-linguistic predictions of Brody’s focus theory, among others. According to Brody (1990, 1995), the FocP projection is also present in languages with no overt focus movement. In such languages, e.g. in English, the [þf] feature of the head of FocP is claimed to be weak, triggering movement into Spec,FocP only at LF. If the prosodic focus of e.g. the English sentence ends up in LF in the same position as the structural focus of the Hungarian sentence, they are expected to be associated with the same type of interpretation in LF; in fact, however, the English prosodic focus lacks the exhaustivity of the Hungarian structural focus. Horvath (2000) also points out a problematic aspect of Brody’s account of Hungarian focus facts. The [þf] feature is sometimes associated with a subconstituent of the Hungarian focus phrase which cannot license pied-piping. According to Webelhuth (1992), a modifier, for example, is not a pied piper. That is, if a modifier has a feature to be checked in the specifier of a functional projection, e.g. it has a wh-feature to be checked in Spec,CP, the modifier cannot move there on its own (because of violating Subjacency), and it cannot take along the modified phrase, either. The claim that modifiers cannot pied-pipe is also confirmed by the following facts of Hungarian: (5)

a. *A szıńe´szno´´, [ne´hańy akiro´´l ´ırt ko¨nyvet] la´ttam a the actress some who-about written book-ACC saw-I the polcon shelf-on ‘The actress some books written about whom I saw on the shelf’ b. *[Ne´hańy kiro ´´l ´ırt ko¨nyvet] la´tta´l a polcon? some who-about written book-ACC saw-you the shelf-on ‘Some books written about whom did you see on the shelf?’

Whereas a [þwh] modifier cannot pied-pipe the noun phrase containing it into Spec,CP or Spec,FocP, the very same modifier seems to licence pied-piping into Spec,FocP when associated with a [þf] feature: (6)

ko¨nyvet] la´ttam a polcon. [Ne´hańy MARILYN MONROE-ro´l ´ırt some Marilyn Monroe-on written book saw-I the shelf-on ‘It was some books written about MARILYN MONROE that I saw on the shelf.’


205

In view of the different grammaticality of (5a,b) and (6), it seems doubtful that it is indeed the [þf] feature of Marilyn Monroe-ro´l ‘about Marilyn Monroe’ that triggers the movement of the maximal noun phrase in (6). Horvath (2005) also presents an argument against the V-to-Foc movement assumed by Brody. The argument is based on the following ellipsis facts: (7)

KIT hı´vott meg? whom invited-he PRT ‘Who did he invite?’ b. (Azt mondta, hogy) PE´TERT. said-he that Peter-ACC it-ACC ‘[He said that he invited] Peter.’

a.

The answer in (7) involves sluicing, which is claimed by Horvath to consists in the deletion of a maximal projection. In Brody’s framework, however, the V-initial string is an intermediate Foc’ projection, which is not a possible target of deletion. E´. Kiss (2003) raises both syntactic and semantic problems in connection with Brody’s proposal. The syntactic problems are the following: Brody’s theory predicts that in a focus construction, the verbal particle immediately follows the verb raised to Foc. In fact, however, any number of constituents can intervene between the verb and the particle, i.e., (8a) and (8b) are equally grammatical: (8)

a. JAŃOS veszett o¨ssze E´va´val. John fell out Eve-with ‘It was John who fell out with Eve.’ b. JAŃOS veszett E´va´val o¨ssze.

It is also unclear how negation can be handled in Brody’s framework. If particle stranding is evidence of V movement, then predicate negation involves V-to-Neg movement: (9)

Jańos nem veszetti o¨ssze ti John not fell out ‘John did not fall out with Eve.’

E´va´val. Eve-with

The string in (9) suggests that the negative particle sits in Spec,NegP, and the V occupies the empty Neg position. In the presence of a focus constituent, however, the V required to move to Foc also takes the negative particle along. This is only possible if the negative particle originally occupies the Neg position, then the V is rightadjoined to Neg, and, finally, the NegþV complex is raised to Foc. The problematic step is the right-adjunction of the V to Neg, instead of the left-adjunction typical of head movement.1


206

(10)

TopP Spec

Top’

Top

FocP Spec

Foc’ Foc

NegP Neg

VP V+ V PRT

XP

XP XP

Jánosi ÉVÁTj [nem hívtak]l t l t k meg vacsorára t j John Eve-ACC not invited PRT dinner-to ‘It was Eve who John did not invite to dinner.’

ti

A further problem of the framework is that NegP can be inserted both above VP and above FocP, but the Neg head attracts the V only in the former case. If NegP dominates FocP, the verb is stranded in Foc position. The different behavior of the lower and the higher Neg heads seems unmotivated. Compare: (11)

a. Jańos nem hı´vta meg E´va´t vacsora´ra. PRT Eve dinner-to John not invited ‘John didn’t invite Eve to dinner.’ b. Jańos nem E´VA´T hı´vta meg vacsora´ra. ‘It was not Eve who John invited to dinner.’2

A semantic problem pointed out by E´. Kiss (2003) has also been raised in a somewhat different form by Wedgwood (2002). They both argue that the constituent moved into the Spec,FocP position of the Hungarian sentence is not always


207

associated with an exhaustive interpretation. Wedgwood’s examples involve a modified numerally quantified noun phrase, e.g.: (12)

fe´lre a ke´rde´st. TO¨BB MINT HAT DIA´KUNK e´rtette more than six student-our understood astray the question ‘More than six students of ours misunderstood the question.’

Because of the lack of exhaustivity, Szabolcsi (1997) places modified numerally quantified noun phrases into a preverbal position other than Spec,FocP (called PredOp by Szabolcsi). Wedgwood demonstrates that the preverbal constituent in (12) does, in fact, occupy the same position that exhaustive foci occupy. He claims that exhaustivity is not part of the truth-conditional semantics associated with the FocP projection; it involves no features and/or operators; it derives from inferential pragmatics relying on the context. An exhaustive answer is the optimal answer when something new is asserted in a restricted context, and is therefore always expected. An exhaustive answer may have different contextual effects depending on whether it merely establishes new assumptions, or also eliminates existing ones. Numerals do exhaustively identify a numeral from among a set of numerals – merely in such cases, the exclusion of alternative numerals is pragmatically trivial (unless there is a specific, contextually given numeral to be contradicted).3 Wedgwood’s proposal actually does not clearly discard the FocP projection, although it does refer to the possibility of a semantically underspecified preverbal position hosting alternatively a verbal particle or a focus. Of course, if FocP is maintained, his findings raise the question what kind of an A-bar position Spec,FocP is; what motivates movement into it. E´. Kiss (2003) presents indefinite foci that do not require an exhaustive interpretation, for example: (13) A beto¨ro´´ EGY ARANY NYAKLAŃCOT E´S EGY EZU¨ST ZSEBO´RA´T the burglar a golden necklace and a silver pocket-watch tala´lt a laka´sban. found the apartment-in ‘The burglar found A GOLDEN NECKLACE AND A SILVER POCKETWATCH in the apartment.’ According to the exhaustivity test of Szabolcsi (1981), (13) is not exhaustive – because (14) is a logical consequence of (13): (14)

A beto¨ro´´ EGY ARANY NYAKLAŃCOT tala´lt a laka´sban. the burglar a golden necklace found the apartment-in ‘The burglar found A GOLDEN NECKLACE in the apartment.’


208

Interestingly, an indefinite focus assumes an obligatorily exhaustive reading if the sentence contains a verbal particle. Compare with (13): (15)

A beto¨ro´´ EGY ARANY NYAKLAŃCOT E´S EGY EZU¨ST ZSEBO´RA´T the burglar a golden necklace and a silver pocket-watch tala´lt meg a laka´sban. found PRT the apartment-in ‘It was a golden necklace and a silver pocket-watch that the burglar found in the apartment.’

(15) and (16) cannot be true simultaneously in the same world, hence their foci are exhaustive: (16)

A beto¨ro´´ EGY ARANY NYAKLAŃCOT tala´lt meg a laka´sban. the burglar a golden necklace found PRT the apartment-in ‘It was a golden necklace that the burglar found in the apartment.’

The objects in (15) and (16) are [þspecific] in the sense of En¸c (1991); they have a partitive reading; e.g. (15) means ‘Of the jewels, the burglar found a golden necklace and a silver pocket watch in the apartment’. As was made clear in chapter 2, the verbal particle, a secondary predicate predicated of the theme, requires that the theme argument functioning as its subject be [þspecific]. That is, a focussed constituent is understood as exhaustive if it is [þspecific]. A focussed constituent with a [þ/specific] feature, like that in (13) or (14), need not be exhaustive (although the exhaustive interpretation is always possible). At the same time, it is not the case that a [specific] focus (e.g. a focussed bare plural) is necessarily non-exhaustive; it is [þ/exhaustive], e.g.: (17)

A beto¨ro´´ ARANY NYAKLAŃCOKAT tala´lt a laka´sban. the burglar golden necklaces found the apartment-in ‘The burglar found GOLDEN NECKLACES in the apartment./It was golden necklaces that the burglar found in the apartment.’

In Brody’s framework, a focussed constituent is always associated with the same [þf] feature, i.e., the theory cannot predict the [þexhaustive] or [þ/exhaustive] nature of a focus.

4.

I ’ S T H E O RY O F F O C U S S I N G A S SZENDRO STRESS-DRIVEN MOVEMENT

From time to time, languages with no visible focus movement inspire focus theories in which focus is a pragmatic notion encoded in prosody. A recent theory of this type, by Reinhart (1995), has been adapted by Szendro´í (2003) to Hungarian. The key


209

element of Reinhart’s theory is the following principle: (18)

Stress–Focus Correspondence Principle The focus set of a clause consists of the constituents containing the main stress of the clause as determined by the regular stress rule of the given language.

Szendro ´í’s version of the theory is based on the claim that in Hungarian main stress falls on the left edge of the maximal extended projection of the V. (Topics are exempt from stress assignment because they are preposed by adjunction, which renders them extrametrical.) In Szendro ´í’s view, focus movement to Spec,FocP is stress-driven movement: it is triggered by the Stress–Focus Correspondence Principle, i.e., it serves the purpose of the focussed constituent occupying the position where main stress is assigned. This solution is exempt from several stipulations necessary for Brody’s analysis to work. Namely, it need not be stipulated that the [þf] feature of the head of FocP is strong in order to account for the overtness of focus movement; movement for stress cannot but be overt. It need not be stipulated, either, that Tense also has a [þf] feature in need of checking in order to account for V movement to F; the V must move so as to license the head position of FP, in other words, to activate a functional projection above VP. In fact, Szendro ´í need not even stipulate a [þf] feature; she merely needs a functional projection capable of extending the VP (FP means ‘Functional Projection’ for Szendro ´í). It also falls out that verb focussing does not trigger Vto-F movement. Szendro ´í’s theory of focus movement for stress predicts that focus movement is movement to the left in Hungarian; and it cannot be iterated. Horvath (2005) refutes important elements of Szendro´í’s theory – though she accepts the relevance of Reinhart’s Stress–Focus Correspondence Principle in Hungarian. She points out that in Szendro´í’s framework, syntax must look ahead into prosody, and must find out what the prosodic structure is about to be like. Namely, V-to-F movement is performed in syntax only if stress-driven focus movement is about to take place, and movement to Spec,FP is triggered only if the constituent to receive main stress occupies a position other than the leftmost position of the extended VP. That is, it must be clear in syntax where the left edge of IntP, the locus of main stress, is about to be. Among the empirical problems, Horvath mentions that the preverbal focus position is not necessarily the position of main stress. Universal quantifiers and is phrases can bear main stress, nevertheless, they cannot appear in the preverbal focus slot – see (19a). They can, however, precede the preverbal focus position, in which case the focussed constituent can be completely destressed. This is what happens to the focussed Marit ‘Mary-ACC’ in (19b). (19)

a. *PE´TER IS ke´rte fel Marit tańcolni, nemcsak a bara´tja. Peter too asked PRT Mary to.dance not-only her friend ‘It was Peter, too, who asked Mary for a dance, not only her friend (did so).’


210

b. PE´TER IS Marit ke´rte fel tańcolni, nemcsak a bara´tja. Peter too Mary-ACC asked PRT to-dance not-only her friend ‘PETER, TOO, asked MARY for a dance, not only her friend (did so).’ Whereas these types of stressed quantified elements cannot appear in the immediately preverbal main stress position, they can stand postverbally, unlike e.g. a focus modified by csak. Compare: (20)

a.

Mari el -ke´sett me´g AZ ESKU¨VO JE´RO L is. also Mary PRT missed even her wedding ‘Mary was late even FOR HER WEDDING.’

cf. b. *Mari el -ke´sett CSAK A FOGADA´SRO´L. Mary PRT missed only the reception ‘It was only the reception that Mary was late for.’ c. *Mari CSAK A FOGADA´SRO´L el-ke´sett. d. Mari CSAK A FOGADA´SRO´L ke´sett el. Szendro´í’s theory cannot derive focus movement into the unstressed preverbal position in (19), and, in general, it cannot predict the distribution of grammaticality in (19)–(20). The Reinhart–Szendro´í theory does not exclude the possibility of an in situ focus. The requirement of Stress–Focus Correspondence can be observed not only by focus movement into the main stress position, but also by stress shift to the constituent to be focussed. What Szendro´í’s theory cannot predict is why preverbal and postverbal foci differ with respect to exhaustivity, why a preverbal focus, like that in (21b), is exhaustive, and why a postverbal focus, e.g. that in (21c), is not. (21)

hı´vott ma´r meg Anna? a. KIKET who-PL-ACC invited already PRT Anna ‘Who has Ann invited yet?’ E´S PE´TERT hı´vta meg. b. KATIT Kati-ACC and Peter-ACC invited-she PRT ‘It is Kati and Peter that she has invited.’ c. Meg-hı´vta KATIT E´S PE´TERT. ‘She has invited KATI and PETER.’

Horvath’s argument against V-to-F movement, based on the sluicing facts in (7), also applies to Szendro´í’s theory. Consider example (22): if sluicing is XP-deletion, as is assumed by Horvath, (22a) should be grammatical. It is not; its grammatical variant is (22b). (22)

KI ugrotta a´t a kerı´te´st? Who jumped across the fence?


a.

*PE´TER ugrotta. Peter jumped ‘Peter did.’ 5.

211

b. PE´TER.. Peter

T H E E I P T H E O RY O F H O RVAT H ( 2 0 0 0 , 2 0 0 5 )

Horvath’s theory of focus aims to assimilate Hungarian structural focus to the notion of focus valid crosslinguistically, manifest in main stress interpreted as ‘new, nonpresupposed information’. She attains this goal by divorcing exhaustive identification from focus. She claims that Hungarian has ‘‘grammaticalized’’ the notion of exhaustive identification. The Hungarian sentence contains a quantificational, phonologically null Exhaustive Identification (EI) operator. The EI feature, projecting a clausal functional head, is strong, thereby it attracts a constituent with an EI-operator feature into Spec,EIP. The EI operator – entering into a checking relation with the strong feature of EI – is merged into the structure as the specifier of a DP (or PP, VP). This invisible EI-operator is assumed to appear in the specifier position of an XP because specifiers are pied-pipers; thus under this assumption it falls out that EI-movement pied-pipes the whole phrase whose specifier the EI operator occurs in. Focus, on the other hand, is claimed by Horvath (2005) not to be encoded syntactically; it is claimed to be a stress-based interface phenomenon universally. Horvath (2005) adopts Reinhart’s (1995) Stress–Focus Correspondence Principle, and she assumes – in accordance with Szendro´í (2003) – that the position of main stress is at the left edge of the extended VP. The relation of the EI-operator to Focus is indirect, a case of ‘‘association with focus’’: the EI operator requires the presence of prosodic focus within its c-command domain. This is the structure that Horvath (2000, 2005) assigns to a Hungarian ‘‘focusconstruction’’:

CP

(23) C

EIP EI’

DPi

EI-op.

DP EI

IP … ti

…


212

The proposed structure eliminates all the problems brought up by Horvath against Brody’s and Szendro´í’s theories. Thus it explains the different behavior of the ‘‘focussing’’ particles csak ‘only’ and is ‘also’, me´g . . . is ‘even’. Csak-phrases are marked as [þEI], hence they will land in Spec,EIP. The semantics of (me´g) . . . is phrases, on the other hand, is incompatible with that of exhaustive identification, hence they are [-EI], excluded from Spec,EIP. Both csak and (me´g) . . . is are focussing operators in Hungarian, too, requiring a prosodic focus in their c-command domain.4 Horvath’s theory also explains why only a preverbal focus is exhaustive, why a postverbal is not. Her theory – she claims – can also accomodate the cases in which the preverbal constituent, though expressing exhaustive identification, is not focussed prosodically, e.g.: (24)

Marit ke´rte fel tańcolni. MINDEN fiu´ every boy Mary-ACC asked PRT to.dance ‘EVERY boy asked MARY for a dance,’

Here Marit occupies Spec,EIP, therefore it expresses exhaustive identification; the main stress marking the focus, on the other hand, is on the universal quantifier in Spec,DistP. In fact, (24) is problematic for Horvath’s theory, as the EI operator in ´ . Horvath solves this Spec,EIP does not c-command the focussed MINDEN FIU problem by the assumption of a stress shift or stress deletion. For me, stress deletion seems more likely. I assume that every maximal projection extending the VP requires the assignment of a ‘strong’ stress, and the stacking of ‘strong’ stresses – shown in (25b) – may trigger stress reduction on one or the other of the constituents.

(25) a.

DistP Spec

b. s

EIP Spec

s

VP

w s

w

V The stress pattern in (26a), corresponding to (25b), and the two stress patterns derivable from it by means of stress deletion, shown in (26b,c), are equally grammatical. (Stress is marked by the symbol ’.) (26)

a. ‘MINDEN fiu´ ‘MARIT ke´rte fel tańcolni. b. ‘MINDEN fiu´ MARIT ke´rte fel tańcolni. c. MINDEN fiu´ ‘MARIT ke´rte fel tańcolni.

Horvath argues that movement to Spec,EIP is a subtype of Q-Raising, which is overt in Hungarian.


213

Horvath’s theory eliminates many of the problems pointed out in Brody’s and Szendro´í’s theories; however, it does not eliminate all of them, and it also raises a new question. If focussing is the movement of a phrase containing an Exhaustive Identification operator into Spec,EIP, one would not expect question–answer pairs of the following type: (27)

hı´vta´l meg? a. KIT whom invited-you PRT ‘Who did you invite?’ b. To¨bbek ko¨zo¨tt PE´TERT (hı´vtam meg). among others Peter-ACC (invited-I PRT) ‘Peter, among others.’

Why should Pe´tert in (27b) be simultaneously supplied with an EI operator, and with a phrase cancelling the EI operator? How can the [þexhaustive] feature of a phrase moved to Spec,EIP in order to check the [þexhaustive] feature of the EI head be cancelled? The constituents that land in Spec,FocP in Brody’s theory all land in Spec,EIP in Horvath’s framework – so the problem why preverbal non-specific indefinites do not obligatorily express exhaustive identification is even more disturbing for Horvath’s theory than it is for Brody’s approach; after all, it is phrases supplied with an EIoperator, sitting in Spec,EIP that lack exhaustive interpretation. Horvath’s focus/EI theory also leaves important questions open. She does not clarify what kind of internal structure she assigns to the IP. The possibility of V-to-EI movement across the particle is excluded, and so is the possibility of analyzing the EI-phrase and the verbal particle as alternative fillers of the same slot. It remains unclear how the reverse ‘V, particle’ order arises in the presence of a constituent in Spec,EIP (let alone the order ‘V, XP, particle’). 6.

F O C U S S I N G A S P R E D I C AT I O N ( E´ . K I S S 2 0 0 3 )

The proposal put forth in E´. Kiss (2003) has three main ingredients. Namely: (i) It is argued that the Hungarian structural focus is a predicate predicated of the open proposition constituted by the post-focus part of the sentence. It functions as a specificational predicate, referentially specifying the set denoted by the subject of predication. Exhaustivity is a property associated with specificational predicates. (ii) It is argued that the focus and the verbal particle are alternative fillers of the same preverbal position identified as Spec,PredP. (iii) The Reinhart (1995) – Szendro´í (2003) approach to information focus is adopted, and Reinhart’s Stress–Focus Correspondence Principle is maintained. It is assumed that primary stress is assigned to the leftmost element of the extended projection of the VP, i.e., to the constituent in Spec,PredP, or in lack of a PredP projection, to the V.5 Hence it is the constituent in Spec,PredP (or the whole PredP projection) that functions as the information focus of the sentence.

214


Claim (i), concerning the predicative nature of focus, is based on the observation that the Hungarian structural focus is an equivalent of the English cleft and pseudo-cleft constituents, sharing their semantic interpretation and their key syntactic properties (cf. E´. Kiss 1998b), hence Higgins’s (1973) analysis of pseudoclefts and Huber’s (2000) analysis of clefts can be adopted to the Hungarian construction. Higgins’s description of the English pseudo-cleft construction (1973) is based on a detailed analysis of nominal predicates. Higgins distinguishes three types of them, illustrated in (28): (28)

a. John is a teacher. b. That is Mary Brown. c. The winner is the Hungarian team.

The nominal predicate of (28a) is a predicational predicate; it predicates a property about a referential subject. A predicational predicate is usually represented by an indefinite noun phrase or an adjective phrase. The predicate of (28b) is an identificational predicate; its function is to teach the name of a referential subject. (28c) contains a specificational predicate. In (28c), the order of the subject and the predicate can be reversed. In this sentence type, neither the subject, nor the predicate is referential. In Higgins’s metaphorical terminology, the subject functions as the superscript of a table, the heading of a list, and the predicate specifies what makes up the list. Or, putting it differently, the subject of predication delimits a domain, and the specificational predicate identifies the members of that domain. A specificational predicate is claimed by Higgins to imply exhaustivity, which, however, can be cancelled. This is what happens in (29): (29)

What I bought was a pen and a pencil, among other things.

That is, although the specificational predicate serves to list the members of the subject set exhaustively, the listing can be broken off – provided the fact of interruption is explicitly indicated. The subject of a specificational predicate is associated with an existential presupposition. When the specificational predicate is negated, the existence of the domain delimited by the subject is not questioned; merely the given specification is denied. Thereby a different specification is implied, i.e., a continuation with a but phrase of the negated predicate is expected: (30)

What I bought was not a pen and a pencil (but a rubber and a ruler).

Higgins (1973) argues that pseudo-cleft sentences are specificational sentences. Their What . . . clause functions as the subject of predication, and the predicate complement to the copula functions as the specificational predicate. With the subject presupposed, the predicate functions as the focus of the sentence. The list reading and the exhaustivity of the predicate are claimed to be implications of the specificational function.


215

Huber (2000) analyzes Swedish and German cleft sentences in the framework of Higgins’s theory of specification. In Huber’s terminology, the subject of a specificational sentence denotes a set, which the predicate characterizes through another set, by listing the individuals that make it up. A specificational predicate implies that its specification of the individuals that make up the set denoted by the subject is exhaustive, that is, other alternatives are excluded. The subject of predication is associated with an existential presupposition because only the content of an existing set can be listed. The semantic characterization of specificational predicates exactly fits Hungarian identificational foci, as well; therefore, E´. Kiss (2003) concludes that the so-called identificational focus of Hungarian is a type of specificational predicate. The specificational analysis of Hungarian structural focus is claimed to be exempt from all the semantic problems raised by other focus theories. In the framework of Szendro´í (2003), it remains unclear why a preverbal focus implies exhaustive identification, unlike a postverbal one. In this framework, on the other hand, exhaustive identification is not a function of main stress; it is associated with a constituent in predicate position predicated of the VP. Crucially, a postverbal stressed constituent is not in predicate position; it sits among the arguments. In the framework of Horvath (2000, 2005), the exhaustive identification interpretation of a preverbal constituent follows from the Exhaustive Interpretation operator in its specifier. It remains unclear how the EI feature of a constituent supplied with an EI operator can be cancelled with a to¨bbek ko¨zo¨tt ‘among others’ phrase – while it is needed to check the uninterpretable EI feature of the head of the EIP projection. In the framework of the specificational predicate analysis of identificational focus, on the other hand, exhaustivity is not an operator; it is not asserted; it is merely implicates. Whereas none of the previous theories of focus have been able to predict the optionality of exhaustive interpretation in the case of non-specific foci, it receives a natural explanation in the framework of E´. Kiss (2003). Chapter 1 argued that the verbal particle and the bare nominal complement are alternative fillers of the same preverbal predicative position. Whereas the verbal particle is a secondary predicate predicated of the theme argument, the bare nominal complement is a secondary predicate predicated of the implicit internal argument. For example: (31)

a. Jańos [PredP meg-vettei [VP ti a ko¨nyvet a gyerekeknek]] PRT bought the book the children-for John ‘John bought the book for the children.’ a gyerekeknek]] b. Jańos [PredP ko¨nyveti [VP ti vett bought the children-for John book-ACC ‘John bought some book(s) for the children.’

The bare nominal complement, e.g. that in (31b), is a predicational predicate. (31b) expresses that what John bought for the children is of the kind ‘book’, without referring to the number of books he bought. It can be paraphrased as follows: (32)

John bought x for the children and x is of the kind ‘book’.

216


E´. Kiss (2003) assumes that the landing site of the identificational focus/specificational predicate is the Spec,PredP position that can also be filled by the bare nominal complement. Then a bare nominal complement moved to Spec,PredP is, in fact, ambiguous; it can be understood either as a predicational element (predicated of the theme argument), or as a specificational element (predicated of the open sentence constituted by the VP). In the latter case, it implies exhaustive listing. This is, indeed, what we find. (31b) can mean, on the one hand, that ‘John bought something for the children, and it is of the kind book’; and on the other hand, it can also mean that ‘what John bought for the children is of the kind book’. The latter reading is exhaustive, implying that John bought nothing but book(s). When the constituent in Spec,PredP is a definite or a specific indefinite noun phrase, the reading under which it is a predicational predicate, predicated of the implicit internal argument, is not available; the specificational, exhaustive interpretation is obligatory: (33)

a. Jańos A HA´BORU´ E´S BE´KE´T vette meg a gyerekeknek. John the War and Peace bought PRT the children-for ‘It was War and Peace that John bought for the children.’ b. Jańos egy orosz szo´ta´rt vett meg a gyerekeknek. the children-for John a Russian dictionary bought PRT ‘It was a Russian dictionary that John bought for the children.’

If, on the other hand, the preverbal constituent is an indefinite noun phrase allowing either a specific or a non-specific interpretation, the non-specific reading will always be ambiguous between a predicational and a specificational reading. This is the case in (34), where there is no verbal particle to enforce the specific reading of the theme. (The stressing of the postverbal argument a gyerekeknek helps to elicit the predicational reading, whereas its destressing helps to elicit the specificational reading – even if there is no strict correlation between destressing and exhaustivity.) (34)

Jańos egy orosz szo´ta´rt vett a gyerekeknek. John a Russian dictionary-ACC bought the children-for ‘John bought the children A RUSSIAN DICTIONARY.’

That is, the proposed analysis correctly predicts the presence or absence of exhaustivity in the various cases of Hungarian focus constructions. Notice that in the proposed framework, the specificational predicate status is always assigned to the maximal phrase in the preverbal slot, therefore, the pied-piping problems noted by Horvath (2000, 2005) in connection with Brody’s theory do not arise.6 Claim (ii) of the theory of E´. Kiss (2003) returns in some respects to the earliest focus theories, according to which the verbal particle, the bare nominal complement, and the identificational focus are alternative fillers of the same preverbal position, analyzed in Chapter 2 as Spec,PredP. So as to account for the coordination fact noticed by Farkas (1986), illustrated in (3) above, and the coordination and ellipsis


217

phenomena illustrated in (35)-(37) below, the structures assigned by E´. Kiss (2003) to a ‘verbal particle/bare nominal, V, XP*’ string and an ‘identificational focus, V, XP*’ string are actually not completely identical. Compare: (35)

a.

!b. (36) a.

!b. (37) a.

!b.

Jańos [ki [nyitotta az ajto´t]] e´s [ki [vette a leveleket a John out opened the door and out took the letters the postala´da´bo´l]] letter-box-from ‘John opened the door and took the letters out of the letter-box.’ *Jańos [ki [nyitotta az ajto´t] e´s [vette a leveleket a postala´da´bo´l]] Jańos [kezet [fogott Zsuzsa´val]] e´s [kezet [ra´zott Pe´terrel]] John hand-ACC clasped Susan-with and hand-ACC shook Peter-with ‘John clasped hands with Susan and shook hands with Peter.’ *Jańos [kezet [fogott Zsuzsa´val] e´s [ra´zott Pe´terrel]] levelet az anyja]] e´s [PE´TERNEK [ku¨ldo¨tt [PE´TERNEK [ı´rt Peter-DAT wrote letter the mother-his and Peter-DAT sent csomagot az apja]] parcel-ACC the father-his ‘It was to Peter that his mother wrote a letter and it was to Peter that his father sent a parcel.’ [PE´TERNEK [ı´rt levelet az anyja] e´s [ku¨ldo¨tt csomagot az apja]]

That is, the V-initial string is a possible target of coordination only if it follows a focus. This fact is derived by assuming V-to-Pred movement in the presence of a verbal particle or bare nominal in Spec,PredP, but no V-movement in the presence of a focus: (38)

ajto´t]]]] a. [TopP Jańos [PredP ki [Pred’ nyitottai [VP ti az John out opened the door b. [TopP Jańos [PredP kezet [Pred’ fogotti [VP ti Zsuzsa´val]]]] Susan-with John hand-ACC clasped c. [PredP PE´TERNEK [VP ´ırt levelet az anyja]]] wrote letter his mother Peter-DAT

The facts of ellipsis also indicate an asymmetry between the focus and the verbal particle/bare nominal; the explanation of this asymmetry, however, does not fall out in such a straightforward fashion. a. Ki -nyitotta Jańos az ajto´t? out opened John the door-ACC ‘Did John open the door?’ b. Ki -nyitotta. c. Ki. ‘He did.’ ‘He did.’ (40) a. Kezet fogott Jańos Zsuzsa´val? hand-ACC clasped John Susan-with ‘Did John clasp hands with Susan?’ (39)

218


b. Kezet fogott. c. ?Kezet. ‘He did.’ ‘He did.’ az anyja levelet? (41) a. PE´TERNEK ´ırt Peter-DAT wrote his mother letter ‘Was it to Peter that his mother wrote a letter?’ c. PE´TERNEK. b. *PE´TERNEK ´ırt. ‘It was.’ ‘It was.’ If ellipsis, like syntactic operations, can only target maximal projections, the grammaticality of (39b), (40b) and (41c), and the ungrammaticality of (41b) fall out, but the grammaticality of (39c) and (40c) remains unexplained. (39c) and (40c) can only be accounted for if ellipsis is operative in PF, and if PF-operations can also target intermediate constituents. The proposed syntactic structure is claimed to be exempt from the main syntactic problem raised by Brody’s FocP theory. Unlike structure (1), the proposed framework does not predict an immediately postverbal position for the verbal particle. In the Hungarian sentence, the order of postverbal constituents is free. Hence, if the verbal particle in a focus construction is not assumed to occupy the specifier of a functional projection, it is predicted – correctly – to appear in a random order postverbally. Thus in the proposed framework, none of the following word order variants is unexpected: (42)

E´va´val o¨ssze. a. JAŃOS veszett tegnap John fell yesterday Eve-with out ‘It was John who fell out with Eve yesterday.’ b. JAŃOS veszett tegnap o¨ssze E´va´val. c. JAŃOS veszett o¨ssze tegnap E´va´val.

It is assumed that NegP can be merged into the sentence either above the VP layer or above the PredP layer of the predicate phrase. No V-movement is triggered, so the negative particle will be adjacent to the verb only in the former case, given that the Spec,VP position intervening betwen them is empty (see E´. Kiss 2002: 27): (43)

a. Jańos [NegP nem [VP hı´vta meg E´va´t vacsora´ra]] John not invited PRT Eve dinner-to ‘John did not invite Eve to dinner.’ b. Jańos [NegP nem [PredP E´VA´T hı´vta meg vacsora´ra]] ‘It was not Eve who John invited to dinner.’ c. Jańos [NegP nem [PredP E´VA´T [NegP nem [VP hı´vta meg vacsora´ra]]]] ‘It was not Eve who John did not invite to dinner.’

The V-to-Pred movement assumed in the case of particle verbs and bare nominal plus V complexes would be blocked by an intervening negative particle; therefore such constructions are correctly ruled out. (44a) is ungrammatical because the V-toPred movement has not been performed, and (44b) is ungrammatical because the V


219

has been moved non-locally, across a non-empty Neg head. The sentence can only be saved if no PredP is generated, as in (43a). (44)

a. *Jańos [PredP meg [NegP nem [VP hı´vta E´va´t]]] PRT not invited Eve John b. *Jańos [PredP meg-hı´vta [NegP nem [VP t E´va´t]]]

At the same time, this theory cannot explain why (45) is ungrammatical – as opposed to (43b), in which negation precedes a focus in Spec,PredP: (45)

vacsora´ra]] ?*Jańos [NegP nem [PredP meg- hı´vta E´va´t PRT invited Eve-ACC dinner-for John not ‘It was not Eve who John invited to dinner.’

The other fact undermining the complementary distribution claim of E´. Kiss (2003) is Csirmaz’s analysis of Hungarian viewpoint aspect in chapter 3 of this book, according to which AspP immediately dominates the PredP projection harboring the verbal particle, but is internal to the focus projection. In fact, it is internal to the open sentence representing the subject of the specificational predicate called focus. Observe the examples in (46): (46)

a. Jańos ke´t o´ra alatt/*ke´t o´raíg ki-takarı´tott. John two hour in two hour.for PRT cleaned ‘John cleaned [the apartment] in two hours/*for two hours.’ b. JAŃOS takarı´tott ki ke´t o´ra alatt/*ke´t o´raíg. ‘It was John who cleaned [the apartment] in two hours/*for two hours.’

According to the evidence of the time adverbials, the predicate of (46a) is perfective. In (46b), on the other hand, the main predicate is an aspectually unspecified (or perhaps imperfective) identificational focus. Still, the perfectivity of the lower PredP is preserved (the time adverbial indicative of imperfective aspect remains ungrammatical). The fact that this time adverbial representing the reference time, associated with AspP, must be placed in the post-focus section of the sentence, as seen in (46b), is also evidence of AspP being located below the focus projection. (47) is only grammatical with a rising, so-called contrastive topic intonation, associated with narrow-scope constituents c-commanded by the V at LF. (47)

%Ke´t o´ra alatt JAŃOS takarı´tott ki. 7.

THE PROPOSAL

The problems of the focus theory of E´. Kiss (2003) concern claim (ii), i.e., the complementary distribution of the focus and the verbal particle/bare nominal. The validity of claim (ii), however, does not affect the validity of the other two claims, namely, claim (i), concerning the predicate function of identificational focus, and assumption (iii), adopted from Szendro´í (2003), concerning the information focus


220

status of the different types of preverbal elements. Therefore, I propose to change claim (ii) while preserving assumptions (i) and (iii). I suggest to place the focus in a predicative projection above AspP, whose head mediates a specificational predication relation between its specifier and its complement. For expository purposes, we might call this higher PredP projection FocP, following Brody (1990, 1995). The FocP elicits V-movement. Horvath (2000, 2005) and E´. Kiss (2003) argued that the focus and the V adjacent to it cannot occupy the specifier and the head of one and the same projection. Olsvay (2000) also presented arguments to the same effect. For example, the verb-initial string following a focus can be subjected to coordination, which is only licensed for maximal projections. In view of this, the landing site of V cannot be the Foc head; it must be a functional head between AspP and FocP – presumably the head of a CP-type projection. In the theory of Olsvay (2000), this projection is called a Non-Neutral Phrase (NNP). The NNP is also projected in the presence of negation; it serves to turn an extended verb phrase into the argument of an operator. This is the structure assigned to a focus construction:

(48)

FocP Spec

Foc’ Foc

NNP NN

AspP Spec

Asp’

Asp

PredP Spec

Pred’ Pred

VP V

…..

The main reason why I earlier discarded the V-movement hypothesis was that it wrongly predicted a fixed postverbal word order. In fact, the postverbal constituents of the Hungarian sentence are unordered; more precisely, if they represent


221

constituents of different weight, they are ordered according to Behaghel’s (1932: 6) Gesetz der wachsenden Glieder (law of the growing constituents), so that the lightest constituent follows the verb immediately, and the heaviest one stands at the right edge. Compare: (49)

a. JAŃOS sza´molta meg gyorsan a leadott szavazatokat. votes John counted PRT quickly the cast ‘It was John who quickly counted the votes cast.’ b. JAŃOS sza´molta jo´l o¨ssze a leadott szavazatokat. John counted well PRT the cast votes ‘It was John who well counted the votes cast.’ c. JAŃOS sza´molta a peńzt o¨ssze nagy ero´´feszı´te´ssel. great effort-with John counted the money PRT ‘It was John who counted the money with great effort.’

The postverbal orders of these sentences represent the most unmarked permutations possible. The most unmarked string is a ‘Particle, Manner Adverbial, Object DP’ string in (49a), a ‘Manner Adverbial, Particle, Object DP’ string in (49b), and an ‘Object DP, Particle, Manner Adverbial’ string in (49c). What is invariant in these strings is the order ‘lightest, medium weight, heaviest’ – in observance of Behaghel’s law. Chomsky’s theory of Derivation by Phase (2001b) allows us to assume focus movement into Spec,FocP and V-movement into the Non-Neutral head, and, at the same time, to account for the freedom of word order in the postverbal section of the sentence. Both theoretical considerations (see Chomsky 2001), and empirical (e.g. prosodic) considerations internal to Hungarian suggest that the maximal extended verbal projection of the Hungarian sentence (whether it be an AspP or a FocP) represents a phase. In phase theory, a crucial role is attributed to the head of the phase, which, together with its left periphery, remains active for further syntactic operations. Let us assume that the head of a phase must be overt; thus it is always the highest overt head in the phase. Then in a phase represented by a FocP, the head is the V sitting in the head position of Non-Neutral P. When no FocP is projected, the extended v*P phase of the Hungarian sentence is represented by an AspP. Recall from chapter 6 that AspP has an EPP feature, as a consequence of which Spec,AspP is filled with the verbal particle/bare nominal in Spec,PredP, with Asp remaining phonologically empty. The highest overt head, representing the head of the phase, is the V sitting in Pred in this case. The post-head domain of the phase is argued by Chomsky (2001) to be closed for further syntactic computation; it is handed over to Spell-Out; it enters PF, the phonological component. Apparently, in Hungarian (and possibly in other partially free word order languages, as well) the post-head domain of the extended v*P phase is rearranged in accordance with Behaghel’s law when passed on to the phonological component.


222

Structure (48) also provides an adequate framework for the analysis of negation. The negative particle can merge with NNP or with FocP, or with both. For example: (50)

a. Jańos [NegP nem [NNP hı´vtai [AspPP megj [PredP tj ti [VP ti E´va´t PRT Eve John not invited vacsora´ra]]]]] dinner-to ‘John did not invite Eve to dinner.’ b. Jańos [NegP nem [FocP E´VA´T [NNP hı´vtai [AspPP megj [PredP tj ti [VP ti vacsora´ra]]]]]] ‘It was not Eve who John invited to dinner.’ c. Jańos [NegP nem [FocP E´VA´T [NegP nem [NNP hı´vtai [AspPP megj [PredP tj ti [VP ti vacsora´ra]]]]]]] ‘It was not Eve who John did not invite to dinner.’

Not only the word order of negative sentences is derived correctly; the lack of Vmovement to the higher Neg head represents no problem, either, any more. To preserve the generalization that a raised V is the highest overt head in the extended v*P projection, and thereby, it represents the head of the phase, we have to assume that the negative particle is a specifier, not a head.

8.

S U M M A RY; T H E PA R A L L E L P R O P E RT I E S O F T H E V E R B A L PA RT I C L E A N D T H E F O C U S

This chapter has aimed to disentangle and explain the similarities and the differences in the syntactic behavior of the focus and the verbal particle/bare nominal. It has been claimed, on the one hand, that foci and verbal particles/bare nominals display a functional similarity. Not only the verbal particle and the bare nominal are predicates predicated of the (explicit or implicit) theme argument of the verb, but the identificational focus is also a predicate expressing specification, predicated of the open proposition constituted by the post-focus section of the sentence. The exhaustivity associated with the Hungaran structural focus has been shown to be an implication of specificational predication. It has been argued, on the other hand, that part of the similarities in the syntactic behavior of the focus, the verbal particle, and the bare nominal derive from the fact that they play similar roles in phase-structure. FocP, with an identificational focus in Spec,FocP, and PredP, with a verbal particle/bare nominal in Spec,PredP, represent alternative realizations of the extended v*P phase of the Hungarian sentence. More precisely, the properties of PredP are similar to those of FocP only when PredP has the V sitting in its head. When a FocP or a NegP is also projected, and the V moves on to the Non-Neutral head, the PredP with no overt head does not function as a phase. Apparently, the extended v*P phase of the Hungarian sentence must be headed by the overt V; this is what motivates V-movement. The postverbal domains


223

of the FocP and PredP phases share the property of being reordered according to Behaghel’s Law of the Growing Constituents when handed over to Spell-Out. NOTES 1

2

A theory of negation along these lines was proposed by Olsvay (2000). For arguments against the derivation in (8), and for an alternative proposal, see Surańyi (2002a, 2002b). (11b) is not an instance of constituent negation. The pre-focus negative quantifier triggers the same type of negative concord as the preverbal negative particle. Compare: (i)

(ii)

3 4

[hı´vtam meg]]]] [DistP Senkit [NegP nem [FocP EŃ I invited PRT nobody-ACC not ‘Nobody was invited BY ME.’ [TopP Eń [DistP Senkit [NegP nem [hı´vtam meg]]]] I nobody-ACC not invited PRT ‘I did not invite anybody.’

For an analysis of numerally quantified foci, see also E´. Kiss (2001) and Surańyi (2000a), (to appear). In fact, there are counterexamples to this claim. Horvath herself discusses the following counterexample in a different context: csak a´t t. (i) A KERI`TE´SEN ma´sztunk the fence-on climbed-we only across ‘We only climbed across the FENCE.’

5

6

It is unclear how the claim that csak must c-command the focus can be extended to this case. If we assume, contrary to Szendro ´í (2003), that Q-Raising is adjunction to the maximal extended projection of the V, then we can formulate a stress rule which assigns primary stress to the leftmost element of every layer of the maximal extended projection of the V. That is, in the case of a Quantifier, Focus, V . . .. string, both the quantifier and the focus receive primary stresses – one of which can undergo destressing. Sentences of the following type would require a special treatment in every type of focus theory: vagy A HA´ZIGAZDA aludt A fo¨ldo¨n? (i) A VENDE´G ment el sza´lloda´ba, the guest went off hotel-to or the host slept the floor-on ‘Was it the guest who went to a hotel, or was it the host who slept on the floor?’ In the framework proposed in this paper, sentences of this type are exceptional in that their subject of predication is not associated with an existential presupposition, e.g. in the case of the first clause of (i) there may be noone who went to a hotel. I assume that in such cases it is not the constituent in Spec,PredP but the whole PredP that functions as a specificational predicate. The subject of predication, associated with an existential presupposition, is implicit; it corresponds roughly to ‘What happened’, i.e., the meaning of (i) could be paraphrased as follows: (ii)

Was it that the guest went to a hotel, or that the host slept on the floor, what happened?

ANIKO CSIRMAZ

A S P E C T, N E G AT I O N A N D Q U A N T I F I E R S

This chapter considers the effect of negation on aspect. It is shown that negation licenses the subinterval property for all types of event descriptions, independently of their lexical or derived aspectual properties. Negation is not an aspectual operator; the subinterval property does not arise from the event description being converted into an atelic event description or state. Rather, I argue that negation results in a predicate which takes the original event description and a interval, the reference time as arguments. The resulting complex predicate has the subinterval property, which can be modified by a for-adverbial or a Hungarian equivalent -a´t or -ig adverb.1 A for-adverbial can thus modify not only the event description and the event time, but also a higher predicate and time interval. In addition, I show that apart from negation, other downward entailing elements also license the subinterval property. The behavior of structurally case-marked equivalents of for-adverbs is also addressed. This chapter is organized as follows. Section 1 outlines the basic assumptions about the temporal intervals in sentences and adverbial modification. Section 2 discusses the aspectual properties of negated event descriptions and derives their adverbial modification properties. In section 3, I show that decreasing quantifiers and only phrases have the same effect on adverbial modification as negation. Finally, section 4 offers an account for the different behavior of structurally case-marked and postpositional/inherently case-marked temporal adverbs in Hungarian. 1.

T I M E I N T E RVA L S A N D F O R - M O D I F I C AT I O N

Before laying out the core proposals of the chapter, let us outline the temporal structure of finite clauses, including the specific time intervals involved and their position in the clausal structure. After introducing the relevant time intervals, I turn to adverbial modification by the Hungarian equivalents of for-adverbs, -ig and a´t adverbs, which are illustrated below. (1)


ke´t e´s two and ran for two ke´t e´s two and ran for two

fe´l half hours fe´l half hours

o´raíg futott2 hour-until ran and a half’ o´rań a´t futott hour-on across ran and a half’ 225

´ Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 225–253. Katalin E. 2008 Springer.

226

ANIKO CSIRMAZ

1.1.

Time intervals

I follow the insight of Reichenbach 1947 and Klein 1994, and the specific implementation of Iatridou et al. 2001 and von Stechow 2002 in adopting a system of three time intervals, which can be characterized as follows. The time intervals are the event time, reference time and utterance time. The event time is the runtime of the event. The reference time is a time interval that can stand in a variety of relations to the event time, according to the viewpoint aspect (perfective or imperfective) specification of the event description. Finally, the reference time is ordered with respect to the utterance time; this ordering is established by tense. With past tense, the reference time precedes the utterance time; with present tense, the reference time includes the utterance time. The motivations for assuming an event time and a time of utterance are clear; these times are determined by the event and timing of the utterance. Reichenbach 1947, who introduces the notion of reference time, does not offer a definition. Reference time is generally viewed, however, as the time under discussion, about which a claim is made (e.g. Klein 1992). The past tense example (2), for instance, does not entail that Jańos is not asleep any longer. Rather, the past tense locates the reference time in the past, and asserts that at that past reference time, Jańos was asleep. Since past tense does not directly order event time, Jańos could still be asleep at the time of utterance. (2)

She walked into the room. Jańos was asleep

The interpretive properties of the reference time have been described in various ways. Reichenbach 1947 and Hornstein 1990, among others, argue that the reference time can be observed in complex perfect tenses, where it is distinct from both the event time and the utterance time. According to Demirdache and Uribe-Etxebarria 2000, 2004, reference time (their assertion time) can be used to encode not only the perfect, but also imperfective aspect of event descriptions. They argue that both aspect and tense specifications order two times. Tense establishes an ordering between the time of utterance and reference time, while aspect orders reference time and event time. According to Demirdache and Uribe-Etxebarria 2000, 2004, the account is advantageous because it requires a minimal number of stipulations, since both aspect and tense can establish the same orderings among the appropriate time intervals: both can establish either an after relation (past tense or perfect aspect) or a within relation (present tense or imperfective aspect)3. Demirdache and Uribe-Etxebarria 2000, 2004 treat perfect and imperfective aspect as different orderings established by the same aspectual head. This treatment predicts that only these two aspects can be specified. As argued by Klein 1994, Iatridou et al. 2001, von Fintel and Iatridou 2002, the reference time can also be ordered with respect to the event time to encode perfective aspect, which they describe as the reference time containing the event time. This option is not available for Demirdache and Uribe-Etxebarria 2000, 2004, since they restrict aspect to perfect


227

and progressive, as noted above. The presence of perfective aspect is required, for instance, by languages which encode perfective – rather than imperfective – aspect overtly. Among others, Smith 1991 argues that perfective aspect can be marked in Chinese, and a similar conclusion is reached, for example, by Gimba 2000 for Bole, a Chadic language. Overt perfective aspect requires an aspectual system where in addition to imperfective and perfect, perfective can also be represented. Klein 1994, Iatridou et al. 2001, von Fintel and Iatridou 2002 propose and adopt such a system, where perfective and imperfective are two orderings of the reference time and the event time. I follow these authors in this treatment of perfective and imperfective as well as in assuming that perfect is encoded by a different functional head.4 I assume that tense is encoded in the T head and perfective or imperfective aspect in the Asp functional head in the sentence structure. Following Iatridou et al. 2001 and von Fintel and Iatridou 2002, I assume that event time is associated with vP, the reference time with AspP and the utterance time with TP. Specifically, I treat these time intervals as arguments of V, Asp and T, respectively, which can be realized syntactically as complements or specifiers.5 (3)

TP utterance time T

AspP

reference time Asp

vP predicate with event time argument

I assume that Asp is always present in finite clauses, and it can either appear as a perfective or imperfective Asp head. Both Asp heads take two arguments: a predicate of times complement, which contains the event time, and a time interval syntactically realized as a specifier. The latter is the reference time, which either contains the event time (with perfective aspect) or is contained in the event time (if the aspect is imperfective).6 Finally, the utterance time is an argument of T. Tense, located in the T head, orders the utterance time and reference time as described above. As noted above, perfective and imperfective heads take two arguments: a time interval i of type s and a predicate which takes a time interval argument (Iatridou et al. 2001, von Fintel and Iatridou 2002, with similar definitions proposed by Klein 1994 and von Stechow 2002). The event time is existentially quantified, as shown below.7 (4)

a. [[perfective]] ¼ ‚P<st>. ‚i . $i¢ . [i¢ i & P(i¢)] b. [[imperfective]] ¼ ‚P<st>. ‚i . $i¢ . [i i¢ & P(i¢)]

228

ANIKO CSIRMAZ

I assume that both T and Asp must be present in a finite clause, which is thus either perfective or imperfective. Finite clauses then contain three time intervals: event time, reference time and the utterance time.8 The next section considers how -ig and a´t adverbs interact with these time intervals. 1.2.

Modification of time intervals

I assume that -ig and a´t adverbs take three arguments: a measure phrase, a time interval and a predicate with a time interval argument: (5)

a. Jańos ke´t e´s fe´l o´raíg / ke´t e´s fe´l o´rań two and half hour-until / two and half hour-on J-NOM a´t aludt across slept ‘Jańos slept for two hours and a half’ b. for (two hours) (Jańos sleep) (t) c. [[ke´t e´s fe´l o´raíg]] ¼ [[ke´t e´s fe´l o´rań a´t]] ¼ ¼ ‚Pst . ‚is . P(i) & |i| ¼ 2 hours & 8i¢ i . P(i¢) & $i¢¢ i . P(i’’)9

The adverb can move to the specifier of some functional projection housing adverbs, as discussed in Csirmaz (chapter 8). A durative adverb, in general, can move either to Spec,Aspdurative of Cinque 1999, or to the specifier of a projection above PredP, but below Cinque’s AspPdurative and the AspP of perfective and imperfective aspect. The distribution of the Hungarian equivalents of for-adverbs, -ig and a´t adverbs is syntactically unconstrained. The adverbs, however, impose semantic restrictions on the predicate argument. The first restriction dictates that the predicate argument P have the subinterval property, defined based on Dowty 1979. (6)

A predicate of times P has the subinterval property iff whenever P(i) for an interval i, then for all i¢ i, P(i¢)

According to the definition above, a predicate of times has the subinterval property if whenever it holds for an interval, it also holds for all subintervals of that interval. The definition applies not only to event descriptions, but also to other predicates of times. Hungarian -ig and a´t adverbs, as well as their English equivalents, also require a non-trivial satisfaction of the subinterval property: in addition to the time interval argument, the predicate must also hold for a proper subinterval of the interval. This requirement forces the time interval argument to be durative, to have a non-atomic duration.10 On the one hand, it ensures that no instantaneous event description can be modified by -ig and a´t adverbs, since these predicates are only true for an atomic time interval. On the other hand, no time interval argument of these adverbs can be atomic. Since atomic time intervals lack a proper subinterval, the proper subinterval requirement on the predicate argument cannot be satisfied.

229


Given the restriction that -ig and a´t adverbs place on the arguments of predicate of times, it is predicted that these adverbs can equally measure the duration of event times, reference times and the utterance time. This is, however, not the case: the adverbs are restricted to modifying the event time or the reference time. The ban on modifying the utterance time has two sources. First, -ig and a´t adverbs only modify durative intervals, as noted above. Let us adopt the idea of Bennett and Partee 1978 that the utterance time is an instantaneous moment. If the utterance time is instantaneous, then it cannot serve as an argument of the adverbs in question. In addition to the specific instance of modification by an -ig or a´t adverb, the modification of the utterance time is excluded in general. Hornstein 1990 argues that the utterance time cannot be modified due to its deictic nature. In general, deictic elements cannot be modified, as the following examples (partly adapted from Hornstein 1990) show. (7)

Jańos ke´t o´rakor e´rkezett meg J-NOM two hour-at arrived perf ‘Jańos arrived at two o’clock’ [*At two o’clock, when the sentence is uttered, Jańos arrived (at some previous time)] b. *Itt, ami Ko ´´szeg mellett van, mindig here which K beside is always esik az eso´´ fall the rain-NOM *‘Here, which is beside Ko´´szeg, it’s always raining’ c. *Eń, akit Jańos tisztel, mindig e´hes vagyok I who-ACC J-NOM respects always hungry be-1sg *‘I, who Jańos respects, am always hungry’ a.

Since the modification of utterance time is ruled out, -ig and a´t adverbs are expected to modify the event time or the reference time. Illustrating event time modification is straightforward. The event time of any atelic event description (state or activity, in the Vendlerian terms) can be modified, as illustrated below.11 (8)

a. Jańos ke´t e´s fe´l he´tig J-NOM two and half week-until he´ten a´t beteg volt week-on across sick was ‘Jańos was sick for two weeks and a b. Jańos ke´t e´s fe´l o´raíg J-NOM two and half hour-until o´rań a´t futott hour-on across ran ‘Jańos ran for two hours and a half’’

/ /

ke´t two

half’ / ke´t / two

e´s and

e´s and

fe´l half

fe´l half

230

ANIKO CSIRMAZ

If the adverbs modify the reference time, then the event description may lack the subinterval property, since the two predicates of times are independent. An instance where the reference time can be modified by an -ig or a´t adverb is provided by imperfective event descriptions. Recall from (4b), repeated below, that the reference time of the imperfective is included in the event time. This entails that every subinterval of the reference time is also included in the event time; thus the predicate of times which applies to the reference time has the subinterval property. (9)

[[imperfective]] ¼ ‚P<st>. ‚t . $t¢ . [t t¢ & P(t¢)]

It is predicted that the imperfective predicate of times and the reference time can serve as arguments of an -ig or a´t adverb. This prediction is borne out, as shown in the following examples. The event description is telic, as the incompatibility of an -ig or a´t adverb and the perfective event description in (10b) shows. Whenever the event description is imperfective, as the postverbal position of the particle le (‘down’) indicates, modification by these adverbs is grammatical. (10)

a. Jańos tı´z percig / tı´z percen a´t J-NOM ten minute-until / ten minute-on across ment le a le´pcso´ń went down the stair-on ‘Jańos was going down the stairs for ten minutes’ b. Jańos tı´z perc alatt /*tı´z percig / J-NOM ten minute under / ten minute-until / /*tı´z percen a´t le ment a le´pcso´ń ten minute-on across down went the stair-on ‘Jańos went down the stairs in ten minutes’

To summarize the discussion so far: I suggested that a predicate of times can serve as the argument of an -ig or a´t adverb if it satisfies the subinterval property. In addition, the subinterval property must be satisfied non-trivially; the predicate must also be true of the time interval argument as well as a proper subinterval of the interval. Both requirements are encoded in the formulation of -ig or a´t adverbs in (5c). In the next section I show that the proposed account of these adverbs carries over to adverbial modification of negated event descriptions. Negation licenses the subinterval property for the predicate applying to the reference time, and thus allows modification by -ig or a´t adverbs. Negation is not a stativizer; the aspectual properties of the event description are not affected. 2.

A D V E R B M O D I F I C AT I O N

This section argues that in addition to imperfective aspect, negation also allows the reference time to be modified by -ig and a´t adverbs: negated perfect event descriptions yield a predicate of times that has the subinterval property, as required by the time adverbs in question.


2.1.

231

Negated event descriptions and adverbial modification

The possibility of -ig or a´t adverb modification with negated event descriptions is illustrated in the following examples. (11)

a. Jańos ke´t e´s fe´l o´raíg / ke´t e´s fe´l J-NOM two and half hour-until / two and half o´rań a´t nem e´rkezett meg hour-on across not arrived perf ‘For two hours and a half, Jańos did not arrive’ b. Jańos ke´t e´s fe´l o´raíg / ke´t e´s fe´l J-NOM two and half hour-until / two and half o´rań a´t nem ment a´t az utcań hour-on across not crossed across the street-on ‘For two hours and a half, Jańos did not cross the street’

In (11), the event descriptions are telic. (11a) contains an instantaneous event description, and (11b), a durative telic one. The telicity of the event description can be shown by adverbial modification of the event time, which can appear with the temporal modifier alatt, the equivalent of an English in-adverb. (11)

b¢. Jańos ke´t e´s fe´l o´raíg nem ment a´t J-NOM two and half hour-until not went across az utcań o¨t perc alatt the street-on five minute under ‘For two hours and a half, Jańos didn’t cross the street in five minutes’ (¼ for an interval which is two hours and a half in duration, there was no event of Jańos crossing the street in five minutes during that interval)

In absence of negation, modification of these event descriptions by -ig or a´t adverbs is ungrammatical, showing that it is indeed negation which licenses adverbial modification. (12)

a. *Jańos ke´t e´s fe´l o´raíg / ke´t e´s fe´l J-NOM two and half hour-until / two and half o´rań a´t meg e´rkezett hour-on across perf arrived *‘For two hours and a half, Jańos arrived’ b. *Jańos [ke´t e´s fe´l o´raíg] / [ke´t e´s fe´l J-NOM two and half hour-until / two and half o´rań a´t] a´t ment az utcań hour-on across across went the street-on *‘For two hours and a half, Jańos crossed the street’

232

ANIKO CSIRMAZ

The -ig or a´t modification of negated event descriptions is predicted by the previous assumptions. First, if the negated event description is imperfective, then the reference time modification by these adverbs is straightforward independently of negation, as shown above.12 The predicate of times which applies to the reference time can still have the subinterval property if the event description is perfective. For (11a), if i is a 2.5 hourlong interval within which there is no event of Jańos arriving, then it is also true for all subintervals i’ of i that there is no event of Jańos arriving within i’. Since the predicate has the subinterval property, the reference time can be modified by -ig and a´t adverbs. Thus if the adverb is merged above negation, then -ig or a´t adverbial modification is possible for all event descriptions, independently of the properties of the event description.13 2.2.

Adverb position and interpretation

Let us consider in detail how the adverbial modification of negated event descriptions is derived. As proposed above, -ig and a´t adverbs can always modify the reference time of negated event descriptions. Reference time is present in all finite event descriptions, since Asp is always merged in finite clauses and the reference time is merged as an argument of Asp. Concerning the structure of negation, E´. Kiss (chapter 2) argues that clausal negation is merged with vP. PredP and AspP, the projections which trigger the movement of particles and encode perfective or imperfective aspect, respectively, are merged on top of negation:

(13)

AspP reference time Asp

PredP

Pred

NegP nem

vP

predicate with event time argument


233

An adverb modifying the reference time must be interpreted above AspP and NegP. In this position, the predicate taking the reference time as an argument can serve as the predicate argument of the temporal modifier, and it also has the subinterval property, as required. To account for for-adverbial modification licensed by negation, Bennett and Partee 1972, Dowty 1979, Verkuyl 1993 argue that negation affects the situation aspect properties of event descriptions. According to these accounts, addressed in more detail in section 2.3, negation converts all event descriptions to states, allowing the temporal adverbs to modify the event time of the predicate. As proposed above, negation leaves the situation aspect properties unaffected, contrary to the previous theories. Negation licenses the subinterval property for the predicate applying to the reference time, but does not affect the stativity or telicity properties of the event description itself. The -ig and a´t adverbs licensed by negation must be interpreted fairly high in the structure, above AspP. The position in the sentence structure interacts with the adverb interpretation with respect to other adverbs in predictable ways. Scope relations are reflected in the surface structure in Hungarian. It is not surprising then that the position of the temporal adverbs correlates with their interpretation, specifically, with whether they modify the event time or the reference time. With neutral intonation, the adverbs modifying the reference time appear in a preverbal position, while those that measure the event time are postverbal.14 (14)

Jańos ke´t e´vig nem ´ırta meg year-until not wrote perf J-NOM two a ha´zi feladata´t fe´l o´ra alatt the home exercise-ACC half hour under ‘For two years, Jańos didn’t write his homework in half an hour’ b.? Jańos fe´l o´ra alatt nem ´ırta meg J-NOM half hour under not wrote perf a ha´zi feladata´t ke´t e´vig the home exercise-ACC two year-until c. Jańos ke´t e´vig nem ´ırta meg fe´l o´ra alatt J-NOM two year-until not wrote perf half hour under a ha´zi feladata´t the home exercise-ACC d.? Jańos fe´l o´ra alatt nem ´ırta meg hour under not wrote perf J-NOM half ke´t e´vig a ha´zi feladata´t two year-until the home exercise-ACC a.

The position of the adverbials is consistent with the account proposed, which assumes that event time is merged lower than reference time. The postverbal position of event time modifying adverbs follows if the verb is the leftmost constituent within

234

ANIKO CSIRMAZ

the vP (E´. Kiss 2002). In the examples above, the event description lacks the subinterval property and can be modified by an alatt adverb. The reference time, in contrast, has the subinterval property and so is amenable to modification by -ig or a´t adverbs. A marked, contrastive topic intonation, which is characterized by a brief fall followed by a rising contour (E_ . Kiss 2002) can yield non-surface scope relations. If a preverbal adverb has contrastive topic intonation, then it can be interpreted as measuring the event time. A postverbal adverb can also measure the reference time if it has nuclear stress (K. E´. Kiss p.c.). In this respect adverbs behave similarly to other scope-taking elements; the linear order corresponds to the scope relations, but marked intonation can yield inverse scope effects (E´. Kiss 2002). As expected, the position of the adverb has a disambiguating effect with atelic event descriptions, where it can modify either the event time or the reference time. If the intonation is neutral and involves neither contrastive topic intonation nor postverbal nuclear stress, then the -ig adverb in (15a) measures the reference time, and in (15b), the event time. (15)

a. Jańos ke´t e´s fe´l o´raíg nem aludt not slept J-NOM two and half hour-until ‘For two hours and a half, Jańos didn’t sleep’ b. Jańos nem aludt ke´t e´s fe´l o´raíg J-NOM not slept two and half hour-until ‘Jańos didn’t sleep for two hours and a half’

With neutral intonation then, the position of the adverb correlates with the time interval measured. Adverbs modifying the event time are postverbal and those measuring the reference time precede the verb.15 The correlation between adverb position and the time interval modified supports the proposed account. Negation licenses the subinterval property for the predicate applying to the higher reference time, but not the lower event time. 2.3.

Negation and stativity

This section discusses the purported stativizing effect of negation (Bennett and Partee 1972, Dowty 1979, Dowty 1986, Verkuyl 1993 and others) in more detail. The approach, which I dub stativity hypothesis, maintains that clausal negation stativizes all event descriptions. Under this hypothesis, negation yields a negative state, where the derived negative state is characterized by the non-occurrence of a specific type of event. I argued above, in line with Moltmann 1991 and Zucchi 1991, and contra Verkuyl 1993, de Swart and Molendijk 1999 and others, that negation is not an aspectual operator. It does not affect the aspectual properties of event descriptions, and so fails to convert events into states. It is not some derived state which allows -ig and a´t modification of event descriptions, but the subinterval property of a higher predicate, which applies to the reference time.


235

To provide support for the non-stativity of negated event descriptions, I consider how various diagnostics of stativity fare with negated event descriptions. I show that negation does not affect the properties of the event description. Negation yields, however, homogeneity at the level of reference time. 2.3.1 Arguments for non-stativity This section presents several environments where negated event descriptions pattern unlike states, providing a convincing argument against the stativity hypothesis. Present tense interpretation States and perfective event descriptions (activities, accomplishments and achievements) differ in the interpretation of the present tense form. With present tense morphology, only states and imperfectives can denote an event that is ongoing at the utterance time. Perfective present tense event descriptions have either a habitual or a future interpretation. In Hungarian, an ongoing interpretation is possible for states and those event descriptions which are ambiguous between perfective and imperfective aspect (16). With unambiguously perfective event descriptions, only a futurate or habitual interpretation is possible (17). a. Jańos beteg J-NOM sick ‘Jańos is sick’ b. Jańos alszik J-NOM sleeps ‘Jańos is sleeping’ c. Jańos olvas J-NOM reads ‘Jańos is reading’ (17) Jańos el olvassa a the J-NOM away reads ‘Jańos reads the book’

(16)

ko¨nyvet book-ACC

What is unexpected, given the hypothesis that negation is a stativizer, is that negation fails to yield an ongoing interpretation for all event descriptions. Negated perfective event descriptions are unlike states; they do not denote an ongoing event. Rather, similarly to their affirmative counterparts, they only allow a habitual or futurate interpretation. (18)

Jańos nem olvassa el reads away J-NOM not ‘Jańos does not read the book’

a the

ko¨nyvet book-ACC

236

ANIKO CSIRMAZ

Discourse structure States and perfective event descriptions are also distinguished by their role in narratives. While perfective states advance narrative time, states typically fail to do so (Dowty 1986, Kamp and Reyle 1993)16. The two events in (19a) are consecutive: Jańos’s leaving is interpreted as following (and possibly being a consequence of) Mari’s looking at him. In contrast, (19b) is interpreted as Jańos’s sleeping having begun (and possibly continuing after) the time when Mari looked at him. (19)

a. Mari Jańosra ne´zett. Jańos el indult M-NOM J-onto looked J-NOM away left ‘Mari looked at Jańos. Jańos left’ b. Mari Jańosra ne´zett. Jańos aludt J-onto looked J-NOM slept M-NOM ‘Mari looked at Jańos. Jańos was asleep’

A negated event description also advances narrative time, similarly to its affirmative counterpart. The second sentence in (20) is interpreted as stating that Jańos did not leave as a consequence of Mari’s looking at him; that is, the expected reaction of Jańos leaving did not happen. (20)

Mari Jańosra ne´zett. Jańos nem looked J-NOM not M-NOM J-onto ‘Mari looked at Jańos. Jańos did not leave’

indult el left away

The consecutive reading of the negated event description can be contrasted with the interpretation of a negated state. The overlapping interpretation of the negated state is also parallel to its affirmative counterpart. (21)

Mari Jańosra ne´zett. Jańos nem aludt looked J-NOM not slept M-NOM J-onto ‘Mari looked at Jańos. Jańos wasn’t asleep’

Negation thus fails to affect the interpretation of event descriptions in discourse structure. Agentive adverbs A test for identifying agentive event descriptions is modification by agentive adverbs such as szańde´kosan (‘deliberately’) or akarattal (‘voluntarily’). These adverbs require the presence of an agent argument, as the following contrast between agentive (22) and non-agentive event descriptions (23) shows. (22)

Jańos szańde´kosan o¨ssze to¨rte J-NOM deliberately together broke ‘Jańos broke the glass deliberately’

a the

poharat glass-ACC

237


(23)

a. #a poha´r szańde´kosan o¨ssze the glass-NOM deliberately together ‘the glass broke deliberately’ b. #Jańos szańde´kosan izzadt J-NOM deliberately sweated ‘Jańos sweated deliberately’

to¨rt broke

States lack agent arguments, and thus cannot be modified by agentive adverbs. (24)

#Jańos szańde´kosan aludt J-NOM deliberately slept ‘Jańos slept deliberately’

If negation yielded derived states, then negated event descriptions would not be amenable to modification by agentive adverbs. This is, however, not the case, as the following examples show. The agentive adverb modifies the negated event description, a scope relation which is mirrored by the linear order of the adverb and negation. (25)


szańde´kosan nem olvasta el deliberately not read away deliberately didn’t read the sign’ szańde´kosan nem javı´totta ki deliberately not corrected out deliberately didn’t correct the mistake’

a the

felira´st sign-ACC

a the

hiba´t mistake-ACC

Negation can also make modification by an agentive adverb felicitous, if the lexical predicate does not assign an agent theta role. In this case, adverbial modification is only possible if the adverb takes scope over negation. (26)

a. #Jańos szańde´kosan fel ismerte a hitelezo´´je´t J-NOM deliberately up recognized the creditor-POSS-ACC #‘Jańos deliberately recognized his creditor’ b. Jańos szańde´kosan nem ismerte fel up J-NOM deliberately not recognized a hitelezo´´je´t the creditor-POSS-ACC ‘Jańos deliberately did not recognize his creditor’

As the preceding examples show, negated event descriptions can have an agent argument, since they can be modified by agentive adverbs. Modification by these adverbs fails to constitute a conclusive argument against the stativity of negated event descriptions, since some states and other non-agentive

238

ANIKO CSIRMAZ

predicates can also be modified by agentive adverbs: (27)


szańde´kosan hallgatott deliberately was.silent deliberately kept silent’ szańde´kosan el esett deliberately away fell fell on purpose’

I assume that the agentive adverb can coerce states, where the lexically non-agentive subject is reinterpreted as the causer of the event. It may be argued then – in face of the preceding arguments – that negated event descriptions are derived states, which can be coerced similarly to some states. Building on the previous observations, it can be shown, however, that stativity is not necessary for modification by -ig or a´t adverbs. It is possible for the negated event description to be modified by a temporal and agentive adverb at the same time: (28)

a. Jańos pa´r percig szańde´kosan nem ismerte fel J-NOM few minute-until deliberately not recognized up a hitelezo ´´je´t the creditor-POSS-ACC ‘For a few minutes, Jańos deliberately didn’t recognize his creditor’ b. Jańos pa´r percig szańde´kosan nem J-NOM few minute-until deliberately not javı´totta ki a hiba´t corrected out the mistake-ACC ‘For a few minutes, Jańos deliberately didn’t correct the mistake’

As the grammaticality of both temporal and agentive adverbials shows, the modification of negated event descriptions by -ig and a´t is not tied to stativity. This conclusion provides further support to the claim that negation yields homogeneity, but not stativity of the event description. 2.3.2 A non-argument against stativity Imperatives are often cited as a stativity diagnostic (Lakoff 1970, Dowty 1979), but I argue that they fail to provide a convincing argument against the stativity hypothesis. The behavior of imperative negated event descriptions is compatible, however, with the proposed account, which also predicts the different behavior of states and negated event descriptions. Generally, non-stative event descriptions can appear as imperatives and statives cannot: (29)

a. #Szeresd a like-IMP.2SG the ‘Like broccoli!’

brokkolit! broccoli-ACC


b.

Edd meg a eat-IMP.2SG perf the ‘Eat the broccoli!’

239

brokkolit! broccoli-ACC

Similarly to agentive adverbs, imperatives diagnose agentivity rather than stativity, since the imperative requires an addressee who has the event under their control. If all negated event descriptions are states, then they are not expected to occur as imperatives. This is not the case, however, as the following negative imperatives show. (30)

a. Ne kiaba´lj! not shout-IMP.2SG ‘Don’t shout!’ b. Ne edd meg a brokkolit! not eat perf the broccoli-ACC ‘Don’t eat the broccoli!’

Negative imperatives express a command that requires the addressee not to do something. At first blush, this may show that negated event descriptions are not states. The possibility of negated imperatives does not, however, provide a convincing argument against the stativity hypothesis. Once again, the correlation between the presence of a lexically determined agent argument and the possibility of imperatives is not perfect. A number of non-agentive predicates can also appear as imperatives: (31)

a. Gurulj le a domboldalon! roll-IMP.2SG down the hill.side-on ‘Roll down the hill!’ b. E´rj oda ido´´ben! arrive-IMP.2SG there time-in ‘Arrive on time!’

I assume that a coercion process accounts for the reinterpretation subjects, just as with agentive adverb modification: the lexically non-agentive subject argument is reinterpreted as the causer of the event. The possibility of coercion can also account for the fact that some states, but crucially not all of them (as shown in (29a)), can appear as imperatives. (32)

a. Legye´l csendben! be-IMP.2SG quiet ‘Be quiet!’ b. Aludj! sleep-IMP.2SG ‘Sleep!’

240

ANIKO CSIRMAZ

Since non-agentive subjects can be coerced by imperatives, the acceptability of negated event descriptions as imperatives fails to provide an argument against the stativity hypothesis. The acceptability of imperative negated event descriptions is compatible, however, with the proposed account of negation. In fact, the present account predicts that these imperative forms are uniformly acceptable, and no similar claim is made for states. In contrast, imperative negated event descriptions pose a problem for the stativity account. Under the latter theory, the difference between negated event descriptions and states – where the former are felicitous as imperatives, while only a handful of states can appear as such – remains to be accounted for. The different behavior of states and negated event descriptions provides arguments against the stativity hypothesis and for the account proposed above. Negation is not an aspectual operator; it fails to affect the properties of the event description. Rather, it yields the subinterval property for a higher predicate of times, making it possible for -ig and a´t adverbs to modify the reference time. 3.

Q U A N T I F I E R S A N D T H E S U B I N T E RVA L P R O P E RT Y

As noted earlier, the necessary condition for modification by -ig and a´t adverbs is the subinterval property. Negation licenses the modification of all event descriptions, since the predicate which applies to the reference time has the subinterval property. Given that the subinterval property is required for adverbial modification, it is predicted that negation does not constitute the only environment that can license -ig and a´t modification of telic event descriptions. Downward entailing quantifiers and arguments modified by csak (‘only’) also license this adverb modification, since they yield the subinterval property for the reference time. 3.1.

Downward entailing quantifiers

With an instantaneous event description, the quantificational kevesebb, mint tı´z ember (‘fewer than ten people’) yields the subinterval property for the predicate applying to the reference time (33a). Modification by -ig and a´t adverbs are thus possible. Nondecreasing quantifier subjects, as in (33b) and (33c), fail to license the subinterval property. (33)

ke´t e´vig / ke´t e´ven a´t kevesebb mint two year-until / two year-on until fewer than tı´z ember ve´gzett gyo´gytorna´sz szakon ten people-NOM finished physiotherapy major-on ‘For two years, fewer than ten people majored in physiotherapy’ b. *ke´t e´vig / ke´t e´ven a´t (pontosan) tı´z two year-until / two year-on until exactly ten ember ve´gzett gyo´gytorna´sz szakon a.


241

people-NOM finished physiotherapy major-on ‘For two years, (exactly) ten people majored in physiotherapy’ c. *ke´t e´vig / ke´t e´ven a´t to¨bb mint tı´z two year-until / two year-on until more than ten ember ve´gzett gyo´gytorna´sz szakon people-NOM finished physiotherapy major-on ‘For two years, more than ten people majored in physiotherapy’ The subinterval property holds for (33a), since if fewer than ten people (including possibly none) graduated within a time interval i, then fewer than ten people graduated within any subinterval of i as well. Not so for (33b); if exactly ten people graduated within a time interval i, then it does not necessarily hold for all subintervals i¢ of i that exactly ten people graduated within i¢. In the previous examples, the -ig and a´t adverbs modify the reference time. As argued in connection with negation, the event description can lack the subinterval property even when the adverb modification is licensed in these examples. The lack of subinterval property for the event description is shown below by the modification by an alatt adverb. The -ig and alatt adverbs modify the reference time and the event time, respectively. (34)

Tı´z e´vig kevesebb, mint tı´z ember [ve´gzett ten year-until fewer than ten people-NOM finished gyo´gytorna´sz szakon ke´t e´v alatt] physiotheraphy major-on two year under ‘For ten years, fewer than ten people received a degree in physiotherapy in two years’ (there may have been any number of people who graduated in some other time interval)

In addition to subjects, other arguments, including objects and goals also license the subinterval property. As in the preceding examples, -ig and a´t adverbial modification is licensed only by decreasing quantifiers. (35)

a. Jańos ke´t e´s fe´l o´raíg / ke´t e´s fe´l o´rań J-NOM two and half hour-until / two and half hour-on a´t kevesebb, mint tı´z kagylo´t tala´lt across fewer than ten shell-ACC found ‘For two hours and a half, Jańos found fewer than ten shells’ (although he hoped to find more) b. Jańos ke´t he´tig kevesebb, mint than J-NOM two week-until fewer o¨t embernek mondta el a hı´rt

242

ANIKO CSIRMAZ

five people-DAT told away the news-ACC ‘For two weeks, Jańos told the news to fewer than five people’ (but afterwards he told it to everyone he met) c. Jańos pa´r percig kevesebb, mint ha´rom embert than three people-ACC J-NOM few minute-until fewer ismert fel recognized up ‘For a few minutes, Jańos recognized fewer than three people’ In the examples involving decreasing quantifiers, the subinterval property holds for the predicate which applies to the reference time. To yield the subinterval property, it is necessary for the adverb to be merged higher than the position where the quantifier is interpreted. If the decreasing quantifier is interpreted below the –ig or a´t adverb, then the subinterval property is satisfied for the predicate that serves as the argument of the adverb. In the following example, the perfective event description has the subinterval property. (36)

ke´t e´s fe´l o´raíg kevesebb, mint o¨t vende´g two and half hour-until fewer than five guest-NOM e´rkezett meg arrived perf ‘For two hours and a half, fewer than five guests arrived’

The reference time in (36) is a time interval within which there was an event of fewer than five guests arriving. Given this reference time, it is true for all subintervals of the reference time that fewer than five guests arrived within that subinterval. If the quantifier takes scope above the adverb, then the subinterval property is not licensed in these cases. As expected, the argument then behaves as its nonquantificational counterparts with respect to the subinterval property. It was noted above that with neutral intonation, the linear order of the quantifier and the adverb corresponds to their relative scope in Hungarian. As predicted, (37) is ungrammatical, unlike (34)–(36), since the -ig adverb is interpreted as modifying the event time.17 (37)

?? kevesebb, mint o¨t vende´g e´rkezett meg ke´t e´s fewer than five guest-NOM arrived perf two and fe´l o´raíg half hour-until ‘For two hours and a half, fewer than five guests arrived’

It is worth noting that the reading discussed above is different from the one described by Moltmann 1991. Moltmann argues that for-adverbial modification – which I assume to be parallel to –ig and a´t modification – is possible if an argument is


243

modified by a vague quantifier, and impossible with absolute quantifiers. Two sets of examples (her (43) and (44)) are reproduced below in (38) and (40). Comparable Hungarian examples are provided in (39). (38)

a. For several years John took a lot of pills/few pills b. #For several years John took those pills/all the pills (39) a. Jańos jo´pa´r e´vig sok tabletta´t / keve´s tabletta´t pill-ACC J-NOM several year-until many pill-ACC / few szedett took b. #Jańos jo´pa´r e´vig ezeket a tabletta´kat/ J-NOM several year-until these-ACC the pills-ACC/ mindegyik tabletta´t szedte each pill-ACC took (40)

a. For several years John had a lot of success / little success b. #For several years John had that success / all success

Moltmann notes that (38a) – and similarly, (39a) – is true if for every contextually relevant part i of an interval i¢ that is several years long there are a lot of / few pills relative to i that John took. Similarly, (40a) is true if for every relevant subinterval i of several years, John had a lot of / little success relative to i. Crucially, the relevant quantities of pills and success in Moltmann’s examples are determined relative to the subintervals of the larger event, which lasts several years. Let us consider for several years, John took few pills. In this case, even though the sum of pills taken during each subinterval is few, the overall sum of the pills can be rather large. For instance, if John takes two pills each day for several years (days being the relevant subintervals), then he ends up having taken a large number of pills over the course of these years. Moltmann’s examples involving vague quantifiers are unlike the quantificational arguments discussed above. In the latter examples, the quantifier arguments have a cumulative interpretation; the quantifier is interpreted with respect to the maximal interval. Furthermore, I showed that rather than vague quantifiers, it is decreasing quantifiers which license this reading. The relevant difference between the quantifier examples discussed in this paper and those in Moltmann 1991, I believe, is the length of the time interval. The examples discussed by Moltmann involve iterative or habitual interpretation. These interpretations are plausible, given the length of the for-interval relative to the event time. Moltmann 1991 does not account for the ungrammaticality of absolute quantifiers in these types of examples, but iterative or habitual interpretation may play a role in establishing this difference. In contrast, the examples of -ig and a´t modification with decreasing quantifiers do not involve iteration or habituality. Finding a total of fewer than ten shells during the two-hour interval in (41) is sufficient.

244 (41)

ANIKO CSIRMAZ

Jańos ke´t o´raíg kevesebb, mint tı´z kagylo´t than ten shell-ACC J-NOM two hour-until fewer ‘For two hours, Jańos found fewer than ten shells’

tala´lt found

The decreasing quantifier yields the subinterval property for the predicate of times, which can serve as the argument of the adverb. To conclude, a decreasing quantifier argument can license -ig and a´t adverb modification of the reference time, irrespectively of the properties of the event description. The precondition of adverb licensing is that the quantifier take scope below the adverb.

3.2.

Arguments modified by csak (‘only’)

Constituents modified by csak, the Hungarian equivalent of only, also license -ig and a´t modification of the reference time. Unlike downward entailing quantifiers, however, only phrases do not license subinterval property as described above. As an illustration of the semantic effect of only, consider the sentence only Jańos graduated in physiotherapy. Given a time interval i that is two years long and for which it holds that only Jańos graduated in physiotherapy within i, it is not the case that only Jańos graduated in physiotherapy is true for all subintervals of i. Specifically, for those subintervals i’ of i that do not contain an event of John graduating in physiotherapy, the assertion will not be true. This is illustrated in the following scenario. (42)

(43)

(44)

Fred graduated in physiotherapy on August 30, 2001. Jańos graduated in physiotherapy on August 31, 2002 Peter and Frank graduated in physiotherapy on August 30, 2003 For two years (between August 30, 2001 and August 30, 2003) only Jańos graduated in physiotherapy (i is the time interval between August 30, 2001 and August 30, 2003) For some subintervals i¢ of i, such as the time span between November 30, 2001 and December 31, 2001, it is not the true that only Jańos graduated in physiotherapy. In fact, no one graduated in physiotherapy during i¢

In order to account for the acceptability of -ig and a´t adverbial modification in examples with csak, I introduce the notion of Strawson entailment (von Fintel 1999). Strawson entailment can be extended to define the Strawson subinterval property, which can, in turn, be used to account for adverbial licensing by csak (‘only’). I argue that only-phrases yield a predicate of times that has the Strawson subinterval property. The -ig and a´t adverbs require the predicate argument to have the Strawson

245


subinterval property rather than the proper subinterval property. In this way, arguments modified by csak, like decreasing quantificational arguments, license adverbial modification. Von Fintel 1999 proposes an account of the downward entailing property of only, which carries over to Hungarian csak. Von Fintel assumes that downward entailment is only checked for those conclusions that have a semantic value that is defined. If the conclusion is not defined, then it is not relevant for Strawson downward entailment. In the example below, the assertion that only Fred ate fish, since the conclusion Fred ate flounder is defined, entails that only Fred ate flounder. Wherever the conclusion is not defined, it is irrelevant whether entailment holds or not. (45)

a. b. c. d.

flounder fish Only Fred ate fish (assertion) Fred ate flounder (defined) Only Fred ate flounder (entailment)

By the notion of Strawson entailment, only and Hungarian csak are Strawson downward entailing. On the assumption that (English) negative polarity items must be in a Strawson downward entailing environment, licensing by only, shown below, is expected. (46)

Only John ate any fish

Strawson entailment can not only be applied to negative polarity licensing, among others, but also to the notion of subinterval property, as suggested above. The Strawson subinterval property can be defined as given below. (47)

Strawson subinterval property A predicate P of times has the Strawson subinterval property iff whenever P(i) for an interval i, then for all i¢ i, such that P is defined at i¢, P(i¢)

In order to account for adverbial modification licensing by csak and only, it is necessary to adopt the notion of Strawson subinterval property. It must also be assumed that for-adverbs and their Hungarian counterparts require their predicate argument to have Strawson subinterval property. As the adoption of the Strawson subinterval property predicts, only phrases – and their Hungarian equivalent csak phrases – license modification of the reference time by -ig and a´t adverbs. In absence of only, these event descriptions cannot be modified by -ig and a´t adverbs. (48)

a.

Ke´t two

e´vig year-until

/ /

ke´t two

e´ven a´t year-on across

csak only

Jańos J-NOM

ve´gzett finished

246

ANIKO CSIRMAZ

gyo´gytorna´sz szakon physiotherapy major-on ‘For two years, only Janos graduated in physiotherapy’ b. Ke´t e´vig / ke´t e´ven a´t csak o¨t two year-until / two year-on across only five ember ve´gzett gyo´gytorna´sz szakon people-NOM finished physiotherapy major-on ‘For two years, only five people graduate in physiotherapy’ c. #Ke´t e´vig / ke´t e´ven a´t (pontosan) two year-until / two year-on across exactly o¨t ember ve´gzett gyo´gytorna´sz szakon five people-NOM finished physiotheraphy major-on ‘For two years, (exactly) five people graduated in physiotherapy’ In (48a,b) the predicate applying to the reference time has the Strawson subinterval property. For a reference time i, if it is true that only Jańos graduated in physiotherapy at i, then it will also be true for all subintervals of i where John graduated in physiotherapy is defined. In order for the account to yield the desired results, a further ingredient is necessary: an appropriate view of csak and only. It must be ensured that even for an instantaneous predicate, once an argument of the predicate is modified by csak or only, as in (49),18 the resulting predicate can hold of a durative interval. (49)

Jańos pa´r percig / pa´r percen a´t J-NOM several minute-until / several minute-on across csak az elno¨ko¨t ismerte fel only the president-ACC recognized up ‘For several minutes, Janos only recognized the president’

The event description Jańos recognize the president and its Hungarian equivalent are instantaneous. In contrast, Jańos recognized only the president is durative in English as well as in Hungarian; it is true at all intervals that include the (instantaneous) event time of recognizing the president. The denotation of the only-example, which carries over to Hungarian, is given below.19 (50)

Jańos recognized only the president ‚i¢ . only the president [‚x . $t i¢ . Jańos recognized x at i]

I have shown that negation, monotone decreasing quantifiers as well as csak and only license the Strawson subinterval property, and hence the modification of the reference time by for-adverbs and its Hungarian equivalents. In addition to negation, downward


247

entailing quantifiers and only phrases also license the (Strawson) subinterval property for the predicate which applies to the reference time, but do not affect the situation aspect properties of the predicate. While all of these cases are handled successfully by the proposed account, the stativity hypothesis, which assumes that negation affects situation aspect, could not be obviously extended to handle the data introduced above. In general, then, -ig and a´t adverbs can measure the duration of either the event or the reference time, if the predicate argument has the subinterval property. The next section shows that not all of the temporal adverbs which require the subinterval property can measure both the event time and the reference time, and offers an independent account of the restriction. 4.

A S Y N TA C T I C R E S T R I C T I O N O N T E M P O R A L A D V E R B S

This section presents arguments for the general freedom of modification by foradverbs and their crosslinguistic correlates. It is shown that durative adverbs with structural case marking are restricted to modifying the event time and cannot measure the duration of the reference time. I argue that reference time modification is banned independently of the semantic properties of the adverb. The modification of reference time is excluded by syntactic restrictions; that of the locality of adverbial modification on the one hand, and the locality of case checking on the other. 4.1.

Adverbs restricted to event time modification

It was argued earlier that for, -ig and a´t adverbs can modify either the event time or reference time. One of the arguments of these modifiers is a predicate of times, which is required by the adverb to have the (non-trivial) subinterval property. In other words, the only restriction on the predicate argument is semantic in nature. In addition to the predicate of times, the adverbs take a time interval argument. Depending on the position of the adverb, the time argument is either the event time or the reference time. Modification of each of these time intervals is illustrated below. (51)

Event time modification Jańos fe´l o´raíg / fe´l o´rań a´t futott J-NOM half hour-until / half hour-on across ran ‘Jańos ran for half an hour’ (52) Reference time modification Jańos tı´z percig nem ismerte fel J-NOM ten minute-until not recognized up az elno¨ko¨t the president-ACC ‘For ten minutes, Jańos didn’t recognize the president’

248

ANIKO CSIRMAZ

Other durative adverbs which can modify the event time of an atelic event description, have a more restricted distribution: they cannot modify the reference time even if the predicate argument has the subinterval property. The durative adverbs subject to this restriction are those with structural case marking. Temporal adverbs with structural structural case marking can modify only the event time in a number of languages, including not only Hungarian, but also Finnish, Greek, Korean and Russian. Event time modification is illustrated below for Hungarian, where the accusative adverb can modify a state or activity, but cannot modify the event time of a telic event description. Considering event time modification only, the accusative time adverb patterns with -ig and a´t adverbs. (53)

Jańos ke´t o´ra´t aludt J-NOM two hour-ACC slept ‘John slept for two hours’ b. Jańos ke´t o´ra´t futott J-NOM two hour-ACC ran ‘John ran for two hours’ c. #Jańos ke´t o´ra´t meg J-NOM two hour-ACC perf ‘John arrived for two hours’ a.

e´rkezett arrived

As noted, modification of the reference time is not possible, even if the predicate applying to the reference time has the subinterval property. In the following example, event time modification is impossible, since the event description is instantaneous. Clausal negation yields a predicate of times with the subinterval property, yet no modification of the reference time is possible. (54)

a. #Jańos ke´t o´ra´t nem e´rkezett arrived J-NOM two hour-ACC not ‘For two hours, John didn’t arrive’ b. #Jańos ke´t o´ra´t nem tala´lta found J-NOM two hour-ACC not ‘For two hours, John didn’t find the key’ 4.2.

meg perf meg perf

a the

kulcsot key-ACC

The source of the restriction

The restriction of case-marked temporal adverbs to event time modification arises as a consequence of the locality of adverb modification and the locality of case checking. Accounts of structural case licensing or case checking require that some local relation hold between the case-marked constituent and the case licensor. I assume a


249

minimalist approach to case checking, where the head checking the case feature must be local to the constituent with the feature to be checked. The structurally case marked adverbs appear with accusative case marking, as in the previous examples.20 The accusative case of the adverbs is checked by v, similarly to that of theta marked arguments. I assume that the constraints on case checking are the same for theta marked and non-theta marked constituents. In Chomsky 2000, 2001a,b, feature checking occurs as a result of an Agree relation that holds between the probe – which checks the feature in question – and the goal with the feature to be checked. The search space of the probe, in this case v, is restricted to its c-command domain. If an appropriate goal is found, which satisfies further conditions on locality and activity, the Agree relation is established between them. If accusative temporal adverbs are c-commanded by v, then they can serve as goals of the probe v, which checks their case feature. It can be shown that the durative adverbs, as well as thematic objects, are merged within the vP, and are thus ccommanded by v. First, durative adverbs are freely ordered with respect to the object within vP, in the postverbal field (E´. Kiss 2002a). (55)

Jańos nem pre´selte nedvesen ke´t o´ra´t two hour-ACC J-NOM not pressed wet-on a be´lyeget the stamp-ACC ‘Jańos did not press the stamp wet for two hours’

In addition, the durative adverb is elided with vP-deletion, along with other vPinternal material. (56)

Jańos ke´t o´ra´t futott, Pe´ter viszont nem P-NOM but not J-NOM two hour-ACC ran [vP futott ke´t o´ra´t] ran two hour-ACC ‘Jańos ran for two hours, but Pe´ter did not’

Since the durative adverb is merged in a position c-commanded by v, it can serve as the goal of the probe v, similarly to objects. Crucially, the case feature of the adverb cannot be checked if it is merged higher than the vP, as it is then outside of the search space of the probe. The locality restriction on case checking thus forces accusative durative adverbs to be merged within vP. The case checking of adverbs also raises some issues with respect to multiple feature checking. The v head can check multiple case features if both an object and an accusative adverb are present, as in (55) above and (57).21

250 (57)

ANIKO CSIRMAZ

Jańos ke´t o´ra´t olvasta a the J-NOM two hour-acc read ‘John read the book for two hours’

ko¨nyvet book-ACC

In order for the v head to check the case feature of both the adverb and the object, it must be allowed that v check multiple case features. The possibility of multiple case checking by a single head has been argued for by Ura 2000 and Bobaljik and Branigan (2006) for ergative patterns. Similarly to these accounts, I assume that multiple case checking by a single head is possible when necessary for convergence. Apart from the locality of case feature checking, the locality of adverb modification also plays a role in restricting the modification of structurally case marked adverbs. Specifically, I assume that adverbs can only modify a time interval which is local at the position where the adverb is externally merged. In other words, if a durative adverb is merged within the vP, then the only time interval available for modification is the event time. In the following structure, the time adverb within the vP can only modify the event time, but not the reference time.22 The locality restriction on adverbs can be stated as follows: (58)

A time adverb can modify a time interval i only if the minimal maximal domain containing the external merge position of the adverb also contains the time interval23

This restriction excludes the modification of a time interval which is introduced by a head higher than the position where the adverb is merged. Specifically, an adverb merged within vP cannot modify the reference time, which is merged as the specifier of AspP. The restriction of accusative durative adverbs to event time modification is not due to an overt restriction on the modification of the adverb. Rather, it arises from the interplay of locality restrictions on accusative case checking and adverbial modification. 5.

S U M M A RY

This chapter examined if negation affects aspectual or event structure. I argued that negation is not an aspectual operator, as it fails to affect the aspectual properties of event descriptions. At first sight negation may appear to affect aspect, since it allows modification of all event descriptions by -ig and a´t adverbs, which are sensitive to the subinterval property. I showed that negation does not affect properties of the event description, but licenses the subinterval property for the predicate applying to the reference time. Thus -ig and a´t adverbs are free to modify the reference time, but not necessarily the event time. Further arguments against the aspectual operator nature of negation included the different distribution of states and negated event descriptions as well as adverbial


251

modification licensed by quantificational arguments. I showed that negated event descriptions do not pattern with states. In terms of discourse interpretation and the availability of agent arguments, negated event descriptions behave like their affirmative counterparts. I also offered data which show that in addition to negation, decreasing quantifiers and constituents modified by csak (‘only’) also license -ig and a´t modification of the reference time. These facts are straightforwardly predicted by the account proposed in terms of the subinterval property of the predicate with the reference time argument. In contrast, an aspectual operator account of negation cannot be easily extended to account for these cases. Negation thus fails to affect properties of the event description, but affects the predicate which applies to the reference time.

NOTES

1

2

3

4

5

6

7

8

9

10

11

For discussions and comments about this chapter, I am greatly indebted to Kai von Fintel, Sabine Iatridou, David Pesetsky and Danny Fox, as well as to Katalin E´. Kiss, Christopher Pinõń, Boldizsa´r Eszes, Bala´zs Surańyi, Csaba Olsvay and Annama´ria Bene. While the discussion centers on Hungarian and occasional reference is made to English, I believe the account to have crosslinguistic relevance (with the Hungarian-specific assumptions addressed in section 2. In Hungarian, temporal adverbs with the suffix -ig are ambiguous between a durative and a punctual interpretation. If the complement of -ig is durative, then the adverb is the equivalent of a for-adverb and measures temporal duration. Whenever the complement is punctual, then the adverb is interpreted as an until-phrase. The adverbs discussed in this chapter appear with a durative complement of -ig. With an after ordering, the reference time follows the event time for perfect aspect, and the time of utterance follows reference time with past tense. If the ordering is within, then with present tense, the time of utterance is contained within the reference time. With imperfective aspect, the reference time is contained within the event time. The perfect head introduces its own time, which precedes the time of utterance and follows the reference time. Other functional projections, including focus, non-neutral, negation and PredP projections, can also appear in the (Hungarian) sentence structure. A more detailed structure is discussed below and in E´. Kiss (chapter 9). The structure is also similar to that assumed by Demirdache and Uribe-Etxebarria 2000, 2004, but their assumptions concerning the interpretation of Asp is different, as noted above. As argued in E´. Kiss (chapter 2) and Csirmaz (chapter 8), particles and other verb modifiers can move to Spec,AspP. To accommodate both particles and time intervals, I assume that AspP has multiple specifiers. As noted in Csirmaz (chapter 8), the reference time of an imperfective event description cannot contain an endpoint of the event time. This restriction is in line with the view of viewpoint aspect in Smith 1991. And a perfect time as well, if an optional perfect head is also present. Hungarian lacks a perfective aspect, but this option is available for a number of languages. The part relation is not strictly interpreted, but allows for the granularity effect, as noted in Csirmaz (chapter 8). The atomicity of the interval is not understood in an ontological sense, since time is dense. Instead, a time interval can be treated as atomic within the linguistic system. The time interval described as 3:27 or the event time of an achievement is, for instance, treated as atomic. If the (non-trivial) subinterval property is taken to be the defining notion of atelic event descriptions, then semelfactive event descriptions (including cough, peck and kick and their Hungarian equivalents)

252

12

13

14

15

ANIKO CSIRMAZ

do not qualify as atelic, contrary to Smith 1991. Semelfactives are classified as telic by approaches that define atelic event descriptions as homogeneous (e.g. Krifka 1989, 1992), rather than an event description without a designated telos, a result or outcome (Smith 1991). I assume that negation can apply to imperfective as well as perfective event descriptions (see Kiefer 1994 for discussion). The negation of imperfectives is pragmatically marked, but is acceptable with an appropriate context. The proposed account shares a number of similarities with Zucchi 1991 and Moltmann 1991. The treatment of adverbial modification is essentially the same in Zucchi 1991 and the present account. Unlike Zucchi 1991, I specifically identify the time intervals involved in the adverbial modification. I also present an account of the modification of negated event descriptions which is not specific to negation; this allows a virtually identical treatment of negated event descriptions and event descriptions with a decreasing quantifier argument or one modified by only (cf. section 3). Moltmann 1991 proposes a definition of for-adverbs which is remarkably similar to that of Zucchi 1991 and the one assumed here. Crucially, however, Moltmann 1991 requires runtimes of events rather than predicates of times to hold at each subinterval of the time modified. Since an event description must hold at each subinterval, Moltmann’s definition of for-adverbs can only be applied to the event time, but not the reference time. If the preverbal adverbs are contrastive topics in (14b,d), then the indicated reading is grammatical (K. E´. Kiss, p.c.). For a discussion of the intonation and interpretive properties of contrastive topics, see chapter 8 and the discussion below. In absence of negation, the event time modifying adverbs precede the verb and any non-specificational constituent in Spec,Pred, rather than appear postverbally: (i) Jańos (ke´t e´s fe´l o´ra´ja) aludt (?ke´t e´s fe´l o´ra´ja) two and half hour-until J-NOM two and half hour-until slept ‘Jańos slept for two hours and a half’ (ii) Jańos (ke´t o´ra alatt) meg ´ırta a ha´zi feladata´t J-NOM two hour under perf wrote the home exercise-POSS-ACC (? ke´t o´ra alatt) two hour under ‘Jańos wrote his homework in two hours’ If negation or focus (a specificational Spec,Pred) is present, then the adverb is postverbal: ´ ZI ´T HA FELADATA ´ırta meg (iii) Jańos (? ke´t o´ra alatt) A two hour under the home exercise-POSS-ACC wrote perf J-NOM (ke´t o´ra alatt) two hour under ‘It was his homework that Jańos wrote in two hours’

16

The distribution of the event time modifying adverbs is similar to that of manner adverbs (E´. Kiss 2002). E´. Kiss argues that manner adverbs are generated in the postverbal field (among other arguments and adjuncts) and are optionally adjoined to AspP (PredP) in the derivation (a position which may be reinterpreted as the specifier of some dedicated functional projection). It appears that the movement to the higher, AspP adjoined position is not optional. Rather, manner adverbs and event time adverbs must move to the higher position whenever a non-specificational constituent moves to Spec,PredP. I leave further discussion of the adverb positions for further research. Hinrichs 1986, Hatav 1997 and Borik 2004, among others, argue that it is not stativity, but rather delimitedness of the event descriptions which is relevant. Non-delimited event descriptions (as in (i)) give rise to a possibly overlapping interpretation and delimited event descriptions (shown in (ii)) are interpreted consecutively. (i) [It was a lovely performance]


253

Az elo ´ádo´ vicceket mese´lt, eńekelt e´s tańcolt sang and danced The performer-NOM jokes-ACC told ‘The entertainer told jokes, sang and danced’ (ii) [It was a lovely performance] Az elo ´ádo´ tı´z percig vicceket mese´lt, the performer-NOM ten minute-until jokes-ACC told fe´l o´raíg eńekelt e´s hu´sz percig tańcolt half hour-until sang and twenty minute-until danced ‘The entertainer told jokes for ten minutes, sang for half an hour and danced for twenty minutes’ These examples involve conjoined event descriptions, which may show a behavior different from the types of examples discussed above, where the two event descriptions appear as separate sentences. Even in these environments, negated event descriptions can behave as delimited, and describe consecutive events when conjoined with another event description. In the following example, the negated event description is interpreted most naturally as following the event in the first conjunct. (iii)

17

18 19

Mari besze´lt a befektete´si tanaćsado´val e´s nem adta el the investment advisor-with and not sold away M-NOM talked a re´szveńyeit the stocks-POSS-ACC ‘Mari talked to her investment advisor and did not sell the stocks she owned’

Decreasing quantifiers in Hungarian are restricted to the preverbal position, and may appear postverbally only if the preverbal slot is filled by some other element (Szabolcsi 1997). In the framework argued for by E´. Kiss (chapter 2), the preverbal position can be equated with Spec,PredP. More precisely, if the first argument of csak is interpreted as an argument of its second argument. Those Hungarian (preverbal) foci which are not modified by csak (‘only’) do not license the subinterval property, and so do not allow -ig or a´t modification either. ¨ KO ¨T AZ ELNO ismerte fel (i) #Jańos pa´r percig J-NOM few minute-until the president-ACC recognized up ‘#For a few minutes, it was the president that Jańos recognized’

20

21 22

23

This difference between identificational foci (E´. Kiss 1998, but see E´. Kiss (chapter 9) on the types of foci) and foci modified by csak. If foci modified by cask, but not those without it, are Strawson downward entailing, it is predicted that only the former allows –ig or a´t modification of the reference time. See Csirmaz (chapter 8) for a detailed discussion of non-theta marked accusative constituents in Hungarian, including arguments for structural case marking. Multiple case checking is also discussed in Csirmaz (chapter 8). I am assuming the account of durative time adverbs outlined in section 1. Event time is merged within vP, and reference time is merged as the specifier of AspP. Thompson 1996 suggests that adverbs must be in the modification domain of the constituent that they modify. According to Thompson, the modification domain of includes only the constituents adjoined to the maximal projection of .

´ ZS SURA ´ NYI BALA

P R E D I C ATE S , N E G ATI V E Q U A N T I F I E R S A N D F O CU S : S P E C I F I C I T Y A N D Q U A N T I F I C AT I O N A L I T Y O F N-WORDS 1.

INTRODUCTION

This chapter sets out to examine the interpretational difference involved in the following pairs of sentences:* a. Nem e´rkezett senki not arrive-PAST-3SG nobody-NOM ‘There hasn’t arrived anybody.’ b. Nem e´rkezett meg senki not arrive-PAST-3SG PRT nobody-NOM ‘Nobody has arrived.’ (2) a. Nem ke´szı´tettem semmit not prepare-PAST-1SG nothing-ACC ‘I haven’t prepared anything’ b. Nem ke´szı´tettem el semmit not prepare-PAST-1SG PRT nothing-ACC ‘I’ve prepared nothing.’ (1)

Recall that without a particle, verbs expressing existence, coming into being (among them, coming into being in a certain location, e.g. e´rkezik ‘arrive’, and creation verbs, e.g. ke´szı´t ‘prepare’) and consumption are definiteness effect predicates, discussed extensively in chapter 4 (by Pinõń), where it is argued that these predicates all introduce a novel discourse marker (cf. also Szabolcsi (1986), E´. Kiss (1995) and references cited there). Consequently, definiteness effect predicates require their internal argument to be non-specific / non-presuppositional, as illustrated in (3) below. (3)

E´rkezett arrive-PAST-3SG ‘There arrived b. #E´rkezett arrive-PAST-3SG a.

egy vende´g a guest-NOM a guest.’ a(z) (egyik) vende´g the (one) guest-NOM

/ /

a legto¨bb vende´g the most guest-NOM

As argued in chapter 2 (by E´. Kiss), verbal particles like meg and el in (1b) and (2b) – which in the perverbal [Spec,PredP] position determine the event type (situation aspect) – function as secondary predicates, predicating of the surface theme 255 ´ Kiss (ed.), Event Structure and the Left Periphery Studies on Hungarian, 255–285. Katalin E. 2008 Springer.


256

argument of the verb. This is claimed to be responsible for a specificity1 / presuppositionality requirement that verbal particles impose on the theme argument (causing the verbs at hand to not function as definiteness effect predicates; see chapter 4 for a polysemy based approach to this alternation). In (4a) the person who arrived must be a member of a presupposed set of guests (this is not a requirement in (3a) above). Whereas overt partitives are excluded in (3b), they are fine in (4b). (4)

e´rkezett egy vende´g arrive-PAST-3SG a guest-NOM ‘A guest arrived.’ b. Meg e´rkezett a(z) (egyik) vende´g / PRT arrive-PAST-3SG the (one) guest-NOM / / a legto¨bb vende´g / the most guest-NOM ‘One of the guests / Most guests arrived.’ a. Meg PRT

This specificity / presuppositionality requirement results in what appears to be a specific / presuppositional reading of the n-word functioning as the theme in (1b) and (2b). Given that the particle-less alternatives in (1a) and (2a) are definiteness effect predicates, only a non-specific / non-presuppositional reading is available there. It is contended in this chapter that the specificity / non-specificity contrast in (1) and (2), triggered by the requirements of the verbal predicate, is to be analysed in terms of an ambiguity in the interpretation of the negative pronouns (n-words) themselves, namely between a quantificational and a non-quantificational lexical meaning. (1b) and (2b) involve an n-word that is universally quantified and scopes above negation, while (1a) and (2a) involve a Heimian indefinite n-word, which is existentially closed in the scope of negation (cf. Heim 1982). Although the two interpretations (‘8x :’ and ‘: $x’) are logical equivalents, they are compositionally different and they behave differently in natural language (for a detailed discussion of this issue, see Giannakidou (2000, 2005)). The simultaneous presence of these two options within the same language (confirming a prediction of Giannakidou (2000: 518)) furnishes solid evidence that the same two options must be available within a cross-linguistic typology of Negative Concord.2 The chapter is structured as follows. In order to show that the ambiguity in (1) and (2) above is one of quantificational force, in section 2 the behaviour of Hungarian n-words is examined in various test environments. One battery of diagnostic tools will be shown to point to a construal of Hungarian n-words involving an existential indefinite in the scope of negation. At the same time, other tests suggest that n-words also have the interpretation of a universal quantifier scoping above negation. Various factors may rule out one or the other of the two interpretations in a given sentence, thus resulting in disambiguation. In fact (1) and (2) exemplify such cases. The ‘universal scoping above negation’ interpretation satisfies the specificity requirement introduced by the verbal particle in (1b) and (2b), hence only that reading is available there. In (1a) and (2a), in contrast, a non-specificity condition is

P R E D I C AT E S , N E G AT I V E Q U A N T I F I E R S A N D F O C U S

257

imposed by the particle-less variety of the verb on the theme, hence only the ‘existential indefinite in the scope of negation’ interpretation survives. Section 3 investigates the distribution of the availability of the two readings in (various domains of) the clause. N-words can in fact be syntactically focussed, and focussing affects their interpretation – an issue taken up in section 4. When in focus, the (morphosyntactically optionally realised) ‘even’ component in the semantics of Hungarian n-words plays an active role (cf. Lee and Horn (1994), Lahiri (1998) and Horn (2000)). N-words in focus are shown to be of the Heimian indefinite variety: what is in semantic focus is their bare predicative restriction. This analysis correctly predicts the general optionality of specificity-inducing particles with focussed n-words. 2.

Q U A N T I F I C AT I O N A L I T Y O F N - W O R D S

After a note on the distribution of n-words, we begin by applying various diagnostic tools to detect their quantificational status. N-words in this language may remain postverbal, or alternatively, they may be fronted to a preverbal position above negation, cf. (5a,b).3 Since this choice is available to each of the n-words in an NC clause, it is possible to front more than one n-word too (5c) (still with a Negative Concord reading).4 (5)

a. Nem jo¨tt el senki not come-PAST-3SG PRT nobody-NOM ‘Nobody came along’ b. Senki nem jo¨tt el nobody-NOM not come-PAST-3SG PRT ‘id.’ c. Senki sehova nem jo¨tt nobody-NOM nowhere-to not come-PAST-3SG ‘Nobody came along anywhere.’

el PRT

The issue of the quantificational nature of n-words in general has been a matter of debate in the literature on Negative Concord, and it has been resolved variously for different language types. According to one approach, n-words (in some languages) are a case of Negative Polarity Items (NPI) and are interpreted as existentially quantified indefinites in the scope of negation (cf. Ladusaw (1992, 1994), Acquaviva (1993, 1997)). Another influential approach (Zanuttini (1991), Haegeman and Zanuttini (1991), Haegeman (1995); Giannakidou (2000)) holds that n-words (in some languages) are universal quantifiers, scoping above negation. The two interpretations are logical equivalents; however, they give rise to discrete linguistic effects. Giannakidou (2005) provides an informative review of approaches to the question of the quantificational properties of n-words cross-linguistically. The literature on Hungarian for a considerable period now has been assuming Hungarian n-words to be universal quantifiers (licensed in some form by negation)

258


(cf. e.g. Szabolcsi 1981, 1997; E´. Kiss 1987, 1994, 2002a, 2002b; Puska´s 1998; Olsvay 2000a). This view has an obvious descriptive appeal: the symmetry found between the distribution of n-words and universals like every-NPs fall out directly.5 For, both n-words and every-NPs can appear both pre- and postverbally (potentially taking wide scope, depending on stress), and in the former case they are positioned to the left of preverbal focus and to the right of topics. I will argue now that appropriate diagnostics reveal that in some cases n-words in Hungarian can indeed be interpreted as universal quantifiers, yet in other cases probes indicate that they can be interpreted as existentially quantified.6 2.1.

Modification

One test often applied is almost-modification (see Dahl 1970, Horn 1972, Zanuttini 1991, van der Wouden and Zwarts 1993). Almost (and absolutely) are taken to be able to modify universal quantifiers, but not existential indefinites (see Horn and Lee (1995) for a refinement).7 As Puska´s (2000: 341) also points out, n-words can be modified by almost in Hungarian ((6) is adapted from her (70c)): (6)

Tegnap (majdnem senkivel) ’nem besze´lt Yesterday almost nobody-with not talk-PAST-3SG (majdnem senkivel) almost nobody-with ‘Yesterday Zeta talked to almost nobody.’

Zeta Z.-NOM

However, further qualification is in order. In reality, not all n-word occurrences allow almost-modification freely. Witness (7): (7)

*?Nem tala´ltam majdnem semmit a ha´zi dolgozata´ban not found-1SG almost nothing-ACC the home paper-POSS-3SG-in ‘I found almost nothing in her home paper.’

We will return to the conditions of almost-modification in section 3 below. The lesson at this point is that the correct generalisation is that n-words sometimes may, sometimes may not be modified by almost. This should mean that some, but not all occurrences of Hungarian n-words are interpreted as universal quantifiers. An existential interpretation (in the immediate scope of negation, cf. Note 7) may be responsible for the degraded status of occurrences that do not tolerate modification by almost. Modification by egya´ltalań ‘whatsoever / at all’ also gives mixed results. ‘Whatsoever’ and ‘at all’ reinforce NPI any in the scope of negation in English, but they cannot modify universals. Consider now (8): (8)

a. Nem vitt el egya´ltalań senkit moziba nobody-ACC cinema-to not take-PAST-3SG PRT at.all ‘He didn’t take anybody at all to the cinema’


259

b. *?Egya´ltalań senkit soha nem visz el moziba at.all nobody-ACC never not take-3SG PRT cinema-to ‘He doesn’t ever take anybody at all to the cinema.’ It appears that some n-word occurrences behave like existential NPI-s, others do not. We return to the distribution of modifiability by egya´ltalań ‘whatsoever / at all’ in section 3; for the moment, the implication of the test is that besides patterning with universal quantifiers, Hungarian n-words can also align with existential NPI-s. 2.2.

Donkey anaphora

A second test involves donkey anaphora. It is well known from dynamic semantics that universal quantifiers do not normally support anaphora appearing outside the sentence that they appear in8, whereas existentials do. An illustration from English is the following: (9)

a. I saw a boy. He was tall. b. I saw every boy. *He was tall.

Giannakidou (2000) argues that Greek emphatic n-words must be universals because they do not support donkey anaphora (whereas non-emphatic k-words occurring in non-veridical contexts do, so she analyses the latter class as existentials). A complication with the argument is that if Greek emphatic n-words were in reality existentials, to get the correct interpretation, they would be placed in the scope of negation; however, existentials in the scope of negation are known not to support donkey anaphora (being inaccessible, in terms of DRT; for the same point, cf. Richter and Salier (1998)) ((10b)¼Giannakidou’s (2000) (39)): (10)

a. I didn’t see a boy. (: > $) *He was tall. b. *The students that didn’t buy any/some book should show it now.

Now, Giannakidou (2000: 476) goes on to argue that in directive sentences like (11) below an anaphoric link can be established between the pronoun and the anyphrase; in other words, negated directives allow donkey anaphora in the case of existentials in the scope of negation ((11)¼Giannakidou, 2000 (40a)). Then, the appropriate test case in Hungarian is (12), a negated directive. (11)

Don’t check any book out from that (Satanic) library; reading it might warp your mind. (12) Ne fogja´l meg semmit (sem) a laborato´riumban! not touch-IMP-2SG PRT nothing-ACC (SEM) the laboratory-P-in Me´g pro megra´zhat possibly pro PRT-give.shock ‘Don’t touch anything in the lab. It could give you a shock.’

260


The grammaticality of such examples suggests that n-words in Hungarian can be interpreted existentially. 2.3.

Predicative nominals

Giannakidou and Quer (1995) note that just like universal quantifiers of the every-NP type, n-words in Greek cannot function as predicate nominals. In contrast, existential indefinites can. Thus Greek n-words side with every-NPs in this respect. NC languages seem to be split in this regard: Italian, Spanish and French n-words follow the same pattern as Greek, while Russian, Polish and Serbian n-words disallow a predicative use (cf. Giannakidou 2005). Hungarian appears to be able to use n-words predicatively:9 (13)

a. Nem lesz semmi baj not will.be nothing problem ‘There won’t be any problem.’ b. Ez a zaj nem volt semi a tegnapihoz this the noise-NOM not was nothing the to-yesterday’s ke´pest in.comparison ‘This noise was nothing compared to yesterday’s.’ c. Nem volt semmi ko¨ze hozza´ not was nothing business-POSS-3SG it-to ‘He had nothing to do with it.’ These data suggest that Hungarian n-words may be interpreted as existentials.10 2.4.

Particles

Giannakidou (2000) points out that in Greek, ke ‘and’ is a modifier of existential quantifiers, and n-words and universal quantifiers are incompatible with it, thus forming a natural class in this respect. A similar consideration may turn out to be relevant in Hungarian as well. Hungarian has a paradigm of weak negative polarity items (cf. To´th 1999) licensed in a range of nonveridical contexts. This paradigm of weak NPI has the morphological structure valaþWhþis, where vala- is ‘some’ (14a). Here Wh stands for a bare indefinite that functions as a wh-pronoun when on its own, and is is (homophonous with) is ‘also/even’.11 Now vala- ‘some’ and minden- ‘every’ combine with the bare wh-indefinites to make the paradigms of existential indefinite and universal pronouns, cf. (14b). It is commonly accepted that weak NPI-s are interpreted existentially. Importantly, though is can modify weak NPI-s and wh-phrases (which are taken to be existentially quantified), cf. (14a), it cannot modify positive universal quantifiers, cf. (14c): (14)

a. valaki is / ki is some-who also/even / who also/even


b. valaki some-who c. *mindenki every-who

261

/ mindenki / ki / every-who / who is also/even

As noted in section 2 (cf. Note 3), n-words may be modified by the particle sem. Historically sem is a combination of is ‘also/even’þnem ‘not’. This means that the nword paradigm and the existential weak NPI paradigm at an abstract level share the property of being modified by is. That n-words pattern with an existential weak NPI is suggestive evidence of the availability of an existential interpretation.12 2.5.

Incorporation

A further relevant observation concerns incorporation. Bare singulars in Hungarian undergo ‘incorporation’ to the verb, cf. (15).13 In this incorporated position, bare singulars have an existential reading. They are scopally inert (cannot scope over any operator that has the predicate in its scope), and need not be in the scope of any operator. Ordinary weak existential complements too can, and with some verbs must, be incorporated (16a/a’). Universals, however, cannot be in the incorporated position, cf. (16b). Also, incorporated bare plurals lack a generic interpretation, but only have an existential one.14 (15)

(16)

a. Jańos re´gi be´lyeget gy u˝ jt J.-NOM old stamp-ACC collect-3SG ‘John collects old stamps.’ b. *Jańos gy u˝ jt re´gi be´lyeget old stamp-ACC J.-NOM collect-3SG a. Jańos valami hı´res embert alakı´t J.-NOM some(thing) famous person-ACC act-3SG ‘John plays the part of a famous person.’ a¢. *Jańos alakı´t valami hı´res embert J.-NOM act-3SG some(thing) famous person-ACC b. *Egy szıńe´sz minden hı´res embert alakı´t an actor-NOM every famous person act-3SG ‘An actor plays the part of every famous person.’

In this light consider now the data below. An n-word within an infinitival clause is licensed by matrix negation in both (17) and (18). As the acceptability contrast between the (a) and (b) examples show, the obligatory fronting of the n-word is due to incorporation (and not quantifier fronting, which, as we have seen, is optional). The fronted n-word must be adjacent to the verb – a trait of incorporation again. (17)

a. Nem szeretne´k Pa´lnak semmi hu¨lyese´get mondani holnap not like-COND-1SG P-DAT nothing stupid-ACC tell-INF tomorrow ‘I wouldn’t like to tell Paul anything stupid tomorrow.’


262

b. ?*Nem szeretne´k Pa´lnak mondani semmi hu¨lyese´get nothing stupid-ACC not like-COND-1SG P-DAT tell-INF holnap tomorrow (18) a. Nem szeretne´k semmi ku¨lo¨no¨snek la´tszani not like-COND-1SG nothing particular-DAT seem-INF ‘I wouldn’t like to seem anything particular.’ b. *Nem szeretne´k la´tszani semmi ku¨lo¨no¨snek not like-COND-1SG seem-INF nothing particular-DAT Given that incorporated nominals in Hungarian can only be quantified existentially, but not universally, these examples too demonstrate that n-words in Hungarian can receive an existential interpretation. 2.6.

Existential import and split readings

It is well known that universal quantification has a pragmatic implicature of existence in natural language (cf. e.g. Strawson 1952).15 If an n-word is interpreted as a universal quantifier scoping above negation (8 > :), then such existential import is predicted, unlike if an n-word is interpreted as an existential indefinite in the scope of negation. As Giannakidou (2000) demonstrates, existential import invariably gets generated in the case of Greek emphatic n-words (which she analyses as universal quantifiers). Hungarian n-words, however, appear to be different: they are not always presuppositional. Witness (19) (IRR.¼irrealis marker): (19)

a. Nem fedeztem fel semi nyoma´t, nothing trace-POSS-3SG-ACC not discover-PAST-1SG PRT hogy ott ja´rt volna that there go- PAST-3SG IRR. ‘I didn’t discover any trace (suggesting) that he’d been there.’ b. Nem la´tom semi e´rtelme´t not see-1SG nothing sense-POSS-3SG-ACC ‘I don’t see any point (in it).’

These examples are perfectly felicitous. Equivalent sentences in Greek are pragmatically odd (cf. Giannakidou 2000: 505), because Greek n-words are invariably presuppositional. The speaker in (19b) asserts that there is no point in it, thus a conflict would arise if the n-word had to have an existential import. On the most prominent reading of (19a), the speaker is committed to assuming that there was no trace of him having been there earlier; if the n-word were presuppositional, the sentence would result in pragmatic oddity. n-words in such examples cannot be universals, but may be existentials in the scope of negation. At this point we can return to our initial examples in (1a) and (2a). So-called ‘definiteness effect’ verbs (like ‘find’, ‘arrive’, ‘prepare’) (when lacking a particle) require a non-specific, hence non-presuppositional, internal argument (cf. Szabolcsi


263

1986, E´. Kiss 1995). Among others, existential indefinites can, while universals cannot combine with such verbs. As noted in E´. Kiss (2002a, 2002b) and Surańyi (2002a), these verbs allow an n-word as internal argument ((20)¼(1a)). (20)

Nem e´rkezett senki not arrive-PAST-3SG nobody-NOM ‘There hasn’t arrived anybody.’

Then n-words in such sentences cannot be universal quantifiers, but must be existentials in the scope of negation. Examples like (7) above confirm the accuracy of this conjecture. What renders (7) degraded is the conflicting requirements of almost (enforcing the universal reading) and the verb ‘find’ (enforcing an existential reading). The availability of so-called ‘split’ readings with modal verbs (cf. e.g. Jacobs 1991, Rullmann 1995, de Swart 1996) once again points to the same conclusion. A German sentence like (21a), or an English example like (21b), has three distinct readings. (21)

a. Die Firma muss keinen Angestellten no-ACC employee-ACC the firm-NOM must ‘The company must fire no employee.’ b. One is allowed to fire no nurses

feuern fire-INF

For (21b), these readings are the de re (‘there are no nurses such that one is allowed to fire them / for every nurse, one is not allowed to fire her’), the de dicto (‘what one is allowed to do is: not fire any nurse / what one is allowed to do is: for every nurse, not to fire her’) and the ‘split’ interpretation (‘one is not allowed to fire any nurses at all’). To see that the de re and the split readings are truth-conditionally distinct, consider the sentence The hospital needs to fire no nurse in a context in which the management of a hospital has decided that due to cutbacks a certain number of nurses need to be made redundant. In that context, the de re reading is true (i.e. there is no nurse such that she needs to be fired), while the split interpretation is false (i.e. it is false that it is not necessary for the hospital to fire any nurse). As Giannakidou (2000) shows, the ‘split’ reading, which has been taken to be decomposed as ‘: > modal > $’, is unavailable with the Greek counterpart (because Greek n-words translate as universal quantifiers). However, the Hungarian equivalent admits this reading rather easily: (22)

Nem kell elbocsa´jtani egy a´polońo´´t sem / senkit (sem) not need PRT-fire-INF a nurse-ACC SEM / nobody-ACC (SEM) ‘One need not fire any nurses / anybody.’

Hence we have evidence that Hungarian n-words may get interpreted as existential quantifiers.16

264


2.7.

Topicalisability

Universal quantifiers denote familiar discourse entities. This renders them suitable topics, and in a number of languages they can even undergo syntactic topicalisation (e.g. Italian, Greek; cf. Cinque 1990; Giannakidou 2000). As noted in Surańyi (2002a), in Hungarian too, when descriptively sufficiently rich to make a topic (e.g. when modified by a relative clause)17, universal quantifiers can undergo syntactic topicalisation, cf. (23a) where the modified universal precedes another topic. Under the same proviso, n-words can also be syntactically topicalised, cf. (23b). (23)

a. Mindenkinek aki ott volt Pe´ter ajańde´kot ad gives everybody-DAT who-NOM there was P.-NOM gift-ACC ‘Peter will give a gift to everybody who was there.’ b. Senkinek aki ott volt Pe´ter nem ad ajańde´kot nobody-DAT who-NOM there was P.-NOM not gives gift-ACC ‘Peter will give a gift to nobody who was there.’

Given that an indefinite in the scope of negation is analysed as novel (e.g. Heim 1982) (and as such can be used in out of the blue contexts), it cannot topicalise. Additionally, it would also be problematic to analyse topicalised instances of n-words as indefinites for an independent reason. As argued in Surańyi (2002a, 2002b, 2006) n-words not modified by a sem particle are non-negative; logical negation is contributed by the negation particle nem. When topicalised, indefinite n-words would escape the scope of their licensing negation, yielding the ‘$ > :’ (i.e., not the attested) scope relations. On a universal quantifier construal, however, topicalisation produces the unproblematic ‘8 > :’ reading. Then, topicalisability doubly argues that Hungarian n-words can receive a universal quantifier interpretation. 2.8.

Licensing

Licensing conditions of existential NPI demand that the NPI be in the immediate scope of the licensor, for our purposes, negation: no other (non-NPI-licensing) operator may intervene (cf. Linebarger’s (1987) Immediate Scope Constraint). If Hungarian n-words were invariably existentially quantified strong NPI-s (strong in the sense of Zwarts (1996)), they would be expected to obey this condition. As (24) shows, this is not the case. In (24a), the sentence is well-formed whether or not a universal quantifier, or a quantificational adverb occupying a preverbal focus position within the infinitival clause intervenes between the licensing negation in the matrix and the n-word in the infinitival clause. In (24b) we have the same configuration of ‘negation > focus > n-word’, only this time within one clause. The n-word needs to be stressed in both cases. (24) a. Nem akarok mindenkinek / KE´TSZER elmondani 0 semmit (sem) PRT-say-INF nothing-ACC SEM not want-1SG everybody-DAT / twice ‘I don’t want to say anything twice / to everybody.’


265

b. Nem MA akarok elmenni ’sehova (sem) not today want-1SG PRT-go-INF nowhere SEM ‘I don’t want to go anywhere TODAY.’ This pattern can be explained if these n-word occurrences are interpreted as universal quantifiers, and as such, covertly move above negation, where they are licensed locally (cf. Giannakidou (2000) for the licensing of universally quantified nwords). Indeed in Hungarian accent on a postverbal universal too marks wide scope of the quantifier with respect to a preceding operator (cf. e.g. E´. Kiss 1994); in Surańyi (2002a) it is suggested (rejecting various other alternatives) that stress marks covert movement of the quantifier to achieve wide scope. In the case of postverbal occurrences of n-words in (24), this operator is negation itself. Notice that in light of section 2.6 the prediction is that these n-words can only be interpreted as de re. Indeed, this expectation is borne out. The same consideration extends to explain the unacceptability of sentences analogous to (24) containing a verb that requires a non-presuppositional / non-specific argument: (25)

Nem fog (*?minden nap) to¨rteńni every day happen-INF not will-3SG ‘There won’t happen anything every day.’

semmi nothing-NOM

(25) is unacceptable with an intervening quantificational adverb, precisely because the presence of that adverb excludes the existential strong NPI construal, and enforces the ‘universal over negation’ construal instead (in effect the de re / presuppositional reading). However, given that the verb selects for a non-specific / non-presuppositional argument, the result is at best severe pragmatic degradation. We have confirmation then that n-words in (24) above can only be interpreted as universal quantifiers, but not as existentials below negation. 2.9.

Distributivity

Predicates like o¨sszegy u˝ lik ‘gather’, sze´tszo´ro´dik ‘scatter’, ko¨ru¨lvesz ‘surround’, etc. require one of their arguments to be interpreted collectively (cf. Dowty 1987).18 Such predicates cannot be modified by adverbs like egyenkeńt ‘individually’ and do not tolerate distributive QPs in their argument position:19 (26)

a. Az emberek (*egyenkeńt) ko¨ru¨lvette´k a ha´zat individually PRT-surround-PAST-3PL the house-ACC the people-NOM ‘The people (*individually) surrounded the house.’ b. *Minden ember ko¨ru¨lvette a ha´zat Every person-NOM PRT-surround-PAST-3SG the house-ACC ‘*Every person surrounded the house.’


266

With this in mind, consider the examples involving collective predicates occurring with n-words below: (27)

a. Nem veszi ko¨ru¨l a ha´zat senki sem not take-3SG around the house-ACC nobody-NOM SEM ‘Nobody surrounds the house.’ ´ ZAT b. *Senki sem A HA veszi ko¨ru¨l the house-ACC take-3SG around nobody-NOM SEM ‘Nobody surrounds THE HOUSE.’

It appears that the n-word argument can only receive a distributive reading in (27b), while that reading is not forced in (27a). Assuming that when n-words are interpreted as universal quantifiers, they are distributive universals of the every-NP type, we have evidence here that at least in some cases n-words must be interpreted as universal quantifiers (as in (27b)). 2.10.

Results

In this section I have applied a number of diagnostics to detect the quantificational status of n-words in Hungarian. As demonstrated in sections 2.2, 2.3, 2.4, 2.5 and 2.6, n-words in this language can have an existential indefinite interpretation in the scope of negation. Sections 2.7, 2.8 and 2.9 have provided evidence that they can be interpreted as universals scoping above negation. As I pointed out in the introduction to this section, the long-standing assumption that n-words in Hungarian are universal quantifiers also straightforwardly explains their distributional and prosodic similarities (but see section 3 below). Finally, section 2.1 yielded mixed results: both a universal and an existential interpretation appear to be necessary; indeed this appears to be the conclusion from the whole of section 2 as well. A question that has been avoided so far concerns the syntactic distribution of the two readings. This matter is taken up in the next section, where it will be shown that the two readings have distinct, albeit overlapping, distributions. 3.

T H E D I S T R I B U T I O N O F T H E T W O I N T E R P R E TAT I O N S

I turn now to investigate under what syntactic conditions the two interpretations are available to n-words. Before embarking on this task, however, we need to have a rough picture of the syntactic positions in the clause that n-words are able to occupy. 3.1.

The distribution of n-words

As illustrated in (5) at the beginning of the chapter, n-words can in general either remain postverbal, in which case they can be assumed to be in situ, or they can be fronted (including fronting of multiple instances). As pointed out in the introduction to section 2 above, when fronted, n-words are positioned to the left of a preverbal


267

focus and to the right of syntactic topics. This field is identical with what is often referred to as the ‘quantifier field’ of Hungarian clause structure (cf. E´. Kiss 1994), which can be targeted by universal quantifiers and other increasing distributive QP-s, and which Szabolcsi (1997), in applying Beghelli and Stowell’s (1994, 1995, 1997) approach to Hungarian, analyses as housing (recursive) DistP.20 A fundamental question that has been left open thus far in the discussion is whether n-words can function as a syntactic focus. The issue is non-trivial, since simple word order considerations are not revealing. This is because verb–particle inversion, which is the crucial linear indicator of syntactic focalisation in Hungarian, is triggered not only by focus fronting, but also in the presence of sentential negation, i.e. in all of the cases at hand. The issue is not settled in studies of Hungarian NC, and is not considered in overviews of Hungarian clause structure where NC is also discussed in some detail (E´. Kiss 1992, 1994, 1998a, 2002a).21 I argue now that n-words in Hungarian are syntactically focussable, albeit not invariably in focus when preverbal. Preverbal focus in Hungarian is marked by emphatic accent which is normally followed by a deaccenting of the immediately right-adjacent inverted verb, or the nemþverb sequence. Such a stress pattern is readily available with preverbal nwords, cf. (28a). On the other hand, if the n-word is not emphatic but bears neutral phrasal stress as determined by the Nuclear Stress Rule, then the following verb may also receive stress, cf. (28b). (Capitals indicate emphatic stress, ’marks neutral stress of phonological phrase, while 0 full stress reduction.) (28)

0 nem 0szavazott Jańosra a. SENKI noone-NOM not vote-PAST-3SG J.-for ‘Noone voted for John.’ 0 nem 0szavazott Jańosra b. 0 Senki

Prosodic evidence then suggests that n-words are focusable, though not necessarily focussed when preverbal. This latter conjecture is verified by the fact that while syntactic focus is unique in the preverbal domain, multiple occurrences of n-words can appear to the left of the verb, where in any case the not immediately preverbal occurrences cannot be focus, but must belong to the quantifier field. Similarly, the non-focussed n-word in (23b) above too is located in a quantifier position. The conclusion that n-words are focussable is also confirmed by the distribution of postverbal focus. A focussed expression is normally only licensed postverbally by a preverbal occurrence of focus. Observe now that a postverbal focus can also be licensed by a preverbal n-word, cf. (29a), but only if it bears emphatic stress and the stress of the negation and the verb is reduced, cf. (29b). (29)

0 ´ NOSRA nem 0szavazott ve´gu¨l CSAK JA a. SENKI vote-PAST-3SG finally only J.-for noone-NOM not ‘Finally nobody voted for only John.’ 0 nem 0szavazott ´ NOSRA ve´gu¨l CSAK JA b. *? 0 Senki


268

Finally, emphatic n-words in the immediately preverbal position have a domain widening effect. According to Kadmon and Landman (1993), NPI elements like anyphrases in English are interpreted like corresponding indefinites, with the difference that the domain of quantification over which the indefinite with any ranges is wider (along a contextually defined dimension) than the domain of a common noun (cf. also Israel 1996). Krifka (1995) suggests that this effect of signalling reduced acceptance of exceptions takes place only in case the NPI is used emphatically (as noted by Chierchia (2002), domain widening is not necessary in all cases: it may easily happen that the base domain of quantification is the largest reasonable domain of quantification). Indeed a domain widening effect obtains with emphatic n-words in the focus slot (30b). Such widening is not available in a postverbal position (30a). (30)

Marinak ege´szse´gileg nem volt semmi baja, health-wise not was no(thing) problem-POSS-3SG M.-DAT csak alva´szavarai voltak only sleep.disorder-PL were b. #Marinak ege´szse´gileg SEMMI BAJA M.-DAT health-wise no(thing) problem-POSS-3SG 0 nem volt, csak alva´szavarai voltak not was only sleep.disorder-PL were ‘Mary had no problems health-wise, she only had a sleep disorder.’ a.

The general conjecture is that n-words are in fact focussable with the regular prosodic and syntactic consequences. (Note that the focus position is identified as FocP in Chapter 9 (by E´. Kiss), which corresponds to the FocP of mainstream clausal architectures, cf. Rizzi 1997). In sum, three different syntactic fields need to be distinguished as far as n-words are concerned: the postverbal (in situ) position, the quantifier field, and the preverbal focus slot. 3.2.

Interpretations and syntactic positions

In order to examine the availability of the two detected readings as a function of syntactic position, the distribution of three properties will be examined, each of which have been correlated in the preceding section with one or the other of the two interpretations: modifiability by ‘almost’ and by ‘whatsoever / at all’, existential import, and split readings. First, modifiability. ‘Almost’ can modify n-words both in their in situ position and in the quantifier field (cf. (6) above). That suggests that the universal quantifier reading is available in these positions. ‘Almost’ can modify n-words even in the focus position: (31)

(?)

0 0 Majdnem SENKI nem besze´lt not eat-PAST-3SG almost nobody-NOM ´ VAL CSAK A SZOMSZE´DJA only the neighbour-POSS-3SG-WITH ‘Almost nobody talked only to his neighbour.’


269

I will return to this latter observation in the next section to discuss how a potential problem it creates can be resolved. Recall that egya´ltalań ‘whatsoever / at all’ is able to modify existential indefinites in the scope of negation, but cannot modify universals. In this respect, the relevant fact is that egya´ltalań can modify n-words in the postverbal field, as well as in the focus slot, though in the latter position some speakers report a degree of degradation.22 N-words modified by egya´ltalań ‘whatsoever/at all’ are disallowed in the quantifier field; see (8b) above, as well as (32). (32)

´ NOS visz (*?Egya´ltalań) senkit nem JA el moziba J.-NOM take-3SG PRT cinema-to at.all nobody-ACC not ‘JOHN doesn’t take anybody (at all) to the cinema.’

Since modification by egya´ltalań ‘whatsoever/at all’ allows the existential but not the universal interpretation, the existential reading is apparently ruled out in the quantifier field. As shown in section 2.6, n-words need not have existential presupposition when postverbal, cf. (19). However, in the quantifier field existential import is generated, and this is what makes (33) odd. (33)

0 soha # 0 Semmi e´rtelme´t nothing sense-POSS-3SG-NOM never ‘I have never seen any point (in it).’

nem not

la´ttam see-PAST-1SG

When the n-word is focussed, the sentence becomes perfectly acceptable again, cf. (34). (34)

SEMMI E´RTELME´T nothing sense-POSS-3SG-ACC ‘I didn’t see any point in it.’

0

nem not

0

la´ttam see-PAST-1SG

In short, there is indication that an existential reading is available postverbally, while such a reading is ruled out in the quantifier field. Focussed n-words do not generate existential presuppositions. As demonstrated in (20) (¼(1a)) above, verbs requiring a non-specific / nonpresuppositional argument are compatible with a postverbal n-word. In contrast, the quantifier field appears to enforce the universal interpretation, which is incompatible with the non-specificity requirement of the verb, whence the oddity of (35a). Focussed instances of n-words, however, are impeccable. (35)

a. #Senki sehova nem e´rkezett arrive-PAST-3SG nobody-NOM nowhere-to not ‘There hasn’t arrived anyone anywhere.’ 0 nem e´rkezett b. SENKI

270


The conclusion we can draw from this is that the quantifier field enforces the universal interpretation, in contrast with the postverbal domain and the focus slot. (36) provides confirmation for the former conjecture: (36)

Senki sehova nem nobody-NOM nowhere-to not ‘Nobody has arrived anywhere’

e´rkezett arrive-PAST-3SG

meg PRT

(36), forming a minimal pair with (35a) involves a particle, which requires the theme argument of the verb to be specific (cf. Chapter 2). If, as we have concluded, the quantifier field (housing the theme argument in (36)) enforces a universally quantified interpretation, then the alteration in acceptability is wholly expected: the universal quantifier is in conformity with the specificity requirement. In section 2.6 above, postverbal n-words were seen to allow a split reading, as well as a de re reading. As for the quantifier field, only the de re reading obtains there. This reading is derived if n-words in the quantifier field are universally quantified and their fronting is put down to universal quantifier movement, (cf. 37a, b). With the n-word in syntactic focus, as in (37c), the sentence is apparently ambiguous between describing a situation involving employees of a particular company and the split reading (‘: > modal > $’), i.e. in this case, a general ethical statement. In other words, (37c) does not require a presupposed set of individuals, being similar in this respect to (22) above, where the n-word is postverbal. (37)

a. Senkit soha nem kell elbocsa´jtani nobody-ACC never not need PRT-fire-INF ‘Nobody ever needs to be fired.’ b. Senkit nem EMIATT kell elbocsa´jtani nobody-ACC not BECAUSE.OF.THIS need PRT-fire-INF ‘For nobody is it the case that s/he needs to be fired because of THIS.’ 0 nem 0lehet elbocsa´jtani INDOKOLATLANUL c. SENKIT may PRT-fire-INF unjustifiably nobody-ACC not ‘Nobody may be fired UNJUSTIFIABLY.’

Taking the de re reading to be generated in Hungarian by the ‘universal over negation’ construal, the quantifier field appears to be compatible only with a universal reading. A focussed n-word has either a reading with a presupposed set of individuals, or the split reading.23 It was demonstrated in section 2.9 that n-words are compatible with a collective predicate when in the postverbal position (27a), but not when in the quantifier field (27b). N-words in the focus slot can also occur with a predicate interpreted

271


collectively: (38)

0 SENKI nem 0veszi take-3SG nobody-NOM not ‘Nobody surrounds the house.’

3.3.

ko¨ru¨l around

a ha´zat the house-ACC

Results and discussion

The findings of the previous discussion are summarised in the table below. Table 1 Summary of results Implication 1. Modification by ‘almost’ possible 2. Modification by ‘at all’ possible 3. Necessary existential presupposition 4. Compatible with definiteness effect verbs 5. ‘Split’ reading available 6. Incompatible with collective predicate

Quantifier field

Focus position

Postverbal field

Can be universally quantified Can be existentially quantified Must be universally quantified Can be existentially quantified

Yes

Yes

Yes

No

Yes

Yes

Yes

No

No

No

Yes

Yes

Can be existentially quantified Must be universally quantified

No

Yes

Yes

Yes

No

No

Given these results, the conclusion to draw for the quantifier field is that there nwords must be universally quantified. In the postverbal field, both existentially quantified and universally quantified n-words are allowed (and the latter, similarly to positive universal QPs, undergo covert quantifier raising (to a position above negation)). I discuss the implications of these findings directly, and then I return to the results obtained for the focus slot. If the conclusions thus far are correct, then a quantificational ambiguity approach to Hungarian n-words is required. Such an approach has been proposed in various forms in the literature (cf. van der Wouden and Zwarts (1993), Dowty (1994), Ladusaw (1994), Richter and Salier (1998), and Herburger (2001)); the ambiguity is generally taken to one between a negative universal and a non-negative NPI indefinite construal.24 The indefinite NPI treatment of the existential interpretation of n-words was first proposed in Ladusaw (1992), and was developed in Ladusaw (1994), Acquaviva (1993, 1997), Giannakidou and Quer (1995, 1997), Giannakidou (1997) and Richter and Salier (1998), among others. As discussed extensively by Giannakidou (2000, 2005), in fact the issue of universal/existential quantificational force and the issue of negativity are logically independent (she advocates the view that Greek n-words have universal force, and are logically non-negative).

272


My proposal for Hungarian is a variety of the ambiguity approach: Hungarian nwords are ambiguous between a logically non-negative universal quantifier interpretation, and a logically non-negative Heimian indefinite interpretation. In particular, I adopt the view that indefinite NPI-s are licensed in the (immediate) scope of negation, and they are interpreted as a Heimian indefinite (contributing a variable and a descriptive restriction only; for the particular implementation, see below): their existential interpretation is due to the fact that they come to be bound under existential closure applying to the nuclear scope of negation (cf. Heim 1982, Giannakidou 1997). The existential reading in the postverbal field is now straightforward to account for: here, being in the scope of negation, indefinite NPI-s are subject to existential closure. In particular, following Heim (1982) and Giannakidou (1997), negation is taken to be interpreted either as : or as :$, where the latter interpretation arises in the case of existential closure in the scope of negation. As for the interpretation of indefinites, I follow von Stechow’s (1993, 2004) implementation of the Heimian view of indefinites in assuming the variable and descriptive restriction part of indefinites to yield an expression of type hhetiti (e.g. senki ‘nobody’ is interpreted as k‚P(person¢(x) P(x))k).25 Given that n-words are (non-quantificational) indefinite NPI-s on their existential reading, it also follows that they cannot be fronted to the quantifier field. This is because the quantifier field is essentially the target domain for the fronting of increasing distributive quantifiers (Szabolcsi 1997). Given that non-quantificational indefinite NPI-s do not belong to this class, they cannot undergo such quantifier raising. Note that as far as the test results in the postverbal field are concerned, the ambiguity of n-words could well be one between a universal quantifier and an existential quantifier interpretation. However, what I have proposed is an ambiguity between a universal QP and a non-quantificational indefinite reading. I come to reasons for opting for a non-quantificational indefinite meaning as one of the two lexically available interpretations, instead of an existential quantifier meaning, after presenting an outline of the proposed picture, which I turn to presently. Besides the existential reading, Hungarian n-words have been demonstrated to also exhibit the universal reading. The universal quantifier itself can possibly be identified with the s(e)- morpheme (historically deriving from is ‘also’) shared by the paradigm of n-words as an initial morpheme. Universal n-words are argued in Giannakidou (1998, 2000) to be semantically licensed as polarity sensitive elements by escaping the antiveridical predicate (for the present purposes: a predicate phrase headed by negation) that they are generated inside of (i.e., there is an anti-licensing condition similar to that on positive polarity items), to end up semantically combining with an antiveridical predicate. This semantic requirement explains why universal n-words as polarity sensitive universal quantifiers are not licensed if they only QR to a scope position within the scope of negation. Qua universal quantifiers, the fronting of n-words to their scope position can take place either overtly or covertly: this is what universal quantifiers are independently characterised by in Hungarian (see Surańyi 2002a). In short, the ambiguity entertained here is one between the presence and absence of (universal) quantificational force. To illustrate, consider the sentence and the surface syntactic structure in (39). (Note that negation is high in Hungarian, i.e. NegP is


273

projected above TP; see Surańyi 2002a, b). The n-word senki ‘nobody’ in (39) can be interpreted as a universal quantifier, in which case the LF form (after QR of the n-word, taken here to the n-word to the left edge of NegP) is (39b); alternatively, the n-word can be interpreted as a Heimian indefinite, in which case the LF structure is as in (39c) (in both LF forms, the verb is placed back for sake of simplicity in its base position). (39)

Nem alszik [TP senki ti ]] not sleep-3SG nobody-NOM b. LF1 with a universal N-word: nem [TP tj alszik]] [NegP senkij sleep-3SG nobody-NOM not c. LF2 with a Heimian indefinite N-word: alszik]] [NegP nem [TP senki not nobody-NOM sleep-3SG a. [NegP

For the universal quantifier reading of the senki ‘nobody’ (40a), the corresponding (simplified) semantic derivation is outlined in (41a); for the indefinite reading of the n-word (40b), the (simplified) semantic derivation is as in (41b).

(40) a. ||λP.∀x.(person9(x) → P(x))|| b. ||λP(person9(x) ∧ P(x))||

(41) a.

∀x.(person9(x) → ¬ sleep9(x)) λP.∀x.(person9(x) → P(x))

λy1.¬ sleep9(y1)

λ1

¬ sleep9(y1) ¬

sleep9(y1) y1

b.

λz. sleep9(z)

¬∃x(person9(x) ∧ sleep9(x)) ¬∃

person9(x) ∧ sleep9(x)

λP(person9(x) ∧ P(x))

λz.sleep9(z)

274


Derivation (41a) involves QR of the universal quantifier n-word, while on the indefinite reading, the n-word is existentially closed under negation. The lexical ambiguity is one between the absence and presence of the universal quantificational force. Let us turn to the results obtained for the focus position in Table 1 above. The findings are identical with those in the case of the postverbal field. I show now that treating focussed n-words as universally quantified is not feasible, but focussed n-words cannot be existentially quantified QPs either. If so, then the ambiguity of nwords is not one between a universal QP and an existential QP. Consider first (37c), a sentence with an n-word in focus that is apparently ambiguous between the de re and the so-called ‘split’ (‘: > modal > $’) readings. A complication that would arise if focussed n-words were lexically quantified existentially (as one of two quantificational force options here) is the following. Given that a focussed n-word is situated in an A-bar position above the modal verb, which it c-commands from this A-bar position, therefore obligatorily taking it in its scope, there is no way one can get ‘: > modal > $’ scope relations. Clearly, the universal quantifier interpretation cannot be applied here either, since that would have to generate an existential presupposition, which is simply not the case with n-words in focus. It appears then that paradoxically, when the n-word is in focus, what appears to be the ‘split’ reading does not involve a lexically existentially quantified n-word and is not represented as ‘: > modal > $’, but must be associated with a different semantic description. Quite independently of the apparent ‘split reading’, if a focussed n-word, which is in an A-bar position above negation, is interpreted as an existential quantifier, the result is wrongly predicted that negation is in the scope of existential quantification (‘$ > :’). Once again, it appears that an n-word in focus cannot be an existential quantifier. However, it must be existentially quantified, according to the implication in lines 2, 4 and 5 in Table 1 above (line 2 records the fact that n-words in focus can be modified by ‘at all’, and line 4 registers that they can combine with a definiteness effect verb). Focussed n-words can (with some degree of degradation) be modified by ‘almost’ (cf. (31)), which in the present context has been taken to be a property of universals. This fact is remarkable, since universals (whether containing a count or a mass noun, cf. English ‘every’ or ‘all’) cannot be syntactically focussed in Hungarian:26 (42)

*MINDEN BORT ivott meg drink-PAST-3SG PRT all wine-ACC ‘He drank ALL THE WINE.’

This fact strongly suggests that n-words in focus are not interpreted as universal quantifiers. This is unexpected, given our earlier conclusion that the existential quantifier interpretation is unavailable to focussed n-words: the remaining other option would have been the universally quantified interpretation. That n-words in focus should be invariably universally quantified is also questioned by the observation, illustrated in (35b), that n-words in focus can be internal


275

arguments of ‘definiteness effect’ verbs, i.e. verbs requiring their internal argument to be non-specific. If the conjecture that n-words in focus are not interpreted as universally quantified is correct, it also explains the fact that a ‘definiteness effect’ verb, which normally does not tolerate an n-word modified by ‘almost’ (cf. (7) above), tolerates a similarly modified n-word up to full acceptability: (43)

(?)

0 Majdnem SENKI nem almost nobody-NOM not ‘Almost nobody arrived exactly on

0

e´rkezett arrive-PAST-3SG time.’

PONT exactly

BEN IDO time-in

The bottom line of the foregoing discussion appears paradoxical: n-words in focus are interpreted neither as existential quantifiers nor as universal quantifiers. This evidently raises the question what interpretation focussed n-words in fact have. A resolution of this issue that I offer is based on the proposal above that the lexical ambiguity of n-words is one between a universal quantifier and a non-quantificational indefinite reading. Accepting that universal quantifiers cannot be in syntactic focus in Hungarian (cf. (42)), I suggest that n-words in focus are of the non-quantificational (Heimian) indefinite variety. The particular treatment that I lay out in the remainder of this chapter in section 4 is grounded on the assumption that when in focus the Heimian indefinite variety of n-words receives a scalar interpretation due to a scalar focus particle ‘even’. The view that the existential reading of n-words derives from a Heimian bare indefinite interpretation finds suggestive support in the following pattern: (44)

a. Nem not ‘I didn’t b. *Nem not ‘I didn’t c. Nem not ‘I didn’t

´ırtam write-PAST-1SG write anything.’ ´ırtam write-PAST-1SG write a letter.’ ´ırtam write-PAST-1SG write a letter.’

semmit nothing-ACC egy levelet a letter-ACC levelet letter-ACC

While an n-word can co-occur with a definiteness effect verb postverbally (44a), the same is disallowed for an indefinite DP (with an indefinite article) (44b). At the same time, bare nominal indefinites are just as fine as n-words (44c). Bare nominals are usually treated as being interpreted as contributing a variable and a predicative restriction (cf. van Geenhoven 1995, 1998, Farkas and de Swart 2003, Chung and Ladusaw 2004, and references cited there). Since I have assumed n-words be on a par with bare nominals in this regard, the fact that they pattern together in (44) above is expected.


276 4.

N-WORDS AND FOCUS

In order to understand how syntactic focussing can affect the interpretation of nwords in Hungarian, we need to examine the semantics of indefinite n-words and NPI-s in a little more detail. 4.1.

Scalarity and focus

The interpretation of indefinite n-words/NPI-s has often been likened to that of minimizers, expressions like the ones below. (45)

a. Egy cseppet sem e´rdekli a drop-ACC SEM interest-3SG ‘It doesn’t interest Peter a bit.’ b. Egy fille´rt sem ko¨lto¨tt a penny-ACC SEM spend-PAST-3SG ‘(S)he didn’t spend a penny.’

Pe´tert P-ACC el PRT

Minimizers are polarity elements which denote a minimal quantity or extent. According to Horn (1989: 400), when these elements ‘occur in negative contexts, the negation denotes the absence of a minimal quantity, and hence the presence of no quantity al all.’ Hence, in negative contexts, they act as a means of negative reinforcement (cf. Vallduvi 1994). Significantly, minimizers in Hungarian are normally modified by the sem particle. As pointed out above, sem is historically a morphological combination of is ‘also’/‘even’ plus nem ‘not’, the negation particle. Analogously to what is as argued by Vallduvi (1994) for Catalan, sem contributes ‘not even’ to the interpretation of minimizers: it holds for not even the minimal quantity x (e.g. csepp ‘bit’, fille´r ‘penny’) that p. Hungarian then realizes overtly in the sem scalar additive particle what Fauconnier (1975a, 1975b) argues to be an implicit ‘even’ built into the semantics of minimizers; see also Heim (1984). N-words can optionally be modified by the sem particle (see Note 3). I contend that a modified indefinite n-word in focus is interpreted in much the same way as minimizers. It has been proposed that NPI-s/indefinite n-words are interpreted as the conventionalized extreme element of a scale for which the given property is most likely to hold, cf. Fauconnier (ibid.), who claims English any as well to mark an endpoint on a contextually derived pragmatic scale; see also Lee and Horn (1994) (according to whom any-phrases are combinations of indefinites þ even) and Krifka (1995) for more recent proposals along the same lines, and also Lahiri (1995, 1998) for an essentially similar view of Hindi.27 Hindi (among other languages, e.g. Japanese and Korean) morphologically overtly combine a scalar particle ‘even’ with an indefinite or indeterminate pronoun in their paradigm of NPI-s. Krifka (1995) considers NPI pronouns to denote general predicates, e.g. ‘thing’ or ‘person’. In focus, then, the bottom element is identified as the element for which the property does not hold, and by implicature, the property also fails to hold for


277

all elements higher on the scale (in the case of senki ‘nobody’, it fails to hold for all sets of persons). Hungarian (similarly to Hindi, Japanese, Korean, etc.) appears to grammaticalize the scalar (i.e., alternative set invoking) nature of n-words in exhibiting an n-word paradigm with an overtly appended scalar additive particle. The option of syntactic focussing itself is just another facet of such grammaticalization. The scalar implicature is directly and overtly triggered by sem ‘even’, but it can be generated in the absence of sem as well when the (unmodified) n-word is in syntactic focus. In this latter case, however, the focussed n-word is perceived as more marked. Thus, in fact (46a) is slightly more marked than (46b) with a modified n-word. (46)

0 nem a. SEMMIT not nothing-ACC ‘I didn’t find anything.’ b. SEMMIT SEM nothing-ACC SEM ‘I didn’t find anything.’

0

tala´ltam find-PAST-1SG

0

tala´ltam find-PAST-1SG

This markedness difference is even more pronounced with minimizers: in (47) below the scalar particle sem is missing, hence it involves some extra processing to generate the appropriate implicature. (47)

a. Egy FILLE´RT 0nem 0ko¨lto¨tt not spend-PAST-3SG a penny-ACC ‘(S)he didn’t spend a penny.’ 0 ´T nem 0szo´ltam b. Egy SZO not say-PAST-1SG a word-ACC ‘I didn’t say a word to Peter.’

el PRT Pe´terhez P.-to

That the analogue between minimizers and indefinite n-words in Hungarian is correct (and that n-words involve a separate ‘even’ component) is confirmed by examples from nineteenth century Hungarian like (48). Here the pseudo-object nword itself functions as a minimizer (and in present-day Hungarian it would be replaced by the minimizer kicsit sem, lit. little-ACC SEM ‘not a bit’). The ‘even’ and the ‘not’ components are separate free morphemes: (48)

Az istento´´l pedig mi semmit is nem the God-from TOP-PRT we nothing-ACC even not ‘And as for God, we are not afraid of him a bit.’ [Szila´gyi, 1889, Chapt. 5]

fe´lu¨nk be.afraid-1PL


278

Further, some minimizers allow the suppression of their indefinite article in presentday Hungarian, once again suggestive of a close link between minimizers and bare indefinite n-words: (49)

0 ´ nem 0volt SZO be-PAST-3SG word-NOM not ‘It wasn’t even mentioned.’

4.2.

ro´la about.it

Interpretation in syntactic focus

Let me now characterize the semantic interpretation of sentences with an n-word in the syntactic focus position more explicitly. Kartunnen and Peters (1979) argued that ‘even’ scopes over clausemate negation in cases like Sam doesn’t know even Italian, cf. [even [not [Sam know [F Italian]]], where the scalar focus particle ‘even’ is associated with ‘Italian’ as focus. I will assume the same for the LF of sentences with a focussed n-word: [even [F n-word] [not [ . . . ]] (cf. Wilkinson 1996, Guerzoni 2002, 2004). The focus associated with ‘even’ is the n-word itself, i.e. the extremely general (therefore in the context extremely likely) predicate: focus interpretation lambda-abstracts only this predicate (P in (38a) below), leaving the individual variable behind, where it is safely bound under existential closure. In other words, the analysis of n-words in the focus slot is one of predicative focus: what is in semantic focus is the predicate part of the n-word. A sentence like (50) is interpreted as represented in (51a) (the representation here is given in a somewhat simplified form, and is formulated in a structured meanings approach to focus, but nothing hinges on this choice). The assertion itself is provided in (51c) and the relevant presupposition (a conjunction of an existential and a scalar presupposition) is spelled out in (51d).28 (50)

(51)

0 0 SEMMIT nem tala´lt Jańos not find-PAST-3SG J.-NOM nothing-ACC ‘John didn’t find anything’ a. k even k h k thing¢ k, k ‚P.[:$x.(P(x) found¢(j(x)))] k i b. >‚x.found¢(j)(x) ¼ { hP,Qi | It’s more likely that there is an x that is a P such that John found x, than that there is a y that is a Q such that John found y} c. :$x.(thing¢(x) found¢(j(x))) d. $Q(Q 6¼ thing¢ thing¢ > Q :$y(Q(y) found¢(j(y))) 8Q(Q 6¼ thing¢ ! thing¢ > Q)

(51a) can be paraphrased informally as ‘Even for the (most general and therefore most likely) predicate thing¢ it is not the case that there is something that is (a) thing¢ and that John found it’. According to the presupposition (51d) the following holds: there exist(s) (an)other predicate(s) Q such that it is less likely for John to find a Q than to find a thing¢, and for this/these Q predicate(s) it is also not the case that there is a Q that John found; and in fact thing¢ is the most likely predicate (the upper bound) on the scale defined by the relation given informally in (51b).29


279

Interesting corroboration of the present analysis in terms of a Heimian indefinite treatment of focussed n-words comes from a somewhat archaic construction involving bare wh-pronouns instead of n-words, with an equivalent interpretation. Wh-pronouns function as bare Heimian indefinites in this construction (and according to Lipta´k (2001), in various others in Hungarian). (52)

a. Mit sem haszna´lt what-ACC sem use-PAST-3SG ‘It had no effect whatsoever.’ b. Mit sem va´ltoztat a teńyeken what-ACC sem change-3SG the facts-on ‘It does not change the facts at all.’

This construction is equivalent to corresponding variants with an n-word in place of the bare wh-pronoun precisely because in focus, n-words, just like bare wh-pronouns, are interpreted as pure indefinites.30 Note that what I have argued is that when n-words are in syntactic focus, they are interpreted as containing an ‘even’ element. It appears that when not in syntactic focus, they do not have an interpreted ‘even’ component. The ‘even’ component is morphosyntactically realised as the sem particle. As pointed out above, when the n-word is in syntactic focus, the sem particle can be absent, however, in such cases the sentence becomes more marked since ‘even’ is interpreted but is not morphosyntactically expressed (cf. 46a); and the same is true of minimizers containing the sem particle (cf. 47). The ‘even’ component of meaning is clearly not present on the universal quantifier reading of n-words (indeed the positive polarity counterpart is ‘even’ cannot modify positive universal quantifiers like mindenki ‘everybody’: *mindenki is). In the postverbal field, the appearance of the sem particle with indefinite n-words is completely optional, without any clearly perceivable semantic inference: (53)

Nem tala´ltam semmit (sem) not find-PAST-1SG nothing-ACC SEM ‘I didn’t find anything.’

It appears that the ‘even’ component of meaning is faded from n-words (even in the overt presence of the sem particle), and it is only activated in the focus position. Sem in minimizers is preserved and interpreted across the board when not in focus: their sem particle is obligatory. (54)

Nem ko¨lto¨tt egy fille´rt *(sem) el not spend-PAST-3SG a penny-ACC SEM PRT ‘(S)he didn’t spend a penny.

[cf. (47a)]


280 4.3.

Results in focus reconsidered

With the above interpretation in mind, let us re-examine the behaviour of n-words in focus in light of the results of section 3. Since n-words in focus are interpreted neither as universal nor as existential quantifiers, the paradoxes highlighted in section 3.3 dissolve. Significantly, it is not problematic that n-words can be in syntactic focus while positive universals cannot, as n-words in focus are not universals. This is also the reason why existential presupposition is not necessarily generated by focussed n-words. On the other hand, no scope reversal of negation and existential quantification (‘$ > :’) is generated either, given that n-words in focus are not existentially quantified above negation (existential closure is below negation). It was shown in section 3.2 (cf. (31)) that focussed n-words can be modified by ‘almost’. As pointed out in section 2.1 (cf. Note 7), ‘almost’-modification is basically a diagnostic for end-of-scale (or exact) values. If the two available alternatives are analysing n-words as universal quantifiers or as existential indefinites in the scope of negation, then this diagnostic would suggest that n-words are universals. The reason why it cannot modify NPI existential indefinite nwords in the scope of negation is that ‘almost’ cannot appear in the immediate scope of negation; see Horn (2000) for detailed discussion of this point. As we have seen, however, when functioning as identificational focus, n-words receive an interpretation where their bare (nominal) predicate component is set in semantic focus. Given that bare n-word predicative restrictions are taken to be end-ofscale values when in focus triggering an appropriate scalar implicature, they can be modified by ‘almost’. The fact that ‘almost’ here does not modify a universal QP, but an end-of-scale predicative element, might be responsible for the perceived difference in acceptability (‘almost’-modification of n-words in focus is slightly degraded.)31 ‘Almost’-modification was seen not to render a focussed n-word incompatible with a verb that selects for a non-specific argument, cf. (44). This can be also understood in terms of the present analysis: since n-words in focus modified by ‘almost’ are not universals, but are indefinites, they are expected to be appropriate arguments for ‘definiteness effect’ verbs (cf. also (34)). Modification by ‘at all’ fits the picture in the same manner: ‘at all’ can modify indefinites, hence it can modify the indefinite nword of focus as well.32 Finally, apparent ‘split’ readings with focussed n-words, just as with postverbal n-words, involve existential closure in the scope of the modal operator (‘: > modal > $’): focussing set only the predicative restriction part of the nword in semantic focus (modulo ‘even’), the scope of existential closure is unaltered. In effect, no existential presupposition (here: specific set of employees) is necessary (although the sentence is of course compatible with a discourse involving a given set of employees). In sum, the mysteries encountered in the behaviour of n-words in focus can be resolved in terms of the proposed semantic analysis.


5.

281

CONCLUDING REMARKS

I have argued in this chapter that n-words in Hungarian are ambiguous between a (polarity sensitive) universal quantifier and non-quantificational indefinite interpretation. This is the ambiguity that is involved in the specificity contrast in the pair of examples in (1). Significantly, this result confirms the prediction of Giannakidou (2000: 518) that there must be languages that realize both universal and existential negation with their n-words.33 Hungarian has an existential paradigm of NPI-s (made up of the morpheme valaplus bare wh-indefinites) which are licensed in a range of averidical environments, including the scope of negation. However, there appears to be an anti-locality requirement: vala-NPI cannot be clausemate to negation. This pattern is also found with Slavic languages. If n-words also have an indefinite (existential) interpretation, as I have argued here, then we can understand this as a blocking effect: local existential reading can be expressed by n-words, hence vala-NPI-s are blocked locally to negation. I also demonstrated that n-words in Hungarian can be syntactically focussed, due to the scalar ‘even’ particle overtly appearing in negative n-words paradigm. The interpretation of n-words in focus was argued to crucially depend on the non-quantificational indefinite analysis of the existential reading of n-words. If correct, these results make two significant theoretical points. (i) The two major mechanisms proposed for Negative Concord structures cross-linguistically are both needed in a single language, therefore both are necessary in a typology of NC. (ii) The issues of negativity and quantificationality are independent: both non-quantificational indefinite and quantificational n-words can be logically nonnegative. NOTES

1

2

3

This chapter has greatly benefited from interactions of various kinds with a number of people, including Anastasia Giannakidou, Istvań Kenesei, Katalin E´. Kiss, Bill Ladusaw, Chris Pinõń, Genoveva Puska´s, Tanya Reinhart, Eric Reuland, and Maaike Schoorlemmer, whom I would like to thank for helpful comments, encouragement and criticism. The usual disclaimers, of course, apply. This work was partially supported by project grant no. TS 040 705 of OTKA (Hungarian Scientific Research Fund and grant no. D-048454 of OTKA (Hungarian Scientific Research Fund.) Specificity here, as in chapter 2, is to be understood in the sense of En¸c (1991), i.e., as involving a subset relation to a set in the domain of discourse. Negative Concord (NC) is commonly understood as a descriptive cover term to designate constructions in which there appear several elements each of which is apparently capable of licensing sentential negation interpretation on its own. The term ‘N-word’ is due to Laka (1990). N-words come in two flavours in Hungarian: they may or may not be modified by the particle sem. For the sake of simplicity of presentation, this variation will be ignored for the most part of this chapter (but we will return to the significance of the sem particle in section 4); illustrations will employ the unmodified variety (the presentation of the syntactic distribution of n-words above too is based on unmodified n-words, which exhibit what has become known as ‘negative doubling’ (Giannakidou’s (1998) ‘strict NC’) found also in Slavic and Greek). The related issue of the negativity of Hungarian nwords will also be pushed into the background here; I am assuming Hungarian n-words not to involve

282

4

5

6 7

8


logical negation in their lexical semantics. For differing views on this matter, see Puska´s (2002) and Surańyi (2002a, b, 2006). In the latter works the issue is shown to be closely linked to the question of the distribution of the sem particle; the discussion of that topic here would lead us too far afield. The same options are obtainable when negation precedes a preverbal focus expression: in that case, N-words can be raised past the preverbal focus to end up preceding both focus and negation, or stay postverbal. This latter option requires stress on the postverbal n-word. The similarity extends to their prosody as well; see section 2.8 below. As discussed in section 3, the distributional (and prosodic) symmetry is not complete. Several of the diagnostics to be applied are based on Giannakidou’s (2000) work. Almost can modify determiners with a precise value (e.g. I could solve almost half/50 of the problems), NPI any-NPs in the protasis of (possibly implicit) conditionals, or in the scope of negation if negation is sufficiently far. Horn (2000) argues that what bars almost is in fact scopally adjacent negation. Given that, as also emphasized in Giannakidou (2000), the almost test remains a suitable diagnostics for the case at hand. Except in special, so-called telescopic contexts, as in (Roberts 1989: 717): (i) Each degree candidate walked to the stage. He took his diploma from the Dean and returned to his seat.

9

As (13a) illustrates, n-words can appear in existential constructions, even with the negative existential verb nincs ‘not be’: (i) Nincs itthon semmi not.be-3SG at.home nothing ‘There isn’t anything at home.’

10

11

12

13

14

15

16

Puska´s (2000) also considers this test, but makes note of a caveat, claiming that it cannot be applied to Hungarian. See Surańyi (2002a) for an argument that it can. Puska´s also notes that Hungarian has predicative uses of n-words, but as pointed out in Surańyi (to appear, n. 21), her example is inappropriate to show this. Hunyadi (1981) notes that the morpheme is ‘also/even’ in turn historically derives from the conjunction es ‘and’ (corresponding to Greek ke). This latter form in fact survives today in some dialects. Such modification by an ‘also/even’ element is found in a number of languages. For instance, Hindi NPI-s, which are licensed in non-veridical contexts (incl. long-distance) as well as generic environments (in which contexts they are prone to an indefinite analysis), are also modified by a particle bhii ‘also/even’ (Lahiri 1998). ‘Incorporation’ here is meant as a cover term for the syntactic position of elements that occupy and immediately preverbal position in neutral clauses, often termed the VM position (cf. e.g. Komlo´sy 1994). This position is identified as PredP in Chapter 9 by E´. Kiss. Incorporated nominals may be modified by adjectives, but crucially, cannot have a determiner. For details on Hungarian incorporation, see Farkas and de Swart (2003) and references therein. According to Farkas and de Swart (2003) this is explained by the fact that the bare plural is required to be interpreted within the same minimal DRS box as its predicate, however, a generic operator is accompanied by a box-splitting operation. For van Geenhoven (1998), incorporation triggers type shift of the predicate to a complex predicate type involving the introduction of an existential quantifier. Lawlike statements like All trespassers will be prosecuted, which are conditional in nature, are exceptional in this regard (cf. Stawson 1952, Moravcsik 1991). The de dicto reading of examples in (21) (involving a ‘modal > :’ scope relation) is not available in Hungarian on a Negative Concord interpretation, given that negation is overtly above the modal verb, hence the ‘:modal’ scope relation is fixed. The lexical integrity problem pointed out by Geurts (1996) and de Swart (1996) does not arise with the Hungarian ‘split’ readings, since negation in the construction at hand is a lexeme that is realised overtly and (lexically) independently of the n-word.

P R E D I C AT E S , N E G AT I V E Q U A N T I F I E R S A N D F O C U S 17

18

19

20

283

Descriptive richness is also a condition on topicalisability of universal quantifiers in Greek, and nwords in both Greek and Italian (otherwise the universal quantifier apparently picks up a set too large for a discourse referent). The relevance of such predicates in testing was pointed out to me by Katalin E´. Kiss. It must be noted that apparently, not all collective predicates (e.g. scatter, make a good team, gather) behave completely uniformly in their interaction with n-word arguments. As discussed by Szabolcsi (1997), ‘surround’ is actually ambiguous: estates, for instance, can surround castles in concentric circles too, in which case a distributive reading can be obtained: Every estate surrounds the castle. However, such concentric surrounding is unfeasible with persons, who can only collectively surround a building. In fact, whether a predicate is interpreted as distributive or collective (or ambiguous) is in part determined by pragmatic knowledge. N-words can precede left-peripheral topics only in case they are syntactically topicalised themselves (a rather restricted option); see section 2.7 above. In Szabolcsi (1997), the quantifier field corresponds to the recursive DistP in (i) below, where (recursive) RefP houses topics, and FocP houses fronted foci. Positionally, RefP corresponds to Rizzi’s (1997) TopP projection. (i) [RefP* (topic field) [DistP* (quantifier field) [FocP (focus) [ . . . ]]]

21

22

Puska´s (1998, 2000) analyses preverbal n-words as being focussed (multiple preverbal n-words are multiply adjoined to [Spec,FocP]); however, no convincing argument is offered. Olsvay (2000a) notes two seemingly relevant facts: (i) a preverbal n-word can be modified by a relative clause, and (ii) a certain stylistically marked pattern is possible for n-words (n-word PRT nem V). The crucial point is that neither option is available to preverbal focus. These facts, however, only indicate that there exist nonfocussed preverbal instances of n-words, not that all instances of are non-focussed. This is illustrated by examples such as (i): ´ LTALA ´ N SEMMIT (i) ?EGYA at.all nothing-ACC ‘I didn’t do anything at all.’

23

sem SEM

/

nem not

csina´ltam do-PAST-1SG

Representing the de re reading with an existentially quantified n-word would involve negation scoping over the existential quantifier, which in turn would take the modal in its scope: ‘: $ modal’. This scopal representation is not feasible, since the preverbal surface order is ‘n-word negation modal’, and we have taken Hungarian n-words involved here not to carry logical negation; logical negation is contributed by the negation particle (see Note 3, as well as the references cited there). Note also that in case we find ambiguity between the split reading and a reading involving existentially presupposed set of individuals, this may be explained (i) as an ambiguity between the de re reading with universal quantification outscoping negation and an interpretation involving existential quantification in the scope of negation, or (ii) as vagueness of the existential quantification in the scope of negation reading in neither requiring, nor disallowing the presence of a given set of individuals in the discourse (that is, being able to be either specific or non-specific; see (i) below for illustrations of the latter), or (iii) a combination of (i) and (ii), i.e. as an ambiguity between the de re reading involving a universal over negation and a vague existential-in-the-scope-of-negation interpretation. As will become clear below, postverbal n-words represent case (iii), and focussed n-words represent case (ii) (i.e., focussed n-words are never universally quantified). (i) a. SENKI nem e´rkezett meg arrive-PAST-3SG PRT nobody-NOM not ‘Nobody has arrived.’ b. He wrote five books altogether. If you read any book by him, you’ll know what he’s like.

24

That n-words can be ambiguous between a negative and a non-negative interpretation within a single language has also been entertained; see van der Wouden and Zwarts (1993) and Zwarts (1997) for a


284

25

26

27

28

configurational/contextual ambiguity treatment, and Longobardi (1991), Herburger (2001) and Giannakidou (2005) for suggestions of lexical ambiguity. The particular choice for the interpretation of non-quantificational (Heimian) indefinites contributing a predicative restriction and a free individual variable is immaterial for the present purposes. This restriction is not limited to Hungarian; see, for instance, Rizzi (1997). Note that syntactic focus in Hungarian is identificational, not informational or ‘new-focus’, cf. E´.Kiss (1998b). According to Lahiri (1998), NPI-s in Hindi are inherently focussed. This might be the case for Hungarian n-words modified by sem generally, although this type of focus is clearly different from the one for which Hungarian reserves the immediately preverbal position, and which is characterized by identification and exclusion, a type of contrastive focus (on different kinds of focus, see E´.Kiss (1998b)). Tsimpli and Roussou (1996) suggest that Greek n-words (which are invariably emphatic) are instances of focus. Watanabe (2004) proposes that n-words check a focus feature universally. As for the (universal) quantification over properties in the second line, the set of predicates to be quantified over is to be understood (as usual with ‘even’) as appropriately restricted (e.g. only entities that can be meaningfully found should be relevant). Note that on the approach to the interpretation of focus presented in Chapter 9 (by E´. Kiss), an exhaustive identificational focus expression serves as a specificational predicate: it provides a set that characterises the other set that is denoted by the logical subject of the predication involving the focussed phrase as a predicate. A non-exhaustive immediately preverbal constituent is treated as predicational. The present analysis can also be cast in this approach, where n-words in the focus position, qua indefinite expressions, can function as either of these (the relevant predicate serving as either predicational or specificational focus being thing¢). Recall from chapter 9 that specificational foci allow for the cancellation of the exhaustivity that is implied by the specificational focus interpretation. Similarly, on its specificational (in another terminology, identificational) interpretation, focussed n-words also allow for cancellation of the exhaustivity implication: (i) SENKI nem jo¨tt el, csak Pe´ter nobody-NOM not come-PAST-3SG PRT only P.-NOM / de Pe´ter igen / but P.-NOM yes ‘Nobody came along, except Peter/but Peter did.’

29

In essence, the proposed interpretation of n-words in focus renders them analogous to other bare indefinites in focus which represent endpoints of a scale, for instance the idiomatic example in (i). A paraphrase of (i) would be ‘Even for the lowest predicate on the pragmatically relevant scale DOG it is not the case that John considers Peter to be DOG.’ (i) Jańos kutya´ba sem veszi dog-in SEM take-3SG J.-NOM ~‘John doesn’t (even) take notice of Peter.’

30

Pe´tert P.-ACC

Note that the interpretation for sentences with an n-word in focus is not a third type of interpretation; it arises as a regular result of focussing a predicative indefinite, which in turn is one of the two readings of n-words in Hungarian. Such wh-pronouns modified by sem must be syntactically focussed. A non-focussed, postverbal occurrence is unacceptable: (i) *Nem not

haszna´lt use-PAST-3SG

mit sem what-ACC SEM

[cf. (52a)]

Another construction where n-words are interchangeable with bare wh-indefinites is illustrated in (ii).


285

(ii) Nem tudok / lehet mit csina´lni not can-1SG / may what-ACC do-INF ‘I can’t do anything./Nothing can be done.’ 31

An alternative would be to admit ambiguity even in the focus slot: besides the n-word-predicate-infocus interpretation, one would allow universally quantified n-words to be syntactically focused. Recall that this option was dismissed above in view of the fact that (positive polarity) universal QPs are not syntactically focussable in Hungarian (cf. (42)). One could argue that n-words are negative polarity counterparts of such universal QPs, and this may be responsible for the fact that they, in contrast to positive polarity universal QPs, can be syntactically focussed. This in turn could explain why almostmodification is possible in focus (albeit not why it is slightly degraded). The fact (to be discussed immediately below) that a focused n-word is compatible with definiteness effect verbs would also be explicable. It is a general observation that definite NPs can combine with definiteness effect verbs when in focus, cf. (i). (Note that universal quantifiers are nevertheless uniformly incompatible with definiteness effect verbs in Hungarian. ADO ´ e´rkezett (i) A SZINTAXIS ELO TEGNAP arrive-PAST-3SG yesterday the syntax lecturer-NOM ‘It’s the SYNTAX LECTURER that arrived YESTERDAY.’

32

As noted in section 3.2, Note 22, for some speakers such examples are slightly degraded. This degradation can be explained in terms of a redundancy that exists in such sentences: by placing the n-word in semantic focus qua an extreme element of a scale (operated on by the (overt or covert) ‘even’ focus particle), the widening effect of ‘at all’ is redundant. A similar oddity deriving from an analogous redundancy is witnessed in English with minimizers when ‘even’ is overt: (i) ?He didn’t even spend a penny at all

33

The quantificational ambiguity uncovered here may also be made sense of in the context of Jespersen’s cycle. Jespersen’s description of the historical development of various languages revealed that in many cases n-words enter the language as NPI-s, and over time, potentially going through several stages, they shift to semantically negative expressions (i.e., expressions carrying logical negation) (cf. Jespersen 1917), typically also acquiring quantificational force of their own. Therefore, from this perspective it is not so surprising after all to find languages which in their present stage simultaneously exhibit both the quantificational and the non-quantificational n-word varieties.

REFERENCES

Ackerman F (1984) Verb modifiers as argument taking predicates: complex verbs as predicate complexes in Hungarian. Groninger Arbeiten zur Germanistischen Linguistik 25: 23–71 Acquaviva P (1993) The logical form of negation. A study of operator–variable structures in syntax. Ph.D. dissertation, Scuola Normale Superiore, Pisa (Published by Garland, 1997) Acquaviva P (1997) The logical form of Negative Concord. In: Corblin F, Godard D, Marandin JM (eds) Empirical issues in formal syntax and semantics. Selected papers from the Colloque de Syntaxe et de Se´mantique de Paris, CSSP 1995. pp 13–41 Alberti G (1997) Restrictions on the degree of referentiality of arguments in Hungarian. Acta Linguistica Hungarica 44: 341–362 Alberti G (1998) On passivization in Hungarian. In: Groot C de, Kenesei I (eds) Approaches to Hungarian 6. JATE, Szeged, pp 103–121 Alberti G (2004) Climbing for aspect: With no rucksack. In: E´. Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam, pp 253–290 Bach E (1981) On time, tense and aspect: An essay in English metaphysics. In: Cole P (ed) Radical pragmatics. Academic Press, New York, pp 63–81 Bach E (1986) The algebra of events. Linguistics and Philosophy 9: 5–16 Baker M (1985) The Mirror Principle and morphosyntactic explanation. Linguistic Inquiry 16: 373– 415 Baker M (1988) Incorporation. A theory of grammatical function changing. University of Chicago Press, Chicago Bańhidi Z (1941) A magyar nyelv o¨sszetett igealakjai [Complex verb forms in Hungarian]. A Magyar Nyelvtudomańyi Ta´rsasa´g Kiadvańyai 63, Budapest Bartos H (1999) Morfoszintaxis e´s interpretaćio´: a magyar inflexio´s jelense´gek szintaktikai ha´ttere [Morphosyntax and interpretation: The syntactic background of inflectional phenomena in Hungarian]. PhD dissertation. ELTE, Budapest Bartos H (2000) Az inflexio´s jelense´gek szintaktikai ha´ttere [The syntactic background of inflectional phenomena] In: Kiefer F (ed) Struktura´lis magyar nyelvtan 3. Morfolo´gia [Structural Grammar of Hungarian 3. Morphology]. Akade´miai Kiado´, Budapest, pp 653–761 Beghelli F, Stowell T (1994) The direction of quantifier movement. GLOW Newsletter 32: 56–57 Beghelli F, Stowell T (1995) Distributivity and negation. Ms, University of California, Los Angeles Beghelli F, Stowell T (1997) The syntax of distributivity and negation. In: Szabolcsi A (ed) Ways of scope taking. Kluwer, Dordrecht, pp 71–108 Behaghel Otto (1932) Deutsche Syntax IV. Carl Winters, Heidelberg ´ (1995) Prefixation and discourse. In: Kenesei I (ed) Approaches to Hungarian 5. JATE, Bende-Farkas A Szeged, pp 192–220 ´ (2001) Verb-object dependencies in Hungarian and English: a DRT-based approach. PhD Bende-Farkas A Dissertation, University of Stuttgart ´ rpa´d-kor magyar nyelvu˝ szo¨vegemle´kei [Hungarian-language texts from the age of Benko ´´ L (1980) Az A the Arpads (897–1301)]. Akade´miai Kiado´, Budapest Bennett M, Partee B (1978) Toward the logic of tense and aspect in English. Indiana University Linguistics Club, Bloomington Bennett M, Partee B (1972) Toward the logic of tense and aspect in English. Santa Monica, California: System Development Corporation; reprinted with an Afterword by Indiana University Linguistics Club

287

288

REFERENCES

Bloomington, 1978. Reprinted in Partee, Barbara H. 2004. Compositionality in Formal Semantics: Selected Papers by Barbara H. Partee. Oxford: Blackwell Publishing, pp 59–109. Bertinetto PM (2001) On a frequent misunderstanding in the temporal–aspectual domain: The ‘perfective– telic confusion’. In: Cecchetto C, Chierchia G, Guasti MT (eds) Semantic interfaces. CSLI, Stanford, pp 177–210 Bobaljik JD, Branigan P (2006) Eccentric agreement and multiple case checking. In Johns A, Massam D, Ndayiragije J (eds) Ergativity. Springer, Dordrecht Bohnmeyer J, Swift M (in press) Default aspect: The semantic interaction of aspectual viewpoint and telicity. To appear in: Hout A van, Swart H de, Verkuyl H (eds) Perspectives on aspect. UiLOTS Working Papers, Utrecht Borik O (2002) Aspect and reference time. Ph.D. dissertation, Utrecht University Borik O, Reinhart T (2004) Telicity and perfectivity: two independent systems. In: Hunyadi L, Ra´kosi G, To´th E (eds) LoLa. The Eighth Symposium on Logic and Language: Preliminary papers. The Organizing Committee of LoLa, Debrecen, pp 12–33 Brassai S (1863–65) A magyar mondat [The Hungarian sentence]. Magyar Akade´miai E´rtesı´to ´´. A Nyelve´s Sze´ptudomańyi Oszta´ly Ko¨zlo¨nye 3, pp 3–128, 173–409 Brody M (1990) Some remarks on the focus field in Hungarian. UCL Working Papers in Linguistics 2. University College London, London, pp 201–225 Brody M (1995) Focus and checking theory. In: Kenesei I (ed) Approaches to Hungarian 5, JATE, Szeged, pp 29–44 Brody M (2004) ‘‘Roll-up’’ structures and morphological words. In: E´ Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam, pp 147–172 Chierchia G (1995) Dynamics of meaning: Anaphora, presupposition, and the theory of grammar. The University of Chicago Press, Chicago Chierchia G (2002) Scalar implicatures, polarity phenomena, and the syntax/pragmatics interface. In: Belletti A (ed) Structures and beyond. Oxford University Press, Oxford Chierchia G (2004) A semantics for unaccusatives and its syntactic consequences. In: Alexiadou A, Anagnostopoulou E, Everaert M (eds) The unaccusativity puzzle. Oxford University Press, Oxford, pp 22–59 Chomsky N (2000) Minimalist inquiries: The framework. In: Michaels D, Martin R, Uriagereka J (eds) Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. MIT Press, Cambridge, MA, pp 1–59 Chomsky N (2001a) Beyond explanatory adequacy. MITOPL 20. MITWPL, Cambridge, MA Chomsky N (2001b) Derivation by Phase. In: Kenstowicz M (ed) Ken Hale. A Life in Language. MIT Press, Cambridge, MA Chung S, Ladusaw W (2004) Restriction and saturation. MIT Press, Cambridge, MA Cinque G (1990) Types of A’-depedencies. MIT Press, Cambridge, MA Cinque G (1999) Adverbs and functional heads. Oxford University Press, New York–Oxford Comrie B (1976) Aspect. An introduction to the study of verbal aspect and related problems. Cambridge University Press, Cambridge Croft WA (1986) Surface subject choice of mental verbs. Paper presented at the Annual Meeting of the Linguistic Society of America, New York Croft WA (1991) Syntactic categories and grammatical relations. University of Chicago Press, Chicago Csirmaz A (2004) Particles and phonologically defective predicates. In: E´ Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam, pp 225–252 ¨ (1970) Some notes on indefinites. Language 46: 33–41 Dahl O ¨ (1985) Tense and aspect systems. Blackwell, London Dahl O Deme L (1959) A nyomate´ktalan mondat egy fajta´ja´ro´l. Az ott hata´rozo´szo´ igeko¨to ´´szeru˝ haszna´lata [On a type of non-emphatic sentences. The particle-like use of the adverb ott]. Magyar Nyelv 55: 185–198

REFERENCES

289

Demirdache H, Uribe-Etxebarria M (2000) The primitives of temporal relations. In Martin R, Michaels D, Uriagereka J (eds) Step by Step. Essays in honor of Howard Lasnik. MIT Press, Cambridge, MA, pp 157–185 Demirdache H, Uribe-Etxebarria M (2004) The syntax of time adverbs. In: Gue´ron J, Lecarme J (eds) The syntax of time. MIT Press, Cambridge, MA, pp 143–180 Depraetere I (1995) On the necessity of distinguishing between (un)boundedness and (a)telicity. Linguistics and Philosophy 18: 1–19 Dikken M den (1995) Particles. Oxford University Press, Oxford Dikken M den (2004) Agreement and ‘clause union’. In: E´ Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam, pp 445–498 Dowty DR (1979) Word meaning and Montague grammar. Reidel, Dordrecht Dowty DR (1986) The effect of aspectual classes on the temporal structure of discourse: semantics or pragmatics. Linguistics and Philosophy 9: 37–61 Dowty DR (1987) Collective predicates, distributive predicates and all. In: Marshal et al. (eds) Proceedings of the 3rd Eastern States Conference on Linguistics (ESCOL’86). Ohio State University, Columbus Dowty DR (1991) Thematic proto-roles and argument selection. Language 67: 547–619 Dowty DR (1994) The role of negative polarity and concord marking in natural language reasoning. In: Harvey M, Santelmann L (eds) Proceedings from Semantics and Linguistic Theory IV. Cornell University, Ithaca, pp 114–144 E Abaffy E (1992) Az igemo´d- e´s igeido ´´rendszer [The system of verbal mood and tense]. In: Benko ´´ L (ed) A magyar nyelv to¨rteńeti nyelvtana II/1. [A historical grammar of Hungarian II/1.]. Akade´miai Kiado´, Budapest, pp 120–183 E´ Kiss K (1987) Configurationality in Hungarian. Dordrecht: Reidel E´ Kiss K (1992) Az egyszeru˝ mondat szerkezete [The structure of the simple clause]. In: Kiefer F (ed) Struktura´lis magyar nyelvtan I. Mondattan [A structural grammar of Hungarian I. Syntax]. Akade´miai Kiado´, Budapest, pp 79–179 E´ Kiss K (1993) Wh-movement and specificity. Natural Language and Linguistic Theory 11: 85–112 E´ Kiss K (1994) Sentence structure and word order. In: Kiefer F, E´ Kiss K (eds) The Syntactic Structure of Hungarian. Syntax and Semantics 27. Academic Press, San Diego–New York, pp 1–90 E´ Kiss K (1995) The Definiteness Effect revisited. In: Kenesei I (ed) Approaches to Hungarian 5. JATE, Szeged, pp 63–88 ´ j magyar nyelvtan [New E´ Kiss K (1998a) Mondattan [Syntax]. In: E´ Kiss K, Kiefer F, Sipta´r P: U Hungarian Grammar]. Osiris, Budapest, pp 17–184 E´ Kiss K (1998b) Identificational focus versus information focus. Language 74: 245–273 E´ Kiss K (2001) Focused number phrases. In: Fe´ry C, Sternefeld W (eds) Audiatur Vox Sapientiae. A Festschrift for Arnim von Stechow. Studia grammatica 52. Akademie Verlag, Berlin, pp 259–266 E´ Kiss K (2002a) Hungarian syntax. Cambridge University Press, Cambridge E´ Kiss K (2002b) Negation and specificity. In: Kenesei I, Sipta´r P (eds) Approaches to Hungarian 8. Akade´miai Kiado´, Budapest, pp 39–60 E´ Kiss K (2003) Focusing as predication. Ms, to appear in: Molna´r V, Winkler S (eds) The architecture of focus. Mouton de Gruyter, Berlin E´ Kiss K (2004a) Egy igeko¨to´élme´let va´zlata [Outlines of a theory of verbal particles]. Magyar Nyelv 100: 15–42 E´ Kiss K (2004b) Parallel strategies of verbal complex formation in Hungarian and West-Germanic In: E Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam, pp 335–358 E´ Kiss K (2005a) Event types and discourse linking in Hungarian. Linguistics 43: 131–154 E´ Kiss K (2005b) First steps towards a theory of the verbal particle. In: Pinõn C, Sipta´r P (eds) Approaches to Hungarian 9. Kluwer Akade´miai Kiado´, Budapest, pp 57–88 E´ Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam Enç M (1991) The semantics of specificity. Linguistic Inquiry 22: 1–25

290

REFERENCES

Engelberg S (2002) Intransitive accomplishments and the lexicon: The role of implicit arguments, definiteness, and reflexivity in aspectual composition. Journal of Semantics 19: 369–416 Ernst T (2002) The syntax of adjuncts. Cambridge University Press, Cambridge Erteschik-Shir N (1997) The dynamics of focus structure. Cambridge Studies in Linguistics 84. Cambridge University Press, Cambridge Farkas D (1986) On the syntactic position of focus in Hungarian. Natural Language and Linguistic Theory 4: 77–96 Farkas D, Sadock JM (1989) Preverb climbing in Hungarian. Language 65: 318–338 Farkas D, Swart H de (2003) The semantics of incorporation: From argument structure to discourse transparency. CSLI Publications, Stanford, CA Fauconnier G (1975a) Pragmatic scales and logical structures. Linguistic Inquiry 6: 353–357 Fauconnier G (1975b) Polarity and the scale principle. In: Grossman RE, San LJ, Vance TJ (eds) Proceedings of the 11th meeting of the Chicago Linguistic Society. CLS, Chicago, pp 188–199 Filip H (2000) The quantization puzzle. In: Tenny C, Pustejovsky J (eds) Events as grammatical objects. CSLI Publications, Stanford, CA, pp 36–96 Fintel K von (1999) NPI licensing, Strawson-entailment, and context-dependency. Journal of Semantics 16: 97–148 Fintel K von, Iatridou S (2002) Since since. Ms, MIT Geenhoven V van (1995) Semantic incorporation: A uniform semantics for West Greenlandic noun incorporation and West Germanic bare plural configurations. In: Dainora A, Hemphill R, Luka B, Need B, Pargman S (eds) Papers from the 31st Meeting of the Chicago Linguistics Society. CLS, Chicago, pp 171–186 Geenhoven V van (1998) Semantic incorporation and indefinite descriptions. CSLI Publications, Stanford, CA Geurts B (1996) On ‘‘no’’. Journal of Semantics 13: 67–86 Giannakidou A (1997) The landscape of polarity items. Ph.D. dissertation, University of Groningen Giannakidou A (1998) Polarity sensitivity as (non)veridical dependency. John Benjamins, Amsterdam Giannakidou A (2000) Negative . . . concord. Natural Language and Linguistic Theory 18: 457–523 Giannakidou A (2005) N-words and negative concord. In: Everaert M, Riemsdijk H van (eds) The Blackwell Companion to Syntax. Blackwell, Oxford Giannakidou A, Quer J (1995) Two mechanisms for the licensing of negative indefinites. In: Gabriele L, Hardison D, Westmoreland R (eds) Formal Linguistics Society of Mid-America (FLSM) 6. Indiana University Linguistics Club (IULC), Bloomington, pp 103–114 Giannakidou A, Quer J (1997) Long-distance licensing of negative indefinites. In: Forget D et al. (eds) Negation and Polarity: Syntax and Semantics. John Benjamins, Amsterdam, pp 95–113 Gimba AM (2000) Bole verb morphology. Ph.D. dissertation, UCLA Goldberg A (1991) It can’t go up the chimney down: Paths and the English resultative. Berkeley Linguistics Society 17, pp 368–378 Grimshaw J, Vikner S (1993) Obligatory adjuncts and the structure of events. In: Reuland E, Abraham W (eds) Knowledge and language, vol 2. Lexical and conceptual structure. Kluwer, Dordrecht, pp 143–155 Guerzoni E (2002) Even-NPIs in questions. In: Hirotani M (ed) Proceedings of the North East Linguistic Society 32, vol 1, pp 153–170 Guerzoni E (2004) Even-NPI’s in yes/no questions. Natural Language Semantics 12: 319–34 Haader L (2004) A mu¨ncheni emle´k [The Mu¨nchen relic]. Magyar Nyelv 101: 161–178 Haegeman L (1995) The syntax of negation. Cambridge University Press, Cambridge Haegeman L, Zanuttini R (1991) Negative heads and the Neg-criterion. The Linguistic Review 8: 233–251 Hale K, Keyser SJ (2002) Prolegomenon to a theory of argument structure. MIT Press, Cambridge, MA Harlig J (1989) The interaction of verbal aspect and noun phrase determination in Hungarian. Ph.D. dissertation, The University of Chicago Hartmann K (2002) Focus constructions in Hausa (with some remarks on Hdi and Miya). Paper presented at the Symposium on Information Structure, Lund, Dec. 2002 Hatav G (1997) The semantics of aspect and modality: Evidence from English and Biblical Hebrew. John Benjamins, Amsterdam

REFERENCES

291

Hay J, Kennedy C, Levin B (1999) Scalar structure underlies telicity in ‘‘degree achievements.’’ In: Matthews T, Strolovitch D (eds) SALT IX, CLC Publications, Ithaca, NY, pp 127–144 Heim I (1982) The semantics of definite and indefinite NPs. PhD dissertation, University of Massachusetts, Amherst Heim I (1984) A note on negative polarity and downward entailingness. In: Proceedings of NELS 14, pp 98–107 Heinrich G (ed) (1897) Heltai Ga´spa´r esopusi meseí [The tales of Ga´spa´r Heltai]. Re´gi Magyar Ko¨nyvta´r 4. Franklin Ta´rsulat, Budapest Herburger E (2001) The Negative Concord puzzle revisited. Natural Language Semantics 9: 289–333 Higgins RF (1973) The pseudo-cleft construction in English. Ph.D. dissertation, MIT, Cambridge, MA Hinrichs E (1985) A compositional semantics for Aktionsarten and NP reference in English. PhD dissertation. Ohio State University Hinrichs E (1986) Temporal anaphora in discourses of English. Linguistics and Philosophy 9: 63–82 Hoekstra T (1988) Small clause results. Lingua 74: 101–139 Hoekstra T (2005) Arguments and structure. Mouton de Gruyter, Berlin Horn L (1972) On the semantic properties of logical operators in English. Ph.D. dissertation, University of California, LA Horn, L (1989) A natural history of negation. University of Chicago Press, Chicago, IL Horn L (2000) ANY and EVER: Free choice and free relatives. In: Proceedings of the 15th Annual Conference of the Israeli Association for Theoretical Linguistics (IATL 15), pp 71–111 Horn L, Lee YS (1995) Diagnosing the diagnostics. Ms, Yale University Hornstein N (1990) As time goes by. Tense and Universal Grammar. MIT Press, Cambridge, MA Horvath J (1986) Focus in the theory of grammar and the syntax of Hungarian. Reidel, Dordrecht Horvath J (2000) Interfaces vs. the computational system in the syntax of focus. In: Bennis, H, Everaert M, Reuland E (eds) Interface strategies. Koninklijke Nederlandse Academie van Wetenschappen, Amsterdam Horvath J (2005) Is ‘focus movement’ driven by stress. In: Pinõn C, Sipta´r P (eds) Approaches to Hungarian 9. Kluwer Akade´miai Kiado´, Budapest, pp 131–158 Hout, A van (1996) Event semantics of verb frame alternations. Tilburg University dissertation, Tilburg Huber S (2000) Es-Clefts und det-Clefts. Zur Syntax, Semantik und Informationsstruktur von Spaltsa¨tzen im Deutschen und Schwedischen. Almquist and Wiksell International, Stockholm Iatridou S, Anagnostopoulou E, Izvorski R (2001) Some observations about the form and meaning of the perfect. In: Kenstowicz M (ed) Ken Hale: A Life in Language. MIT Press, Cambridge, MA, pp 189–238. Reprinted in: Alexiadou A, Rathert M, Stechow A (eds) (2003) Perfect explorations. Mouton de Gruyter, Berlin I Gallasy, M (1991) A ne´velo ´´ e´s ne´velo´´-elo´´zmeńy [The article and its predecessor]. In: Benko ´´ L (ed) A magyar nyelv to¨rteńeti nyelvtana I. [A historical grammar of Hungarian I.]. Akade´miai Kiado´, Budapest, pp 461–475 I Gallasy, M (1992) A ne´velo ´´k [Articles]. In: Benko ´´ L (ed) A magyar nyelv to¨rteńeti nyelvtana II/1. [A historical grammar of Hungarian II/1.]. Akade´miai Kiado´, Budapest, pp 716–770 Israel M (1996) Polarity sensitivity as lexical semantics. Linguistics and Philosophy 19: 619–666 Jacobs J (1991) Negation. In: Stechow A von, Wunderlich D (eds) Semantics: An international handbook of contemporary research. Walter de Gruyter, Berlin, pp 560–59 Jespersen O (1917) Negation in English and other languages. AF Hos, Copenhagen Jespersen O (1931) A modern English grammar on historical principles, IV. Allen and Unwin, London Kadmon N, Landman F (1993) Any. Linguistics and Philosophy 16: 353–422 Ka´lmań L (1995) Definiteness effect verbs in Hungarian. In: Kenesei I (ed) Approaches to Hungarian 5: 192–220. JATE, Szeged, pp 221–242 Ka´lmań L, Varasdi K (2005) The Hungarian definiteness effect hoax. Ms, Linguistics Institute of the Hungarian Academy of Sciences, Budapest Kamp H, Reyle U (1993) From discourse to logic. Kluwer, Dordrecht Karttunen L, Peters S (1979) Conventional implicature. In: Oh C, Dineen D (eds) Presuppositions. Syntax and semantics 11. Academic Press, New York, pp 1–56

292

REFERENCES

Kayne RS (1985) Principles of particle constructions. In: Gue´ron J et al. (eds) Grammatical representation. Foris, Dordrecht, pp 101–140 Kennedy C, Levin B (2002) Telicity corresponds to degree of change (unpublished manuscript) Kiefer F (1992) Az aspektus e´s a mondat szerkezete [Aspect and the structure of the sentence]. In: Kiefer F (ed) Struktura´lis magyar nyelvtan I. Mondattan [A structural grammar of Hungarian I. Syntax]. Akade´miai Kiado´, Budapest, pp 797–886 Kiefer F (1994) Aspect and syntactic structure. In: F Kiefer, K E´ Kiss (eds) The syntactic structure of Hungarian. Syntax and Semantics 27. Academic Press, New York–San Diego, pp 415–463 Kiefer F (2000a) Aspektus, akcio´mino ´´se´g, esemeńyszerkezet [Aspect, Aktionsart, event structure]. In: Jelente´selme´let [Theory of meaning]. Corvina Kiado´, Budapest, pp 275–305 Kiefer F (2000b) Jelente´selme´let [Theory of meaning]. Corvina, Budapest. Kiefer F, Ladańyi M (2000) Az igeko¨to´´k [Verbal particles]. In: Kiefer F (ed) Struktura´lis magyar nyelvtan 3. Morfolo´gia [A structural grammar of Hungarian 3. Morphology]. Akade´miai Kiado´, Budapest, pp 453–518 Kiparsky P (2001) Structural Case in Finnish. Lingua 111: 315–376 Klein W (1992) The present perfect puzzle. Language 68: 525–552 Klein W (1994) Time in language. Routledge, London Komlo´sy A (1994) Complements and adjuncts. In: Kiefer F, E´ Kiss K (eds) The syntactic structure of Hungarian. Syntax and Semantics vol 27. Academic Press, New York–San Diego ´ s [Causativisation]. In: Kiefer F (ed) Struktura´lis magyar nyelvtan 3. Komlo´sy A (2000) A muveltete ˝ Morfolo´gia [Structural Hungarian grammar Vol 3. Morphology] Akade´miai Kiado´, Budapest, pp 215–292 Koopman H, Szabolcsi A (2000) Verbal complexes. MIT Press, Cambridge, MA Koster J (1994) Predicate incorporation and the word order of Dutch. In: Cinque G, Koster J, Pollock J-Y, Rizzi L, Zanuttini R (eds) Paths towards Universal Grammar. Studies in Honor of Richard S. Kayne. Georgetown University Press, Washington DC, pp 255–276 Kratzer A (2004) Telicity and the meaning of objective case. In: Gue´ron J, Lecarme J (eds) The syntax of time. MIT Press, Cambridge, MA, pp 389–423 Krifka M (1989) Nominal reference, temporal constitution, and quantification in event semantics. In: Bartsch R, Benthem J van, Emde Boas P van (eds) Semantics and Contextual Expressions. Foris, Dordrecht, pp 75–115 Krifka M (1992) Thematic relations as links between nominal reference and temporal constitution. In: Sag I, Szabolcsi A (eds) Lexical matters. CSLI, Stanford, CA, 29–53 Krifka M (1993) Focus and presupposition in dynamic interpretation. Journal of Semantics 10: 269–300 Krifka M (1995) The semantics and pragmatics of polarity items. Linguistic Analysis 25: 209–257 Krifka M (1998) The origins of telicity. In: Rothstein S (ed) Events and grammar. Reidel, Dordrecht, pp 197–235 Kuroda SY (1972) The categorical and the thetic judgment (evidence from Japanese). Foundations of Language 2: 153–185 Ladusaw W (1992) Expressing negation. In: Baker C, Dowty D (eds) Proceedings of the Second Conference on Semantics and Linguistic Theory (SALT II), pp 237–259 Ladusaw W (1994) Thetic and categorial, stage and individual, weak and strong. In: Proceedings of the Fourth Conference on Semantics and Linguistic Theory (SALT IV). Cornell University, Ithaca, pp 220–229 Lahiri U (1995) Negative polarity in Hindi. In: Simons M, Galloway T (eds) Proceedings of the Fifth Conference on Semantics and Linguistic Theory. Cornell University, Ithaca, pp 168–185 Lahiri U (1998) Focus and negative polarity in Hindi. Natural Language Semantics 6: 57–123 Laka I (1990) Negation in syntax: on the nature of functional categories and projections. Ph.D. dissertation, MIT Lakoff G (1970) Irregularity in syntax. Holt, Rinehart and Winston, New York Larson RK (1988a) On the double object construction. Linguistic Inquiry 19: 335–392 Larson RK (1988b) Light predicate raising. Lexicon Project Working Papers 27. MIT Press, Cambridge, MA Lee YS, Horn L (1994) Any as indefinite even. Ms, Yale University

REFERENCES

293

Levin B, Rappaport Hovav M (1995) Unaccusativity: At the syntax – lexical semantics interface. MIT Press, Cambridge, MA Lin J (2004a) Fine-grained lexical semantic representations and compositionally-derived events in Mandarin Chinese. In: Proceedings of the HLT/NAACL 2004 Workshop on Computational Lexical Semantics, May 2004, Boston Lin J (2004b) Event structure and the encoding of arguments: The syntax of Mandarin and English verb phrases. PhD dissertation, MIT, Cambridge MA Linebarger M (1987) Negative polarity and grammatical representation. Linguistics and Philosophy 10: 325–387 Lipta´k A (2001) On the syntax of wh-items in Hungarian. Ph.D. dissertation, Universiteit Leiden Longobardi G (1991) In defense of the Correspondence Hypothesis: Island effects and parasitic constructions in Logical Form. In: Huang C-T J, May R (eds) Logical Structure and Linguistic Structure. Kluwer, Dordrecht, pp 149–198 Maleczki M (1995) On the definiteness effect in Hungarian (a semantic approach). In: Kenesei I (ed) Approaches to Hungarian 5. JATE, Szeged, pp 261–284 Maleczki M (2001) Indefinite arguments in Hungarian. In: Kenesei I (ed) Argument structure in Hungarian. Akade´miai Kiado´, Budapest, pp 157–199 Maleczki, M (2005) A hata´rozottsa´gi korla´toza´s [The definiteness restriction]. Ms, University of Szeged Maling J (1989) Adverbials and structural case in Korean. In: Kuno S et al. (eds) Harvard Studies in Korean Linguistics III, pp 297–308 Marantz A (1984) On the nature of grammatical relations. MIT Press, Cambridge, MA Massam D (1990) Cognate objects and thematic objects. Canadian Journal of Linguistics 35:161–190 McNally L (1998) Existential sentences without existential quantification. Linguistics and Philosophy 21: 353–392 Mittwoch A (1998) Cognate objects as reflections of Davidsonian arguments. In: Rothstein S (ed) Events and Grammar. Kluwer, Dordrecht, pp 309–332 Moens M, Steedman M (1988) Temporal ontology and temporal reference. Computational Linguistics 14: 15–28 Moltmann F (1991) Measure adverbials. Linguistics and Philosophy 14: 629–660 Morzycki M (2001) Almost and its kin, across categories. In: Hastings R, Jackson B, Zvolenszky Z (eds) Proceedings of SALT XI. CLC Publications, Ithaca, NY, pp 306–325 Muskens R (1996) Combining Montague semantics and discourse representation. Linguistics and Philosophy 19: 143–186 Olsen MB (1997) A semantic and pragmatic model of lexical and grammatical aspect. Garland, New York Olsvay C (2000a) A syntactic analysis of negative universal quantifiers in Hungarian. MA thesis. Eo¨tvo¨s Lorańd University, Theoretical Linguistics Program, Budapest Olsvay C (2000b) Forma´lis jegyek egyeztete´se a magyar nemsemleges mondatokban [The agreement of formal features in non-neutral Hungarian sentences]. In: Bu¨ky L, Maleczki M (eds) A mai magyar nyelv leı´ra´sańak u´jabb mo´dszerei 4. JATE, Szeged, pp 119–151 Olsvay C (2004) The Hungarian verbal complex: An alternative approach. In: E´ Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam, pp 290–333 Perrot J (1966) Adale´kok a meg igeko¨to´´ funkcio´jańak vizsga´lata´hoz a mai magyar nyelvben [Contribution to the analysis of the function of the verbal particle meg in present day Hungarian]. Nyelvtudomańyi E´rtekeze´sek 52. Akade´miai Kiado´, Budapest Pesetsky D, Torrego E (2001) T-to-C movement: Causes and consequences. In: Kenstowicz M (ed) Ken Hale: A life in language. MIT Press, Cambridge, MA, pp 355–426 Pesti G (1536/1950) Ezo´pus fabulaí [The fables of Aesop]. Ko¨zoktata´su¨gyi Kiado´, Budapest Pinõń C (1995) Around the progressive in Hungarian. In: Kenesei I (ed) Approaches to Hungarian V. JATE, Szeged, pp 153–189 Pinõń C (2001a) A finer look at the causative-inchoative alternation. Proceedings of SALT XI, Hastings R – Jackson B – Zvolenszky Z (eds) CLC Publications, Cornell University, Ithaca, NY Pinõń C (2001b) To¨prengtem egyet: azon, hogy mit jelent az egyet [I wondered about the meaning of one´ jabb tanulmańyok a struktura´lis magyar ´ -Nagy M, Bańre´ti Z, E´ Kiss K (eds) U ACC]. In: Bakro

294

REFERENCES

nyelvtan e´s a nyelvto¨rteńet ko¨re´bo ´´l [New studies on structural Hungarian grammar and the diachrony of Hungarian]. Osiris, Budapest Puska´s G (1998) On the Neg-criterion in Hungarian. Acta Linguistica Hungarica 45: 167–213 Puska´s G (2000) Word order in Hungarian: The syntax of A-bar positions. John Benjamins, Amsterdam Puska´s G (2002) On negative licensing contexts and the role of n-words. In: Kenesei I, Sipta´r P (eds) Approaches to Hungarian 8. Akade´miai Kiado´, Budapest, pp 81–106 Ramchand GC (1997) Aspect and predication. Clarendon Press, Oxford Rapp I, Stechow A von (1999) Fast ‘almost’ and the visibility parameter for functional adverbs. Journal of Semantics 16:149–204 Rappaport Hovav M, Levin B (2001) An event structure account of English resultatives. Language 77:766–797 Reichenbach H (1947) Elements of symbolic logic. Macmillan, London Reinhart T (1995) Interface strategies. OTS Working Papers OTS-WP-TL-95002, Utrecht Reuland E (2000) Agreement and Case: Explaining Burzio’s Generalization. John Benjamins, Amsterdam Richter F, Salier M (1998) LF constraints on expression of Ty2: An HPSG analysis of negative concord in Polish. In: Borsley R et al (eds) Slavic in HPSG. CSLI Publications, Stanford, pp 247–282 Ritter E, Rosen ST (1998) Delimiting events in syntax. In Butt M, Geuder W (eds) The projection of arguments. CSLI, Stanford, CA, pp 135–164 Rizzi L (1997) The fine structure of the left periphery. In: Haegeman L (ed) Elements of grammar. Kluwer, Dordrecht, pp 281–337 Rizzi L (2002) Locality and the left perihpery. In: Belletti A (ed) Structures and beyond. The cartography of syntactic structures. Oxford University Press, Oxford, vol. 3 Roberts C (1989) Modal subordination and pronominal anaphora in discourse. Ph.D. dissertation, University of Massachusetts, Amherst Rosen C (1984) The interface between semantic roles and initial grammatical relations. In: Perlmutter D, Rosen C (eds) Studies in relational grammar 2. University of Chicago Press, Chicago, pp 38–77 Rothstein S (1985) The syntactic forms of predication. Ph.D. dissertation. MIT, Cambridge, MA Rullmann H (1995) Geen eenheid. Tabu 25: 194–197 Ryle G (1949) The concept of mind. Barnes and Noble, London Simpson J (1983) Resultatives. In Levin L, Rappaport M, Zaenen A (eds) Papers in Lexical – Functional Grammar. Indiana University Linguistics Club, Bloomington, pp 143–158 Schu¨tze C (1997) INFL in child and adult language: Agreement, case and licensing. Ph.D. dissertation, MIT Smith C (1991) The parameter of aspect. Kluwer, Dordrecht (2nd ed. 1997) Stechow A von (1993) Die Aufgaben der Syntax. In: Stechow A von, Sternefeld W, Vennemann T (eds) Ein internationales Handbuch zeitgeno¨ssischer Forschung, vol 91. de Gruyter, Berlin, pp 1–88 Stechow A von (2002) Temporal prepositional phrases with quantifiers: Some additions to Pratt and Francez. Ms, Universita¨t Tu¨bingen Stechow A von (2004) Schritte zur Satzsemantik. Ms, Universita¨t Tu¨bingen Strawson PF (1952) Introduction to logical theory. Methuen, London Strawson PF (1971) Identifying reference and truth values. In: Steinberg DD, Jakobovits LA (eds) Semantics. Cambridge University Press, Cambridge, pp 86–114 Surańyi B (2002a) Multiple operator movements in Hungarian. LOT, Utrecht Surańyi B (2002b) Negation and an ambiguity account of n-words: the case of Hungarian. In: Kenesei I, Sipta´r P (eds) Approaches to Hungarian 8. Akade´miai Kiado´, Budapest, pp 107–132 Surańyi B (2004) The left periphery and cyclic spell-out: the case of Hungarian. In: Adger D, Cat C de, Tsoulas G (eds) Peripheries. Syntactic edges and their effects. Studies in Natural Language and Linguistic Theory. Kluwer, Dordrecht, pp 49–73 Surańyi B (2006) Quantification and focus in negative concord. Lingua 116:272–313 Svenonius P (2002) Case is uninterpretable aspect. In: Proceedings of Perspectives on Aspect. University of Utrecht, Utrecht. Also available at http://www.uilots.let.uu.nl/conferences/PerspectivesonAspect/ PoAprocpapers.html

REFERENCES

295

Swart H de (1996) Scope ambiguities with negative quantifiers. In: Heusinger K von, Egli U (eds) Proceedings of the Workshop on Reference and Anaphoric Relations. Universita¨t Konstanz, Konstanz, pp 145–164 Swart H de (1998) Aspect shift and coercion. Natural Language and Linguistic Theory 16:347–385 Swart H de (2000) Tense, aspect and coercion from a cross-linguistic perspective. In: Butt M, King TH (eds) Proceedings of the Berkeley Formal Grammar Conference. University of California, Berkeley Swart H de, Molendijk A (1999) Negation and the temporal structure of narrative discourse. Journal of Semantics 16: 1–42 Szabolcsi A (1981) The semantics of Topic-Focus articulation. In: Groenendijk J et al. (eds) Formal methods in the study of language. Matematisch Centrum, Amsterdam Szabolcsi A (1986) From the definiteness effect to lexical integrity. In: Abraham W, Meij S de (eds) Topic, Focus, and configurationality. John Benjamins, Amsterdam, pp 321–348 Szabolcsi A (1992) A birtokos szerkezet e´s az egzisztencia´lis mondat [Possessive constructions and existential sentences]. Akade´miai Kiado´, Budapest Szabolcsi A (1997) Strategies for scope taking. In: Szabolcsi A (ed) Ways of scope taking. Kluwer, Dordrecht, pp 109–155 Szendro ´í K (2003) A stress-based approach to the syntax of Hungarian focus. The Linguistic Review 20: 37–78 Szendro ´í K (2004) A stress-based approach to climbing. In: E´ Kiss K, Riemsdijk H van (eds) Verb clusters. A study of Hungarian, German and Dutch. John Benjamins, Amsterdam, pp 205–225 Szent Margit e´lete [Saint Margaret’s life] (1510/1990) Re´gi magyar ko´dexek 10. MTA Nyelvtudomańyi Inte´zet, Budapest Szili K (2001) A perfektivita´s mibenle´te´ro ´´l a magyar nyelvben a meg igeko¨to ´´ funkcioí kapcsań [On the nature of perfectivity in Hungarian, with reference to the functions of meg.] Magyar Nyelv 97: 263–282 Talmy L (1985) Lexicalization patterns: Semantic structure in lexical forms. In: Shopen T (ed) Language typology and syntactic description, vol. 3 Grammatical categories and the lexicon. Cambridge University Press, Cambridge Tenny C (1994) Aspectual roles and the syntax – semantics interface. Kluwer, Dordrecht Thompson E (1996) The syntax of tense. Ph.D. dissertation, University of Maryland To´th I (1999) Negative polarity item licensing in Hungarian. Acta Linguistica Hungarica 46: 119–142 Tsimpli IM, Roussou A (1996) Negation and polarity items in Greek. The Linguistic Review 13: 49–81 Ura H (2000) Checking Theory and Grammatical Functions. Oxford University Press, Oxford ¨ ro¨gdi B (2003) Feature doubling, aspectual structure and expletives. In: Kadowaki M, Kawahara S (eds) U Proceedings of NELS 33. GLSA, Amherst Vainikka A (1989) Deriving syntactic representations in Finnish. Ph.D. dissertation, University of Massachusetts at Amherst Vallduvi E (1994) Polarity items, n-words, and minimizers in Catalan and Spanish. Research paper HCRC/ RP 55, HCRC Publications, University of Edinburgh Vendler Z (1967) Verbs and times. In: Linguistics in Philosophy. Cornell University Press, Ithaca Verkuyl HJ (1972) On the compositional nature of the aspects. Reidel, Dordrecht Verkuyl HJ (1993) A theory of aspectuality. Cambridge University Press, Cambridge Wacha B (1976) Az igeaspektusro´l [On verbal aspect]. Magyar Nyelv 72: 59–69 Wacha B (1978) Az ige´s-szerkezeti aspektus katego´ria´ja a magyar nyelvben [The category of verbal aspect in Hungarian]. MA thesis, University of Budapest Wacha B (1989) A folyamatos–nem-folyamatos szembena´lla´sro´l [On the continuous–non-continuous ´ ltalańos Nyelve´szeti Tanulmańyok 17: 279–328 opposition]. A Watanabe, A (2004) The genesis of negative concord. Linguistic Inquiry 35: 559–612 Webelhuth G (1992) Principles and parameters of syntactic saturation. Oxford University Press, New York/ Oxford Wechsler S, Lee Y-S (1996) The domain of direct case assignment. Natural Language and Linguistic Theory 14: 629–664.

296

REFERENCES

Wedgwood D (2002) ‘‘Focus position’’ and quantification: the relevance of pragmatic theory. In: Kenesei I, Sipta´r P (eds) Approaches to Hungarian 8. Kluwer Akade´miai Kiado´, Budapest, pp 241–261 Wilkinson K (1996) The scope of even. Natural Language Semantics 4: 193–215 Williams E (1980) Predication. Linguistic Inquiry 11: 210–238 Winkler S (1996) Focus and secondary predication. Walter de Gruyter, Berlin Woolford E (2003) Burzio’s generalization, markedness, and locality constraints on nominative objects. In: Brandner E, Zinsmeister H (eds) New perspectives on case theory. CSLI, Stanford, pp 301–329 Wouden T van der, Zwarts F (1993) A semantic analysis of negative concord. In: Proceedings of SALT 3. Cornell University, Ithaca Yip M, Maling J, Jackendoff R (1997) Case in tiers. Language 63: 217–250 Zaenen A (1993) Unaccusativity in Dutch: Integrating syntax and lexical semantics. In Pustejovsky J (ed) Semantics and the lexicon. Kluwer, Dordrecht, pp 129–161 Zanuttini R (1991) Syntactic properties of sentential negation: a comparative study of Romance languages. Ph.D. dissertation, University of Pennsylvania Zanuttini R (1994) Re-examining negative clauses. In: Cinque G, Koster J, Pollock J-Y, Rizzi L, Zanuttini R (eds) Paths towarlds Universal Grammar. Georgetown University Press, Washington DC, 427–452 Zucchi A (1991) Negation and aspect. ESCOL 90, pp 344–363 Zwarts F (1996) A hierarchy of negative expressions. In: Wansing H (ed) Negation: A notion in focus. Walter de Gruyter, Berlin, pp 169–194 Zwarts F (1997) Three types of polarity. In: Hamm F, Hinrichs E (eds) Plurality and quantification. Kluwer, Dordrecht, pp. 177–237

I N D E X WO R D S TO E V E N T S T R U C T U R E A N D T H E L E F T P E R I P H E RY. S T U D I E S O N H U N G A R I A N

accomplishment 2, 12, 19, 27, 40, bare nominal 7, 9, 23, 35, 42, 43, 44, 58–61, 68, 75, 91–106, 108, 129, 146, 45, 46, 48, 49, 51, 52, 55, 147, 149–152, 154, 156, 235 154n10, 201, 202, 215, 218, 219, accusative case 159–199, 248, 250 221, 222, 275 achievement 2, 18, 21, 27, 58–61, 66, 68, 69, 75, 76, 108, 124–126, 129, change of state 2, 3, 10, 17–26, 29, 31, 134, 145, 146, 149–154, 171, 32, 34, 37, 69–73, 149 179, 235 mental change-of-state verbs 70 adverbial modification 14, 127n3, 164, Croft’s generalisation 61–64 176, 225, 230–233, 237, 240–242, cumulative reference 65 244, 245, 247, 250, 252n13 adverbial participle 154, 156 definiteness effect agreement ~verbs 75–90, 106n6, 262, 275, object~141, 193n1 280, 285n31 subject~9, 141 degrees 60, 67, 68, 70, 71 argument structure 13, 62, 83, 165, 172, delimiting 41, 51, 62, 63, 126, 153, 181, 182, 187, 188 155, 156, 157, 169, 171, 179, article 4, 21, 24, 53, 144, 147, 275 180, 185 aspect difference value 65–68 neutral~107, 124, 126 discourse referent 82, 84–89, 90n12, ongoing~235 232n17 situation~4, 6, 37, 40–42, 52, Dutch 49, 121, 162 58–60, 107, 110–116, 118, 119, dynamic 65, 82–89, 90n13, 94, 108, 121, 126, 128n10, 129–131, 143, 112, 113, 127n1/6, 188, 259 147, 148, 150, 233, 246, 247 viewpoint 3, 4, 37–42, 58, 59, English 3, 47, 58–61, 63, 64, 90n15, 91, 107, 108, 111, 112, 114–121, 92, 114, 115, 135, 138, 157n26, 159, 123–126, 127n8, 130, 131, 136, 160, 162, 163, 166, 168, 173, 176, 144, 147, 148, 150, 152, 153, 187, 191, 193n2, 202, 204, 214, 228, 155, 156, 226 231, 245, 246, 251n1, 258, 259, 263, aspectualizer 119, 120, 123 268, 274, 276 atelic(icity) 3, 4, 5, 18, 23, 27, 43, 58, event structure 1, 2, 3, 5, 6, 10, 59, 63, 65, 68, 109, 110, 114, 115, 127n3/4, 72, 160, 164, 166, 171, 192, 250 129, 143, 144, 146, 149, 151, 152, event time 109, 111, 112, 114, 123, 126, 156, 172, 177–179, 185, 188, 197n30, 127n2/8, 128n10/19, 133, 134, 135, 225, 229, 234, 247, 251n11 175, 176, 177, 184, 225, 225–31, 233, 297

298

INDEX

234, 241, 242, 243, 246, 247, 248, 250, 251n3/7/10, 252n13/15 eventuality description modifiers 59 exhaustive, exhaustivity, exhaustive identification 1, 42, 43, 45, 82, 202, 207, 208, 210–16, 284n28 existential closure 272, 278, 280 fake reflexive 160–163, 165, 166, 168, 169, 172, 173, 176 focus identificational~201, 203, 215, 216, 217, 219, 222, 280, 284n28 ~movement 54n15/16/17, 201, 204, 208, 209, 210, 221 framing effect 113, 115, 116 German 106n9/12, 128n19, 146, 148, 162, 215, 263 Germanic 17, 165, 193n3 grammaticalization 130, 148, 277 habitual interpretation 117, 235, 243 Heimian indefinite 256, 257, 272, 273, 275, 279, 284n25 imperfective/imperfectivity 4, 5, 9, 37–41, 43, 51, 111–128, 131, 133– 135, 138–140, 145–156, 226–228, 230, 232, 235, 251n3, 252n12 incremental theme 2, 29, 30, 97, 196n23

n-word 255–281 negation 205, 206, 218, 219, 222, 225–253 negative concord 7, 10, 223n2, 256, 257, 281n2, 282n16 negative polarity item 89, 90n7/16, 245, 257, 260 novelty 84 perfective/perfectivity 4, 5, 9, 17, 23, 37–41, 43, 76, 111–126, 127n8, 128n19, 132, 135, 139, 144, 146, 148–151, 153, 155, 156, 196n23, 226–228, 230, 232, 235, 236, 242 phase (theory) 221 predication 18, 22, 29, 34, 50, 213–216, 220, 222, 223n6 presupposition ~of existence 214, 215, 269, 274, 280 psychological causation 72, 73 psychological verbs (psych verbs) 57, 73 quantificational ambiguity 271, 285n33 quantifier 253 quantized reference 65, 97, 106n8 reference time 111, 112, 123, 132, 133, 135, 136, 139, 175, 225–230, 232–234, 240, 242, 244, 246, 247, 250, 251 resultative 17–22, 24, 27, 28, 37, 41, 110, 130, 160–162

Latin 129 measure phrase 68, 159, 169, 171–174, 176, 177, 179, 184, 192, 194n17/18, 196n22/23, 199n40/41/43, 228 measuring-out 60, 62, 63, 64, 170, 173 Mirror Principle 112, 237, 194n17 multiplicative 159, 167, 169, 170, 171, 177–192, 197n27/28/29/31/34/36/38/ 39/40, 200n49

scales 68 semantics dynamic~259 event~84, 94 situation ~aspect – see aspect ~delimiter 23, 123, 159, 168, 169, 171, 172, 177–180, 182, 185, 186, 188, 191, 192, 197n31

299

INDEX

Slavic 17, 58, 148, 181, 281n3 specific(ity) 18, 21, 22, 23, 25, 27, 29, 31, 32, 36, 53n6/7, 54n12, 77, 157n24, 185, 198n40, 216, 255–257, 270, 280, 281, 283n23 specification(al predicate) 213–216, 219, 223n6, 284n28 state, 41, 58, 65, 69–73, 85, 108, 149, 150, 154, 181, 234, 238, 240, 248 stativity hypothesis 234, 235, 239, 240, 247 Strawson ~entailment 244, 245 ~subinterval property 244, 245, 246 telic(icity) 2–5, 10, 17–19, 21–25, 27, 29, 37–42, 66–69, 109, 110, 114, 115, 118, 121, 129, 130, 143–56, 169–172, 185 temporal adverb 72, 175, 225, 247, 248, 249, 251n2 temporal modification 197 tense past imperfective~124, 131, 133, 134, 138, 140, 145, 147 past perfect~131, 135, 138, 140, 143, 147

present perfect~113, 131, 132, 137, 147 simple past~4, 131, 132, 133, 139, 144, 147 simple present~131, 139 unaccusative verb 11, 18, 150, 165 unergative verb 23, 24, 28, 53n8, 150, 165 universal quantification 79, 262, 272, 274 Vendler 2, 3, 15n1, 18, 58, 108, 179, 198n36, 229 verb (V) ~of appearance 1, 2, 30 ~of coming-into-being 30 ~of creation 2, 32, 66, 95, 146, 198n40 ~of existence 201, 262 ~of spatial configuration 17, 18, 32, 33, 34, 35, 37, 41, 52, 201 verb movement (V-movement) 8, 17, 201, 202, 217, 218, 220, 221, 222, 128n17 viewpoint aspect – see aspect

S T U D I E S I N N AT U R A L L A N G U A G E A N D L I N G U I S T I C T H E O RY Managing Editors Liliane Haegeman, University of Geneva Joan Maling, Brandeis University James McCloskey, University of California, Santa Cruz

Publications 1. L. Burzio: Italian Syntax. A Government-binding Approach. 1986 ISBN Hb 90-277-2014-2; Pb 90-277-2015-0 2. W.D. Davies: Choctaw Verb Agreement and Universal Grammar. 1986 ISBN Hb 90-277-2065-7; Pb 90-277-2142-4 3. K. E´. Kiss: Configurationality in Hungarian. 1987 ISBN Hb 90-277-1907-1; Pb 90-277-2456-3 4. D. Pulleyblank: Tone in Lexical Phonology. 1986 ISBN Hb 90-277-2123-8; Pb 90-277-2124-6 5. L. Hellan and K. K. Christensen: Topics in Scandinavian Syntax. 1986 ISBN Hb 90-277-2166-1; Pb 90-277-2167-X 6. K. P. Mohanan: The Theory of Lexical Phonology. 1986 ISBN Hb 90-277-2226-9; Pb 90-277-2227-7 7. J. L. Aissen: Tzotzil Clause Structure. 1987 ISBN Hb 90-277-2365-6; Pb 90-277-2441-5 8. T. Gunji: Japanese Phrase Structure Grammar. A Unification-based Approach 1987. ISBN 1-55608-020-4 9. W. U. Wurzel: Inflectional Morphology and Naturalness. 1989 ISBN Hb 1-55608-025-5; Pb 1-55608-026-3 10. C. Neidle: The Role of Case in Russian Syntax. 1988 ISBN 1-55608-042-5 11. C. Lefebvre and P. Muysken: Mixed Categories. Nominalizations in Quechua 1988. ISBN Hb 1-55608-050-6; Pb 1-55608-051-4 12. K. Michelson: A Comparative Study of Lake-Iroquoian Accent. 1988 ISBN 1-55608-054-9 13. K. Zagona: Verb Phrase Syntax. A Parametric Study of English and Spanish 1988 ISBN Hb 1-55608-064-6; Pb 1-55608-065-4 14. R. Hendrick: Anaphora in Celtic and Universal Grammar. 1988 ISBN 1-55608-066-2 15. O. Jaeggli and K.J. Safir (eds.): The Null Subject Parameter. 1989 ISBN Hb 1-55608-086-7; Pb 1-55608-087-5 16. H. Lasnik: Essays on Anaphora. 1989 ISBN Hb 1-55608-090-5; Pb 1-55608-091-3

S T U D I E S I N N AT U R A L L A N G U A G E A N D L I N G U I S T I C T H E O RY 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37.

S. Steele: Agreement and Anti-Agreement. A Syntax of Luisenõ. 1990 ISBN 0-7923-0260-5 E. Pearce: Parameters in Old French Syntax. Infinitival Complements. 1990 ISBN Hb 0-7923-0432-2; Pb 0-7923-0433-0 Y. A. Li: Order and Constituency in Mandarin Chinese. 1990 ISBN 0-7923-0500-0 H. Lasnik: Essays on Restrictiveness and Learnability. 1990 ISBN 0-7923-0628-7; Pb 0-7923-0629-5 M. J. Speas: Phrase Structure in Natural Language. 1990 ISBN 0-7923-0755-0; Pb 0-7923-0866-2 H. Haider and K. Netter (eds.): Representation and Derivation in the Theory of Grammar. 1991 ISBN 0-7923-1150-7 J. Simpson: Warlpiri Morpho-Syntax. A Lexicalist Approach. 1991 ISBN 0-7923-1292-9 C. Georgopoulos: Syntactic Variables. Resumptive Pronouns and A’ Binding in Palauan. 1991 ISBN 0-7923-1293-7 K. Leffel and D. Bouchard (eds.): Views on Phrase Structure. 1991 ISBN 0-7923-1295-3 C. Tellier: Licensing Theory and French Parasitic Gaps. 1991 ISBN 0-7923-1311-9; Pb 0-7923-1323-2 S.-Y. Kuroda: Japanese Syntax and Semantics. Collected Papers. 1992 ISBN 0-7923-1390-9; Pb 0-7923-1391-7 I. Roberts: Verbs and Diachronic Syntax. A Comparative History of English and French. 1992 ISBN 0-7923-1705-X A. Fassi Fehri: Issues in the Structure of Arabic Clauses and Words. 1993 ISBN 0-7923-2082-4 M. Bittner: Case, Scope, and Binding. 1994 ISBN 0-7923-2649-0 H. Haider, S. Olsen and S. Vikner (eds.): Studies in Comparative Germanic Syntax. 1995 ISBN 0-7923-3280-6 N. Duffield: Particles and Projections in Irish Syntax. 1995 ISBN 0-7923-3550-3; Pb 0-7923-3674-7 J. Rooryck and L. Zaring (eds.): Phrase Structure and the Lexicon. 1996 ISBN 0-7923-3745-X J. Bayer: Directionality and Logical Form. On the Scope of Focusing Particles and Wh-in-situ. 1996 ISBN 0-7923-3752-2 R. Freidin (ed.): Current Issues in Comparative Grammar. 1996 ISBN 0-7923-3778-6; Pb 0-7923-3779-4 C.-T. J. Huang and Y.-H. A. Li (eds.): New Horizons in Chinese Linguistics 1996 ISBN 0-7923-3867-7; Pb 0-7923-3868-5 A. Watanabe: Case Absorption and WH-Agreement. 1996 ISBN 0-7923-4203-8

S T U D I E S I N N AT U R A L L A N G U A G E A N D L I N G U I S T I C T H E O RY 38. 39. 40. 41. 42.

43. 44. 45. 46. 47. 48. 49. 50. 51. 52.

53. 54.

55. 56.

H. Thraínsson, S.D. Epstein and S. Peter (eds.): Studies in Comparative Germanic Syntax. Volume II. 1996 ISBN 0-7923-4215-1 C. J. W. Zwart: Morphosyntax of Verb Movement. A Minimalist Approach to the Syntax of Dutch. 1997 ISBN 0-7923-4263-1; Pb 0-7923-4264-X T. Siloni: Noun Phrases and Nominalizations. The Syntax of DPs. 1997 ISBN 0-7923-4608-4 B. S. Vance: Syntactic Change in Medieval French. 1997 ISBN 0-7923-4669-6 G. Mu¨ller: Incomplete Category Fronting. A Derivational Approach to Remnant Movement in German. 1998 ISBN 0-7923-4837-0 Studies in Natural Language and Linguistic Theory A. Alexiadou, G. Horrocks and M. Stavrou (eds.): Studies in Greek Syntax 1998 ISBN 0-7923-5290-4 R. Sybesma: The Mandarin VP. 1999 ISBN 0-7923-5462-1 K. Johnson and I. Roberts (eds.): Beyond Principles and Parameters. Essays in Memory of Osvaldo Jaeggli. 1999 ISBN 0-7923-5501-6 R. M. Bhatt: Verb Movement and the Syntax of Kashmiri. 1999 ISBN 0-7923-6033-8 A. Neeleman and F. Weerman: Flexible Syntax. ATheory of Case and Arguments 1999 ISBN 0-7923-6058-3 C. Gerfen: Phonology and Phonetics in Coatzospan Mixtec. 1999 ISBN 0-7923-6034-6 I. Paul, V. Phillips and L. Travis (eds.): Formal Issues in Austronesian Linguistics. 2000 ISBN 0-7923-6068-0 M. Frascarelli: The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. 2000 ISBN 0-7923-6240-3 I. Landau: Elements of Control. Structure and Meaning in Infinitival Constructions. 2000 ISBN 0-7923-6620-4 W. D. Davies and S. Dubinsky (eds.): Objects and other Subjects. Grammatical Functions, Functional Categories and Configurationality. 2001 ISBN 1-4020-0064-2; Pb 1-4020-0065-0 J. Ouhalla and U. Shlonsky (eds.): Themes in Arabic and Hebrew Syntax. 2002 ISBN 1-4020-0536-9; Pb 1-4020-0537-7 E. Haeberli: Features, Categories and the Syntax of A-Positions. Cross-Linguistic Variation in the Germanic Languages. 2002 ISBN 1-4020-0854-6; Pb 1-4020-0855-4 J. McDonough: The Navajo Sound System. 2003 ISBN 1-4020-1351-5; Pb 1-4020-1352-3 D.E. Holt (ed.): Optimality Theory and Language Change. 2003 ISBN 1-4020-1469-4; Pb 1-4020-1470-8

S T U D I E S I N N AT U R A L L A N G U A G E A N D L I N G U I S T I C T H E O RY 57.

58. 59. 60.

61. 62. 63. 64. 65. 66. 67. 68.

J. Camacho: The Structure of Coordination. Conjunction and Agreement Phenomena in Spanish and Other Languages. 2003 ISBN 1-4020-1510-0; Pb 1-4020-1511-9 I. Toivonen: Non-Projecting Words. A Case Study of Swedish Particles. 2003 ISBN 1-4020-1531-3; Pb 1-4020-1532-1 D. Adger, C. de Cat and G. Tsoulas (eds). Peripheries. Syntactic Edges and their Effects. 2004 ISBN 1-4020-1908-4; Pb 1-4020-1909-2 C. Goria: Subject Clitics in the Northern Italian Dialects. A Comparative Study Based on the Minimalist Program and Optimality Theory. 2004 ISBN 1-4020-2736-2; Pb 1-4020-2737-0 V. Dayal and A. Mahajan (eds.): Clause Structure in South Asian Languages 2004 ISBN 1-4020-2717-6; Pb 1-4020-2718-4 P. Kempchinsky and R. Slabakova (eds.): Aspectual Inquiries. 2005 ISBN 1-4020-3032-0; Pb 1-4020-3035-5 M. Arad: Roots and Patterns. Hebrew Morpho-syntax. 2005 ISBN 1-4020-3243-9; Pb 1-4020-3245-5 ˇ P. Stekauer and R. Lieber (eds.): Handbook of Word-Formation. 2005 ISBN 1-4020-3595-0; Pb 1-4020-3597-7 A. Johns, D. Massam and J. Ndayiragije (eds.): Ergativity. Emerging Issues ISBN Hb 1-4020-4186-1; Pb 1-4020-4187-X 2006 F. Lee: Remnant Raising and VSO Clausal Architecture. A Case Study from ISBN 1-4020-4300-7 San Lucas Quiavini Zapotec. 2006 O. Misˇeska Tomic´: The Balkan Sprachbund Morpho-syntactic Features. 2006 ISBN 1-4020-4487-9 ´ Kiss: Event Structure and the Left Periphery. Studies on Katalin E. ISBN 978-1-4020-4753-4; Pb 978-1-4020-4754-1 Hungarian. 2008

Event Structure and the Left Periphery: Studies on Hungarian

Multinationals on the Periphery

Multinationals on the Periphery

ON HUNGARIAN MORPHOLOGY

On Hungarian morphology (Linguistica)

Datives And Other Cases: Between Argument Structure And Event Structure (Studies in Languages Companion, SLCS 75)

Thunder on the Left

Centre and Periphery: Comparative Studies in Archaeology

Some remarks on subject positions and the architecture of the left periphery in Spanish

The Event Structure of Perception Verbs

Complex Predicates: Cross-linguistic Perspectives on Event Structure

Second Grave on the Left

Second Grave on the Left

Studies on Old High German Syntax: Left sentence periphery, verb placement and verb-second (Linguistik Aktuell Linguistics Today)

Ocular Periphery and Disorders

Islamic Identity and Development: Studies of the Islamic Periphery

Totalitarian Imperialism Reflections On The Hungarian Revolution

The Left and the Jews: The Jews and the Left

Event

The Left Periphery: The Interaction of Syntax, Pragmatics and Prosody in Czech

The Event

Event

Hungarian Language Contact Outside Hungary: Studies on Hungarian as a Minority Language (IMPACT: Studies in Language & Society)

The Hungarian Cookbook

The Hungarian Revolution 1956

On the Periphery of the Periphery: Household Archaeology at Hacienda San Juan Bautista Tabi, Yucatán, Mexico

Arguments and Structure: Studies on the Architecture of the Sentence (Studies in Generative Grammar)

Nation-Building and Contested Identities: Romanian and Hungarian Case Studies

Lexical Semantics, Syntax, and Event Structure (Oxford Studies in Theoretical Linguistics)

Shakespeare and the Theatrical Event

Event Structure and the Left Periphery: Studies on Hungarian

Multinationals on the Periphery

Multinationals on the Periphery

ON HUNGARIAN MORPHOLOGY

On Hungarian morphology (Linguistica)

Datives And Other Cases: Between Argument Structure And Event Structure (Studies in Languages Companion, SLCS 75)

Thunder on the Left

Centre and Periphery: Comparative Studies in Archaeology

Some remarks on subject positions and the architecture of the left periphery in Spanish

The Event Structure of Perception Verbs

Complex Predicates: Cross-linguistic Perspectives on Event Structure

Second Grave on the Left

Second Grave on the Left

Studies on Old High German Syntax: Left sentence periphery, verb placement and verb-second (Linguistik Aktuell Linguistics Today)

Ocular Periphery and Disorders

Islamic Identity and Development: Studies of the Islamic Periphery

Totalitarian Imperialism Reflections On The Hungarian Revolution

The Left and the Jews: The Jews and the Left

Event

The Left Periphery: The Interaction of Syntax, Pragmatics and Prosody in Czech

The Event

Event

Hungarian Language Contact Outside Hungary: Studies on Hungarian as a Minority Language (IMPACT: Studies in Language & Society)

The Hungarian Cookbook

The Hungarian Revolution 1956

On the Periphery of the Periphery: Household Archaeology at Hacienda San Juan Bautista Tabi, Yucatán, Mexico

Arguments and Structure: Studies on the Architecture of the Sentence (Studies in Generative Grammar)

Nation-Building and Contested Identities: Romanian and Hungarian Case Studies

Lexical Semantics, Syntax, and Event Structure (Oxford Studies in Theoretical Linguistics)

Shakespeare and the Theatrical Event

Recommend Documents