On the Syntactic Composition of Manner and Motion (Linguistic Inquiry Monographs)

50985Zubizarreta

1/9/07

9:55 AM

Page 1

The MIT Press Massachusetts Institute of Technology Cambridge, Massachusetts 02142 http://mitpress.mit.edu 0-262-74029-X 978-0-262-74029-6

Zubizarreta and Oh

Linguistic Inquiry Monograph 48

“Zubizarreta and Oh give us an exemplary study of how to approach crosslinguistic phenomena with explanatory tools. Their work establishes compelling connections between a construction of theoretical significance on the one hand, and the independent properties of syntactic computation and grammatical variations on the other. It is a remarkable contribution to theories of syntax and semantics, with important implications for the study of morphology and language acquisition as well.” —Charles Yang, University of Pennsylvania, author of Knowledge and Learning in Natural Language and The Infinite Gift

On the Syntactic Composition of Manner and Motion

This crosslinguistic study of the structure of motion predicates argues for the universal syntactic nature of the composition of manner and motion within the verbal constituent. In serial verb languages, manner and motion are overtly represented as two distinct morphosyntactic units, sequentially ordered. Zubizarreta and Oh argue that the same analysis into two units holds for nonserial verb languages, albeit at a more abstract level. They argue further that this abstract level is part of the syntactic component of the grammar. The authors support their argument with a wealth of empirical data and a discussion of significant theoretical issues. Unlike many books and articles that discuss the relation between constructional meaning and the lexicon, On the Syntactic Composition of Manner and Motion examines one phenomenon in detail: the articulation of manner and motion, in three distinct language families— Germanic, Korean, and Romance. The authors’ defense of the syntactic approach to constructional meaning will be of interest to linguists and psycholinguists both inside and outside the generative tradition, and to scholars of Romance, Germanic, and Korean languages.

Maria Luisa Zubizarreta is Professor of Linguistics at the University of Southern California. She is the author of Levels of Representation in the Lexicon and in the Syntax and Prosody, Focus, and Word Order (MIT Press, 1998). Eunjeong Oh is a Postdoctoral Fellow in Linguistics at Korea University.

48

On the Syntactic Composition of Manner and Motion Maria Luisa Zubizarreta and Eunjeong Oh

Linguistic Inquiry Monograph Forty-Eight

On the Syntactic Composition of Manner and Motion Maria Luisa Zubizarreta and Eunjeong Oh

On the Syntactic Composition of Manner and Motion

Linguistic Inquiry Monographs Samuel Jay Keyser, general editor 6. 10. 12. 13. 15. 16. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48.

Some Concepts and Consequences of the Theory of Government and Binding, Noam Chomsky On the Nature of Grammatical Relations, Alec Marantz Logical Form: Its Structure and Derivation, Robert May Barriers, Noam Chomsky Japanese Tone Structure, Janet Pierrehumbert and Mary Beckman Relativized Minimality, Luigi Rizzi Argument Structure, Jane Grimshaw Locality: A Theory and Some of Its Empirical Consequences, Maria Rita Manzini Indefinites, Molly Diesing Syntax of Scope, Joseph Aoun and Yen-hui Audrey Li Morphology by Itself: Stems and Inflectional Classes, Mark Arono¤ Thematic Structure in Syntax, Edwin Williams Indices and Identity, Robert Fiengo and Robert May The Antisymmetry of Syntax, Richard S. Kayne Unaccusativity: At the Syntax–Lexical Semantics Interface, Beth Levin and Malka Rappaport Hovav Lexico-Logical Form: A Radically Minimalist Theory, Michael Brody The Architecture of the Language Faculty, Ray Jackendo¤ Local Economy, Chris Collins Surface Structure and Interpretation, Mark Steedman Elementary Operations and Optimal Derivations, Hisatsugu Kitahara The Syntax of Nonfinite Complementation: An Economy Approach, Zˇeljko Bosˇkovic´ Prosody, Focus, and Word Order, Maria Luisa Zubizarreta The Dependencies of Objects, Esther Torrego Economy and Semantic Interpretation, Danny Fox What Counts: Focus and Quantification, Elena Herburger Phrasal Movement and Its Kin, David Pesetsky Dynamic Antisymmetry, Andrea Moro Prolegomenon to a Theory of Argument Structure, Ken Hale and Samuel Jay Keyser Essays on the Representational and Derivational Nature of Grammar: The Diversity of Wh-Constructions, Joseph Aoun and Yen-hui Audrey Li Japanese Morphophononemics: Markedness and Word Structure, Junko Ito and Armin Mester Restriction and Saturation, Sandra Chung and William A. Ladusaw The Linearization of Chains and Sideward Movement, Jairo Nunes The Syntax of (In) Dependence, Ken Safir Interface Strategies, Tanya Reinhart Asymmetry in Morphology, Anna Maria Di Sciullo Relators and Linkers, Marcel den Dikken On the Syntactic Composition of Manner and Motion, Maria Luisa Zubizarreta and Eunjeong Oh

On the Syntactic Composition of Manner and Motion

Maria Luisa Zubizarreta and Eunjeong Oh

The MIT Press Cambridge, Massachusetts London, England

6 2007 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. MIT Press books may be purchased at special quantity discounts for business or sales promotional use. For information, please e-mail [email protected] .edu or write to Special Sales Department, The MIT Press, 55 Hayward Street, Cambridge, MA 02142. This book was set in Times New Roman on 3B2 by Asco Typesetters, Hong Kong and was printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Zubizarreta, Maria Luisa. On the syntactic composition of manner and motion / by Maria Luisa Zubizarreta and Eunjeong Oh. p. cm.—(Linguistic inquiry monographs) Includes bibliographical references and index. ISBN-13: 978-0-262-24052-9 (hardcover : alk. paper) ISBN-13: 978-0-262-74029-6 (pbk. : alk. paper) 1. Grammar, Comparative and general—Verb. 2. Motion in language. 3. Grammar, Comparative and general—Syntax. I. Oh, Eunjeong. II. Title. P281.Z83 2007 415—dc22 2006030108 10 9 8 7

6 5 4 3 2

1

To the memory of Ken Hale Longfellow’s ‘‘Psalm of Life’’ Lives of great men all remind us We can make our lives sublime, And, departing, leave behind us Footprints on the sands of time; Footprints, that perhaps another, Sailing o’er life’s solemn main, A forlorn and shipwrecked brother, Seeing, shall take heart again.

Contents

Series Foreword Acknowledgments

ix xi

Chapter 1 Preliminaries to the Lexicon-Syntax Relation and the Variable Behavior of Manner-of-Motion Verbs 1

1.1 The Motion Construction: The Main Issues 2 1.2 A Lexical Approach: Levin and Rappaport 1995; Rappaport and Levin 1998 4 1.3 Some Syntactic Approaches: Borer 1994, 2005; Ritter and Rosen 1998; Folli 2001 8 1.4 The Hale and Keyser Model and Some Elaborations 13 1.5 Event Boundedness and Temporal Boundedness 25 1.6 A Crosslinguistic Analysis of Manner-of-Motion Constructions: A Preview of Chapters 2 and 3 30

Chapter 2 The Serial-Verb Construction and Verbs of Motion in Korean 57

2.1 Serial-Verb Constructions (SVCs) in Korean 57

viii

Contents 2.2 Serial Verbs of Motion in Korean 77 2.3 Summary

108

Appendix 2.1 Some Remarks on the Aspectual Properties of -(e)ci 109 Appendix 2.2 On Distinguishing the ‘‘Prospective Possessor’’ Construction from the ‘‘Directed-Motion’’ and ‘‘Benefactive’’ Constructions 115 Chapter 3 Motion Verbs in Germanic and Romance 127

3.1 Germanic

128

3.2 Romance

152

3.3 Summary and Some Final Remarks 187 Notes

191

References Index

225

215

Series Foreword

We are pleased to present the forty-eighth in the series Linguistic Inquiry Monographs. These monographs present new and original research beyond the scope of the article. We hope they will benefit our field by bringing to it perspectives that will stimulate further research and insight. Originally published in a limited edition, the Linguistic Inquiry Monographs are now more widely available. This change is due to the great interest engendered by the series and by the needs of a growing readership. The editors thank the readers for their support and welcome suggestions about future directions for the series. Samuel Jay Keyser for the Editorial Board

Acknowledgments

An early version of this work was presented at the Universitat Auto`noma de Barcelona in March 2003. We thank our hosts Maria Teresa Espinal, Carmen Picallo, Jaume Mateu, Maria Luisa Hernanz, and Gemma Rigau. Parts of chapter 2 were presented at the Sixth Seoul International Conference on Generative Grammar (SICOGG 6) at Hansung University in August 2003 and at the LSA in January 2004, and parts of chapter 3 at the Linguistic Symposium on Romance Languages (LSRL) at Rutgers University in March 2006. We have benefited greatly from discussions with Jean-Roger Vergnaud, Jaume Mateu, Jay Keyser, Karen Zagona, and Victor Manfredi. We are also grateful to many scholars who have taken the time to provide enlightened answers to our queries on Dutch (Marcel den Dikken and Henk van Riemsdijk), on English (Bill Rutherford and Tom Roeper), on Italian (Elena Guerzoni, Anna Cardinaletti, and Mario Saltarelli), on French (Jean-Roger Vergnaud and Alain Rouveret), on Spanish (Mo´nica Cabrera, Roberto Mayoral, and Violeta Demonte), and on Korean (Hongjoong Kim, Seonkyung Jeon, Soyoung Park, Hyuna Byun, Yongjoon Cho, and Seungyoon Lee). Last but not least, we thank three anonymous reviewers for valuable comments and suggestions.

Chapter 1 Preliminaries to the LexiconSyntax Relation and the Variable Behavior of Manner-of-Motion Verbs

The most noteworthy development in the area of the lexicon-syntax interface since the 1980s has been the realization that there are ‘‘constructional’’ meanings, which are independent of the particular lexical items that make up the sentence. For an excellent overview, see Goldberg 1995. The notion of ‘‘construction’’ varies across theories. In e¤ect, the notion of ‘‘construction’’ assumed by Goldberg 1995 is distinct from the one assumed in generative grammar. Nevertheless, they share a common insight, which may be stated pretheoretically in general terms along the following lines: (1) The meaning of an expression is to be attributed to the superimposition of the meaning of grammatical closed-class items and the meaning of open-class items. See Fillmore 1988. In other words, there are ‘‘structures’’ that carry meaning and these ‘‘structures’’ are flagged by ‘‘closed-class items.’’ Theories vary as to the nature of these ‘‘structures,’’ as well as to the grammatical status of ‘‘closed-class items.’’ Nevertheless, common to many di¤erent theories of the lexicon-syntax interface is the insight that linguistic expressions are associated with structured meaning that is independent of the particular open-class lexical items they contain. Interestingly, as acknowledged by Goldberg 1995, even so-called lexicalist theories, such as Pinker 1989 and Levin and Rappaport 1995, have bits of ‘‘constructional’’ assumptions, in the general sense of the term, as defined in (1). Indeed, the notion of construction defined in (1) is very general and uncontroversial. It is equally compatible with very di¤erent views of the lexicon-syntax interface.1 It is compatible with a lexical-based account, such as the one put forth by Levin and Rappaport 1995 and Rappaport and Levin 1998, as well as with a syntax-based account, put forth by Hale and Keyser 1993, 2002,

2

Chapter 1

as well as many other sources (e.g., Borer 1994, 2005; Folli and Ramchand 2001; Folli 2001; Harley 2002; Mateu and Rigau 2002; Mateu 2002, 2005; Megerdoomian 2002; Ramchand 1995, 2001, 2002; Ritter and Rosen 1998, 2000). In this chapter, we briefly present the lexicalist approach (section 1.2) and some of the syntactic approaches (section 1.3) and assess their merits with respect to the main topic of this book, the much-discussed mannerof-motion construction (section 1.1). We present the framework put forth by Hale and Keyser 2002 (section 1.4), on which our proposal builds, and briefly discuss the aspectual notion of ‘‘boundedness’’ (section 1.5), a notion highly relevant to the understanding of constructions headed by motion verbs. And finally, we give a brief summary of the analysis defended in subsequent chapters (section 1.6), where we discuss in detail the properties of the motion construction in Korean, Germanic, and Romance. 1.1

The Motion Construction: The Main Issues

Manner-of-motion verbs have attracted the attention of many scholars due to their variable behavior (e.g., Hoekstra and Mulder 1990; Borer 1994; Ritter and Rosen 1998; Folli 2001; Mateu and Rigau 2002; Mateu 2002). The variable behavior of that class of verbs is systematic in English and Dutch, lexically restricted in Italian, and quasi-nonexistent in French and Spanish. Thus, in Dutch (and English), any manner-of-motion intransitive verb can express an activity, in which case its auxiliary in the perfect tense is hebben and its syntactic properties are that of an unergative verb; see (2) below. On the other hand, any manner-of-motion verb in Dutch (and English) can appear in sentences that express an accomplishment (i.e., movement toward a goal), in which case its auxiliary in the perfect tense is zijn and its syntactic properties are those of an unaccusative verb; see (3). (2) a.

dat Jan naar Groningen twee uur lang heeft gewandeld. that Jan to Groningen two hours long has walked ‘. . . Jan walked in the direction of Groningen for two hours.’ b. ??dat Jan twee uur lang naar Groningen heeft gewandeld. that Jan two hours long to Groningen has walked

(3) dat Jan in twee uur naar Groningen is gewandeld. that Jan in two hours to Groningen is walked ‘. . . Jan walked to Groningen in two hours.’

Preliminaries to the Lexicon-Syntax Relation

3

While in French and Standard Spanish, the directed-motion usage of intransitive manner-of-motion verbs is rare, in Italian a small subclass of manner-of-motion verbs clearly shows variable behavior. Compare (4b) with (5b). (4) a. Maria a corso (fino a casa). Maria has run-3rd p.s.masc. (to the house) ‘Maria has run (to the house).’ b. Maria e` corsa *(fino a casa). Maria is run-3rd p.s.fem. *(to the house) ‘Maria has run to the house.’ (5) a.

Maria a camminato (fino a casa). Maria has walked-3rd p.s.fem (to the house) ‘Maria has walked (to the house).’ b. *Maria e` camminata (fino a casa). Maria is walked-3rd p.s.fem. (to the house)

In serial-verb (SV) languages, we find that the manner meaning and the directed-motion meaning are expressed independently by two distinct verbs; see the example in (6) from Igbo (a head-initial language) and the example in (7) from Korean (a head-final language). ´ gba`-ra o´so ga-a ahya´. (6) O ´ market.Gen 3s do-rV race go-A ‘He ran and went to the market.’/‘He ran to the park.’ (Cited in De´chaine 1993, 239.) (7) John-i kongwen-ey kel-e ka-ss-ta. John-Nom park-Loc walk-L go-Past-Decl ‘John went to the park running.’/‘He walked to the park.’ The case of Nicaraguan Sign Language (NSL), which also seems to be an SV language, is particularly illuminating because it has recently evolved from a pidgin to a creole. In the September 2004 issue of Science, Sen¨ zyu¨rek report that the early generation of NSL, which ghas, Kita, and O is gestural and iconoclastic, represents manner and motion simultaneously, but that in later generations, when NSL develops into a linguistic system with discrete units and combinatorial rules, manner and path are represented sequentially. This evolution shows that the compositional nature of human languages is truly universal and suggests to us that, despite appearances, in the Dutch example in (3) and the Italian example in (4b),

4

Chapter 1

the verb does not simultaneously encode manner and directed motion, but that at some abstract level, these two are represented independently. As we will see below, the lexical approach, like the syntactic approach, can readily capture the compositionality of the phenomenon under discussion. The true challenge for any approach is to provide a principled account of the linguistic variation mentioned above and discussed more fully in chapters 2 and 3 of this book. A principled account is one in which the di¤erences in the behavior of manner-of-motion verbs among languages can be related to other well-established di¤erences among them. 1.2 A Lexical Approach: Levin and Rappaport 1995; Rappaport and Levin 1998 These authors (L&R) put forth an analysis in which verb meaning is decomposed into two parts: the structural and the idiosyncratic. The lexical decomposition associated with verbs encodes both aspects of lexical meaning. While the idiosyncratic part is encoded in terms of constants (i.e., a phonological string), the structural part is encoded in terms of a small number of lexical-semantic templates formed via the combination of primitive predicates such as ACT, CAUSE, BECOME, STATE, and in some cases the modifiers of such predicates (such as MANNER and INSTRUMENT), as well as their variable arguments (or event participants). A verb’s grammatical meaning consists of the association of a constant with a particular lexical-semantic template, also referred to as ‘‘event-structure template.’’2 The basic inventory of event-structure templates is given in (8). The predicate STATE can be replaced by the predicate PLACE, depending on whether the meaning of the constant expresses a state or a location. Thus, if the constant paired with the event-structure template in (8d) encodes a state, the resulting event denotes a change of state (e.g., bloom, blossom, decay, flower, rot, rust, sprout), and if the constant paired with the event-structure template in (8d) encodes a place, the resulting event denotes a change of location (e.g., bag, box, cage, crate, garage, pocket). Likewise, in (8a), the modifier MANNER can be replaced by the modifier INSTRUMENT, depending on whether the meaning of the constant expresses the manner of the activity (e.g., creak, jog, run, whistle) or the instrument used in the activity (e.g., brush, hammer, saw, shovel ). Two types of causative events are recognized: internally caused and externally caused events. Internally caused

Preliminaries to the Lexicon-Syntax Relation

5

events are associated with the template in (8d) (e.g., bloom, rot, rust, sprout), and externally caused events are associated with the template in (8e) (e.g., break, dry, melt, open). (8) a. i. ii. b. i. ii. c. i. ii. d. i. ii. e. i.

[x ACThMANNERi ] (Activity) [x ACThINSTRUMENTi ] [x hSTATEi] (State) [x hPLACEi] [BECOME [x hSTATEi]] (Achievement) [BECOME [x hPLACEi]] [x CAUSE [BECOME [y hSTATEi]]] (Accomplishment) [x CAUSE [BECOME [y hPLACEi]]] [[x ACThMANNERi ] CAUSE [BECOME [y hSTATEi]]] (Accomplishment) ii. [[x ACThMANNERi ] CAUSE [BECOME [y hPLACEi]]]

The Canonical Realization Rule (CRR) ensures that the minimal elements of meaning encoded in the constants are syntactically expressed. This is achieved by associating a constant (i.e., a phonological string) with a particular element in the template. For example, the Manner CRR replaces MANNER in (8ai) with a constant such as creak, jog, run, whistle (as exemplified in (9)); the constant in this case modifies the activity. The Externally Caused State CRR replaces STATE in (8e) with a constant such as break, dry, harden, melt, open (as exemplified in (10)); the constant in this case names the end state of the change (but provides no information regarding the causing subevent; i.e., MANNER remains unspecified). (9) Run: [x ACThRUNi ] (10) Break: [[x ACThMANNERi ] CAUSE [BECOME [y hBROKENi]]] Within this theory, ACT, CAUSE, BECOME, and STATE belong to the closed-class items of the language, while RUN and BROKEN belong to the open class. Such a theory is then in line with the general and quite uncontroversial assumption in (1). L&R put forth two well-formedness conditions on the syntactic realization of lexical-event structures. One is the Subevent Identification Condition: (11) Each subevent in the event structure must be identified by a lexical head (e.g., a V, an A, or a P) in the syntax.

6

Chapter 1

L&R (1998, 112) mention that ‘‘the Subevent Identification Condition allows for a single verb in the syntax to identify more than one subevent when a Canonical Realization Rule associates the constant with a complex event structure template.’’ Thus, in a sentence like John broke the vase, both the subevent ACT and the subevent BECOME-STATE in (10) are identified by the verb break. The other well-formedness condition is the Argument Realization Condition (cf. the Theta Criterion stated in Chomsky 1981): (12) Argument Realization Condition a. There must be an argument XP in the syntax for each structure participant in the event structure. b. Each argument XP in the syntax must be associated with an identified subevent in the event structure. Furthermore, Linking Rules are postulated to capture the generalizations of which variable participant in the event template is linked with which grammatical function in the syntax. As mentioned earlier, L&R distinguish two types of causatives: internally caused and externally caused. Internally caused verbs describe an eventuality in which ‘‘some property inherent to the argument of the verb is ‘responsible’ for bringing about the eventuality,’’ as is the case with verbs of emotion (blush, tremble), verbs of emission (glitter, gush, smell, whistle), and agentive verbs ( play, speak). On the other hand, externally caused verbs ‘‘imply the existence of an ‘external cause’ with immediate control over bringing about the eventuality described by the verb: an agent, an instrument, a natural force, or a circumstance’’ (L&R 1995, 91–92). L&R appeal to this distinction to characterize the class of alternating verbs. Alternating verbs are precisely those in which the causer does not depend on the internal properties of the verb and therefore can remain unspecified. The unspecified nature of the causing event in externally caused verbs is shown by the wide range of external argument types it can take. Compare the externally caused transitive break in (13a) with the internally caused transitive cut in (14a–b). If (and only if ) the causer is unspecified, it can be existentially bound in the lexical-semantic representation and remain syntactically unrealized. This accounts for the contrast between (13b) and (14c). (13) a. The vandals/The rocks/The storm broke the windows. b. The windows broke.

Preliminaries to the Lexicon-Syntax Relation

7

(14) a. The baker/that knife cut the bread. b. *The lightning cut the clothesline. c. *The bread cut. (On the interpretation ‘The bread came to be cut.’) L&R (1995) acknowledge the variable behavior of manner-of-motion verbs in Germanic and give an account based on polysemy: such verbs optionally take a PP goal. The crosslinguistic variation is also acknowledged but no account is provided. In their 1998 article, L&R enrich their system with a process referred to as Template Augmentation, which allows for ‘‘more complex event structure templates to be built on simpler ones.’’ (15) Template Augmentation Event-structure templates may be freely augmented up to other possible templates in the basic inventory of event-structure templates.3 To exemplify, the basic event structure in (16), to which activity-denoting events like (17) are associated, can be augmented as in (18). The latter is an accomplishment-denoting event structure, which underlies sentences like (19).4 (16) [x ACThSWEEPi y] Cf. (8ai) (17) Phil swept (the floor). (18) [x ACThSWEEPi y] CAUSE [BECOME [y hSTATEi]]] Cf. (8ei) (19) Phil swept the floor clean. Possibly the variable behavior of manner-of-motion verbs could be analyzed in terms of Template Augmentation. More precisely, the basic activity template in (20) could be augmented into an accomplishment, as shown in (21). (20) [x ACThRUNi ] Cf. (8ai) (e.g., John ran) (21) [[x ACThRUNi ] CAUSE [BECOME [y hPLACEi]]] Cf. (8dii) (e.g., John ran to the store) Indeed, Folli 2001 reports unpublished work by Fong and Poulin, who put forth such a proposal in Fong and Poulin 1997. While the activitydenoting event is identified by the verb, the accomplishment-denoting event is identified by the preposition, thus complying with the Subevent

8

Chapter 1

Identification Condition in (11). In the serial-verb languages, the two events can be said to be identified by two distinct verbs, although in such cases we have the added complication of determining the subordination relation between the two verbs; see De´chaine 1993 for extensive discussion, as well as chapter 2. As mentioned earlier, the challenge is to provide a principled account of the crosslinguistic variation in the behavior of manner-of-motion verbs. Recall that in Romance, the variable behavior of manner-ofmotion verbs is not a general phenomenon. It is lexically restricted, and in some languages (like Spanish and French) more so than in others (namely, Italian). How then would a lexical theory account for this variability? One could stipulate that verbs in Romance do not allow for Template Augmentation. The question then arises as to how to treat those cases in which manner-of-motion verbs do seem to express directed motion, such as the Italian example in (4b). One could stipulate that Template Augmentation applies to some manner-of-motion verbs, but not to others. Yet the question of what accounts for the typological variability remains unanswered. More specifically, what other independently attested properties are the above-mentioned crosslinguistic di¤erences related to? 1.3 Some Syntactic Approaches: Borer 1994, 2005; Ritter and Rosen 1998; Folli 2001 Several authors have defended the view that the notion of event is structurally defined in the syntax (see Borer 1994, 2005; Folli 2001; Hale and Keyser 2002; Ramchand 2001, 2002; Ritter and Rosen 1998). This is often referred to as the constructional approach. A particular version of the constructional approach was put forth by Borer 1994, and more extensively by Borer 2005; we will denote this as the functional-based constructional approach. While we cannot address the di¤erent versions of this approach in any detail here, we will briefly examine how these accommodate the variable behavior of manner-of-motion verbs. The functional-based approach (put forth by Borer 1994) argues that the interpretation of DP arguments is assigned by functional projections (via Spec-head agreement). It is not at all dependent on the properties of the verb. Thus, intransitive verbs are not lexically categorized as unaccusative or unergative. The unergative versus unaccusative behavior of verbs depends on which Spec position the argument occupies. There is a higher F category that licenses the meaning of originator and a lower one that

Preliminaries to the Lexicon-Syntax Relation

9

licenses the meaning of undergoer of the event (i.e., the delimiter of the event). Ritter and Rosen (1998) develop a mixed theory, which assumes (with Borer 1994, 2005) that the interpretation of arguments depends on which functional projection licenses the argument, but it also assumes that the choice of functional projection is determined by the lexical properties of the verbs. More concretely, they put forth the following general idea: (22) ‘‘The addition of a secondary predicate is constrained by the basic lexical representation of the main verb, but does not involve augmentation of this representation to derive a new lexical entry. Rather, an interpretation is assigned to lexical items postsyntactically on the basis of their lexically listed semantic content in conjunction with semantic content encoded in the syntactic structure. For example, when a secondary resultative predicate is added in the syntax, the delimiting FP is also added in the syntax, giving rise to a D[elimited]-eventive interpretation at LF’’ (p. 153). Thus, a sentence like John walked to the store would have the structure below (John is both initiator and delimiter of the event in this theory). (23)

10

Chapter 1

We can see that the theory developed by Ritter and Rosen is (with some variation) a syntactic rendition of the Template Augmentation proposed by L&R. In e¤ect, L&R (1998) address the debate between their own lexical approach (which they refer to as the ‘‘projectionist’’ approach) and the syntactic (or constructional) approach in the following terms: Most current lexical semantic theories recognize a dichotomy in verb meaning and distinguish what we have . . . referred to as the structural and the idiosyncratic aspects of verb meaning. . . . In the projectionist approach the structural aspect of meaning is usually encoded in some sort of skeletal event structure representation, while the idiosyncratic element of meaning is represented by the constant. The constructional approach recognizes the same distinction, although it is represented di¤erently. In the constructional approach, the idiosyncratic component of meaning itself constitutes the lexical representation of the verb, while the structural aspects of meaning do not reside in the lexical entries of individual verbs but rather are associated with certain basic syntactic structures, those which are associated with skeletal event interpretations. Since both projectionist and constructional approaches recognize this basic distinction among the elements of verb meaning, the major di¤erence between the two concerns whether the association between the constant and the template is registered in the lexicon or not. (p. 129)

Like L&R’s theory, Ritter and Rosen’s proposal does not address the typological di¤erences across languages with respect to the variable behavior of manner-of-motion verbs. Borer (2005) develops a more radical version of the functional-based constructional approach. Within this view, DP arguments are not arguments of the verb at all; they are solely arguments of aspectual functional projections. The verb, according to this view, is a modifier of the event structure. In e¤ect, the radical theory put forth by Borer (in which DP arguments are completely severed from the verb) is the position that naturally ensues from the functional-based approach, when carried out to its ultimate logical extreme. Within this theory, verbs are not categorized into unbounded activities and bounded change-of-location events. Any verb can in principle receive any kind of eventive interpretation depending on its syntactic context. Thus, if the verb combines with an aspectual functional projection specified as Q (for quantifiable event), it will acquire the relevant interpretation and the argument in the Spec of the FQ node will be interpreted as the undergoer/delimiter participant of the event.5 Quantifiable events are þdivisive; they are constituted of countable subintervals. Still, it is the case that the presence of the directional PP is crucial

Preliminaries to the Lexicon-Syntax Relation

11

in licensing the quantifiable interpretation. This fact can be accommodated by stating that verbs are (to a large extent)
Q and some element (such as the directional PP in the case of intransitives) is needed to license FQ .6 In such a case, it is said that the FQ ‘‘quantifies into the verb,’’ in the same way that a generic adverb is said to quantify into bare nouns in examples like cows are usually vegetarians. In the absence of an Fprojection marked as Q, the event is interpreted as an unbounded activity. H. Borer (personal communication) suggests that the typological distinction between Romance and Germanic may be stated in the following terms: (24) The intransitive verbs in Romance languages do not allow a þQ Asp node to ‘‘quantify into’’ the verb. Borer (personal communication) furthermore suggests that the property in (24) might be a property not only of Romance verbs, but also of Romance nominals. Bare plural DPs in subject position in Spanish and Italian lack a generic interpretation. If generic bare plurals are analyzed along the lines of Heim 1982—that is, as variables bound by a sentential generic operator—then a cross-categorial generalization can indeed be made.7 The question then arises as to how to accommodate examples such as the Italian (4b), which needs a PP to denote a þQ type of event. Another issue is the assumption within such a theory that all verbs (like all nouns) are grammatical equals. Although in Germanic the class of verbs that imply directed motion seems to be relatively small compared to the class of activity-denoting manner-of-motion verbs, a grammatical characterization of such verbs is still necessary; see note 6. Folli 2001 adopts the lexical-based constructional approach developed by Ramchand 2001, which in turn is a particular rendition of the influential approach developed by Hale and Keyser in numerous manuscripts and which culminated in Hale and Keyser 2002. Within this approach, the syntax generates the maximal event structure in (25), where v is interpreted as introducing a causative event, V specifies a change or process, and Rv introduces the telos of the event.

12

Chapter 1

(25)

Like Ritter and Rosen, Folli also proposes to derive the accomplishment usage of manner-of-motion verbs by adding a result-denoting phrase. In this sense, Folli’s proposal is in the same general spirit as the lexicalist proposal based on the process of event template augmentation. But unlike the previously discussed works, Folli attempts to characterize the di¤erences between Germanic and Romance. That author locates the typological di¤erence between the two sets of languages in the morphological (and eventive) makeup of prepositions. The proposal is based on the observation that morphologically simple prepositions in Italian (and other Romance languages) are point-denoting—that is, they denote a location. In these languages, only morphologically complex prepositions are dynamic; in other words, they denote both path and location. But in English (and other Germanic languages), morphologically simple prepositions can also be dynamic.8 For this reason, morphologically simple prepositions in Germanic can be adjoined to a VP headed by an activitydenoting verb and give rise to a directed motion. On the other hand, in Romance, a morphologically simple preposition (which is unambiguously point-denoting) can combine with a verb to give rise to a directed-motion reading if and only if that verb selects for a RvP (a result phrase); see the structure in (25). This is the case of correre ‘run’, which is analyzed as ambiguous—that is, as either denoting an activity or an accomplishment (with a resultative PP complement). Caminare ‘walk’, on the other hand, is unambiguous; it only denotes an activity. In chapter 3, we will argue that the di¤erence between the Germanic and the Romance morphologically simple prepositions cannot fully explain the di¤erences between the two sets of languages. Furthermore, the question arises as to whether it is

Preliminaries to the Lexicon-Syntax Relation

13

desirable to analyze the Italian manner-of-motion verbs as polysemous between an activity-denoting verb and an accomplishment that simultaneously encodes manner and directed motion. We do not exclude the possibility that there might be some cases of polysemy of this type, but they are very rare, as can be appreciated in Spanish and French. Indeed, the phenomenon appears to be much more robust in Italian than in Spanish and French. While there is no property that is shared by all the mannerof-motion verbs that can appear in the directed-motion structure in Italian (i.e., the verbs that belong to this class must be stipulated in any account), it would be desirable if we could derive the directed-motion structure via some mechanism that is independently needed to account for some other phenomenon in the language. Before we summarize the account that we will defend in the following chapters, we briefly present the Hale and Keyser model (section 1.4), which will serve as a framework for our own proposal, and we briefly discuss the notions of event boundedness and temporal boundedness (section 1.5), since they are relevant to an understanding of manner-of-motion verbs. 1.4

The Hale and Keyser Model and Some Elaborations

Hale and Keyser (H&K), in a series of articles that culminated in their 2002 monograph, developed a model, very much inspired by Larson 1988, in which predicative categories are associated with syntactic structures referred to as l-structures. In this model, the eventive properties of predicates are syntactically decomposed. Thus, like the works by Ritter and Rosen, Ramchand, and Folli cited earlier, H&K’s model shares important insights with the lexicalist approach proposed by L&R. While the latter proposal encodes eventive structures in terms of relations between some basic semantic predicates (CAUSE, BECOME, ACT, PLACE, STATE) and argument variables, the H&K proposal encodes eventive structures in the syntax in terms of specifier-head and head-complement relations, in which the basic building blocks are the traditional part-ofspeech categories (V, P, N, A). 1.4.1 Intransitives At the heart of the theory developed by these authors is the syntactic distinction between unaccusative and unergative verbs. More specifically, they propose that unaccusatives, unlike unergatives, are associated with an l-structure that contains a specifier position.

14

Chapter 1

(26) [VP D [VP V XP]] According to H&K, the subject of an unergative structure is not an argument of the verb at all. It originates in the sentential part of the syntax (or s-syntax). H&K furthermore argue that unergatives are denominal verbs associated with the general structure in (27a). Thus, the unergatives laugh and sleep arise from the structures in (27b) and (27c), respectively. More concretely, these structures give rise to a process of ‘‘conflation,’’ whereby the head projects the categorial feature and the complement provides the phonological content for the derived V. (27) a. [V V N] b. [V V [laugh]] c. [V V [sleep]] Probably not all unergatives are denominal verbs. Note that mannerof-motion verbs are systematically ambiguous between an (activitydenoting) unergative use and an (accomplishment-denoting) transitive use, comparable to verbs of consumption and creation, as recognized by Tenny 1987, 1994, 1995. (28) John ran (a mile). (29) a. John ate (an apple). b. John drew (a circle). A possible way around polysemy here would be to assign the same kind of analysis for (28) that has been advanced for (29), namely, that such structures are in fact hidden transitives with an indeterminate object incorporated into the verb. More precisely, in the case of (28), the incorporated object would be an indeterminate distance-denoting object (where distance is understood as an abstract path). More concretely, the object could be analyzed as a distance classifier; consider John ran a long distance and John ran a distance of two miles. This would mean that run, walk, swim, and other manner-of-motion verbs are not denominal verbs (in contrast with laugh, work, sleep); see chapter 3 for further discussion. While the details of the analysis of unergative verbs will not be essential for what we have to say here, we are committed to the assumption that unergative verbs lack a specifier and that they have branching l-structures, as proposed by H&K. We will endorse the view that the presence of an external argument is not regulated by the verb per se. However, following Marantz 1984,

Preliminaries to the Lexicon-Syntax Relation

15

Chomsky 1993, and Hale and Keyser 1993, we will assume the existence of a vP, headed by a functional head v and generated immediately above the l-structure of the verb. The category v is comparable to that referred to as ‘‘Voice’’ by Kratzer 1996 and as ‘‘Tr(ansitive)’’ node by Collins 1997; it serves the purpose of licensing an external argument and allows us to structurally define the notion of transitivity, which is so intimately related to Case assignment, in terms of the sequence ‘‘v V.’’9 The node v is introduced by a general rule of syntactic composition: (30) a. v may be freely merged with a VP. b. If VP lacks a Specifier, v must be merged with VP. Given (30b), the structures in (27) should be as in (31). Such structures are interpreted as activity-denoting events. (31) a. [vP D [v [VP V D]]] b. [vP D [v [VP V [laugh]]]] c. [vP D [v [VP V [sleep]]]] We may furthermore assume that (32) In the unmarked case, only one v per l-structure is allowed. Returning to the unaccusative structure in (26), two subtypes can be distinguished, depending on the nature of the XP complement. If the complement denotes a path with an endpoint, the structure encodes movement toward a telos (i.e., directed motion). Note that the path may be a scale and changes along such a scale can be conceptualized as directed motion. See Goldberg 1995; Jackendo¤ 1990, 1996; Krifka 1989; Tenny 1987, 1994, 1995. Similarly, change of possessor (e.g., give a book to Mary) can be represented as a case of directed motion; see Gruber 1965 and Jackendo¤ 1996. It is therefore not surprising that possessors (e.g., (33b)) can appear as the goal-denoting complement of go, alongside locatives (e.g., (33a)). And note that state-denoting predicates can do so, too (e.g., (33c)). (33) a. John went to the park. b. The prize goes to Mary. c. The milk went sour. On the other hand, if the XP complement in the unaccusative structure in (26) denotes a location or state, then the structure can be interpreted as stative:

16

Chapter 1

(34) a. The book is on the table. b. John is in the park. c. The milk is sour. The question arises as to how to structurally distinguish the complements in (33) from the complements in (34). Let us consider first the cases with a PP complement. H&K and others (e.g., Koopman 1997; Folli 2001; den Dikken 2003) have suggested that PPs that denote a directed path have complex syntactic structures. They consist of a (locative) P embedded under a (directional) P, as shown in (35a). On the other hand, a point-locating (or locative) preposition consists of one single P, as shown in (35b). (35) a. [ P P [ P P D]] b. [ P P D] The examples in (33a) and (34b) will then have the structure in (36) and (37), respectively. (36) a. [D [V [P [P [D]]]]] b. [VP John [V went [ PP to [ PP P [the park]]]]] (37) a. [D [V [P [D]]]] b. [VP John [V is [ P in [the park]]]] Evidence for the syntactic complexity of path-denoting prepositions is provided by the existence of morphologically complex prepositions, illustrated in (38). In such examples, the locative preposition is morphophonologically realized (in); it is cliticized onto the path-denoting preposition (to). (38) John went into the room. Note that the structure in (37), associated with a stative meaning, is also an unaccusative structure. Recall that an unaccusative l-structure is identified as one that contains a Spec-head relation. In the cases discussed above, the di¤erence between the process-denoting and the stative unaccusatives is attributed to a di¤erence in complementation (i.e., presence or absence of a path-denoting preposition). The unaccusative status of both types of events is confirmed by languages like Dutch and Italian, in which unaccusative structures select the auxiliary be. (39) a. Jan is/*heeft in de tuin gegaan. Jan is/*has in the garden gone ‘Jan has gone to the garden.’

Preliminaries to the Lexicon-Syntax Relation

17

b. Jan is/*heeft in de tuin geweest. Jan is/*has in the garden been ‘Jan has been in the garden.’ The challenge to a constructional approach is posed by predicates that contain adjectival and nominal complements. Consider the unaccusatives with an adjectival complement. The question is whether we can structurally distinguish the adjectival complement in dynamic events, such as (33c), from the adjectival complement in stative events, such as (34c). If we follow the analogy with the prepositional structure, then adjectival complements in the change-of-state construction should be embedded under a path-denoting category as well, as shown in (40). The adjective sour names the telos of the path, in the same way that the park names the telos of the path in (36a). On the other hand, the sentence in (34c) would be associated with the simple structure in (41), which would be unambiguously associated with a stative meaning. (40) a. [D [V [Xpath [A]]] b. [VP The milk [V went [Xpath [ AP sour]]] (41) a. [D [V [A]]] b. [VP The milk [V is [ AP sour]]] Sentences like (42) would also be associated with the structure in (40). (42) a. The milk became sour. (Cf. The milk came to be sour.) b. John became famous. (Cf. John came to be famous.) The problem with the approach outlined above, which we may refer to as the full decompositional approach, is that there is no evidence for an abstract path-denoting category in (40), in contrast with PPs, where morphologically overt complex Ps testifiy to the existence of a path-denoting P; see (38). An alternative to the full decompositional approach is provided by H&K’s proposal that stativity be represented as a direct relation between a nominal specifier and an adjectival or nominal complement (a relation that H&K refer to as ‘‘central coincidence’’). More precisely, P will have a specifier, as shown in (43a). In this structure, P establishes a relation of ‘‘central coincidence’’ between the specifier D and the complement D. In the case of the adjective, it is proposed that an abstract category a introduces a specifer for A, as shown in (43b). While this category is abstract, at least it can be identified with some other abstract category with the same type of function, namely ‘‘little v,’’ discussed earlier, and for this

18

Chapter 1

reason, we will refer to it as ‘‘little a,’’ as in (43c). While ‘‘little v’’ introduces the originator of the event, ‘‘little a’’ introduces the relation of ‘‘central coincidence.’’ (43) a. [ P D [P D]] b. [D [a A]] c. [D [a A]] If we combine the above proposal with the assumption that verbs like be and seem are raising predicates, we can then distinguish the dynamic events in (33c) and (42) from the stative events in (34c) and (44) in structural terms. The latter will be associated with the structure in (45) and the former with the structure in (46). (44) The milk seems sour. (45) [VP The milk [V is/seems [ aP (the milk) [a [ AP sour]]]] (46) [VP The milk [V became/went [ AP sour]]] The structure in (46) lacks a category that introduces the relation of ‘‘central coincidence.’’ In such a case, A is associated with the following default interpretation: (47) If A is not immediately dominated by ‘‘little a,’’ then A is interpreted as denoting the endpoint of a path. Given the general interpretative assumption in (48), the VP in (46) will be correctly interpreted as a process-denoting event. (48) If V immediately dominates a category that denotes the endpoint of a path (A or N), V denotes a change of state or location. The analysis sketched above abandons the full decompositional approach and acknowledges that some aspects of eventive meanings are interpretational in nature (based on the immediate structural context). The question that then arises is how to handle stative verbs other than the light verb be or semimodal seem, namely, measure verbs such as cost/ weigh, possessor verbs such as have/own, and experiencer verbs like love/ fear. The experiencer verbs can actually be reduced to the possessor category (where the object of possession is an abstract emotion). A possible approach is to extend the proposal put forth by Freeze 1992 for the verb have and adopted by Kayne 1993, den Dikken 1995, and others. These authors have proposed that structures headed by the verb have are actually copular structures with a locative P complement and that be þ P is

Preliminaries to the Lexicon-Syntax Relation

19

spelled out as have. The relation is more obvious in a language like French, in which the copula þ P structure actually surfaces, as illustrated in (50). (49) a. John has the book. b. The book is Ploc John. (be þ Ploc ! have) (50) a. Marie a le livre. b. Le livre est a` Marie. Following this proposal, we could analyze the complement of stative verbs that denote a point in a scale (like cost and weigh) as a locative PP—for example, the metal weighs one pound has a meaning comparable to ‘‘the metal is at point one pound in the scale.’’ More precisely, we could argue that the cases of stative measure verbs involve subject-tosubject raising, as illustrated in (51). (51) [the metal [weighs [(the metal ) [P [one pound]]]]] The case of experiencer verbs is more complex (e.g., John feared the war). We might analyze it as a case of subject control with a dyadic complement structure (comparable to John promised Mary to leave). Furthermore, the object in the ‘‘central coincidence relation’’ would be the emotion-denoting nominal—that is, fear in the example below; consider ‘‘John is at point fear in the emotion scale.’’ (Fear ends up in V via head-to-head movement.) (52) [Johni [V [the war [PROi [P [fear]]]]]] This analysis would account for why the stative measure verbs do not passivize, while the experiencer verbs do so; compare One pound was weighed by the metal versus the war was feared by John. In e¤ect, as we know, subject-to-subject raising structures do not passivize, while control structures do. We will not attempt to develop this line of analysis any further and will not address the issue of the representation of stative predicates again in this work. We return to structures that encode change of state or location. Assuming that the line of analysis outlined earlier for predicates of ‘‘change of state or location’’ is on the right track, it entails the following: (53) The verbs go and come do not have intrinsic meaning. Other than deixis, the meaning of go and come is purely eventive in nature. To the extent that the syntactic structure itself encodes that eventive meaning, such verbs do not contribute anything of major relevance to

20

Chapter 1

the meaning of the sentence. Following Goldberg’s (1995) terminology, we may then say that (54) The light verbs go, come, become are the morphological signature of a constructional meaning. On the other hand, unlike Goldberg, we assume that the constructional meaning arises solely from the composition of syntactic categories, as outlined above. To recapitulate, the verbs go and come are the morphological signature of the aspectual meaning associated with structure (36a) (repeated in (55a)). Become is the morphological signature of the aspectual meaning associated with structure (41a) (repeated in (55b)). In some restricted cases, go can function as the morphological signature of the aspectual meaning associated with structure (41a) (repeated in (55b))— that is, in cases where A dominates certain specified lexical items (crazy, sour, stale, . . .). Thus, we have John went crazy, the milk went sour (alongside John became crazy, the milk became sour), but not John went famous or the water went frozen. Compare the latter with the well-formed John became famous and the water became frozen.10 (55) a. [VP D [V [P [P [D]]]]] (Change of location) b. [VP D [V A]] (Change of state) Within the theory outlined here, what does (54) mean exactly? Before we answer this question, we must make precise the notion of ‘‘lexical item’’ as we intend to use it here. A lexical item is composed of several types of information: a pointer to a concept (call it C), a bundle of phonological features (call it P-features), and possibly some formal features such as tense in the case of verbs (call it FF) (e.g., Chomsky 1994, 1995). Thus, run, kill, eat, book, love, fear, sad, and so on are just convenient shorthand labels for complex objects. And not all the properties of lexical items are relevant for the syntactic composition of phrases. In particular, P-features and C-features are not. Marantz (1997) has suggested that roots are unspecified for category type, the latter being dependent on the syntactic environment in which the lexical item is inserted. We will assume that roots are generally unspecified (or underspecified) for category type, but perhaps more so in a language like English (which has a prolific use of so-called backformation) than in a language like Spanish. Following the framework put forth by Hale and Keyser, we assume that a lexical item also specifies the type of l-structure that it instantiates and it is this property of a lexical item (along with its FF) that plays a fundamental role in the syntactic computation.11 Thus, the lexical items

Preliminaries to the Lexicon-Syntax Relation

21

freeze and melt are category-neutral roots. These category-neutral roots are specified as being able to instantiate the l-structure in (55b), which expresses change of state (i.e., movement toward an endpoint state). More precisely, it is inserted under A and then moves to V, thus acquiring the categorial specification of a verb. (56) a. [The water [V froze [ A ( froze)]]] b. [The butter [V melt [ A (melt)]]] We now return to the case of go/come and become. The proposal here is that these items are not part of the lexicon and therefore cannot be part of the lexical array that serves as input to the syntax. The morphophonological form of the verbs go/come can only be determined after the composition of the phrasal structure in (55a). More precisely, go/come is the spell-out of V in the context of (55a). It will be go if the P expresses movement away from the speaker (i.e., endpoint) and it will be come if the P expresses movement toward the speaker (i.e., source point). The question then arises at what point in the derivation such spell-out occurs. We will adopt the assumption in (57). Categories defined as ‘‘phases’’ determine the domain of spell-out; see Chomsky 1999/2001. More precisely, Chomsky proposes that the complement of the head of a phasal category constitutes the domain of spell-out. On the other hand, Fox and Pesetsky (F&P) (2005) argue that the domain of spell-out should be identified with the phase itself, a proposal we adopt here.12 (57) CP and vP (or VP in the absence of v) are phases and phases constitute the domain of spell-out. Furthermore, we put forth the hypotheses in (58) and (59): (58) If VP in (55a) is a phase and its head V is lexically unspecified, the V is spelled out as go/come. (59) If VP in (55b) is a phase and its head V is lexically unspecified, the V is spelled out as become. In (55), the domain of spell-out is the VP. If the head of the VP in (55a) is empty, it will be spelled out as go. If the head of the VP in (55b) is empty, it will be spelled out as become. Thus, in (56a), if the root corresponding to froze remains in A, it will be spelled out as frozen and the phonologically empty V will be spelled out as become; see the water became frozen.13 As mentioned earlier, when A dominates specific lexical roots (such as crazy, stale, sour, and so on), an empty V in structure (55b) can also be

22

Chapter 1

spelled out as go (e.g., he went crazy, the bread went stale, the milk went sour). In section 1.6, we will summarize the consequences of (57) and (58) for the manner-of-motion constructions across di¤erent language types (Korean, Germanic, and Romance). 1.4.2 Transitives As discussed in the previous subsection, unergatives, unlike unaccusatives, are associated with an l-structure that lacks a Specifier. In the case of unergatives, the external argument is licensed by a vP shell; see (31). (We will refer to this level of structure as the ‘‘extended’’ l-structure of the verb.) On the other hand, the l-structure of unaccusatives does contain a specifier. Consider next the addition of a vP shell to the unaccusative structures (55a) and (55b).14 (60) [vP DP v [VP DP [V [P [P [DP]]]]] (61) [vP DP v [VP DP [V [AP]]] The above structures are associated with a causative meaning. This is illustrated by the sentences in (62) and (63), which are associated with the structures in (60) and (61), respectively. (62) a. b. c. d. e.

John sent the package to Paris. John carried/took the package to Paris. John gave the package to Mary. John brought the package from Paris. This medicine will up your blood pressure.

(63) a. b. c. d.

The cold froze the water. John broke the window. Mary lowered the temperature. Mary lengthened the pants.

To exemplify, sentences (62a) and (63a) would have the structures in (64a) and (64b), respectively. (64) a. [vP John [v sent v [VP the package [V (sent) [ PP to [P [Paris]]]]]]] b. [vP The cold [v froze v [VP the water [V (froz-) [ A (froz-)]]]]] To recapitulate, in the present theory, the causative meaning does not arise from any (abstract) lexical item. It is a meaning associated with the construction itself, which can be summarized as in (65). (65) [vP DP v [VP DP [V XP]]]

Preliminaries to the Lexicon-Syntax Relation

23

In this structure, the Spec of v is interpreted as bringing about (or causing) the event denoted by the embedded VP. To (54), we can therefore add the following generalization:15 (66) The causative meaning is not associated with some abstract lexical item. In the unmarked case, causation is the meaning derived from the structure in (65).16 Indeed, in the unmarked case, the causative meaning is derived from a structure formed by merging little v with an unaccusative structure. The above generalization follows from the fact that the Specifier of the embedded VP is interpreted as the Causee. Since the embedded VP in an unergative structure lacks a Specifier, there is no position that can be interpreted as the Causee of the embedded subevent.17 While all languages allow lexical causatives to be formed with unaccusative verbs and many languages allow causative formation only with unaccusatives, only some languages allow lexical causative formation with unergatives verbs (see Hale and Keyser 2002). It is therefore justified to consider the latter cases as marked. Lexical causatives formed with unergatives may be assumed to involve a higher ‘‘causative’’ structure. More precisely, while the unmarked case is for languages to allow at most one vP shell per VP structure (cf. (32)), some languages allow a recursive vP shell (as in (67)), making it possible for causative structures to be formed on the basis of unergative verbs (such as the clown laughed the children). The structure in (67) gives rise to an interpretation in which the higher vP shell is interpreted as bringing about the activity denoted by the lower vP shell. (67) [vP D [v [vP D [v [VP V D]]]]]

(Marked causative structure)

A final note regarding causative structures: In the unmarked case, the causative construction is formed by merging little v with an unaccusative structure. Therefore, transitives like catch a fly as well as transitives headed by ‘‘verbs of consumption’’ like eat an apple or smoke a cigarette should not be analyzed as causatives because there is no evidence that they contain an unaccusative structure. While some of the situations in the outside world associated with these sentences may be conceptualized as an object moving along a path (e.g., an object moving through the digestive organs in the case of an apple being eaten), it does not imply that they are grammaticalized as such. Tenny 1994 (building on ideas put forth by Dowty 1991 and Krifka 1989, 1992) argues that the object of verbs of

24

Chapter 1

consumption and creation denotes a path that measures out the event denoted by the VP and thus such verbs share some aspectual commonality with verbs of directed motion. In her theory, verbs of directed motion and ‘‘verbs of consumption’’ share a notion of ‘‘core event’’ (which is derived from the common aspectual role ‘‘path’’). Although both eat an apple and go to school give rise to bounded events, we will assume that they have fundamentally di¤erent l-syntax. Indeed, there is no reason to assume that the l-syntax of the bounded VP eat an apple is any di¤erent from that of the unbounded VP eat apples. Such transitive verbs are associated with an l-structure that contains a complement but no specifier: (68) [VP V D] As with unergatives, a v is merged with the structure in (68), giving rise to the transitive structure in (69). Such a structure will not be interpreted as causative because, as mentioned earlier, the notion of causation is derivative in this framework. (69) [vP D [v [VP V D]] To close this section, we note that the VP structures discussed above are themselves embedded under a series of functional projections. We are not going to address in any detail the nature of these functional projections, but it must be mentioned that above the (extended) l-syntax, we find the so-called inflectional categories, in the specifier of which Nominative and Accusative Cases are licensed. It is generally assumed that Nominative Case is licensed in the Spec of Tense. More controversial is the position for licensing of Accusative Case. Some have identified it with an abstract Object Agreement, in line with a proposal put forth by Chomsky 1991. Others have identified it with an ‘‘inner’’ Aspect Phrase projected immediately above the VP (e.g., Borer 1994; Ritter and Rosen 1998; Megerdoomian 2002; Travis 2000).18 As argued by these authors, this assumption allows us to articulate the relation between the aspectual properties of the object and the aspectual properties of the VP. In this work we will endorse the view that T and Asp play a role in licensing Nom and Acc, respectively.19 The general picture that emerges is summarized below. As we will see later, a series of VPs uninterrupted by functional projections is one core property of serial-verb constructions (i.e., a necessary but not su‰cient condition). In a serial-verb construction, a series of VPs share the same set of functional projections: one Asp node and one Tense node immediately above the highest VP.

Preliminaries to the Lexicon-Syntax Relation

25

(70) a. Unaccusative structure [D1 [T . . . [Asp [D [V XP]]]]] b. l-causative structure [D1 [T . . . [D2 [Asp [D1 v [D2 [V XP]]]]]]] c. Unergative structure [D1 [T . . . [Asp [D1 v [V D]]]]] d. Non–l-causative transitive structure [D1 [T . . . [D2 [Asp [D1 v [V D2 ]]]]]] 1.5

Event Boundedness and Temporal Boundedness

In this section, we briefly discuss the notions of event boundedness and temporal boundedness, since they are relevant to the discussion of manner-of-motion verbs. 1.5.1 Disentangling Event Boundedness (Telicity) and Temporal Boundedness This section is based on Zagona 2004, which contains an illuminating discussion of how e(vent) boundedness and t(emporal) boundedness are articulated in the grammar. As mentioned in note 2, four main event types have been recognized (see Vendler 1957): (71) 1. States involve no change throughout an interval. (own a house, know a poem, love a person) 2. Activities involve a process with transitions from one state to another (i.e., it is durative), but lacking a natural endpoint. (run, laugh, drink juice, push a cart) 3. Accomplishments involve a process with transitions from one state to another state toward a natural endpoint. Since it contains substates, it is durative. (run a mile, run to the park, draw a circle, eat an apple) 4. Achievements involve a process that consists of a transition to an endpoint state from an immediately preceding state. It contains no intermediate substates, and is therefore nondurative. (reach the top, notice a problem) Because activities lack an endpoint, they are homogeneous processes. In e¤ect, in the absence of a telos, the substates that compose an activity cannot be di¤erentiated from each other. On the other hand, because

26

Chapter 1

accomplishments have a telos (or endpoint), the substates that compose this type of process are nonhomogeneous.20 In e¤ect, any given substate will be viewed as closer to or farther from the telos than some other substate. Therefore, the substates involved in an accomplishment can be differentiated from each other. The endpoint (or telos) of an event is a property encoded in the l-structure by a PP complement in verbs of directed motion (72a) or by a determinate object in the case of verbs of consumption (72b) and verbs of creation (72c). (72) a. John ran to the park (in an hour). b. John ate the apple (in a minute). c. John drew a circle (in a second). In the Reichenbach framework on ‘‘Times,’’ Speech time (S) and Event time (E) are mediated by Reference time (R). While S (moment of speech) and E (moment of the event) are intuitive notions, R is a more abstract theoretical notion that serves to explicate the concepts of ‘‘more past than normal past’’ (past perfect tense, such as John had left) and ‘‘past in the future’’ (future perfect tense, such as John will have left). According to Zagona 2004, the relation between R and E is an aspectual one, not a tense-ordering relation. In the absence of Viewpoint aspect (i.e., the progressive and the perfect), R and E are simultaneous. In the present tense, R and E are simultaneous with S, while in the simple past, R and E precede S. Since Viewpoint aspect is absent in (72), R and E are simultaneous. Because the examples in (72) denote processes with a natural endpoint or telos (they are accomplishments), the temporal adjunct introduced by the preposition in can measure R, which coincides with E. The in-temporal phrase thus measures the temporal span of the event. On the other hand, the examples in (73) denote processes that lack an endpoint; they are activities. Therefore, these cannot be modified by a telic intemporal phrase, but they can be modified by the durative adverb introduced by the preposition for. (73) a. John ran (for an hour/*in an hour). b. John ate (for an hour/*in an hour). c. John laughed (for an hour/*in an hour). Zagona points out that even in the absence of Viewpoint aspect, there are certain types of verb phrases in which there are systematic mismatches between the duration of R and of E. Zagona discusses two types of mis-

Preliminaries to the Lexicon-Syntax Relation

27

matches. Below, we will discuss one of them, namely, cases that involve iterated events within an interval modified by a durational adverb. The iterative reading arises in the presence of a bare plural object or subject (see Verkuyl 1993 and the references cited there): (74) a. John broke glasses for an hour. b. John read abstracts for an hour. (75) a. Trees fell over for an hour. b. People voted for an hour. The above examples involve iterative events, in which each event is a subinterval of R. Each event (which corresponds to a subinterval in the R-timeline) is bounded, but R (modified by the durative adverb) is unbounded. The mismatch can be best appreciated with an example such as (76). It involves iterative events of picture framing. Each pictureframing event constitutes a subinterval in the R-timeline. On the other hand, each picture-framing event is bounded. While the durative phrase modifies the unbounded set of subintervals, the in-temporal phrase modifies each subinterval that, as just mentioned, corresponds to a telic event. (76) John framed pictures in an hour for years. As is well known, determinate plurals block the type of mismatch described above: (77) Fred ate the apples in a minute/*for a minute. Yet Zagona points out that if the event ‘‘ate the apples’’ is individuated with the help of the modifier ‘‘one-by-one,’’ then the iterative reading becomes available, giving rise to an unbounded event: (78) Fred ate the apples one by one for an hour. Similarly, note that activities can be individuated by the modifier ‘‘in bouts of ’’ and each individuated bout of activity can be modified by a telic temporal phrase (e.g., ten minutes). This is illustrated by the example below, which involves iterative bouts of laughter, each of which lasts ten minutes. Each individuated laughing event constitutes a subinterval in the R-timeline, which is unbounded (as indicated by the temporal adverb for an hour). (79) John laughed in bouts of ten minutes (for an hour). To conclude, the temporal adverbs (i.e., the telic in-phrase and durative for-phrase) modify the Reference time or R. In the absence of Viewpoint

28

Chapter 1

aspect, the (un)boundedness of R is determined by the properties of the event. If the event is singular and telic, R will be bounded. If there is a plurality of events and if each event can be individuated and mapped onto a subinterval in R, then each subinterval of R will be bounded (although the iteration of such subintervals is unbounded). 1.5.2 Some Clarification on the Notion of Endpoint At this point, it will be useful to clarify the grammatical status of endpoint. When we say that an l-structure specifies a telos, it means that it specifies a reference point in terms of which the endpoint can be computed. Thus, in the directed-motion reading of (80), the bridge is said to be the grammatical telos because it is the reference point in terms of which the endpoint of the movement is computed, namely, any point that is perceived as being on the other side of the bridge. (80) The boat floated under the bridge. Furthermore, we must distinguish between having a telos and reaching a telos. Directed motion requires a telos toward which movement is directed, but it does not necessarily assert that the telos has been reached; compare (81a) and (81b). (See note 16.) (81) a. The boat floated toward the bridge. b. The boat floated up to the bridge. Scalar verbs like increase/decrease, lengthen/shorten, and many others are also worth mentioning because they have led some scholars to reject the idea that telicity is a relevant grammatical concept. In particular, Hay, Kennedy, and Levin (1999) argue that the endpoint of an event may be entirely determined on the basis of world knowledge. These authors give the paradigm in (82)–(85) to illustrate the dependency of telicity on world knowledge. They use an entailment test (attributed to Vendler 1957 and Dowty 1979) to determine (a)telicity. Atelic predicates are entailed by their progressive forms, while telic predicates are not. The di¤erence between (82)–(83) and (84)–(85) lies in the fact that we have a conventionalized notion of ‘‘lengthened pants’’ and ‘‘lowered blind,’’ but not of ‘‘lengthened commute’’ and ‘‘lowered heat.’’ Therefore, the endpoint is flexible in (84)–(85) and any amount of lengthening or lowering (whatever it might be) counts as the endpoint. (82) The tailor is lengthening my pants 6) The tailor has lengthened my pants

Preliminaries to the Lexicon-Syntax Relation

29

(83) Kim is lowering the blind 6) Kim has lowered the blind (84) The tra‰c is lengthening my commute ) The tra‰c has lengthened my commute (85) Kim is lowering the heat ) Kim has lowered the heat As Hay and colleagues show, the fact that the notion of endpoint can be vague or relative to the discourse situation is also true for verbs of creation and consumption: (86) a. She ate the sandwich, but as usual she left a few bites. b. She drew a house, but it was missing a door. If a speaker is happy to accept a variable endpoint in (84)–(85), that speaker should be equally willing to take that variable endpoint as determining the temporal span measured by the in-phrase. For such a speaker, (87b) should be as acceptable as (87a): (87) a. Kim lowered the blind in a minute. b. Kim lowered the heat in a minute. Similarly, if a speaker is willing to accept (86) as noncontradictory, that speaker should be equally willing to accept (88): (88) a. She ate the sandwich in a minute, but as usual she left a few bites. b. She drew a house in a minute, but it was missing a door. What the above discussion shows is that the notion of ‘‘telos’’ is not absolute. Yet this notion is crucial in the grammatical characterization of directed motion. The l-structure of verbs of directed motion, including scalar verbs, encodes movement toward a telos, where the telos can be entirely specified, or it can be vague, or it can have a variable interpretation and be subject to pragmatic considerations. In the case of scalar predicates, movement is along an abstract path in the direction specified by a variable argument ‘‘x-amount more’’ or ‘‘x-amount less.’’ This variable measure can be left unspecified by the grammar, in which case pragmatics can ‘‘fill in’’ the information (as in the cases discussed earlier). Alternatively, the grammar provides an adverb that specifies the variable measure, as in (89). The status of the measure phrase is thus comparable to that of the implicit agent in verbal passives. Indeed, it has long been recognized

30

Chapter 1

that in verbal passives there is a covert indefinite argument, which can remain unspecified or can be modified by an overt by-phrase; see (90). (89) a. The tailor lengthened my pants by 3 inches. b. Kim lowered the heat by 3 . (90) a. My pants were lengthened (by the tailor). b. The heat was lowered (by Kim). 1.5.3 Duration as an Interpretational Feature A final comment with regard to the event types summarized in (71). Note that both accomplishments and achievements involve movement toward a telos; the di¤erence between them is that accomplishments are þdurative (the process is composed of internal substates) and achievements are
durative (the process lacks internal substates). In the system proposed in section 1.4, achievements and accomplishments are not structurally di¤erentiated—that is, they are associated with the same type of l-structure. The duration component is purely interpretational. More precisely, we will assume that achievements (such as arrive, shatter, glimpse, reach) are lexically specified as [
durative], but accomplishments and activities are unspecified for duration. In other words, in the unmarked case, a process is interpreted as [þdurative]. More precisely: (91) A process will be automatically interpreted as [þdurative], unless it is lexically specified as [
durative]. It follows from (91), in conjunction with (58) and (59), that the structures headed by go/come and become will receive a þdurative interpretation. Indeed, since these items are not part of the lexicon (i.e., they are the spell-out of V in a certain structural configuration), they cannot be associated with any type of lexical feature. The prediction is correct for the directed-motion constructions in English, as well as for the directedmotion constructions (headed by the light verbs ka-, o-, -(e)ci) and the causative constructions in Korean; see chapter 2 for extensive discussion. 1.6 A Crosslinguistic Analysis of Manner-of-Motion Constructions: A Preview of Chapters 2 and 3 The line of analysis that we will develop here is highly inspired by the Nicaraguan Sign Language (NSL) mentioned earlier in this chapter. Recall that at the pidgin stage in NSL, ‘‘manner’’ and ‘‘directed motion’’

Preliminaries to the Lexicon-Syntax Relation

31

were expressed simultaneously, but that at the creole stage, they get to be realized sequentially (i.e., in a serial-verb fashion). This strongly argues for the view that ‘‘manner’’ and ‘‘directed motion’’ are generally expressed compositionally in human language. We will articulate an analysis of these constructions, in which ‘‘manner’’ and ‘‘directed motion’’ are expressed compositionally at the syntactic level. The formal mechanism that we will develop is close in spirit (but not in details) to the one put forth by Mateu 2002. We first introduce the subject of serial verbs, their structure and how they are formed, with particular attention to Korean (section 1.6.1). We then discuss Germanic (section 1.6.2) and Romance, and in particular the di¤erences between Italian, on the one hand, and French and Spanish, on the other hand (section 1.6.3). Finally, we conclude and recapitulate the theoretical implications of the present analysis (section 1.6.4). 1.6.1 Serial-Verb Constructions (SVCs) and Directed Motion in Korean In the minimalist approach outlined first in Chomsky 1994 (see also Chomsky 1995), the general operation Merge is the basic mechanism by which phrase structure is constructed. Merge can apply to any two syntactic constituents to form a new syntactic object. The resulting syntactic object must receive a label and, given the compositional nature of language, this label must be calculated from the labels of its parts. In particular, no new features can be added in the course of the derivation (cf. the Inclusiveness Condition). The label of a constituent is ultimately identified with a particular lexical item that it dominates. Recall that a lexical item is defined as a set of features: phonological (P), a pointer to a concept (C), categorial features, and some set of formal features (FF). This set of features constitutes the label. There are therefore three possible options to determine the label of a constituent Z created by Merge of X and Y: (92) a. The label of Z ¼ the label of X b. The label of Z ¼ the label of Y c. The label of Z ¼ the union or the intersection of X and Y The option in (92a) gives rise to a phrase in which X is the head, the option in (92b) gives rise to a phrase in which Y is the head, and the option in (92c) is assumed to give rise to a biheaded phrase. Chomsky (1994) rejects (92c) as a possible option on the grounds that a phrase with such a label will not be able to undergo further computation. Note that under

32

Chapter 1

this theory, there is no intrinsic di¤erence between the label of a head and that of its phrasal projection. They have the same label and can only be distinguished contextually. Contra Chomsky, Baker and Stewart (1999) argue that (92c) is indeed an option, in particular, in languages with serial-verb constructions (SVCs), like the West African languages. A signature property of SVCs is as follows:21 (93) SVCs consist of a succession of verbs and their complements (if any) in a single clause with one Tense or Aspect value and one subject (e.g., De´chaine 1993; Collins 1997). Baker and Stewart (1999) put forth the following logic. In SVC languages, two verbs or verb phrases can be merged and the label of the output category is an intersection of the features of the two Vs. Indeed, since P-features and C-specification are irrelevant to the syntactic computation, they can be ignored. The output label can, in principle, be constituted solely by the categorial features and the formal features (FF) or a subset of these. Thus, if the two objects merged have the same categorial features and the same relevant FF, then (92c) is indeed an option—that is, X W YL ¼ X X YL ¼ XL ¼ YL . The question that then arises is why SVC is available in some languages but not others. Baker and Stewart suggest that the relevant parameter is whether or not the verb carries inflectional morphology, and in particular, tense specification. They argue that in the Kwa languages, and in particular, in Edo, the verb in most tenses is not morphologically specified for Tense. In such cases, tense is morphologically realized by an independent morpheme. Therefore, T does not attract V (overtly or covertly). On the other hand, in a language in which V is inflected for tense, T does attract V. In such a case, a biheaded verbal structure is impossible because T would not be able to find a unique head to attract. Baker and Stewart note that in Edo, there is one tense in which verbs are inflected. This is the case of the past perfective and as expected, serial verbs cannot appear in this tense. The formulation of Attract in (94) is proposed to achieve the desired result. It is furthermore assumed that functional heads between T and V (like Voice and v) are ignored by Attract. (94) X attracts a head Y i¤ Y can check a feature of X, and for all Z such that Z is not equal to Y and Z can check this feature of X, Y asymmetrically c-commands Z.

Preliminaries to the Lexicon-Syntax Relation

33

Baker and Stewart discuss three types of SVCs: Covert Coordination, Consequential SVC, and Resultative SVC. ` zo´ gha´ gbe´ e`we´ khie`n u`hu`nmwu`n (95) a. O ` zo´ FUT hit goat sell head O e´re`n. (Covert Coordination) its ` zo´ will kill the goat and sell its head.’ ‘O ` zo´ gha´ gbe` e`we´ khie`n. b. O (Consequential SVC) ` zo´ FUT hit goat sell O ` zo´ will kill the goat and sell it.’ ‘O ` c. Ozo´ gha´ gbe` e`we´ wu`. (Resultative SVC) ` zo´ FUT hit goat die O ` zo´ will strike the goat dead.’ ‘O Baker and Stewart assume that Voice (and not v) introduces the external argument. They argue that Covert Coordination (CC) arises when two Voice Phrases are merged, that the Consequential SVCs (CSVCs) arise when two vPs are merged, and that the Resultative SVCs (RSVCs) arise when two Vs are merged. Stewart 1996, on the other hand, argues that the CC involves two distinct Aspect categories, one associated with the first V and another associated with the second V, and that it is the presence of two distinct AspPs that allows for the licensing of two overt objects. We will follow Stewart 1996 and furthermore assume that the second Asp is merged with the first V, giving rise to a subordination (rather than a coordination) structure, as shown in (96). (96) [T [Asp [v [V . . . [Asp [V . . . ]]]]] Korean and Japanese both have CC and CSVCs, but they lack RSVCs. These languages also have another type of SVC (not discussed by Baker and Stewart but discussed by De´chaine 1993, among others), namely, the Simultaneous or Coevents SVC (or SSVC). We will not discuss the CC any further here and will restrict our attention to the other types— CSVC, RSVC, and SSVC. We note that Korean and Japanese verbs do inflect for tense, unlike Edo and other Kwa languages. Therefore, the analysis proposed by Baker and Stewart for these languages does not extend to Korean and Japanese. We conclude (contra Baker and Stewart) that the parameter that distinguishes the SVC languages from the nonSVC ones is not morphological. Following Larson 1991 and Nishiyama

34

Chapter 1

1998, we assume that such SVCs involve verbal subordination. More precisely, we suggest that22 (97) An SVC arises when a language uses the particular Generalized Transformation (GT) below: a. Merge a verbal l-structure with the head of another verbal l-structure. b. Merge a verbal lexical item with the head of a verbal l-structure. To exemplify, consider the Korean example in (98), which expresses a consequential relation between the two Vs. The reader must keep in mind that the English translation does not do entire justice to the meaning of the CSVC. It is crucial that the two events be connected. In the example below, the event of ‘‘gripping the rope’’ renders possible the event of ‘‘pulling the rope.’’ In other words, the first event is a necessary (although not su‰cient) condition for the second event to take place. See chapter 2 for further discussion of this point. (98) John-i cwul-ul cap-a tangki-ess-ta. John-Nom rope-Acc grip-L pull-Past-Decl ‘John gripped and then pulled the rope.’ Both cap- ‘grip’ and tangki- ‘pull’ instantiate the same type of l-structure, as shown below. For the sake of convenience, we will continue to use the old notation, whereby the syntactic category labels the lexical item. But we must keep in mind that the lexical item (cap- and tangki- in the case under discussion) is simply a shorthand notation for a set of features, among them the categorial feature (V in this case) and the formal features (FF). (99)

Preliminaries to the Lexicon-Syntax Relation

35

The SVC in (98) arises when the GT in (97a) merges the V-projection in (99b) with the V-head in (99a). The resulting structure is shown below, where the V in bold is the category created by the GT: (100)

Recall that little v is not part of the l-structure of verbs; it is the extended l-structure. It is inserted via the constructional convention in (30), giving rise to the structure in (101). (101)

We furthermore assume the following interpretative convention: (102) In an SVC, the DP in the Spec of the highest V is interpreted as the originator of the event denoted by each one of the Spec-less Vs that the highest V dominates. The structure in (101) is then merged with the relevant functional projections—for example, Neg, Asp, and Tense. Ultimately, Nom DP

36

Chapter 1

subjects are licensed by T and Acc DP objects are licensed in Asp via the operation ‘‘Agree’’ (see note 24). To the extent that there is one AspP, only one object argument will be overtly realized in the SVC. Suppose we assume with Baker and Stewart that the label of a category that results from Merge need only specify features relevant to the syntactic computation of phrases, such as categorial features and FF (but not Pand C-features). On the other hand, unlike Baker and Stewart, we will assume that there is no absolute notion of head. Headedness is relativized with respect to the type of computation involved. Therefore the root V in (100) is structurally ambiguous as to whether it is the projection of the first or second V. The morphosyntactic head is the one that realizes the tense (and other inflectional) feature(s) of the clause. Technically speaking, this will depend on which head checks the features of T and this is determined by the Minimality Condition on movement operations. Given Minimality, the highest V (with a phonological label) will be attracted to T.23 In a left-branching language like Korean, the highest V will be the last V, and indeed, the last V in the SV sequence morphologically realizes the tense feature in this type of language; see tangki- in (98).24 While the final V in an SV sequence is the morphosyntactic head in a left-branching language like Korean, it remains open as to which V gets computed semantically as the matrix event and which head gets computed as the subordinate event. Indeed, the semantic ambiguity in this type of structure was noted and extensively discussed by De´chaine 1993, a phenomenon she refers to as the ‘‘bivalency’’ property of SVCs. With this in mind, let us turn back to the CSVC. In the CSVC, the antecedent event (A) is perceived as a necessary condition for the consequent event (B) to take place. In other words, the worlds that contain B are a subset of the worlds that contain A. The relation between the two is then comparable to the well-known cause-result structures, where the cause constitutes the matrix event and the result the subordinate event. This suggests that in the CSVC, the consequent is subordinate to the antecedent (see Carstens 2001). If this is indeed the case, we must conclude that the first V in (100) is the semantic head of the SVC, while the second V is semantically subordinate. We then have a mismatch between the morphosyntactic head (the second V) and the head that introduces the matrix event (the first V). In chapter 2, we provide empirical evidence for the semantic headedness of the first V in the CSVC, based on adverbial modification. We note furthermore that the CSVC is not the only case where we have a mismatch between the notion of morphosyntactic head and the notion of

Preliminaries to the Lexicon-Syntax Relation

37

sem(antico)-syntactic head. As we will see in section 1.6.3, this is also the case in verbal structures with an auxiliary verb. The alternative order of verbs in (98) is unacceptable; see (103). But this is not due to a syntactic ill-formedness. Indeed, there is no syntactic principle that can exclude this particular order of verbs. (103) *John-i cwul-ul tangki-e cap-ass-ta. John-Nom rope-Acc pull-L grip-Past-Decl The ill-formedness of (103) is due to a PF interface condition, namely, the Temporal Iconicity Condition or TIC (Muysken 1988; Li 1993): (104) Temporal Iconicity Condition (TIC) In an SVC, if the events denoted by the SVs are sequential, the surface order of Vs must reflect the temporal ordering of events. We turn next to the RSVC. This type of SVC also encodes a consequential relation between the events denoted by V1 and V2: the event denoted by V1 is a necessary condition for the event denoted by V2 to take place. Yet the CSVC and the RSVC are grammatically di¤erent. While the CSVC consists of two transitive Vs, the RSVC consists of a transitive V and an unaccusative V. Interestingly, Baker and Stewart discovered an interpretational di¤erence between the object in the CSVC and the RSVC in Edo. They illustrate the di¤erence in meaning between the two with the following examples: ` zo´ su`a´ e`ha´n khe´rhe´ de`-le´. (105) O RSVC ` Ozo´ push tree few fall-PL ` zo´ pushed a few trees down.’ ‘O ` zo´ de´ e`be´ khe´hre´ tie´. (106) O CSVC ` Ozo´ buy book little read ` zo´ bought a few books and read them.’ ‘O ` zo´ bought few books and read all of The CSVC in (106) implies that O ` zo´ bought many them. It is not compatible with a situation in which O books but read only a few of them. The above contrast suggests that in the CSVC there are two events and the quantifier few quantifies into the first one only.25 (107) John v (few (bought x¼books)) & (read x¼books) On the other hand, the RSVC in (105) is compatible with a situation ` zo´ pushed many trees but most of them did not fall. Or in which O

38

Chapter 1

alternatively, it is compatible with a situation in which many trees fell, ` zo´ pushing them. In either situation, it but for some other reason than O ` zo´ pushis the case that there are few trees that fell as a consequence of O ing them. This suggests that in the RSVC, the two Vs conjointly give rise to a complex event ( push and fell ) and the quantifier few quantifies into this complex event: ` zo´ v (few ((pushed & fell) x¼trees)) (108) O Baker and Stewart take the meaning discussed above to suggest that the syntax of RSVC involves a complex V, which is predicated of the same object. Let us assume that this is achieved via the GT in (97b), which adjoins the verbal lexical item su`a´ ‘push’ to the head of the lstructure of de` ‘fall’. The latter is an unaccusative verb that encodes change of state, as shown in (109). Adjunction of su`a´ to the head of (109) gives rise to the structure in (110). The resulting structure is merged with little v, and after further adjunction of su`a´ to v, (111) is obtained. Su`a´ thus gets interpreted as the causing event. Note that the resulting structure obeys the TIC: su`a´ ‘push’ both temporally and linearly precedes de` ‘fall’.26 (109)

(110)

Preliminaries to the Lexicon-Syntax Relation

39

(111)

As noted earlier, Korean and Japanese lack the RSVC (see Li 1993; Nishiyama 1998). In these languages a transitive V cannot be combined with an unaccusative V, as illustrated by the ill-formedness of (112). ` zo´-ka yemso-lul ttali-e cwuk-ess-ta. (112) a. *O ` zo´-Nom goat-Acc hit-L die-Past-Decl O ` zo´-ka yemso-lul cwuk-e ttali- ess-ta. b. *O ` zo´-Nom goat-Acc die-L hit-L-Past-Decl O ` zo´ hitting it.’ Intended meaning: ‘The goat died as a result of O Recall that Korean and Japanese are head-final languages. In Korean and Japanese, functional projections, such as T, branch to the left. And this has consequences for an RSVC analysis. Consider the Korean examples in (112). The verb ttali- ‘hit’ must move out of the verbal compound and adjoin to little v; see (111). This verb will then be the highest V in the extended l-structure and will be attracted to T (to check the tense feature of T). Since in this language T is rightmost, ttali- should surface as the rightmost V. This is not the case in (112a) and it is therefore ruled out (as failure of V-to-T movement). In (112b), ttali- ‘hit’ has adjoined to little v and then to T. The syntactic computation is therefore felicitous. The problem with the output form in this case is that it violates the linear ordering required by the TIC: the consequent event cwuk- ‘die’ precedes (rather than follows) the antecedent event ttali- ‘hit’. (See Li 1993 for a similar conclusion within a di¤erent framework of analysis.) As we will

40

Chapter 1

see in chapter 2, the lack of RSVC in Korean has important consequences with regard to the typology of directed-motion constructions found in this language. Indeed, Korean lacks the variety of directed-motion SVCs that Yoruba and Edo have.27 We turn next to the third type of SVC, the Simultaneous (coevents) SVC (or SSVC). A manner-of-motion verb (such as heyemchi- ‘swim’) can be combined with the light verbs ka- ‘go’ or o- ‘come’, as illustrated in (113). As we will see in detail in chapter 2, the Korean manner-ofmotion verbs are unambiguously activity-denoting verbs, with an unergative structure; see (115). Light verbs, on the other hand, are the spell-out of V in the context of the directed-motion construction; see (55a) and (114). All of these structures are formed independently of one another; therefore, each constitutes a phase and a spell-out domain. The GT in (97a) forms the SVC in (113) by adjoining the l-structure in (115) to the head of the l-structure in (114). The morphosyntactic head of the resulting structure is unambiguously the second V, namely, the highest V in the structure. Therefore, it is attracted by T and, consequently, it realizes the tense inflectional feature of the clause. (113) John-i hoswu hanccok-phyen-ey heyemchi-e ka-ss-ta. John-Nom lake one side-side-Loc swim-L go-Past-Decl ‘John swam to one side of the lake.’ (114)

(115)

Preliminaries to the Lexicon-Syntax Relation

41

(116)

As for the semantic relation between the two Vs, the most salient meaning is one in which the first V denotes the manner in which the directed motion ‘‘going to the other side of the lake’’ was achieved; consider ‘John went to the other side of the lake swimming’.28 Under this interpretation, V2 constitutes the matrix event and V1 the subordinate event. Therefore, V2 is not only the morphosyntactic head of this type of SVC, but also the sem-syntactic head. In chapter 2, we will discuss in detail this type of SVC, which overtly exemplifies the compositional nature of manner and directed motion in natural language. 1.6.2 Directed Motion in Germanic In section 1.1, we mentioned that in Germanic, manner-of-motion verbs appear to exhibit a variable behavior systematically. They appear to be able to head a construction that denotes an activity (117) or directed motion (118). (117) John ran/walked/danced/swam for hours/*in an hour.29 (118) a. John ran/walked/danced to the park in an hour. b. John swam to the other side of the lake in ten minutes. Furthermore, it has been argued that the activity-denoting construction has unergative properties, while the directed-motion construction has unaccusative properties. This can best be appreciated in Dutch, in which the two types of constructions select a di¤erent type of auxiliary (heben versus zijn); see the examples in (2)–(3), repeated in (119)–(120). Futhermore, the locative PP in the unergative construction can be shown to have the grammatical status of an adjunct, while the locative PP in the

42

Chapter 1

unaccusative construction can be shown to have the status of a complement; see chapter 3 for detailed discussion. (119) a.

dat Jan that Jan b. ??dat Jan that Jan

naar to twee two

Groningen twee uur lang Groningen two hours long uur lang naar Groningen hours long to Groningen

heeft has heeft has

gewandeld. walked gewandeld. walked

(120) dat Jan in twee uur naar Groningen is gewandeld. that Jan in two hours to Groningen walked is ‘. . . Jan walked to Groningen in two hours.’ As mentioned earlier, postulating polysemy for each manner-of-motion verb would miss a robust generalization. Crosslinguistic data strongly suggests that manner-of-motion verbs do not select a PP directional argument (see the discussion of Korean in chapter 2). If that is indeed the case, then the PP complements in (118) and (120) are not complements of the lexical verb at all. Jackendo¤ (1983, 1990) proposes to analyze the English manner-of-motion construction exemplified in (118) as a case of subordination of manner with respect to an abstract verb GO at the semantic level of representation. Under such a view, GO is the main predicate, which is modified by the manner-of-motion verb. The meaning of the sentences in (118) is thus close (although not identical) to the ones in (121). Many authors who have studied the directed-motion construction in English and other languages have been inspired by Jackendo¤ ’s insight (see, for example, Goldberg 1995; Mateu and Rigau 2002; Mateu 2002; Spencer and Zaretskaya 1998). (121) a. John went to the park running/walking/dancing. b. John went to the other side of the lake swimming. In Germanic, there are also plenty of ‘‘cause–directed motion’’ cases in which directed motion is not entailed by the lexical verb. Indeed, Goldberg 1995 and Goldberg and Jackendo¤ 2004 point out examples like those in (122) and (123) as showing that the complementation properties in such cases cannot be attributed to the lexical properties of the verbs involved. (122) a. The professor talked us into a stupor. (Cf. *The professor talked us) b. The critics laughed the play o¤ the stage. (Cf. *The critics laughed the play)

Preliminaries to the Lexicon-Syntax Relation

43

c. He sneezed the napkin o¤ the table. (Cf. *He sneezed the napkin) (123) a. Bill belched/danced his way out of the restaurant. (Cf. *Bill belched/danced his way) b. Bill elbowed his way through the crowd. (Cf. *Bill elbowed his way) Verbs such as talk, laugh, sneeze are typically intransitive verbs that head a construction denoting an activity. A theory that assumes that complements are systematically arguments of the lexical verb would have to multiply the senses associated with these verbs. On the one hand, such verbs would be associated with an intransitive argument structure (with an activity meaning) and, on the other hand, they would be associated with a ditransitive structure (with a cause-motion-along-a-path meaning). Undoubtedly, some polysemy does exist in natural language. Nevertheless, a parsimonious approach to polysemy would trivialize the issue of the relation between form and meaning. Furthermore, and more worrisome, it would have to assume that languages that lack forms like (122) and (123) (such as the Romance languages) have radically di¤erent lexicons from English (and other Germanic languages). Goldberg 1995 and Goldberg and Jackendo¤ 2004 argue that, while the verbs talk, laugh, sneeze are not causative verbs, there is a causative construction that underlies the sentences in (122). The verbs may be analyzed as modifiers of the construction, giving rise to meanings close (but not identical) to the ones in (124). Similarly, it may be argued that while dance, belch, elbow are not causative verbs, there is a causative construction that underlies the sentences in (123),30 giving rise to meanings close (but not identical) to the ones in (125).31 (124) a. The professor made us go into a stupor by (excessive) talking. b. The critics got the play o¤ the stage by (excessive) laughing. c. He got the napkin o¤ the table by (excessive) sneezing. (125) a. Bill made his way out of the restaurant by belching/dancing. b. Bill made his way through the crowd by elbowing. The apparent variability of manner-of-motion verbs can be readily understood if the contribution of the verb is separated from the grammatical contribution of the construction. The path in (118), (122), and (123) is contributed by the construction, not the lexical verb. The lexical verb modifies the construction; it specifies the manner of the motion or the means by which the motion is brought about.

44

Chapter 1

The constructional approach outlined above distinguishes cases in which verbs ‘‘instantiate’’ a construction (by virtue of their lexical meaning) from cases in which verbs ‘‘modify’’ a construction. In a nutshell, the distinction can be described as follows: Instantiation This is the typical case, where there is a one-to-one correspondence between the arguments of the lexical verbs and the arguments of the construction. For instance, among the verbs that instantiate the cause-directed-motion construction are hand, give, send, throw (e.g., John handed/gave the ball to the boy; John sent the letter to Mary; John threw the towel to the floor). The purest cases of ‘‘instantiation’’ of a construction are the so-called light verbs, in which the verb adds very little meaning beyond that which is encoded by the construction itself (such as deixis) (e.g., go/come; give/get). In Goldberg’s terms, such verbs are the morphological signature of constructional meaning. They belong to the closed class of lexical items, alongside prepositions. Modification In this case the main syntactic verb constitutes the modifier of the construction. Some of the arguments are arguments of the construction alone, and not of the verbal lexical item. In section 1.4, we outlined a structural conception of the constructional approach. Within that framework, we can state the following generalizations for Germanic: (126) In Germanic, there is a compositional analysis for ‘‘manner’’ and ‘‘directed motion.’’ The structure in (55a) (¼ (128)) encodes the meaning of ‘‘directed motion,’’ which is modified by ‘‘manner.’’ (127) In Germanic, there is a compositional analysis for ‘‘means’’ and ‘‘cause-directed motion.’’ The structure in (60) (¼ (129)) encodes the meaning of ‘‘cause-directed motion,’’ and ‘‘means’’ modifies a subevent encoded by this structure. (128) [VP D [V [P [P [D]]]]]

(Directed motion)

(129) [vP DP v [VP DP [V [P [P [DP]]]]]

(Cause-directed motion)

Yet the Germanic languages are not SVC languages, in the sense that they lack (97). How then do they compose manner and directed motion? Germanic has another property that we believe is relevant and that was first identified by Snyder (1995, 2001). This author proposes that the presence/absence of certain types of resultative structures be related to the presence/absence of productive compounding in a given language. More precisely,

Preliminaries to the Lexicon-Syntax Relation

45

(130) Germanic has productive and compositional N-N (root) compounding. Such compounds are created by the following mechanism: (131) Merge two lexical categories of the same categorial type. Thus, English can combine any two nouns freely, and the semantic relation among them is open. Roeper, Snyder, and Hiramatsu (2002) note that this is not the case in Romance. These authors illustrate the difference between the two sets of languages with the following examples. Unlike the English compound, the French compound only has an idiomatic meaning and admits no variation; compare *femme grenouille versus ‘‘frog woman.’’32 (132) a. homme grenouille (lit. ‘man frog’) [¼ ‘undersea diver’]. b. frogman [¼ ‘undersea diver’, or ‘man who collects frogs’, or ‘man resembling a frog’, or ‘man who sells statues of frogs’, etc., ad infinitum]. We note that overt V-V compounding is virtually nonexistent in English and we conjecture that this is due to the fact that in simple tenses, English verbs associate with the functional category Tense in the absence of the dummy auxiliary do.33 V-V compounding would therefore lead to an intolerable ambiguity with respect to attraction by Tense because both members of the compound are equidistant from T; see (94). On the other hand, if one of the Vs in the compound is not a lexical item at all (no Pfeatures and no C-specification), then V-V compounding should be possible, if we make the reasonable assumption that Tense unambiguously associates with the V specified with P-features. Indeed, if V lacks Pfeatures, then there will be no morphological realization of Tense. We suggested earlier (section 1.4) that there are some lights verbs, in particular go and come, that are not lexical items listed in the lexicon. They are the spell-out of V in a certain syntactic context. It is precisely this kind of element that is a candidate to appear as a null V when it merges with a fully specified verbal lexical item. We develop this point below. At this point, we part ways with Beck and Snyder (2001), who put forth a semantic rule for generating semantic complex predicates in Germanic that applies both to word-level compounds and phrasal-level resultative structures.34 Instead, we advance the hypothesis that Germanic makes use of the syntactic Compound Rule (131) to compose directed motion and manner. More precisely, we propose that a manner verb

46

Chapter 1

(such as dance, run, swim, and so on) can modify the directed-motion construction in (128) in Germanic by merging the manner verb with the phonologically empty V in that structure, thus creating verbal compounds such as the one exemplified below.35 (133) [John [VP [V dance V] [to [P the kitchen]]]] Similarly, contact verbs like kick, push, pull, and many others can modify the cause-directed-motion construction in (129) in Germanic. This arises when a lexical item such as kick, push, pull, and so on is merged with little v. (134) [John [kick v [the ball [V V [to [P the garden]]]]]] The question that we must then address is the following. Why is it that the light V in the above structures cannot be lexicalized by the verb go? Indeed, why is it that forms like John danced go to the garden and John kicked the ball go to the garden never surface? The answer lies in the assumption formulated in (57) and (58), and repeated below. (135) CP and the highest verbal phrase in the (extended) l-structure are phases. Phases are the domain of spell-out rules.36 (136) If VP in (55) (¼(128)) is a phase and its head V is morphophonologically empty, then V is spelled out as go/come or become. Consider the derivation of (133). Given the proposal that the light verb go has no inherent lexical meaning and is nothing other than the morphological spell-out of the meaning of a construction, it follows that Merge at the lexical level cannot produce [V dance go]. The lexicon does not contain a verb go; therefore go is not part of the lexical array on the basis of which a verbal compound [V dance go] can be generated via Merge. On the other hand, dance can merge with empty V. The derived compound [dance V ] is then merged with the PP [to P the kitchen], and the output is merged with the specifier John, giving rise to the output in (133). VP is a phase (and therefore a domain for spell-out rules to apply), but its head is not empty. The head of VP is a compound and one of its members has morphophonological content. Assumption (136) therefore fails to apply, and compound forms like [V dance go] fail to be generated at that point in the derivation as well. Likewise, the structure in (134) cannot generate forms like John kicked the ball go to the garden because the lower V projection does not constitute a phase. Indeed, (135) defines the highest V-

Preliminaries to the Lexicon-Syntax Relation

47

projection in the (extended) l-projection as the verbal phase. Therefore, the lowest V fails to be spelled out. Finally, we would like to point out that Merge applies to syntactic objects independently of their morphological status. As mentioned earlier, Snyder (1995, 2001) proposes to relate the existence of resultative structures (which this author analyzes as semantic complex predicates) to the existence of root compounding. While we believe that Snyder is right in relating resultatives to word formation (i.e., compounding in particular), we do not think that the relation should be restricted to inflectional morphology. It should be extended to include derivational morphology as well. Indeed, the mechanism responsible for generating root-root compounding is no di¤erent from the one responsible for generating stem-stem compounding, namely Merge. Russian, which Snyder (1995) considered a counterexample to his own proposal, provides evidence for such a unified view. Consider some Russian examples from Spencer and Zaretskaya 1998, 28. (137) a. One v-bezala v magazin. she V-ran into the shop-ACC ‘She ran into the shop.’ b. Rebenok pod-lez pod stol. baby POD-crawled under table-ACC ‘The baby crawled under the table.’ c. Mjac pere-katilsja cerez dorogu. ball PERE-rolled across the road-ACC ‘The ball rolled across the road.’ Spencer and Zaretskaya note that the prefix is obligatory to obtain the directional meaning, even in the perfective aspect. On the other hand, the accusative marked locative is optional: it has the status of an adjunct, the function of which is to specify further the path meaning encoded by the prefix. These authors put forth an analysis of complex predicates in which the prefix is the main predicate and the verb is the modifier. We propose to modify their analysis slightly in the following way. In Russian the path is morphologically a prefix. The path-denoting prefix merges with a morphologically empty V, giving rise to the structure in (138a). The derived V then merges with a manner-of-motion verbal stem, giving rise to the complex verbal form in (138b). Under this analysis, the prefix is not the semantic head of the word; the semantic head is an empty light V to which the prefix is adjoined. This combination of prefix-V constitutes

48

Chapter 1

a verbal stem that forms a stem-stem compound with the manner-ofmotion verbal stem. (138) a. [VP DP [V a‰x(path) V]] b. [VP DP [V [V a‰x(path) V] V(manner) ]] We illustrate below the resulting structure with a concrete example; see (137b): (139) [ DP Rebenok [V [V pod-V] lez]] baby prefix-V crawled 1.6.3 Directed Motion in Romance As mentioned earlier, the Compound Rule is not productive in Romance. More precisely, (140) In Romance, the Compound Rule is lexically restricted and its output is semantically frozen. Because the Compound Rule in Germanic is lexically unrestricted, it can merge a lexical item with an empty V. On the other hand, because the application of the Compound Rule in Romance is restricted to specified combinations of lexical items, it cannot be extended to cases that involve an empty V—that is, it cannot merge a lexical item with an element that consists only of a category type. Consequently, Romance cannot use the Compound Rule to compose ‘‘manner’’ and ‘‘directed motion’’ in the way Germanic does. As predicted, a close examination of Spanish reveals that mannerof-motion verbs (with a few rare lexical exceptions) cannot appear in the directed-motion construction; see (141) and section 3.2.1.1, where apparent dialectal variations are also discussed. Nor do we find cause-directedmotion cases in Spanish such as the ones in (142). The contrast between English and Spanish has been discussed and analyzed by Talmy 1985, Aske 1989, Morimoto 2001, Mateu and Rigau 2002, and Mateu 2002, among others. French patterns with Spanish; see Bergh 1948, Lamiroy 1983, Boons 1987, and section 3.2.1.2. (141) a. *Juan balio´ a la cocina. ‘Juan danced to the kitchen.’ b. *Los atletas nadaron al barco. ‘The athletes swam to the ship.’ c. *La botella floto´ a la playa. ‘The bottle floated to the beach.’

Preliminaries to the Lexicon-Syntax Relation

49

(142) a. *Juan pateo´ la pelota al jardı´n. ‘Juan kicked the ball to the garden.’ b. *Juan empujo´ la carretilla al granero. ‘Juan pushed the cart to the barn.’ Italian is somewhat di¤erent from Spanish and French. In Italian, there is a subset of manner-of-motion verbs that can appear in the directedmotion construction. Folli 2001 gives the list in (143). The verbs in (143a) are unambiguous manner-of-motion verbs; they can appear only in the unergative, activity-denoting structure. The verbs in (143b) can appear either in the unergative activity-denoting structure or in the unaccusative directed-motion structure.37 (143) a. camminare (walk), galleggiare (float), galoppare (gallop), danzare (dance), nuotare (swim), sciare (ski), passeggiare (walk around), vagabondare (wander) b. correre (run), rotolare (roll), rimbalzare (bounce), scivolare (glide, slide), gattonare (crawl), saltare ( jump), volare (fly), saltellare (hop) The di¤erence between the two classes of verbs is clearly indicated by the choice of auxiliary and by the type of complementation (the correre class, unlike the camminare class, can take a PP complement headed by the preposition a). Particularly noteworthy is the fact that the correre class takes the auxiliary essere ‘be’ when it appears in the directed-motion construction, which is typical of unaccusative constructions. This is an especially robust fact, not shared by Spanish (which uses the auxiliary haber across all contexts), nor by French (which no longer exhibits ‘‘restructuring’’ properties—a relevant fact, as we suggest below). Maria e` corsa a casa. Maria is run-3rd p.s.fem. to house ‘Maria has run to the house.’ b. *Maria e` camminata a casa. Maria is walked-3rd p.s.fem to house Cf. Maria a camminato fino a casa. ‘Maria has walked up to the house.’

(144) a.

How then can we account for Italian? One possibility is that the verbs in (143b) are ambiguous; they can instantiate either an unergative or an unaccusative (resultative) structure. This is the line developed by Folli 2001. It is likely that some of these verbs (namely, the nonagentive roto-

50

Chapter 1

lare, rimbalzare, and scivolare) can indeed instantiate an unaccusative directed-motion construction, but this is unlikely in the case of agentive verbs such as correre, gattonare, saltellare, saltare. It is therefore worth raising the question as to whether the variable behavior of these verbs in Italian could be related to some other property of the language. In section 3.2.2, we explore the following hypothesis: (145) Italian recruits the auxiliary position designated for a class of restructuring verbs in order to compose ‘‘directed motion’’ and ‘‘manner’’ (in some lexically restricted cases). We suggest that the choice of auxiliary is what makes such a recruitment process robust and gives a clear signal to the learner that a position designated for a certain class of restructuring verbs has been extended to a particular manner-of-motion verb. We develop this idea below. In section 1.6.1, we presented the idea that the notion of headedness is a relative one. More precisely, it was suggested that there are structures in which a category C n functions as the morphosyntactic head while another category C m functions as the sem-syntactic head. We argued that such head ambiguity is found in certain SVCs. Another case in point is that of verbal auxiliaries. Indeed, in the structure below, either the verbal auxiliary or the lexical verb can be interpreted as the head of the highest V node. Because VAux is the highest V, it is attracted by T. This requires that the tense features be realized on VAux . We may therefore say that VAux functions as the morphosyntactic head of the verbal structure. On the other hand, VAux (by definition) is not associated with any l-structure, but Vlex is. Semantically, VAux is the modifier (i.e., specifier) of the verbal structure, while Vlex is the head of that verbal structure.38 We will therefore refer to Vlex as the sem-syntactic head of the maximal V projection above it. (146)

As is well known, Italian and Spanish (but not Modern French) have a class of modals (e.g., potere ‘can’, dovere ‘must’, volere ‘want’), aspectual

Preliminaries to the Lexicon-Syntax Relation

51

verbs (e.g., comenciare ‘begin’, continuare ‘continue’, solere ‘used to’), and motion verbs (e.g., venire ‘come’, andare ‘go’, tornare ‘return’) that trigger a phenomenon known as restructuring. We will refer to these as the Rverbs (or VR ). In both Italian and Spanish, these verbs give rise to clitic climbing (i.e., placement of the clitic on the auxiliary verb instead of the lexical verb) and long object preposing in middle constructions. But even more remarkably, in Italian, these verbs trigger an ‘‘auxiliary switch.’’39 To illustrate this phenomenon, we first state the generalization regarding Aux-selection in Italian. Whatever the ultimate account of Auxiliary choice in Italian might be, the generalization at the observational level is as follows:40 (147) a. Unaccusative VPs select Auxpp essere. b. If the node Auxpp has the impersonal si clitic or the reflexive clitic adjoined to it, Auxpp is essere. c. Otherwise, Auxpp is avere. The generalization in (147a) seems to be pretty much exceptionless (although judgments appear to be of a gradient nature with some subclasses of unaccusatives; see Soracce 2000). The contrast below (from Rizzi 1978) illustrates the phenomenon of auxiliary switch in ‘‘restructuring’’ contexts. In (148b), clitic climbing indicates that we are dealing with a ‘‘restructuring’’ context, and in such a context, the nature of Vlex is relevant in determining the choice of auxiliary. Indeed, venire heads an unaccusative structure and therefore the choice of auxiliary is essere. Compare, on the one hand, (148a) and (148b) and, on the other hand, (148b) and (148c).41 (148) a.

Maria ha dovuta venici molte volte. Maria has modal comeþloc.Cl many times ‘Maria has had to come there many times.’ b. Maria c’e` dovuta venire molte volte. Maria loc.Clþis modal come many times c. *?Maria ci ha dovuta venire molte volte. Maria loc.Clþhas modal come many times

Cinque 2004 argues extensively for the functional or auxiliary status of R-verbs. (See also Strozer 1976; Picallo 1985, 1990; Rochette 1988, 1990.) Cinque develops a very fine-grained hierarchy of functional projections for the clause and locates di¤erent semantic subclasses of R-verbs within such a hierarchy. We will not dwell on that issue here, but will assume at least three main VR positions (some of which are recursive): VMod (for

52

Chapter 1

modal verbs), VAsp (for aspectual verbs), and VMT (for motion verbs).42 (We ignore temporal categories in the schema given below.) (149) [VMod [VAsp [VMT [Vlex . . . ]]]]

(Italian and Spanish)

An example that illustrates the presence of all three types of R-verbs is given below (from Rizzi 1978): (150) Maria li avrebbe potuti stare per Maria acc.Clþwould-have-been able (Mod) be-on-the point (Asp) andare a prender lei stessa. go (MT) to get herself ‘Maria would have been able to be on the point of going to get them herself.’ We now return to the manner-of-motion cases—for example, (144a). Since correre and the other verbs in (143b) are unergatives, they cannot license the presence of the Auxpp essere, nor can they license the presence of a goal-denoting argument. The presence of these elements indicates the presence of a directed-motion structure headed by an empty V. Where then is correre located in the structure? We suggest that correre, and other agentive manner verbs in (143b), recruit the VMT position in (149) in order to modify the directed-motion construction. The structure of (144a), prior to undergoing Merge with T, will be as shown in (151). Furthermore, we assume that in Italian the verbal phase (which defines the spellout domain) includes VMT . Therefore, in a structure like (151), the head of the verbal phase is the VMT (corsa), which is morphophonologically specified. Consequently, the embedded V that heads the directed-motion structure remains phonologically unspecified; see the rule in (135). (151)

Preliminaries to the Lexicon-Syntax Relation

53

In section 3.2.2.2, we will discuss evidence for the above analysis, based on the phenomenon of infinitival final -e deletion, discussed and analyzed by Cardinaletti and Sholonsky 2004. It was suggested above that in Italian, the verbal phase can be extended to include VMT , thus giving rise to structures such as (151). The question then arises as to why it cannot be extended beyond VMT to include higher functional categories, such as VMod . One possibility is that if ‘‘phasal extension’’ occurs, it can do so only minimally within the functional hierarchy in (149). More precisely, only the verbal functional category immediately above the lexical verb in the schema in (149) can be included in the verbal phase. Interestingly, Wumbrand 2004 argues that in German there are two types of ‘‘restructuring’’: one functional (in which the ‘‘restructuring’’ verb has a functional status as suggested by Cinque 2004) and one lexical (in which the ‘‘restructuring’’ verb is a lexical verb that takes a small VP complement). According to Wumbrand, in German, only the modal verbs have a functional status. Therefore, for German, the schema in (149) should be replaced by (152) (again ignoring temporal categories): (152) [VMod [ . . . Vlex . . . ]]

(German)

If Wumbrand is correct and if indeed ‘‘phasal extension,’’ when it occurs, can only be extended to include the functional verb immediately above the lexical verb, we can expect to find among the Germanic languages a spell-out domain that includes VMod . The prediction is borne out. van Riemsdijk, forthcoming, has provided evidence for the existence of a phonologically empty verb in the directed-motion construction in Dutch, German, and Swiss German in the context of modals; see the examples below. One of the arguments that van Riemsdijk puts forth in favor of an empty V is that it explains a number of word-order puzzles—for example, the apparent right-edge position of directional PPs in subordinate clauses; see (154). Indeed, if the PPs in these examples are to the left of an empty final V, the observed word order is readily accounted for. (153) a. Du darfst nach Hause. (German) you may to home ‘You may hgoi home.’ b. Moeten wij nog de stad in. (Dutch) must we still the town in ‘Do we still need hto goi to town?’

54

Chapter 1

c. Si sott aber no in challer. (Swiss German) she should but still into-the cellar ‘But she should still hgoi down into the cellar.’ (154) a. . . . wil si het muese i d scheul. because she would-have-had-to into the school ‘. . . because she should have hgonei to school.’’ b. . . . das mer noni hand doorfe hai. that we not-yet have may (p.part) home ‘. . . that we were not allowed hto goi home yet.’ 1.6.4 Some Final Remarks The study of how the notions of ‘‘manner’’ and ‘‘directed motion’’ are concomitantly expressed provides support for the idea that grammatical meaning in natural language is highly compositional in nature. It furthermore suggests that lexical ambiguity (of the type that gives rise to distinct grammatical realizations) is not the norm in natural language, but rather the exception. That this generalization follows from innate principles of the language faculty is supported by the observation that children tend to associate one form with one meaning; consider the Uniqueness Principle put forth by Clark 1987 and Pinker 1989, and the Avoid Synonymy Principle argued for by Carstairs-McCarthy 1999. With Hale and Keyser 2002 and many other authors cited in this introductory chapter, we argue that the compositionality of meaning follows from the compositionality of the syntax (albeit an abstract syntax). What defines syntax is, on the one hand, the vocabulary (i.e., the objects that are manipulated are defined in terms of syntactic categories) and, on the other hand, the rules of syntax—for example, Merge and Move (or double Merge), syntactic distance computed in terms of Minimality, and sisterhood and extended sisterhood relations. The force of such a view does not appear from looking at one language, but rather at a variety of languages. A successful account is one that can account for the properties of each language, as well as for the variability among languages. Furthermore, a successful account of language variation is an account that can relate the di¤erences among languages to other independently established properties of these languages. In the case under discussion, namely the composition of ‘‘manner’’ and ‘‘directed motion,’’ we have identified the following relevant parameters:

Preliminaries to the Lexicon-Syntax Relation

55

(155) a. The grammar of language L makes use of the syntactic GT in (97). b. The grammar of language L makes use unrestrictively of the syntactic Compound Rule in (131). c. The language extends the syntactic VMT position in (151) to some manner-of-motion verbs. SVC languages (like Korean) make use of (155a). Non-SVC languages with lexically unrestricted compounding (like Germanic) make use of (155b). Non-SVC languages with lexically restricted compounding (like Romance) cannot generally compose ‘‘manner’’ and ‘‘directed motion’’ within the verbal projection, although a ‘‘restructuring’’ language with highly robust properties (like Italian) makes use of strategy (155c) for some lexically specified cases. Furthermore, the account put forth can provide an explanation for certain gaps in the ‘‘directed-motion’’ paradigm in Korean (chapter 2), as well as for what appears on the surface to be contradictory behavior on the part of certain manner-of-motion verbs in Romance (chapter 3). Finally, we comment briefly on other analyses in the literature that have argued for a syntactic composition of manner and directed motion: Mateu 2002 and McIntyre 2004. Mateu 2002 endorses Jackendo¤ ’s insight that the manner-of-motion verb is not the main semantic predicate of the construction, but rather the modifier of the construction. Furthermore, this author argues that this ‘‘subordination’’ relation is syntactically encoded and that the typological di¤erence between English and Spanish is syntactic in nature. Mateu assumes an abstract l-syntax in which lexical categories are derived from more abstract eventive features (transitional or change, terminal coincidence or result, and agentive source). Putting details aside, it is assumed that manner-of-motion verbs systematically have the syntax of unergative verbs and that in English, the l-structure of unergative verbs can merge and conflate with the head of the unaccusative construction. Thus, an example like John danced to the kitchen arises by merging the two structures in (156) via a generalized transformation, generating the structure in (157). The null head stands for an eventive (þtransitional) head.43 A process of ‘‘conflation’’ is assumed, by which a lexical item provides phonological content to a sister category; see Hale and Keyser 2002. Via the conflation process, the manner verb provides phonological content to the empty eventive head: q ! dance.

56

Chapter 1

(156) a. [John [q-Tr [to the kitchen]] b. [v dance X] (157) [John [v [v dance X] [q-Tr]] [to the kitchen]] Details aside, the above analysis is very close to the one we propose for SVC languages. But Germanic is not an SVC language, so we conclude that Mateu’s analysis of English is not quite on the right track. As for the crosslinguistic di¤erences between Germanic and Romance, Mateu adopts Talmy’s intuition that these languages cannot conflate ‘‘motion’’ and ‘‘manner’’ because they choose to conflate ‘‘motion’’ and ‘‘path.’’ Mateu’s analysis assumes a dependency between Merge and conflation: if Merge applies, then conflation must also apply. Because path and motion are systematically merged and conflated in Romance, the output cannot merge with manner since the head that encodes motion has already been provided phonological content by path. The problem with such an analysis is that the operations of Merge and conflation are logically independent. Merge does not entail conflation. Both can coexist in a given language, and there is no conceptual reason why conflation of ‘‘path’’ and ‘‘motion’’ should block merging of ‘‘manner’’ and ‘‘path.’’ In fact, this would be the case for Russian (discussed at the end of section 1.6.2), if analyzed in terms of conflation. Note furthermore that Mateu’s analysis denies the existence of phonologically empty null light verbs; it therefore cannot account for the Dutch and Swiss German data analyzed by van Riemsdijk (see the discussion at the end of section 1.6.3). The analysis proposed by McIntyre (2004) for English is very close to the one we propose here for Germanic. Like Mateu, this author assumes that the primitives involved are eventive semantic notions like ‘‘change’’ or ‘‘become’’ and ‘‘initiation’’ or ‘‘originator.’’ (158) a. [ChangeP Ethel [[Change dance þ GO] [into the room]]] b. [ IntP Ethel [[ Int dance þ INT] [ChangeP herself [CHANGE AP]]]] On the other hand, in the present work, we adopt the view that the aspectual interpretation is read o¤ from the composition of syntactic categories (a` la Hale and Keyser). This assumption is crucial if the crosslinguistic variations discussed in detail in the next two chapters (and summarized above) are to be understood in terms of variation in syntactic composition.

Chapter 2 The Serial-Verb Construction and Verbs of Motion in Korean

The first part of this chapter deals with the general structural and semantic properties of the serial-verb construction in Korean. A general understanding of the SVC is crucial if we are to gain some understanding of how ‘‘manner’’ and ‘‘motion’’ are articulated to express directed motion in a serial-verb language like Korean, a topic we address in the second part of this chapter. 2.1

Serial-Verb Constructions (SVCs) in Korean

We first discuss the grammatical characteristics of the SVCs (section 2.1.1) and argue that the Korean SVs are not compounds (section 2.1.2). We then turn to the semantic relations between the events encoded by the SVC in Korean: the Consequential SVC and the Simultaneous/coevents SVC (section 2.1.3) and how such structures are generated (section 2.1.4). 2.1.1 The Grammatical Properties of the SVC Korean serial verbs have the characteristic grammatical properties of SVCs (among others, see Jo 1990; Lee 1992; Chung 1993; Yi 1997; Y. Lee 2003). These properties are1 (159) a. All verbs are interpreted as having the same tense, aspect, mood, and polarity. b. There is only one overtly expressed syntactic subject. c. There is only one overtly expressed syntactic object. We exemplify these properties briefly below. Following the tradition in the literature, we will translate the Korean SVC examples with English coordinate sentences throughout this chapter. We note though that such translations are not accurate; they are merely approximations of the meaning conveyed by the Korean SVC.2

58

Chapter 2

Korean SVCs allow for one single tense specification, which scopes over the entire sequence of Vs:3 (160) John-i sayngsen-ul kwu-(*ess)-e mek-ess-ta. John-Nom fish-Acc broil-Past-L eat-Past-Decl ‘John broiled and ate the fish.’ Korean SVCs contain one and only one overt subject and at most one overt object, as exemplified below. (161) John-i sayngsen-ul (*Mary-ka) kwu-e mek-ess-ta. John-Nom fish-Acc Mary-Nom broil-L eat-Past-Decl ‘John broiled and (Mary) ate the fish.’ (162) John-i sayngsen-ul (*koki-lul ) kwu-e mek-ess-ta. John-Nom fish-Acc meat-Acc broil-L eat-Past-Decl ‘John broiled the fish and ate (the meat).’ Korean SVCs allow for one single polarity. This can best be appreciated in the case of negative polarity. Korean has two types of negation, namely, short-form and long-form negation. Short-form negation is realized as an and immediately precedes the verb; see (163a). Long-form negation is constituted by three morphemes: ci (nominalizer) þ an (negation) þ ha (the light verb ‘do’). For the sake of simplicity, we will use the contracted form of the long-form negation cianh. Long-form negation immediately follows the verb; see (163b).4 (163) a. John-i cwul-ul an cap-ass-ta. John-Nom rope-Acc Neg grip-Past-Decl ‘John did not grip the rope.’ b. John-i cwul-ul cap-cianh-ass-ta. John-Nom rope-Acc grip-Neg-Past-Decl ‘John did not grip the rope.’ In the SVC, short-form negation an immediately precedes the first V and long-form negation cianh follows the last V. Both negative morphemes can scope over the entire sequence of Vs. Thus, short-form negation in (164a), as well as long-form negation in (165a), can be interpreted as negating ‘broil and eat’ and contrasting with another verb, such as ‘throw away’ (e.g., John didn’t broil and eat the fish; he threw it away as soon as he got back from the market). Two negation forms may not co-occur; see (164b) and (165b).5

Serial-Verb Construction and Verbs of Motion in Korean

59

(164) a.

John-i sayngsen-ul an kwu-e mek-ess-ta. John-Nom fish-Acc Neg broil-L eat-Past-Decl ‘John did not broil and eat the fish.’ b. *John-i sayngsen-ul an kwu-e an mek-ess-ta. John-Nom fish-Acc Neg broil-L neg eat-Past-Decl Intended reading: ‘John did not broil and did not eat the fish.’

(165) a.

John-i sayngsen-ul kwu-e mek-cianh-ass-ta. John-Nom fish-Acc broil-L eat-Neg-Past-Decl ‘John did not broil and eat the fish.’ b. *John-i sayngsen-ul an kwu-e mek-cianh-ass-ta. John-Nom fish-Acc Neg broil-L eat-Neg-Past-Decl Intended reading: ‘John did not broil and did not eat the fish.’

While there are many semantic similarities between the subordinate structure introduced by the subordinate marker -ese and the SVC (see section 2.1.3 for discussion), they are syntactically distinct. Subordinate structures are biclausal, and therefore each clause has its own tense node. This is illustrated in (166). Note that in the case of the subordinate clause introduced by the subordinator -ese, the tense is not morphologically realized. Nevertheless, it can be inferred, based on the presence of ecey ‘yesterday’, that there is a tense node present in the subordinate structure, which is specified as ‘past’. Compare it with (160). (166) John-i ecey sayngsen-ul kwu-ese onul mek-ess-ta. John-Nom yesterday fish-Acc broil-Sub today eat-Past-Decl ‘Because John broiled the fish yesterday, he/she ate it today.’ As shown in (167), the subordinate clause can carry its own negative morpheme. (167) a. John-i sayngsen-ul an kwu-ese (kukes-ul) an John-Nom fish-Acc Neg broil-Sub (it-Acc) Neg mek-ess-ta. eat-Past-Decl ‘Because John didn’t broil the fish, he/she didn’t eat it.’ b. John-i sayngsen-ul kwup-cianh-ase (kukes-ul) John-Nom fish-Acc broil-Neg-Sub (it-Acc) mek-cianh-ass-ta. eat-Neg-Past-Decl ‘Because John didn’t broil the fish, he/she didn’t eat it.’

60

Chapter 2

In the subordinate structures, short-form negation bears scope solely over the verb that immediately follows it and long-form negation bears scope solely over the verb that immediately precedes it. Neither short- nor longform negation can bear scope over both Vs in this construction. (168) a. John-i sayngsen-ul an kwu-ese (kukes-ul) mek-ess-ta. John-Nom fish-Acc Neg broil-Sub (it-Acc) eat-Past-Decl ‘Because John didn’t broil the fish, she ate it.’ b. John-i sayngsen-ul kwu-ese (kukes-ul) an mek-ess-ta. John-Nom fish-Acc broil-Sub (it-Acc) Neg eat-Past-Decl ‘Because John broiled the fish, she didn’t eat it.’ (169) a. John-i sayngsen-ul kwup-cianh-ase (kukes-ul) mek-ess-ta. John-Nom fish-Acc broil-Neg-Sub (it-Acc) eat-Past-Decl ‘Because John didn’t broil the fish, she ate it.’ b. John-i sayngsen-ul kwu-ese (kukes-ul) mek-cianh-ass-ta. John-Nom fish-Acc broil-Sub (it-Acc) eat-Neg-Past-Decl ‘Because John broiled the fish, she didn’t eat it.’ Given the biclausal nature of these structures, it is unsurprising that each of the clauses can express its own subject and object as well: (170) a. John-i sayngsen-ul kwu-ese kukes-ul mek-ess-ta. John-Nom fish-Acc broil-Sub it-Acc eat-Past-Decl ‘Because John broiled the fish, she ate it.’ b. John-i sayngsen-ul kwu-ese Mary-ka kukes-ul John-Nom fish-Acc broil-Sub Mary-Nom it-Acc mek-ess-ta. eat-Past-Decl ‘Because John broiled the fish, Mary ate it.’ The data discussed above clearly show that the SVC is structurally di¤erent from the one involving subordination. While the subordinate structure is biclausal, the SVC is unambiguously monoclausal. SVCs are also distinct from coordinate structures. In coordinated clauses, each clause contains a tense marker (although the tense marker in the first clause can be deleted if identical to the tense marker in the second clause) and each clause can have a negative morpheme, as well as a distinct subject and object: (171) John-i sayngsen-ul kwup-(ess)-ko (kukes-ul) mek-ess-ta. John-Nom fish-Acc broil-(Past)-Conj (it-Acc) eat-Past-Decl ‘John broiled the fish and ate it.’

Serial-Verb Construction and Verbs of Motion in Korean

61

(172) John-i sayngsen-ul an kwup-(ess)-ko (kukes-ul) an John-Nom fish-Acc Neg broil-(Past)-Conj (it-Acc) Neg mek-ess-ta. eat-Past-Decl ‘John neither broiled nor ate the fish.’ (173) a. John-i sayngsen-ul kwup-(ess)-ko John-Nom fish-Acc broil-(Past)-Conj mek-ess-ta. eat-Past-Decl ‘John broiled the fish and Mary ate it.’ b. John-i sayngsen-ul kwup-(ess)-ko John-Nom fish-Acc broil-(Past)-Conj ‘John broiled the fish and (he) ate it.’

Mary-ka kukes-ul Mary-Nom it-Acc

kukes-ul mek-ess-ta. it-Acc eat-Past-Decl

The coordinate structure is also distinct from the SVC with respect to the scope of the negative morpheme. In particular, short-form negation can scope over the verb that it immediately precedes but cannot scope over the entire sequence of Vs. Thus, (174) can be associated with the meaning in (175a), but not with the meaning in (175b). (174) John-i sayngsen-ul an kwup-ko mek-ess-ta. John-Nom fish-Acc Neg broil-Conj eat-Past-Decl (175) a. Although John did not broil the fish, he ate it. b. John did not broil and eat the fish (he threw it away). Further evidence for the monoclausal property of SVCs is provided by several syntactic tests, namely, passivization, honorific agreement, and Negative Polarity Item (NPI) licensing; see Lee 1992 and Yi 1997. We will discuss the NPI licensing test here because we will use it throughout this chapter to distinguish the SVC from cases of covert subordination. NPIs in Korean require a clause-mate negation, as illustrated by the examples below (cf. Choe 1988). This requirement is met in (176a–b), but not in (176c–d).6 In the latter cases, the NPI and the negative morpheme belong to di¤erent clauses. This distribution violates the clauseboundedness condition on NPI licensing.7 (176) a.

John-un [amwuto kukos-ey an ka-ss-ta-ko] John-Top anyone there-Loc Neg go-Past-Decl-Comp sayngkakha-n-ta. think-Pres-Decl ‘John thinks that no one went there.’

62

Chapter 2

b.

Amwuto [John-i kukos-ey ka-ss-ta-ko] Anyone John-Nom there-Loc go-Past-Decl-Comp sayngkakha-cianh-nun-ta. think-Neg-Pres-Decl ‘No one thinks that John went there.’ c. ?(?)John-un [amwuto kukos-ey ka-ss-ta-ko] John-Top anyone there-Loc go-Past-Decl-Comp sayngkakha-cianh-nun-ta. think-Neg-Pres-Decl ‘John doesn’t think that anyone went there.’ d. *Amwuto [John-i kukos-ey an ka-ss-ta-ko] Anyone John-Nom there-at Neg go-Past-Decl-Comp sayngkakha-n-ta. think-Pres-Decl ‘Anyone thinks that John didn’t go there.’ As we have seen earlier, in the case of the SVC, short-form negation can precede the first verb or long-form negation can follow the second verb. In either case, an NPI in subject or object position can be equally licensed. This suggests that the subject and object are in the same clause as the negative morpheme, whether the negation precedes the first verb or follows the second verb (cf. Lee 1992; Yi 1997; Y. Lee 2003). (177) a. John-i amwukesto an kwu-e mek-ess-ta. John-Nom anything Neg broil-L eat-Past-Decl ‘John didn’t broil and eat anything.’ b. John-i amwukesto kwu-e mek-cianh-ass-ta. John-Nom anything broil-L eat-Neg-Past-Decl ‘John didn’t broil and eat anything.’ (178) a. Amwuto sayngsen-ul an kwu-e mek-ess-ta. Anyone fish-Acc Neg broil-L eat-Past-Decl ‘No one broiled and ate the fish.’ b. Amwuto sayngsen-ul kwu-e mek-cianh-ass-ta. Anyone fish-Acc broil-L eat-Neg-Past-Decl ‘No one broiled and ate the fish.’ In the case of coordinate structures, if short-form negation precedes the first verb, it may only license an NPI in the first conjunct; if it precedes the second verb, it may only license an NPI in the second conjunct. See the paradigm in (179). If long-form negation follows the last verb, it may

Serial-Verb Construction and Verbs of Motion in Korean

63

only license an NPI in the second conjunct; if it follows the first verb, it may only license an NPI in the first conjunct. See the paradigm in (180). These facts are as expected given the biclausal nature of the coordinate sentences. (179) a.

John-i amwukesto an kwup-ko mek-ess-ta. John-Nom anything Neg broil-Conj eat-Past-Decl ‘John didn’t broil anything and ate it.’ b. John-i sayngsen-ul kwup-ko amwukesto an John-Nom fish-Acc broil-Conj anything Neg mek-ess-ta. eat-Past-Decl ‘John broiled the fish and didn’t eat anything.’ c. *John-i sayngsen-ul an kwup-ko amwukesto John-Nom fish-Acc Neg broil-Conj anything mek-ess-ta. eat-Past-Decl d. *John-i amwukesto kwup-ko an mek-ess-ta. John-Nom anything broil-Conj Neg eat-Past-Decl

(180) a.

John-i amwukesto kwup-cianh-ko mek-ess-ta. John-Nom anything broil-Neg-Conj eat-Past-Decl ‘John didn’t broil anything and ate it.’ b. John-i sayngsen-ul kwup-ko amwukesto John-Nom fish-Acc broil-Conj anything mek-cianh-ass-ta. eat-Neg-Past-Decl ‘John broiled fish and didn’t eat anything.’ c. *John-i sayngsen-ul kwup-cianh-ko amwukesto John-Nom fish-Acc broil-Neg-Conj anything mek-ess-ta. eat-Past-Decl d. *John-i amwukesto kwup-ko mek-cianh-ass-ta. John-Nom anything broil-Conj eat-Neg-Past-Decl

It is very important to keep in mind that the SVC should not be confused with the subordinate construction, especially in view of the fact that the subordinate marker -ese can sometimes be deleted. In the cases of covert subordination, the first V is generally followed by a pause. On the role of the pause between verbs, see Im 1975 and Sohn 1976.8 The

64

Chapter 2

example in (181a) must be a case of covert subordination, since that sentence contains two objects and the SVC allows for only one object. And indeed, short-form negation immediately preceding the first (subordinate) V cannot license an NPI in the matrix clause; see (181b). The paradigm in (181) (from Yi 1997) is therefore identical to (182), which contains an overt subordinating conjunction. (181) a.

John-i oscangmwun-ul yel-e os-ul John-Nom closet door-Acc open-L clothes-Acc kkenay-ss-ta. take out-Past-Decl ‘John opened the closet door and took out clothes.’ b. *Amwuto oscangmwun-ul an yel-e os-ul Anyone closet door-Acc Neg open-L clothes-Acc kkenay-ss-ta. take out-Past-Decl Intended meaning: ‘No one opened the closet door and took out clothes.’

(182) a.

John-i oscangmwun-ul yel-ese os-ul John-Nom closet door-Acc open-Sub clothes-Acc kkenay-ss-ta. take out-Past-Decl ‘John opened the closet door and took out clothes.’ b. *Amwuto oscangmwun-ul an yel-ese os-ul Anyone closet door-Acc Neg open-Sub clothes-Acc kkenay-ss-ta. take out–Past-Decl Intended meaning: ‘No one opened the closet door and took out clothes.’

2.1.2 The Korean SVs Are Not Compounds The issue of whether Korean SVCs are compounds has been controversial; see Yi 1997. In this section, we review some evidence that strongly suggests that the SVs in Korean do not constitute a compound.9 As discussed by Chung 1993, the scalar or contrastive markers -man (only), -to (even), and -nun (contrastive topic) can be attached to the right of the first V and in that case, they take scope over the first V; see (183)– (185).

Serial-Verb Construction and Verbs of Motion in Korean

65

(183) John-i hakkyo-ey (pesu-nun tha-cianh-ko) kel-e-man John-Nom school-Loc (to) bus-Contr take-Neg-Conj walk-L-only ka-ss-ta. go-Past-Decl ‘John went to school, only by walking (not by bus).’ (184) John-i hakkyo-ey (talli-e-nun ka-cianh-ass-eto) John-Nom school-Loc (to) run-L-Contr go-Neg-Past-although kel-e-nun ka-ss-ta. walk-L-Contr go-Past-Decl ‘John went to school by walking (although not by running).’ (185) John-i hakkyo-ey talli-e-to ka-ss-ta. John-Nom school-Loc (to) run-L-even go-Past-Decl ‘John went to school, even by running.’ (186) John-i sayngsen-ul kwu-e-nun mek-ess-ta. John-Nom fish-Acc broil-L-Contr eat-Past-Decl ‘John broiled (not boiled) the fish and then ate it.’ (187) John-i sayngsen-ul kwu-e-to mek-ess-ta. John-Nom fish-Acc broil-L-even eat-Past-Decl ‘John even broiled and ate the fish’ (where ‘‘broiling’’ is the least expected alternative because John generally boils the fish). We note though that if the first V is transitive, such scalar or contrastive markers can also take scope over V1 and its object. We illustrate this case below. Consider the following situation. John works in a store that sells fish, cheese, and so on. Among his chores (making cheese, cleaning fish, and so forth), the one he avoids most is cleaning fish. In the example below, -to takes scope over V1 and its object and gives rise to an interpretation in which John accomplished the least expected chore, besides his other chores. (188) Onul-un John-i chicu-lul mantu-nun kes today-Top John-Nom cheese-Acc make-Rel thing ppwunmananila sayngsen-ul ssis-e-to phal-ass-ta. besides fish-Acc clean-L-even sell-Past-Decl ‘Today, besides making cheese, John even cleaned fish and sold them.’ The example in (188) shows that the first V and its object constitute a semantic unit independent of the second V—that is, the two Vs do not

66

Chapter 2

constitute a semantic complex predicate, as a compound analysis would entail. The SVs discussed above contrast in this respect with so-called lexical compounds, which are said to exhibit morphological integrity. Such forms do not allow delimiters or contrastive elements to intervene between verbs; see (189b).10 Note though that this fact does not show that such cases have a di¤erent structure from the SVs discussed earlier. The examples like the one below are associated with an idiomatic, noncompositional meaning, and the fact that contrastive markers and delimiters cannot be attached to the first member of these forms could be attributed to this semantic factor. Indeed, if the meaning of a subpart is dependent on the whole, it is not possible to generate a semantic contrast set for the subpart. (189) a. Apeci-kkeyse tol-a ka-si-ess-ta. father-Nom(Hon) turn-L go-Hon-Past-Decl ‘The father died.’ b. Apeci-kkeyse tol-a/*to/*man/*nun ka-si-ess-ta. father-Nom(Hon) turn-L/even/only/Contr go-Hon-Past-Decl Further evidence against the compound status of Korean SVCs is provided by the plural marker -tul on V1 ; see Yi 1997 and Chung 1993. The plural marker -tul is a phrasal su‰x, which attaches to the subject.11 Concomitantly, a copy of this plural marker can appear on a clausemate constituent of any type of syntactic category (i.e., AdvP, VP, PP, NP). While -tul attached to the subject denotes plurality, its copy functions as a distributivity marker. It distributes the event or property denoted by that constituent over the plural subject. In the sentence provided below, the plural marker on the verb distributes events of ‘‘picture drawing’’ over the plural subject ‘‘three children.’’ As exemplified below, when the distributive marker is absent, the sentence is ambiguous. Thus, the sentence in (190a) is compatible with a situation in which there is one picture being drawn by three children, or with a situation in which there are three pictures, each drawn by one of the three children. When the distributive marker is present, the sentence is unambiguous. The sentence in (190b) is compatible only with a situation in which each of the three children is drawing a distinct picture. (190) a. Sey-myeng-uy ai-tul-i kulim-ul kuli-ko iss-ta. three-Cl-Gen child-Pl-Nom picture-Acc draw-Prog-Decl ‘Three children are drawing a picture.’

Serial-Verb Construction and Verbs of Motion in Korean

67

b. Sey-myeng-uy ai-tul-i kulim-ul kuli-ko-tul iss-ta. three-Cl-Gen child-Pl-Nom picture-Acc draw-Prog-Pl-Decl ‘Three children are each drawing a picture.’ In the SVC, the plural marker -tul can appear on V1, which points to the presence of an embedded VP boundary. In this case, -tul distributes the lower VP (i.e., the fish-catching event) over the plural subject, giving rise to a meaning in which each of the children caught a di¤erent fish. Note that the distribution does not carry over to the fish-eating event. Indeed, the sentence does not entail that each of the children ate the fish that he or she caught. Rather, the sentence entails that the children (collectively) ate the fish that each of them caught.12 This clearly shows that these SVCs are composed of two distinct (sequential) events. (191) Sey-myeng-uy ai-tul-i sayngsen-ul cap-a-tul mek-ess-ta. three-Cl-Gen child-Pl-Nom fish-Acc catch-L-Pl eat-Past-Decl ‘Three children each caught a fish and ate them.’ The following well-formed sentence (with an NPI object licensed by longform negation on V2) shows that the type of example under discussion is indeed an SVC. (192) Sey-myeng-uy ai-tul-i amwukesto cap-a-tul three-Cl-Gen child-Pl-Nom anything catch-L-Pl mek-cianh-ass-ta. eat-Neg-Past-Decl Compare with the coordinate counterpart, which is ill-formed. (193) *Sey-myeng-uy ai-tul-i amwukesto cap-ko-tul three-Cl-Gen child-Pl-Nom anything catch-Conj-Pl mek-cianh-ass-ta. eat-Neg-Past-Decl 2.1.3 Semantic Relations within the SVC: A Comparison with Coordinate and Subordinate Structures Another essential property of the serial-verb construction is that the Vs it contains must be understood as denoting events that are intrinsically connected. In this section, we discuss the semantic relation between the events in the Consequential SVC (CSVC) and in the Simultaneous or coevents SVC (SSVC).

68

Chapter 2

In the case of the CSVC, the event denoted by the first V precedes the event denoted by the second V and the first event must be perceived as enabling the second event to take place. Indeed, they cannot be perceived as two unrelated sequential events. To illustrate this point, consider the relation between the SVs in the examples below. In (194a–b), John’s catching/broiling the fish enabled John to eat the fish. In (194c), John’s gripping the rope enabled John to pull the rope. (194) a. John-i sayngsen-ul cap-a mek-ess-ta. John-Nom fish-Acc catch-L eat-Past-Decl ‘John caught and (then) ate the fish.’ b. John-i sayngsen-ul kwu-e mek-ess-ta. John-Nom fish-Acc broil-L eat-Past-Decl ‘John broiled and (then) ate the fish.’ c. John-i cwul-ul cap-a tangki-ess-ta. John-Nom rope-Acc grip-L pull-Past-Decl ‘John gripped and (then) pulled the rope.’ This perceived ordering is pragmatic in nature. Indeed, given our knowledge of the world, it must be the case that the event ‘‘catching/broiling the fish’’ precedes the event ‘‘eating the fish’’ and that the event ‘‘gripping the rope’’ precedes the event ‘‘pulling the rope.’’ Interestingly, the surface ordering of the SVs reflects this (pragmatically determined) temporal ordering. The ordering of SVs in the examples in (194) obeys this condition, but the opposite ordering does not, as illustrated below. (195) a. *John-i sayngsen-ul mek-e cap-ass-ta. John-Nom fish-Acc eat-L catch-Past-Decl ‘John ate and (then) caught the fish.’ b. *John-i sayngsen-ul mek-e kwu-ess-ta. John-Nom fish-Acc eat-L broil-Past-Decl ‘John ate and (then) broiled the fish.’ c. *John-i cwul-ul tangki-e cap-ass-ta. John-Nom rope-Acc pull-L grip-Past-Decl ‘John pulled and (then) gripped the rope.’ The temporal-ordering constraint that we see in the SVC also applies to coordinated VPs, if the VPs denote temporally ordered events. This temporal ordering is reflected in the surface order of the verbs in such coordinate structures. Indeed, all of the English translations given in (195) are pragmatically odd ( just like the Korean SVC counterparts). Muysken 1988 and Li 1993 attribute this ordering constraint to the Temporal Icon-

Serial-Verb Construction and Verbs of Motion in Korean

69

icity Condition (or TIC), an interface condition that applies at the level of Phonological Form (or PF). (196) Temporal Iconicity Condition (TIC) In an SVC, as well as in a verbal coordinate structure, if the events denoted by the Vs are sequential, the surface ordering of the Vs must reflect the temporal ordering of events. Although the verbal coordinate structures also obey the TIC, it is different from the SVC in that the coordinated events in the former construction may be unrelated, even if temporally ordered. See the examples in (197). This is not the case in the SVC; see (198). In this respect, the SVC is closer to the Subordinate Construction (SC) than to the Coordinate Construction (CC); see Sung 1972, Song 1976, and Lee 1992. Both in the SVC and in the SC, the matrix and subordinate events must be intrinsically related. Indeed, if ‘‘talking’’ and ‘‘eating’’ are understood as unrelated events, these two verbs cannot give rise to an SVC, or to an SC; see the ill-formedness of (198) and (199). (197) a. John-i mek-ko malhay-ss-ta. John-Nom eat-Conj talk-Past-Decl ‘John ate and talked.’ b. John-i malha-ko mek-ess-ta. John-Nom talk-Conj eat-Past-Decl ‘John talked and ate.’ (198) a. *John-i mek-e malhay-ss-ta. John-Nom eat-L talk-Past-Decl ‘John ate and talked.’ b. *John-i malhay mek-ess-ta. John-Nom talk eat-Past-Decl ‘John talked and ate.’ (199) a. *John-i John-Nom b. *John-i John-Nom

mek-ese malhay-ss-ta. eat-Sub talk-Past-Decl malhay-se mek-ess-ta. talk-Sub eat-Past-Decl

Another example that illustrates the intrinsic-ordering relation between the events in the CSVC is provided by (200a). The SVC nature of this example is shown by (200b), which involves NPI licensing. Recall that NPI licensing is a property of monoclausal structures, which distinguishes SVCs from coordinate and subordinate structures.

70

Chapter 2

(200) a. Totwuk-i (paykhwacem-eyse) panci-lul hwumchi-e Thief-Nom (department store-Loc) ring-Acc steal-L ka-ss-ta. go-Past-Decl ‘The thief stole the ring and (then) left (the department store).’ b. Amwuto (paykhwacem-eyse) panci-lul an hwumchi-e Anyone (department store-Loc) ring-Acc Neg steal-L ka-ss-ta. go-Past-Decl ‘No one stole the ring and (then) left (the department store).’ The semantic relation between the SVs in (200) is a consequential one. Indeed, the stealing event is understood as intrinsically related to the leaving event, in which the former triggers the latter.13 Given our knowledge of the world, namely that robbers leave after they steal, the causal relation in the above example is conventionalized (and needs no context). In other cases, the intrinsic relation between the two SVs is less conventionalized and more context-dependent, such as the relation between ‘pulling out the electric wire’ and ‘leaving the room’ in (201). This sentence would be appropriate in a context in which John’s task is to pull out the electric wire in a given room and having done the job, John leaves the room. The NPI example in (202) shows that we are indeed dealing with an SVC. (201) John-i censen-ul ppop-a ka-ss-ta. John-Nom electric wire-Acc pull (out)-L go-Past-Decl ‘John pulled (out) the electric wire and left (or went away).’ (202) Amwuto censen-ul an ppop-a ka-ss-ta. Anyone electric wire-Acc Neg pull (out)-L go-Past-Decl ‘No one pulled (out) the electric wire and left (went away).’ We turn next to the SSVC. In this case, the relation that the SVs entertain is one of modification, as illustrated below. In this example, ‘walking’ expresses the manner of the directed-motion event denoted by ka-. (203) a. John-i hoswu-lul heyemchi-e kenne-ss-ta. John-Nom lake-Acc swim-L cross-Past-Decl ‘John crossed the lake by swimming.’ b. John-i hakkyo-ey kel-e ka-ss-ta. John-Nom school-Loc walk-L go-Past-Decl ‘John went to school by walking.’

Serial-Verb Construction and Verbs of Motion in Korean

71

When the relation between the SVs expresses manner or means, the events denoted by V1 and V2 are temporally simultaneous and therefore the TIC does not apply. The ordering between the SVs in this case is determined by the syntax. In e¤ect, in Korean verbal modifiers are leftadjoined to the verbal structure that they modify. The ill-formedness of the Korean examples in (204) is comparable to the ill-formedness of their English counterparts (given below the Korean examples).14 (204) a. *John-i hoswu-lul kenne heyemchi-ess-ta. John-i lake-Acc cross swim-Past-Decl *‘John swam by crossing the lake.’ b. *John-i cengwen-ey ka kel-ess-ta. John-Nom garden-Loc go walk-Past-Decl *‘John walked by going to the garden.’ As we discussed earlier, the meanings associated with the SVC can be (approximately) expressed by the SC. Indeed, this construction can express the consequential reading, as illustrated in (205), which is the SC counterpart of the SVC in (194c). (205) John-i cwul-ul cap-ese tangki-ess-ta. John-Nom rope-Acc grip-Sub pull-Past-Decl The SC can express the manner or means meaning as well. In this usage, the clause introduced by -ese describes the means or manner by which the matrix event is brought about. See (206), which is the SC counterpart of the SVC in (203b). (206) John-i kel-ese hakkyo-ey ka-ss-ta. John-Nom walk-Sub school-Loc go-Past-Decl ‘John went to school walking.’ On the other hand, the SC can also have a temporal meaning ‘when’/ ‘while’, as illustrated below (cf. Suh 1996). There is no SVC counterpart to this SC. We take this to be an important and telling di¤erence between the SC and the SVC, because it shows that the semantics of the SC is not identical to that of the SVC after all. (207) John-un elye-se nol-ki-man ha-ss-ta. John-Top young-when play-NL-only do-Past-Decl ‘When he was young, John only played.’

72

Chapter 2

To summarize, in this section, we have made the following points: (208) a. The semantic relations between the subevents encoded by the SVC may be of two kinds: a consequential relation (CSVC) or a manner relation (SSVC). Whether or not we can establish such semantic relations between the two SVs depends on our cognitive understanding of these relations, our knowledge of the world, and the discourse context.15 b. The SVC is semantically distinct from the CC. The coordinated events may be sequentially ordered but semantically unrelated. c. The SVC expresses meanings that are akin to some of the meanings expressed by the SC. Yet the SVC expresses only a subset of the meanings that the SC can express. The SC, but not the SVC, can express two independent events that are temporally concomitant (i.e., the ‘when’/‘while’ relation). d. The Temporal Iconicity Condition governs the surface ordering of SVs (and coordinated VPs) when the events that these Vs denote are temporally ordered. 2.1.4 The Structural Analysis of SVCs As mentioned in chapter 1, the SVC parameter is a syntactic one. A language generates SVC structures when it makes use of one or both of the following Generalized Transformations (GTs): (209) a. Merge a verbal l-structure with the head of another verbal l-structure. b. Merge a verbal lexical item with the head of a verbal l-structure. Korean makes use of the GT in (209a) to generate CSVCs and SSVCs. To illustrate, consider the derivation of the CSVC in (194a), repeated below. (210) John-i sayngsen-ul cap-a mek-ess-ta. John-Nom fish-Acc catch-L eat-Past-Decl ‘John caught and (then) ate the fish.’ The building blocks of this sentence are the l-structures of cap- ‘catch’ and mek- ‘eat’, given in (211) and (212), respectively. As usual, the lexical items are to be understood as a short-term notation for a bundle of features: the phonological (or P-) features, Conceptual pointer (or Cfeature), categorial features (if any), and the formal features (or FF).

Serial-Verb Construction and Verbs of Motion in Korean

73

The structures below are obtained by applying the operation Merge to V (cap- or mek-) and D (sayngsen-ul ). With Baker and Stewart 1999, we assume that when constituents A and B are merged, the resulting node inherits from the head only those features that are relevant for the syntactic composition of phrases. In other words, the label of the resulting constituent consists uniquely of syntactic features. In the case under discussion, these are the categorial feature V and the FF, which include at the very least the functional feature Tense.16 (211)

(212)

The GT in (209a) merges (211) with the head of (212), giving rise to the structure in (213). We may assume that the V created by the GT (marked in bold below) is invisible with respect to predicate-argument structure relations. Thus, the original predicate-argument relations are preserved in the SVC. On the other hand, the V created by the GT may be assumed to encode the semantic relation that holds between the events denoted by its two daughters. We return to this point later. (213)

74

Chapter 2

By convention, the structure in (213) is merged with little v, which introduces the external argument, giving rise to the extended l-structure in (214); see the discussion in section 1.4. (214)

The structure in (214) is then merged with clausal functional projections, such as Polarity, Asp, and Tense. As mentioned in chapter 1 (note 24), we may assume (with Ko 2005) that Case in Korean is checked via the operation ‘‘Agree.’’ In the above structure, Case is checked via Agree, which establishes an agreement relation between T and the Nom DP (John-i) and between Asp and the Acc DP (sayngsen-ul ). But as we know, scrambling is pervasive in Korean; therefore, these DPs may be scrambled out of the verbal phrase by a subsequent movement operation. As noted by Baker and Stewart, to the extent that the two Vs in (214) share the same syntactic features, either V can function as the head of that structure. But unlike Baker and Stewart, we do not conclude that the structure is biheaded. Instead, we assume a relative interpretation of headedness—that is, a category is interpreted as the head of the structure with respect to a certain computation. Consider the issue of morphosyntactic headedness. Which of the two Vs is computed as the morphosyntactic head? By Minimality, the V with P-features that is structurally closer to T (namely, mek-) will be attracted to T, and tense will then be morphophonologically realized on that V. Consequently, in the Korean SVC, V2 (or the last V) functions as the morphosyntactic head of the clause.

Serial-Verb Construction and Verbs of Motion in Korean

75

Let us turn next to the logical relation between the two events introduced by V1 and V2. In the SVC under discussion, the relation is a consequential one. In other words, (215) In the CSVC, the event denoted by V1 is a necessary condition for V2 to take place. Thus, contrary to what the English translation suggests, (210) is associated with the meaning in (216): (216) John ate the fish ! John caught the fish If the semantics is compositionally read o¤ the syntactic structure, which is the view that the present work endorses and argues for, then the meaning stated in (215) must be structurally encoded. In the SVC, there are no conjunctions, but it is reasonable to assume that the relation between the two Vs is structurally encoded by the V node introduced by the GT, namely, the V in bold in (214). We can elaborate such a node to include that semantic information (i.e., V1 V2): (217) V, FF, {V1

V2}

The consequential relation described in (215) is akin to the well-known cause-result relation. Since typically the result is subordinate to the cause at the VP level, it is reasonable to assume that (218) In a CSVC, V1 functions semantically as the matrix event and V2 as the subordinate event. Some evidence for (218) is provided by adverb modification. Degree adverbs (acwu ‘quite’, ‘very’; kkway ‘quite’; wancenhi ‘completely’; cemcem ‘gradually’), temporal adverbs (cacwu ‘frequently’), and manner adverbs (chenchenhi ‘slowly’; ppalli ‘quickly’) are typical VP adverbs in Korean. Interestingly, in the case of the CSVC, these unambiguously modify the first V. This is indeed expected if the first V is semantically the matrix event. The following examples illustrate this point. The verb kwupwuli- ‘bend’ can be modified by cemcem, but not the verb kkunh‘cut’ because it is [
durative]; see (219). When these two verbs are combined into an SVC, modification by cemcem is possible, as shown in (220). The SVC status of such sentences is confirmed by the NPI test in (221).17 (219) a.

John-i chelsa-lul cemcem kwupwuli-ess-ta. John-Nom wire-Acc gradually bend-Past-Decl

76

Chapter 2

b. *John-i chelsa-lul cemcem kkunh-ess-ta. John-Nom wire-Acc gradually cut-Past-Decl (220) John-i chelsa-lul cemcem kwupwuli-e kkunh-ess-ta. John-Nom wire-Acc gradually bend-L cut-Past-Decl ‘John gradually bent and cut the wire.’ (221) Amwuto chelsa-lul an kwupwuli-e kkunh-ess-ta. Anyone wire-Acc Neg bend-L cut-Neg-Past-Decl ‘No one bent and cut the wire.’ The example in (223) illustrates a case in which the adverb ppalli ‘quickly’ unambiguously modifies the first verb of the SVC, despite the fact that either the first verb (cap- ‘catch’) or the second verb (mek- ‘eat’) can be individually modified by this adverb; see (222). Indeed, (ii) is not a possible English translation of the example in (223). (222) a. John-i sayngsen-ul ppalli John-Nom fish-Acc quickly ‘John quickly caught a fish.’ b. John-i sayngsen-ul ppalli John-Nom fish-Acc quickly ‘John quickly ate a fish.’

cap-ass-ta. catch-Past-Decl mek-ess-ta. eat-Past-Decl

(223) John-i sayngsen-ul ppalli cap-a mek-ess-ta. John-Nom fish-Acc quickly catch-L eat-Past-Decl (i) ‘John quickly caught a fish and then ate it.’ (ii) *‘John caught a fish and then quickly ate it.’ The case of the SSVC, which we will discuss in detail in the following sections, is di¤erent from the CSVC with respect to the semantic composition of the two Vs. In the case of the SSVC, V2 functions both as the morphosyntactic head and as the semantic head of the SVC. More precisely, (224) In an SSVC, V1 and V2 are simultaneous events, where V2 denotes the matrix event and V1 modifies V2. Adverb modification provides evidence that V2 is indeed the semantic head in the SSVC. As shown in (225), the adverb wancenhi ‘completely’ is compatible with the SVC heyemchi-e kenne-ss-ta ‘swim cross’, despite the fact that first V is not compatible with such an adverb. Only the second V is; see (226).

Serial-Verb Construction and Verbs of Motion in Korean

77

(225) John-i hoswu-lul wancenhi heyemchi-e kenne-ss-ta. John-Nom lake-Acc completely swim-L cross-Past-Decl ‘John completely crossed the lake by swimming.’ (226) a. *John-i wancenhi heyemchi-ess-ta. John-Nom completely swim-Past-Decl b. John-i hoswu-lul wancenhi kenne-ss-ta. John-Nom lake-Acc completely cross-Past-Decl ‘John completely crossed the lake.’ To conclude, in the Korean SVC, V2 (or the last V) is consistently interpreted as the morphosyntactic head. On the other hand, there is ambiguity as to which V functions as denoting the matrix event. In the case of the CSVC, V1 is interpreted as denoting the matrix event and V2 is interpreted as denoting the subordinate event. In the case of the SSVC, V2 is interpreted as denoting the matrix event and V1 is interpreted as the modifier of the matrix event. As was mentioned in section 1.6.1, the GT in (209b) generates the Resultative SVC (RSVC). This type of SVC is excluded in Korean because it violates the Temporal Iconicity Condition (or TIC). We return to this issue, and its implications, in section 2.2.4. 2.2

Serial Verbs of Motion in Korean

Having introduced the Korean SVCs, their basic properties, and the mechanism by which they are generated, we now return to our main topic: the directed-motion construction. This section is organized as follows. We begin with a general discussion of the aspectual import of the light verb ka-/o- (section 2.2.1). We discuss manner-of-motion verbs and show that these are unambiguously activitydenoting verbs. When these verbs enter into an SVC with ka-, the resulting structure encodes the meaning of directed motion in a certain manner. In such structures, ka- is unambiguously the semantic head of the SVC, while the manner is the modifer of the directed-motion event (section 2.2.2). A structural analysis of these SVCs is proposed within the framework outlined in the previous section (section 2.2.3). We then turn to transitive contact verbs (like kick, push, pull, roll ) and show that these do not imply directed motion. We examine the SVCs formed with transitive contact verbs and ka-, and show that these lack a causative meaning,

78

Chapter 2

which we attribute to the lack of RSVCs in Korean (section 2.2.4). To complete the picture, we examine the SVCs formed with these transitive contact verbs in combination with path verbs, with and without ka(section 2.2.5). This chapter ends with two appendixes that deal with related constructions: the construction headed by the su‰x -(e)ci ‘become’ (appendix 2.1) and the prospective possessor construction (appendix 2.2). 2.2.1 The Aspectual Nature of the Light Verbs ka- and oWe assume that like their English counterparts go and come, the Korean verbs ka- and o- are not lexical items listed in the lexicon. These are the spell-out of the head V in a certain syntactic context, namely in a context in which V takes a directional path complement and a specifier, of which the path is predicated. Such a construction encodes movement along an abstract or physical path (toward an endpoint or away from an initial point); see section 1.4. Thus, ka-/o- are the ‘‘morphological signature’’ of the directed-motion construction in Korean. 2.2.1.1 Directed Motion along an Abstract Path The examples in (227) illustrate cases in which ka- combines with a verb to give rise to movement along an abstract path. In these examples, the verbs denote the endpoint state of that abstract path. In other words, the verb in question functions as an argument of ka-.18 (227) a. John-i cwuk-e ka-ss-ta. John-Nom die-L go-Past-Decl ‘John was on the path to death.’ b. John-i salaci-e ka-ss-ta. John-Nom disappear-L go-Past-Decl ‘John was on the path to disappearance.’ c. Yenkuk-i ttuthna ka-ss-ta. performance-Nom end go-Past-Decl ‘The performance approached the end.’ The verbs cwuk- ‘die’, salaci- ‘disappear’, and ttuthna- ‘end’ are achievements; they are lexically specified as [
durative].19 This is shown by their incompatibility with adverbs like cemcem ‘gradually’, which are only compatible with events that give rise to an interpretation that involves a series of subintervals at the aspectual temporal level (the R line in

Serial-Verb Construction and Verbs of Motion in Korean

79

the Reichenbach framework); see section 1.5.1. This is exemplified in (228a–c).20 (228) a. *John-i cemcem cwuk-ess-ta. John-Nom gradually die-Past-Decl b. *John-i cemcem salaci-ess-ta. John-Nom gradually disappear-Past-Decl c. *Yenkuk-i cemcem ttuthna-ss-ta. performance-Nom gradually end-Past-Decl If the verbs discussed above form an SVC with the light verb ka-, the resulting SVC is compatible with cemcem (cf. Im 2001). See (229) below. This is expected given that ka- is semantically the head and that ka- is the head of an accomplishment-denoting VP. Indeed, accomplishments consist of subevents mapped onto subintervals at the temporal level, and these are modifiable by adverbs such as cemcem; see section 1.5.1. (229) a. John-i cemcem cwuk-e ka-ss-ta. John-Nom gradually die-L go-Past-Decl ‘John was gradually dying.’ b. John-i cemcem salaci-e ka-ss-ta. John-Nom gradually disappear-L go-Past-Decl ‘John was gradually disappearing.’ c. Yenkuk-i cemcem ttuthna ka-ss-ta. performance-Nom gradually end go-Past-Decl ‘The performance was gradually ending.’ Note that while achievement verbs, like the ones discussed above, can combine with ka-, activity verbs such as wus- ‘laugh’, nol- ‘play’, and wul‘cry’ cannot. This is expected because, although they are [
durative], achievement verbs denote a process with an endpoint state. Therefore, they can be interpreted as path-denoting and can function as the path argument of the construction headed by ka-. But this is not the case with activity verbs, which are open-ended dynamic events. (230) a. *John-i John-Nom b. *John-i John-Nom c. *John-i John-Nom

wus-e ka-n-ta. laugh-L go-Pres-Decl nol-a ka-n-ta. play-L go-Pres-Decl wul-e ka-n-ta. cry-L go-Pres-Decl

80

Chapter 2

Ka- can combine with scalar stative predicates, which denote a relative endpoint state (e.g., x-amount higher) and with nonscalar stative predicates, which denote an absolute endpoint state (e.g., flat). In such structures, the stative predicate functions as the endpoint state of the path argument introduced by the construction headed by ka-. (231) a. Kenmwul-i noph-a ka-n-ta. building-Nom high-L go-Pres-Decl ‘The building becomes higher.’ b. Patak-i phyengphyenghay ka-n-ta. floor-Nom flat go-Pres-Decl ‘The floor becomes flatter.’ The progressive marker ko iss- is incompatible with statives; see (232). But when the stative predicate is combined with ka-, the progressive morpheme can modify the resulting predicate; see (233). This is as expected since ka- is the head of a dynamic construction. (232) a. *Kenmwul-i noph-ko iss-ta. building-Nom high-Prog-Decl ‘The building is being high.’ b. *Patak-i phyengphyengha-ko iss-ta. floor-Nom flat-Prog-Decl ‘The floor is being flat.’ (233) a. Kenmwul-i noph-a ka-ko iss-ta. building-Nom high-L go-Prog-Decl ‘The floor is becoming higher.’ b. Patak-i phyengphyenghay ka-ko iss-ta. floor-Nom flat go-Prog-Decl ‘The floor is becoming flatter.’ Stative predicates are incompatible with the adverb cemcem ‘gradually’. Recall that this adverb requires a subinterval interpretation at the aspectual temporal level and stative predicates do not give rise to such an interpretation. When ka- combines with the stative predicate, the resulting event can be modified by cemcem. Again, this is as expected since ka- is the head of an accomplishment construction, which introduces the required subinterval interpretation. (234) a. Kenmwul-i (*cemcem) noph-ta. building-Nom gradually high-Decl

Serial-Verb Construction and Verbs of Motion in Korean

81

b. Patak-i (*cemcem) phyengphyengha-ta. floor-Nom gradually flat-Decl (235) a. Kenmwul-i cemcem noph-a ka-n-ta. building-Nom gradually high-L go-Pres-Decl b. Patak-i cemcem phyengphyenghay ka-n-ta. floor-Nom gradually flat go-Pres-Decl 2.2.1.2 Directed Motion along a Physical Path: Locative PPs and Path Verbs We have illustrated above cases in which a verbal argument functions as the abstract path argument of the construction headed by ka-. Below we illustrate cases in which the construction headed by ka-/odenotes movement along a physical path. As in English, a physical path can be encoded by a PP complement: (236) a. John-i pang-ey ka-ss-ta. John-Nom room-Loc go-Past-Decl ‘John went to the room.’ b. John-i pang-ey o-ass-ta. John-Nom room-Loc come-Past-Decl ‘John came to the room.’ On the other hand, in Korean, a physical path can also be encoded by a path verb—for example, the verb tul- in the examples below. (237) a. John-i pang-ey tul-e ka-ss-ta. John-Nom room-Loc (move) into-L go-Past-Decl ‘John went into the room.’ b. John-i pang-ey tul-e o-ass-ta. John-Nom room-Loc (move) into-L come-Past-Decl ‘John came into the room.’ Korean has four path verbs, which express the directionality of the path (olu-/nayli- ‘(move) up’/‘(move) down’; tul-/na- ‘(move) into’/ ‘(move) out of ’). Each of these also has a causative counterpart, which is formed by attaching the causative morpheme -i to it, as shown in (238). Here we will discuss only the intransitive forms. The transitive ones will be discussed in section 2.2.5.

82

Chapter 2

(238) Path verbs Intransitive path verbs

Transitive path verbs

Olu- ‘(move) up’

Ol-li ((move) up þ causative) ‘cause something to move up’ Nayli-q ((move) down þ causative) ‘cause something to move down’ Tul-i ((move) into þ causative) ‘cause something to move in’ Na-i ((move) out of þ causative) ‘cause something to move out’

Nayli- ‘(move) down’ Tul- ‘(move) into’ Na- ‘(move) out of ’

The intransitive path verbs tul-/na- (‘(move) into’/‘(move) out of ’) cannot stand on their own; they appear to be bound morphemes. Therefore, they consistently appear in combination with ka-/o-.21 See the examples in (239). Note that in the context of a path verb, the locative is optional, which suggests that the path-denoting V fulfills the complementation requirements of ka-/o-. In such cases, the optional locative is a specifier of the path-denoting V. The optional status of the locative when ka- is merged with a path verb contrasts with the obligatory status of the locative in the absence of a path verb; see (240). Note that the path verb can in turn be analyzed as a path argument incorporated into the head of the directed-motion construction—that is, [Vpath V]. (239) a. John-i (pang-ey) tul-e *(ka)-ss-ta. John-Nom (room-Loc) (move) into-L go-Past-Decl ‘John went in(to the room).’ b. John-i (pang-eyse) na *(ka)-ss-ta. John-Nom room-from (move) out-of go-Past-Decl ‘John went out(of the the room).’ (240) John-i *(pang-ey) ka-ss-ta. John-Nom (room-Loc) go-Past-Decl ‘John went to the room.’ In the case of the path verbs olu-/nayli- (‘(move) up’/‘(move) down’), there are examples, such as the ones below, in which it can be assumed that the path verb has merged directly with ka- (o)-: (241) John-i i-chung-ey oll-a ka-ss-ta. John-Nom second-floor-Loc (move) up-L go-Past-Decl ‘John went up to the second floor.’

Serial-Verb Construction and Verbs of Motion in Korean

83

The form in (241) is strange in the absence of ka-. However, the strangeness has nothing to do with the morphological status of the path verb. It has to do with the aspectual property of the path verb and our knowledge of the world. More specifically, the path verbs are achievement verbs; they describe punctual—that is, [
durative]—processes; see Choi and Bowerman 1991 and Im 2001. This is illustrated by the contrast between (241) and (242). In the case of (241), John (with his human limitations) could have reached the second floor only with a movement decomposed into subintervals. On the other hand, in (242), where the subject is Superman, the form without ka- becomes natural and it implies that Superman reached the second floor in one instantaneous movement.22 (242) Swuphemayn-i i-chung-ey oll-ass-ta. Superman-Nom second-floor-Loc (move) up-Past-Decl ‘Superman went up to the second floor.’ The [þdurative] property of the construction headed by ka- also explains the following contrast. The forms without ka- are achievements and therefore cannot be modified by cemcem ‘gradually’. On the other hand, the forms with ka- can be modified by such an adverb.23 (The contrast in (245) is adapted from Im 2001.) (243) a. Pihayngki-ka hwalcwulo-ey (*cemcem) airplane-Nom runway-Loc (*gradually) nayli-ess-ta. (move) down-Past-Decl b. Pihayngki-ka hwalcwulo-ey (cemcem) airplane-Nom runway-Loc (gradually) nayli-e ka-ss-ta. (move) down-L go-Past-Decl ‘The airplane gradually landed at the runway.’ (244) a. Chensa-ka kwulum wi-eyse matang-ey (*cemcem) angel-Nom cloud up-from yard-Loc (*gradually) nayli-ess-ta. (move) down-Past-Decl b. Chensa-ka kwulum wi-eyse matang-ey (cemcem) angel-Nom cloud up-from yard-Loc (gradually) nayli-e ka-ss-ta. (move) down-L go-Past-Decl ‘The angel gradually went down to the yard from the cloud.’

84

Chapter 2

(245) a. Kispal-i (*cemcem) oll-ass-ta. flag-Nom (*gradually) (move) up-Past-Decl b. Kispal-i (cemcem) oll-a ka-ss-ta. flag-Nom gradually (move) up-L go-Past-Decl ‘The flag went up gradually.’ To summarize, in this section we have discussed cases in which either a locative PP or a verb functions as the path argument of the directedmotion construction headed by ka-(o-). Depending on the semantic nature of the path argument, the construction is interpreted as directed motion along a physical path or as directed motion along an abstract path. In the following sections, we will discuss SVCs headed by ka-(o-), in which the first V does not function as an argument, but rather as a modifier, of ka-(o-). 2.2.2 Simultaneous SVC (SSVC) with Manner-of-Motion Verbs Manner-of-motion verbs in Korean do not encode directed motion; see Lee 1999 and Chae 1999. Compare the examples below with the ones in (237). The locative -ey can denote the goal of the motion in the context of ka- and o-, but not in the context of a manner-of-motion verb like talli‘run’, heyemchi- ‘swim’, nal- ‘fly’, ki- ‘crawl’, or chwumchwu- ‘dance’.24 (246) a. *John-i kongwen-ey talli-ess-ta. John-Nom park-Loc run-Past-Decl ‘John ran to the park.’ Cf. John-i kongwen-eyse talli-ess-ta. John-Nom park-Loc run-Past-Decl ‘John ran at the park.’ b. *John-i kongwen-ey kel-ess-ta. John-Nom park-Loc walk-Past-Decl ‘John walked to the park.’ Cf. John-i kongwen-eyse kel-ess-ta. John-Nom park-Loc walk-Past-Decl ‘John walked at the park.’ c. *John-i hoswu hanccok-phyen-ey heyemchi-ess-ta. John-Nom lake one side-side-Loc swim-Past-Decl ‘John swam to one side of the lake.’ Cf. John-i hoswu hanccok-phyen-eyse heyemchi-ess-ta. John-Nom lake one side-side-Loc swim-Past-Decl ‘John swam at one side of the lake.’

Serial-Verb Construction and Verbs of Motion in Korean

85

d. *Aki-ka pang-ey ki-ess-ta. baby-Nom room-Loc crawl-Past-Decl ‘The baby crawled to the room.’ Cf. Aki-ka pang-eyse ki-ess-ta. baby-Nom room-Loc crawl-Past-Decl ‘The baby crawled in the room.’ As expected, manner-of-motion verbs are compatible with pathdenoting adjuncts, introduced by -kkaci or -lo:25 (247) a. John-i kongwen-kkaci talli-ess-ta. John-Nom park-up to run-Past-Decl ‘John ran up to the park.’ b. John-i kongwen-kkaci kel-ess-ta. John-Nom park-up to walk-Past-Decl ‘John walked up to the park.’ c. John-i hoswu hanccok-phyen-kkaci heyemchi-ess-ta. John-Nom lake one side-side-up to swim-Past-Decl ‘John swam up to one side of the lake.’ d. Aki-ka pang-kkaci ki-ess-ta. baby-Nom room-up to crawl-Past-Decl ‘The baby crawled up to the room.’ (248) a. John-i hakkyo-lo ttwi-ess-ta. John-Nom school-toward run-Past-Decl ‘John ran toward the school.’ b. John-i hakkyo-lo kel-ess-ta. John-Nom school-toward walk-Past-Decl ‘John walked toward the school.’ The adjunct status of the phrases headed by -kkaci and -lo in the above examples is shown by the fact that they do not a¤ect the eventive properties of the verb. The verbs remain unbounded, activity-denoting predicates, which can be modified by the temporal phrase headed by tongan ‘for’. Such predicates are gradable events and can be modified by gradable adverbs, such as cokum ‘a bit’; see (249). Furthermore, the examples in (250a–b) do not imply that John ran/walked all the way to the park. He could have run/walked only part of the way, and covered the rest of the trajectory in some other way; see (250c). (249) a. John-i (cip-eyse) hakkyo-ccok-ulo cokum John-Nom (house-from) school-direction-toward a bit

86

Chapter 2

ttwi-ess-ta. run-Past-Decl ‘John ran a bit toward the school.’ b. John-i hakkyo-ccok-ulo kkway John-Nom school-direction-toward considerably kel-ess-ta. walk-Past-Decl ‘John walked considerably toward the school.’ (250) a. John-i (cip-eyse) kongwen-kkaci (sip pwun tongan) John-Nom (home-from) park-up to (ten minute for) talli-ess-ta. run-Past-Decl ‘On his way (from home) up to the park, John ran (for ten minutes).’ b. John-i (cip-eyse) kongwen-kkaci (sip pwun tongan) John-Nom (home-from) park-up to (ten minute for) kel-ess-ta. walk-Past-Decl ‘On his way (from home) up to the park, John walked (for ten minutes).’ c. John-i (cip-eyse) kongwen-kkaci sip pwun tongan John-Nom (home-from) park-up to ten minute for talli-ess-ko nameci sikan tongan kel-ess-ta. run-Past-Conj rest time for walk-Past-Decl ‘On his way (from home) up to the park, John ran for ten minutes and walked for the rest of the time.’ Unsurprisingly, when a manner-of-motion verb enters into an SVC with ka-/o-, the resulting predicate denotes directed motion, and it does so unambiguously. Compare the examples below with (246). (251) a. John-i kongwen-ey talli-e ka-ss-ta. John-Nom park-Loc run-L go-Past-Decl ‘John ran to the park.’ b. John-i kongwen-ey kel-e ka-ss-ta. John-Nom park-Loc walk-L go-Past-Decl ‘John walked to the park.’ c. John-i hoswu hanccok-phyen-ey heyemchi-e ka-ss-ta. John-Nom lake one side-side-Loc swim-L go-Past-Decl ‘John swam to one side of the lake.’

Serial-Verb Construction and Verbs of Motion in Korean

87

d. Aki-ka pang-ey ki-e ka-ss-ta. baby-Nom room-Loc crawl-L go-Past-Decl ‘The baby crawled to the room.’ The sentences in (251) can be modified by a temporal phrase headed by man-ey ‘in’, but not by a temporal phrase headed by tongan ‘for’. Recall that the former, but not the latter, is compatible with an event bounded at the temporal level (R-time); see section 1.5.1. Compare the examples in (252) with the ones in (250). (252) a. John-i kongwen-ey sip pwun man-ey/*sip pwun tongan John-Nom park-Loc ten minute in/ten minute for ka/o-ss-ta. go/come-Past-Decl ‘John went/came to the park in ten minutes/*for ten minutes.’ b. John-i kongwen-ey sip pwun man-ey/*sip pwun tongan John-Nom park-Loc ten minute in/ ten minute for talli-e ka-ss-ta. run-L go-Past-Decl ‘John ran to the park in ten minutes/*for ten minutes.’ In section 1.5.1, we have seen that while event boundedness (E-time) and temporal boundedness (R-time) generally coincide, cases of mismatch between event boundedness and temporal boundedness exist. Therefore, while accomplishments are, by definition, bounded events (i.e., they involve movement toward or away from an endpoint), the presence of an atelic postposition can render the event temporally unbounded and therefore compatible with the for-temporal phrase. This is illustrated in the examples below. (253) a. John-i sip pwun tongan kongwen-ccok-ulo John-Nom ten minute for park-direction-toward ka-ss-ta. go-Past-Decl ‘John went toward the park for ten minutes.’ b. John-i sip pwun tongan kongwen-ccok-ulo John-Nom ten minute for park-direction-toward ka-ss-ta. go-Past-Decl ‘John ran toward the park for ten minutes.’ c. John-i sip pwun tongan kongwen-ccok-ulo John-Nom ten minute for park-direction-toward

talli-e run-L

kel-e walk-L

88

Chapter 2

ka-ss-ta. go-Past-Decl ‘John walked toward the park for ten minutes.’ d. John-i sip pwun tongan hoswu hanccok-phyen-ulo John-Nom ten minute for lake one side-side toward heyemchi-e ka-ss-ta. swim-L go-Past-Decl ‘John swam toward one side of the lake for ten minutes.’ The bounded nature of the SVC that consists of a manner-of-motion verb and ka-(o-) can furthermore be shown by their incompatibility with gradable modifiers such as cokum ‘a bit’, kkway ‘considerably’, and manhi ‘a lot’. Thus, the examples in (249) contrast with the examples in (254). This contrast clearly shows that ka-(o-) is the head of such SVCs. (254) a. John-i kongwen-ey (*cokum) talli-e ka-ss-ta. John-Nom park-Loc (a bit) run-L go-Past-Decl ‘John ran a bit to the park.’ b. John-i kongwen-ey (*cokum) kel-e ka-ss-ta. John-Nom park-Loc (a bit) walk-L go-Past-Decl ‘John walked a bit to the park.’ c. John-i hoswu hanccok-phyen-ey (*cokum) heyemchi-e John-Nom lake one side-side-Loc (a bit) swim-L ka-ss-ta. go-Past-Decl ‘John swam a bit to one side of the lake.’ d. Aki-ka pang-ey (*cokum) ki-e ka-ss-ta. baby-Nom room-Loc (a bit) crawl-L go-Past-Decl ‘The baby crawled a bit to the room.’ Intransitive path verbs can enter into an SVC with a manner-of-motion verb. The resulting forms are achievements. Such an SVC can in turn function as the path-complement of ka-. The resulting structure denotes an event composed of subintervals. This can be illustrated with the SVC ttwi-e nayli-, which means ‘jump down’ when it stands by itself (255a) and ‘run down’ when it combines with ka- (255b). We will discuss the structure of manner-of-motionþpathþka in section 2.2.5. (255) a. John-i il-chung-ulo ttwi-e nayli-ess-ta. John-Nom first-floor-toward jump-L (move) down-Past-Decl ‘John jumped down to the first floor.’

Serial-Verb Construction and Verbs of Motion in Korean

89

b. John-i il-chung-ulo ttwi-e nayli-e John-Nom first-floor-toward run-L (move) down-L ka-ss-ta. go-Past-Decl ‘John ran down to the first floor.’ While Korean intransitive manner-of-motion verbs can readily enter into an SVC with the light verbs ka-(o-), intransitive verbs that do not express motion cannot enter into an SVC with ka-(o-): (256) a. aJohn-i kongwen-ey wus-e ka-ss-ta. John-Nom park-Loc laugh-L go-Past-Decl b. aJohn-i kwuncwung-ul kwanthonghay solichi-e John-Nom crowd-Acc through scream-L ka-ss-ta. go-Past-Decl This is reminiscent of the English facts below, discussed by Goldberg and Jackendo¤ 2004: (257) a. aJohn honked down the road. b. aThe dog barked out of the room. c. aBill whistled past the house. Goldberg and Jackendo¤ (2004) discuss the contrast between English examples such as (257) and (258). They note that the di¤erence between the two cases lies in the fact that in the latter examples, but not in the former ones, the emitted sound can be interpreted as the result of the motion, and the emitted sound is coextensive with the motion. (We return to such examples in chapter 3.) (258) a. The trolley rumbled through the tunnel. b. The wagon creaked down the road. c. The bullets whistled past the house. While sound-of-emission verbs are onomatopoetic expressions in Korean and their distribution is more restricted than in English, we have found one case in Korean that allows us to illustrate in that language a contrast comparable to the above-mentioned ones in English. This is the case of whistle. The literal meaning of whistle in (258c) is translated in Korean as ‘blow the whistle’. This cannot appear in the SVC, as illustrated in (259). It can only appear in the corresponding subordinate construction, as shown by the obligatory presence of the conjunctive su‰x, mye ‘while’

90

Chapter 2

(conveying the meaning of simultaneity). On the other hand, the metaphorical meaning of whistle in (258c) can give rise to the meaning whereby the sound is the result of the movement. Under that interpretation, this verb can appear in the SVC; see (260). (Ssayng ssayng is an onomatopoetic expression, which means ‘the sound of high wind’.) (259) Bill-i cip-yeph-ul hwiphalam-ul pwul-*(mye) Bill-Nom house-flank-Acc whistle-Acc blow-while cina-ka-ss-ta. pass-go-Past-Decl (260) Palam-i cip-cwuwi-lul ssayng ssayng pwul-e wind-Nom house-around-Acc sound of high wind blow-L cina-ka-ss-ta. pass-go-Past-Decl ‘The wind whistled around the house.’ The examples below also illustrate that there must be an intrinsic semantic connection between the modifier and the directed motion. The activity-denoting verbs (wus- ‘laugh’, nongtamha- ‘joke’, iyakiha- ‘talk’) cannot function as the modifer of the directed-motion SVC, because there is no intrinsic semantic connection between the two. (Note that the English translations are equally unacceptable; see section 3.1.3.) (261) a. aKutul-un ku kil-ul wus-e kel-e nayli-e they-Top the road-Acc laugh-L walk-L (move) down-L ka-ss-ta. go-Past-Decl a‘They laughed down the road.’ b. aKutul-un ku pang-eyse nongtamhay they-Top the room-from joke na-ka-ss-ta. (move) out of-go-Past-Decl a‘They joked out of the room.’ c. aKutul-un ku cip-ul iyakihay cina-ka-ss-ta. they-Top the house-Acc talk pass-go-Past-Decl a‘They talked past the house.’ To summarize, manner-of-motion verbs in Korean are unambiguously unbounded, activity-denoting verbs. These verbs can enter into an SVC with the head of the directed-motion construction, ka-(o-). In such SVCs,

Serial-Verb Construction and Verbs of Motion in Korean

91

the light verb ka-(o-) is the semantic head of the SVC and the manner-ofmotion verb is the modifier of the directed-motion construction. There is an intrinsic, simultaneous semantic relation between the motion and the manner—that is, they constitute one single grammatical event. 2.2.3 The Structural Analysis of the SSVC To illustrate the structural analysis of the SSVC, consider the derivation of (251a), repeated below. (262) John-i kongwen-ey talli-e ka-ss-ta. John-Nom park-Loc run-L go-Past-Decl ‘John ran to the park.’ The source of the sentence in (262) is the l-structures in (263) and (264). Each of these structures is formed independently and therefore each constitutes a spell-out domain. Recall furthermore that ka- has no lexical entry. Ka- is the spell-out of V in the syntactic context in (264). As for the complex PP that functions as the complement of ka-, we will assume that the locative postposition -ey lexicalizes both the directional (higher) P and the locative (lower) P. (263)

(264)

92

Chapter 2

The GT (209a) takes the l-structure of talli- ‘run’ (263) and adjoins it to the head of the directed-motion construction (264), giving rise to the structure in (265). Recall that, following Baker and Stewart 1999, we have adopted the following assumption. Lexical items consist of a bundle of features, such as formal features (or FF), phonological features (or Pfeatures), a conceptual pointer (or C-features), and categorial features. When lexical items are merged together, the resulting constituent inherits from the head only those features that are relevant to the syntactic computation—that is, categorial features and formal features (or FF). P-features and C-features are therefore not part of the specification of phrasal nodes. (265)

The V node marked in bold is created by the GT that adjoins the lstructure of talli- to the head of the directed-motion construction. It may be assumed that such a node introduces the semantic relation that articulates the relation between the two Vs, namely the relation of modification, whereby V1 modifies V2. (266) V, FF, {V1 Mod V2} As noted in the previous section, there are semantic restrictions on the type of Vs that can modify ka- in the SVC (see also the next section). The verbs must denote a property that is intrinsically related to the motion, namely, one that may be interpreted as the means or manner of the motion. In the structure in (265), ka- is both the morphosyntactic head and the sem(antico)-syntactic head. It is the morphosyntactic head by virtue of being the V ‘‘closest’’ to T. Therefore, ka- adjoins to T and the tense feature is morphologically realized on ka-. As stated in (266), V1

Serial-Verb Construction and Verbs of Motion in Korean

93

modifies V2 (ka-). From the sem(antico)-syntactic point of view, this means that V2 (ka-) is the head and V1 is the Specifier of that structure. Note that the alternative order of SVs in (262) is unacceptable, as shown below: (267) *John-i kongwen-ey ka talli-ess-ta. John-Nom park-Loc go run-Past-Decl The ill-formedness of (267) cannot be attributed to the TIC since the SVs in this case denote coevents. The fact that V2 cannot function as the Specifier of V1 must then follow from the universal principle that a Specifier must precede the head that it specifies.26 2.2.4 SSVCs with Transitive Contact Verbs The examples below show that contact verbs (e.g., cha- ‘kick’, mil- ‘push’, and kkul-/tangki- ‘pull’) do not imply directed motion. Indeed the contact object may be aimed in a certain direction, but motion is not entailed.27 (268) a. John-i (cip-pakk-ulo) kong-ul cha-ss-una John-Nom house-outside-toward ball-Acc kick-Past-but kong-un wumciki-cianh-ass-ta. ball-Top move-Neg-Past-Decl ‘John kicked the ball (toward the outside of the house) but the ball didn’t move.’ b. John-i (cip-pakk-ulo) khathu-lul mil-ess-una John-Nom house-outside-toward cart-Acc push-Past-but khathu-nun wumciki-cianh-ass-ta. cart-Top move-Neg-Past-Decl ‘John pushed the cart (toward the outside of the house) but the cart didn’t move.’ c. John-i (cip-pakk-ulo) khathu-lul kkul-ess-una John-Nom house-outside-toward cart-Acc pull-Past-but khathu-nun wumciki-cianh-ass-ta. cart-Top move-Neg-Past-Decl ‘John pulled the cart (toward the outside of the house) but the cart didn’t move.’ Given that the transitive contact verbs do not imply motion, they cannot instantiate the directed-motion construction. Indeed, they cannot be merged with a path-denoting PP, lexicalized by the locative -ey, as shown below:

94

Chapter 2

(269) a. *John-un ku kong-ul cengwen-ey cha-ss-ta. John-Top the ball-Acc garden-Loc kick-Past-Decl ‘John kicked the ball to the garden.’ Cf. John-un ku kong-ul cengwen-eyse cha-ss-ta. John-Top the ball-Acc garden-Loc kick-Past-Decl ‘John kicked the ball while in the garden.’ b. *John-un ku khathu-lul cengwen-ey mil-ess-ta. John-Top the cart-Acc garden-Loc push-Past-Decl ‘John pushed the cart to the garden.’ Cf. John-un ku khathu-lul cengwen-eyse mil-ess-ta. John-Top the cart-Acc garden-Loc push-Past-Decl ‘John pushed the cart while in the garden.’ c. *John-un ku khathu-lul cengwen-ey kkul-ess-ta. John-Top the cart-Acc garden-Loc pull-Past-Decl ‘John pulled the cart to the garden.’ Cf. John-un ku khathu-lul cengwen-eyse kkul-ess-ta. John-Top the cart-Acc garden-Loc pull-Past-Decl ‘John pulled the cart while in the garden.’ On the other hand, contact verbs can enter into an SVC with ka-/o-, giving rise to a directed-motion meaning:28 (270) a. John-un John-Top b. John-un John-Top c. John-un John-Top

cengwen-ey kong-ul cha ka-ss-ta. garden-Loc ball-Acc kick go-Past-Decl cengwen-ey khathu-lul mil-e ka-ss-ta. garden-Loc cart-Acc push-L go-Past-Decl cengwen-ey khathu-lul kkul-e ka-ss-ta. garden-Loc cart-Acc pull-L go-Past-Decl

The transitive contact verbs contrast with verbs such as tenci- ‘throw’ and kennay- ‘hand’, which do entail directed motion. These verbs can merge with a goal argument, realized as a locative P -ey if the goal is inanimate, and as a dative P -eykey if the goal is animate: (271) a. John-un cengwen-ey kong-ul tenci-ess-ta. John-Top garden-Loc ball-Acc throw-Past-Decl ‘John threw the ball to the garden.’ b. John-un Bill-eykey kong-ul tenci-ess-ta. John-Top Bill-Dat ball-Acc throw-Past-Decl ‘John threw the ball to Bill.’

Serial-Verb Construction and Verbs of Motion in Korean

95

(272) John-un Bill-eykey kong-ul kennay-ss-ta. John-Top Bill-Dat ball-Acc hand-Past-Decl ‘John handed the ball to Bill.’ A closer look at transitive contact verbs reveals what appears to be a contradictory situation. While these verbs do not allow for a locativemarked complement (see (269)), they do allow for the presence of a dative-marked argument. In appendix 2.2 to this chapter, we show that this is only an apparent contradiction. We argue there that the dative argument in the examples in (273) is not a goal argument of the verb, but that it is a prospective possessor argument of an abstract applicative verb (comparable to the dative argument in the English example John kicked Mary the ball ). (273) a. John-un Bill-eykey ku kong-ul cha-ss-ta. John-Top Bill-Dat the ball-Acc kick-Past-Decl ‘John kicked the ball to Bill.’ b. John-un Bill-eykey ku khathu-lul mil-ess-ta. John-Top Bill-Dat the cart-Acc push-Past-Decl ‘John pushed the cart to Bill.’ c. (?)John-unBill-eykey ku khathu-lul kkul-ess-ta. John-Top Bill-Dat the cart-Acc pull-Past-Decl ‘John pulled the cart to Bill.’ If we examine closely the meaning of the sentences in (270), we note that they are not equivalent to the English sentences below (see the approximate paraphrases below the examples): (274) a. John kicked the ball to the garden. (John causes the ball to move to the garden by kicking it.) b. John pushed the cart to the garden. (John causes the cart to move to the garden by pushing it.) c. John pulled the cart to the garden. (John causes the cart to move to the garden by pulling it.) In the English sentences, the entity that undergoes movement is the object. The subject may or may not move. On the other hand, in the Korean examples in (270), the subject does undergo movement, and the object moves along with the subject. Approximate translations of the examples in (270) are given below.

96

Chapter 2

(275) a. John goes to the garden, repeatedly kicking the ball. b. John goes to the garden, repeatedly pushing the cart. c. John goes to the garden, repeatedly pulling the cart. It is clear from the above discussion that the Korean examples in (270) lack a causative meaning. They are unambiguously associated with a modification meaning, which means that they are SSVCs with a structure comparable to the one in (265). To exemplify, consider the derivation of the example in (270a). The GT in (209a) adjoins the l-structure of cha‘kick’ to the head of the directed-motion construction (264), and the structure in (276) is obtained. (276)

As indicated by the rough translations in (275), there is a clear intuition that the motion in the examples in (270) is accompanied by repeated events of ball-kicking, cart-pushing, or cart-pulling. Recall that an accomplishment structure consists of a series of subevents in E (event level) and each subevent is mapped onto a distinct subinterval in R (temporal level); see section 1.5.1. We can then say that in the semantic analysis of the examples in (270), the event denoted by V1 modifies each subevent that composes the event denoted by V2. Since each subevent is then mapped onto a subinterval, we obtain an analysis by which V1 indirectly modifies each subinterval of the directed motion. As we have seen in the preceding section, the modifer in the SSVC must express the means or manner of the motion. Transitive verbs that are not compatible with such meanings are therefore not possible in the SSVC:29

Serial-Verb Construction and Verbs of Motion in Korean

(277) a. *John-i John-Nom b. *John-i John-Nom c. *John-i John-Nom

97

cengwen-ey Mary-lul nolyepo-a ka-ss-ta. garden-Loc Mary-Acc stare-L go-Past-Decl cengwen-ey maykcwu-lul mashi-e ka-ss-ta. garden-Loc beer-Acc drink-L go-Past-Decl hakkyo-ey chayk-ul ilk-e ka-ss-ta. school-Loc book-Acc read-L go-Past-Decl

The question arises as to why the Korean sentences in (270) cannot be associated with a causative meaning. The reason is that Korean lacks Resultative SVCs (or RSVCs). Indeed, as mentioned in section 1.6.1, Korean lacks examples like the ones below.30 (278) *John-i Mary-lul mil-e nemeci-ess-ta. John-Nom Mary-Acc push-L fall-Past-Decl Intended meaning: ‘John pushed Mary down.’ (279) *John-i kangto-lul ttayli-e John-Nom robber-Acc hit-L cwuk-ess-ta. (Kang 1991; cited in Lee 1992) die-Past-Decl Intended meaning: ‘John hit the robber dead.’ (280) *Pawi-ka pang-ulo khep-ul chi-e tul-e rock-Nom room-toward cup-Acc hit-L (move) into-L ka-ss-ta. go-Past-Decl Intended meaning: ‘A rock hit a cup into the room.’ As mentioned in section 1.6.1, the West African languages do have RSVCs. Some examples are given below. (281) Fe´mi tı` Akin subu´. (Yoruba; Lord 1974, Baker 1989) ˙ ˙ Fe´mi push Akin fall ˙ ‘Fe´mi pushed Akin down.’ ˙ (282) Olu´ lu ma`a´lu´ ku´. (Yoruba; Awoyale 1988) Olu´ beat cow die ‘Olu´ beat the cow dead.’ ` zo´ su`a´ a´ga´ de´. (283) O (Edo; Stewart 1998) ` zo´ push chair fall O ` zo´ pushed the chair down.’ ‘O

98

Chapter 2

(284) Ekpe a fo kOpo yi xO-me. (Ewe; Collins 1993) rock FUT hit cup go room-in ‘A rock will hit a cup into the room.’ As expected, the counterparts of the Korean examples in (270) in the West African languages have a causative meaning. We illustrate this with the Yoruba examples in (285) and (286), provided to us by Oladiipo Ajiboye. (285) Jo.o.nu gba bo.olu wo. inu o.gba. John kick ball enter inside garden Cf. Bo.olu wo. inu o.gba. ball enter inside garden (286) Jo.o.nu ti o.mo.kurin jade lati inu yara. John push boy go-out from inside room Cf. O.mo.kurin jade lati inu yara. boy go-out from inside room Baker and Stewart (1999) argue that the serial verbs in the RSVC constitute a complex predicate. Within the framework of analysis assumed here, such complex predicates are generated by the GT in (209b). As mentioned in section 1.6.1, the resulting forms in head-final languages give rise to a violation of the TIC. To illustrate this point, let us consider an RSVC structure for the Korean example in (270a). The GT in (209b) adjoins the lexical item cha- to the head of the directed-motion construction; see (264). The resulting structure is given in (287). This structure is then merged with little v, to which cha- is adjoined, as shown in (288). Thus, cha- is interpreted as the causing event, and the resulting directed motion (ka-) is interpreted as the resulting event. (287)

Serial-Verb Construction and Verbs of Motion in Korean

99

(288)

Cha-, being the highest overt V in the structure in (288), is attracted by T (to check the Tense feature). Since Korean is a head-final language, functional projections in this language, including T, are left-branching. The ordering of verbs, after cha- moves to T, will be the one in (289) and not the one in (270a). The sentence in (289) is ruled out by the TIC (a PF interface condition) because the surface order of verbs does not respect the temporal ordering of events: the cause (cha-) should precede the result (ka-). (289) *John-un cengwen-ey kong-ul ka cha-ss-ta. John-Nom garden-Loc ball-Acc go kick-Past-Decl It has been proposed in the literature that the causative counterpart of Korean ka- ‘go’ is ponay- ‘send’ (see Sohn 2001): (290) John-i cengwen-ey kong-ul ponay-ss-ta. John-Nom garden-Loc ball-Acc send-Past-Decl ‘John sent the ball to the garden (where the boys were waiting to play baseball).’ The transitive contact verbs can enter into an SVC with ponay-, as illustrated in (291). (291) a. John-i cengwen-ey kong-ul cha ponay-ss-ta. John-Nom garden-Loc ball-Acc kick send-Past-Decl ‘John caused the ball to go to the garden by kicking it.’

100

Chapter 2

b. John-i cengwen-ey khathu-lul mil-e ponay-ss-ta. John-Nom garden-Loc cart-Acc push-L send-Past-Decl ‘John caused the cart to go to the garden by pushing it.’ If ponay- is indeed the lexicalization of a cause-directed-motion construction, then the derivation of these sentences will proceed as follows. The lstructure of the transitive contact verb will be adjoined to the head of the directed-motion construction (namely, ka-), giving rise to an SSVC. This SSVC will then be merged with little v by convention, and the output form ‘ka- v’ will be spelled out as ponay-. To illustrate, we give below the structure of the sentence in (291a).31 (292)

To summarize, we have seen in this section that certain transitive verbs, namely contact verbs, can also enter into an SVC with the head of the directed-motion construction. This gives rise to an SSVC, in which the contact verb modifies the directed-motion construction—that is, the contact verb specifies the means of the directed motion. We have seen that, although V1 is transitive, these sentences do not have a causative meaning (with V1 denoting the cause and V2 the result). To have a causative meaning, such sentences would have to be associated with an RSVC. We have shown that this is not possible because the resulting surface order of verbs violates the TIC.

Serial-Verb Construction and Verbs of Motion in Korean

101

2.2.5 The Transitive Path Verbs, Manner-of-Motion Verbs, and the SSVC As mentioned in section 2.2.1.2 and summarized in table (238), intransitive path verbs have a causative counterpart. We repeat the table below. (293) Path verbs Intransitive path verbs

Transitive path verbs

Olu- ‘(move) up’

Ol-li ((move) up þ causative) ‘cause something to move up’ Nayli-q ((move) down þ causative) ‘cause something to move down’ Tul-i ((move) into þ causative) ‘cause something to move in’ Na-i ((move) out of þ causative) ‘cause something to move out’

Nayli- ‘(move) down’ Tul- ‘(move) into’ Na- ‘(move) out of ’ Some examples are given below.

(294) a. John-i i-chung-ey cim-ul John-Nom second-floor-Loc package-Acc ol-li-ess-ta. (move) up-Caus-Past-Decl ‘John caused the package to go up to the second floor.’ b. John-i il-chung-ey cim-ul John-Nom first-floor-Loc package-Acc nayli-q-ess-ta. (move) down-Caus-Past-Decl ‘John caused the package to go down to the first floor.’ (295) a. John-i kesil-ey chayksang-ul John-Nom living room-Loc desk-Acc tul-i-ess-ta. (move) into-Caus-Past-Decl ‘John caused the desk to go into the living room.’ b. John-i kesil-ey chayksang-ul John-Nom living room-Loc desk-Acc na-i-ss-ta. (move) out of-Caus-Past-Decl ‘John caused the desk to go out to the living room.’ In section 2.2.1.2, we showed that the intransitive path verbs are achievements—that is, [
durative] processes. Interestingly, the causative

102

Chapter 2

path verbs are not. More precisely, while the resulting event (i.e., the motion) is [
durative], the causing event is not. This is shown by the fact that transitive path verbs are compatible with adverbs like cemcem ‘gradually’. (Recall that such adverbs can only modify events that contain subintervals.) Compare the examples in (243)–(245) with the ones in (296). (296) a. John-i i-chung-ey cim-ul (cemcem) John-Nom second-floor-Loc package-Acc (gradually) ol-li-ess-ta. (move) up-Caus-Past-Decl ‘John (gradually) caused the package to go up to the second floor.’ b. John-i il-chung-ey cim-ul (cemcem) John-Nom first-floor-Loc package-Acc (gradually) nayli-q-ess-ta. (move) down-Caus-Past-Decl ‘John (gradually) caused the package to go down to the first floor.’ In the theory developed in chapter 1, the di¤erence between achievements and accomplishments is captured in terms of a durative feature. Furthermore, it was proposed that [þdurative] is the unmarked interpretation of a process. In e¤ect, (297) A process is automatically interpreted as [þdurative], unless it is lexically specified as [
durative]. Light verbs are not listed in the lexicon; in other words, they do not have idiosyncratic properties of their own. Therefore, a directed-motion construction headed by a light verb must be interpreted as a durative event. As we saw in previous sections, there is ample evidence that this is indeed the case in Korean. For the same reason, their causative counterparts will also be interpreted as durative events. Recall that a path verb can be merged with ka-, fulfilling the complementation requirements of the directed-motion construction. In section 2.2.1.2, we saw examples where an intransitive path verb functions as the complement of ka-. Examples with a transitive path verb as the complement of ka- are given below. As expected, in the presence of a path verb, the locative PP is optional. In these cases, the path verb (and not the locative PP) functions as the complement of ka-; the locative PP is a specifier of the path verb.32

Serial-Verb Construction and Verbs of Motion in Korean

103

(298) a. John-i (i-chung-ey) cim-ul ol-li-e John-Nom (second-floor-Loc) package-Acc (move) up-Caus-L ka-ss-ta. go-Past-Decl ‘John took the package up (to the second floor).’ b. John-i (il-chung-ey) cim-ul nayli-q-e John-Nom (first-floor-Loc) package-Acc (move) down-Caus-L ka-ss-ta. go-Past-Decl ‘John took the package down (to the first floor).’ (299) a. John-i (kesil-ey) chayksang-ul tul-i-e John-Nom (living room-Loc) desk-Acc (move) into-Caus-L ka-ss-ta. go-Past-Decl ‘John took the desk in(to the living room).’ b. John-i (kesil-ey) chayksang-ul na-i John-Nom living room-Loc desk-Acc (move) out of-Caus ka-ss-ta. go-Past-Decl ‘John took the desk out (to the living room).’ There are significant meaning di¤erences between the forms with kaand the forms without ka-. As indicated by the English translations, in the sentences without ka-, (294)–(295), the object (but not necessarily the subject) undergoes movement. In other words, the path is predicated of the object and not the subject (which is interpreted as the causer). On the other hand, in the sentences with ka-, (298)–(299), the subject moves along with the object.33 This semantic di¤erence between the two sets of sentences is revealed by the contrasts below. In (300a), ka- is predicated of the subject, which is also the causer of the event. Therefore, this sentence entails movement of the causer. This entailment is contradicted by the adjunct clause ‘while John stood still’ in (300a), which accounts for the oddity of the sentence. On the other hand, in (300b), where ka- is absent, movement of the causer is not entailed by the structure. Therefore, in this case, the adjunct clause ‘while John stood still’ does not give rise to a contradiction. (300) a. aJohn-i kamanhi se-se (i-chung-ey) John-Nom still stand-while second-floor-Loc

104

Chapter 2

cim-ul ol-li-e ka-ss-ta. package-Acc (move) up-Caus-L go-Past-Decl a‘While John stood still, he took the package to go up to the second floor.’ b. John-i kamanhi se-se i-chung-ey John-Nom still stand-while second-floor-Loc cim-ul ol-li-ess-ta. package-Acc (move) up-Caus-Past-Decl ‘While John stood still, he caused the package to go up to the second floor.’ In conclusion, the transitive path verbs in (298)–(299) are not SVCs. Recall that SVCs involve adjunction of a phrasal V projection to another verbal head. This is not so in the cases under discussion, in which the path-denoting verbal projection functions as the complement of the head of the directed-motion construction. In (298)–(299), the relation between the subject of ka- and the subject of the path verb is one of control. We illustrate below with the structure that corresponds to the sentence in (299a). (We do not endorse any particular theory of control here; we simply indicate control by rewriting PRO in the controllee position.) Recall that the path verb argument incorporates into the verbal head of the directed-motion construction—for example, [tul- V]. When this complex V adjoins to little v, it is spelled out as a causative form (‘tul-i’ in the example below). (301)

Serial-Verb Construction and Verbs of Motion in Korean

105

On the other hand, a transitive contact verb can enter into an SVC with a transitive path verb, as illustrated in (302). These are generated by the GT (209a), which adjoins the l-structure of the transitive verb (kwulli‘roll’ , mil- ‘push’, kkul-/tangki- ‘pull’) to the structure headed by the transitive path verb.34 (302) a. John-i il-chung-ey cim-ul kwulli-e John-Nom first-floor-Loc package-Acc roll-L nayli-q-ss-ta. (move) down-Caus-Past-Decl ‘John caused the package down to the first floor, by rolling it.’ b. John-i kesil-ey chayksang-ul mil-e John-Nom living room-Loc desk-Acc push-L tul-i-ess-ta. (move) into-Caus-Past-Decl ‘John caused the desk into the living room, by pushing it.’ c. John-i kesil-ey chayksang-ul kkul-e John-Nom living room-Loc desk-Acc pull-L na-i-ss-ta. (move) out of-Caus-Past-Decl ‘John caused the desk to go out to the living room, by pulling it.’ As indicated by the English translation, the sentences in (302) do not imply movement of the Nom subject. This is as expected since those sentences are derived by adjoining the transitive l-structure of the transitive contact verb (kwulli-, mil-, or kkul-) to the head of the l-structure associated with the path verb, via the GT (209a). We exemplify with the structure that corresponds to the sentence in (302b).

106

Chapter 2

(303)

On the other hand, if the SVC (kwulli-e nayli-e, mil-e tul-i-e, kkul-e na-i) is merged as the path argument of ka-, as shown in (304), movement of the Nom subject is entailed. Again these results are as expected, since the Nom subject is in the Specifier of the structure headed by ka-; see (301). (304) a. John-i (il-chung-ey) cim-ul kwulli-e John-Nom (first-floor-Loc) package-Acc roll-L nayli-q-e ka-ss-ta. (move) down-Caus-L go-Past-Decl ‘John went down (to the first floor), rolling the package.’ b. John-i (kesil-ey) chayksang-ul mil-e John-Nom (living room-Loc) desk-Acc push-L tul-i-e ka-ss-ta. (move) into-Caus-L go-Past-Decl ‘John went in(to the living room), pushing the desk.’ c. John-i (kesil-ey) chayksang-ul kkul-e John-Nom (living room-Loc) desk-Acc pull-L na-i ka-ss-ta. (move) out of-Caus go-Past-Decl ‘John went out (to the living room), pulling the desk.’

Serial-Verb Construction and Verbs of Motion in Korean

107

We exemplify below with a structure that corresponds to the sentence in (304b): (305)

That fact that movement of the Nom subject is entailed when ka- is present is shown by the fact that (306a) gives rise to a contradiction. Compare it with (306b), where ka- is absent and no contradiction arises. In this case, John could be lifting the package with a pulley. (306) a. aJohn-i kamanhi se-se (i-chung-ey) John-Nom still stand-while second-floor-Loc cim-ul tul-e ol-li-e ka-ss-ta. package-Acc lift-L (move) up-Caus-L go-Past-Decl a‘While John stood still, he took the package up to the second floor, by lifting it.’ b. John-i kamanhi se-se (tolulay-lo) John-Nom still stand-while (pulley-with) i-chung-ey cim-ul kkul-e second-floor-Loc package-Acc pull-L

108

Chapter 2

ol-li-ess-ta. (move) up-Caus-Past-Decl ‘While John stood still, he moved the package up to the second floor, by pulling it (with a pulley).’ To summarize briefly, we have shown in this section that in Korean, transitive path verbs can enter into an SVC with another transitive (contact) verb, in which case the transitive path verb is semantically and morphologically the head. On the other hand, transitive path verbs can function as the path argument of ka-. When they concomitantly do so, we obtain hybrid structures composed of an SVC, in which the SVC (headed by the path verb) in turn functions as the complement of the directed-motion construction headed by ka-. 2.3

Summary

In this chapter, we have examined in detail the properties of the Korean SVCs, namely, the Consequential SVC (or CSVC) and the Simultaneous/ coevents SVC (or SSVC). SVCs are adjunction structures, generated by a Generalized Transformation that adjoins one verbal structure to the head of another verbal structure. We have provided evidence (based on adverb modifications) that in the CSVC, the first V functions as the semantic head, while in the SSVC, the second V functions as the semantic head. This should follow from the semantic relation that the two verbal structures entertain. In the CSVC, the semantic relation between the two Vs is semantically akin to a causeresult relation, where the event denoted by V1 is a necessary condition for the event denoted by V2 to take place. Therefore, in this type of SVC, the first V functions semantically as the head of the SVC and V2 as subordinate to V1; see Carstens 2001. On the other hand, in the case of the SSVC, V1 entertains a relation of modification with V2, whereby V1 specifies the manner or means of the directed motion denoted by V2. It then follows that V2 is the semantic head of the SSVC. Furthermore, it was noted that the surface ordering of Vs in the CSVC is governed by the Temporal Iconicity Condition (or TIC). This is not the case in the SSVC, since in this type of SVC, the events encoded by the Vs are simultaneous. The fact that the modifier is realized as the first V follows from the fact that modifiers are structurally encoded as Specifiers and Specifiers are universally left-adjoined to the head structure that they modify, possibly due to a revised LCA (Kayne 1994); see note 26.

Serial-Verb Construction and Verbs of Motion in Korean

109

We have seen that Korean provides unequivocal evidence that mannerof-motion verbs are unambiguously unbounded, activity-denoting verbs. To compose a directed-motion meaning with a manner-of-motion verb, Korean uses a mechanism independently available in the grammar of this language, namely the Generalized Transformation in (209), which generates SVCs. We noted that Korean cannot combine causative directed motion with manner or means, unlike the West African SVC languages. This is due to the fact that Korean lacks an RSVC—that is, an SVC generated by adjoining a lexical item to the head of a cause-directed-motion l-structure. Following insights due to Li 1993, it was argued that Korean lacks RSVCs because it is a head-final language (i.e., its functional heads are left-branching). We showed that in a head-final language like Korean, an RSVC gives rise to a violation of the TIC. In the next chapter, we will address the question as to what mechanisms, if any, non-SVC languages, such as English and Italian, have at their disposal in order to compose directed motion and manner of motion. Appendix 2.1: Some Remarks on the Aspectual Properties of -(e)ci Besides ka-, there is another morpheme that can be analyzed as the spellout of V in the directed-motion construction. This is the morpheme -(e)ci. In this appendix, we do not intend to give a full account of this morpheme. We will restrict our attention to cases in which -(e)ci appears to overlap with ka-. First, we remark that ka-, but not -(e)ci, is compatible with a pathdenoting locative. (307) a.

John-i John-Nom b. *John-i John-Nom

(308) a.

sicang-ey market-Loc sicang-ey market-Loc

(kel-e) ka-ss-ta. (walk-L) go-Past-Decl (kel)-eci-ess-ta. (walk)-CI-Past-Decl

John-i i-chung-ey (talli-e) John-Nom second-floor-Loc (run-L) ka-ss-ta. go-Past-Decl b. *John-i i-chung-ey (talli-e) John-Nom second-floor-Loc (run-L)

oll-a (move) up-L

oll-eci-ess-ta. (move) up CI-Past-Decl

110

Chapter 2

As we will see below, -(e)ci is compatible with an abstract path, but there are restrictions.35 In sharp contrast with ka-, -(e)ci cannot take an achievement verb as its path argument. (309) a. John-i cwuk-e ka-ss-ta. John-Nom die-L go-Past-Decl ‘John was on the path to death.’ b. John-i salaci-e ka-ss-ta. John-Nom disappear-L go-Past-Decl ‘John was on the path to disappearance.’ c. Yenkuk-i ttuthna ka-ss-ta. performance-Nom end go-Past-Decl ‘The performance approached the end.’ (310) a. *John-i cwuk-eci-ess-ta. John-Nom die-CI-Past-Decl b. *Yenkuk-i ttuthna-ci-ess-ta. performance-Nom end-CI-Past-Decl c. *John-i salaci-eci-ess-ta. John-Nom disappear-CI-Past-Decl The -(e)ci construction can take scalar predicates as its path argument. And as we have seen in section 2.2.1.1, this is also the case with ka-. (311) a. Nal-i etwuwe-ci-ess-ta. day-Nom dark-CI-Past-Decl b. John-i khi-ka khe-ci-ess-ta. John-Nom height-Nom tall-CI-Past-Decl (312) a. Nal-i etwuwe-ka-n-ta. day-Nom dark-go-Pres-Decl b. John-i khi-ka khe-ka-ss-ta. John-Nom height-Nom tall-go-Past-Decl The above examples show that the constructions headed by ka- and -(e)ci can denote movement along an abstract path or scale; these are dynamic (not stative) predicates. Therefore they are compatible with the progressive marker ko iss-. (313) a. Kang-i kiph-eci-ko iss-ta. River-Nom deep-CI-Prog-Decl b. Nal-i etwuwe-ci-ko iss-ta. day-Nom dark-CI-Prog-Decl

Serial-Verb Construction and Verbs of Motion in Korean

111

(314) a. Kang-i kiph-e ka-ko iss-ta. river-Nom deep-L go-Prog-Decl b. Nal-i etwuwe ka-ko iss-ta. day-Nom dark go-Prog-Decl Since the path-denoting complement in these constructions is scalar in nature, the event that these constructions denote is interpreted as consisting of subintervals at the aspectual temporal level. Therefore, they are compatible with the adverb cemcem ‘gradually’. (315) a. Kang-i cemcem kiph-eci-ess-ta. River-Nom gradually deep-CI-Past-Decl b. Nal-i cemcem etwuwe-ci-ess-ta. day-Nom gradually dark-CI-Past-Decl (316) a. Kang-i cemcem kiph-e ka-n-ta. river-Nom gradually deep-L go-Pres-Decl b. Nal-i cemcem etwuwe ka-n-ta. day-Nom gradually dark go-Pres-Decl The question we must now raise is the following. In cases where V takes a scalar predicate as the path argument, when is V spelled out as ka- and when is V spelled out as -(e)ci? If there is something like a Uniqueness Principle that guides language acquisition (see Pinker 1989), the pressure should be to have languages conform to ‘‘one form–one meaning’’ correspondences. Therefore, we should not expect free alternations of ka- and -(e)ci. Interestingly, if we look more closely at the data, we can detect a clear meaning di¤erence between the two. As we will see below, the construction headed by ka- is interpreted as denoting a movement toward an absolute endpoint in the scale. On the other hand, the construction headed by -(e)ci is interpreted as denoting a movement to a relative endpoint. We propose that this relative-endpoint interpretation arises from the presence of an abstract comparative marker. This proposal is supported by the fact that the -(e)ci construction (but not the ka-construction) is compatible with the comparative phrase headed by tota ‘than’: (317) a. Kang-i han sikan cen tota kiph-eci-ess-ta. river-Nom one hour before than deep-CI-Past-Decl ‘The river became deeper than an hour ago.’ b. Nal-i han sikan cen tota etwuwe-ci-ess-ta. day-Nom one hour before than dark-CI-Past-Decl ‘The day became darker than an hour ago.’

112

Chapter 2

(318) a. *Kang-i han sikan cen tota kiph-e ka-ss-ta. river-Nom one hour before than deep-L go-Past-Decl b. *Nal-i han sikan cen tota etwuwe ka-ss-ta. day-Nom one hour before than dark go-Past-Decl The above-mentioned meaning di¤erence between the -(e)ci construction and the ka-construction can also be appreciated with the following minimal pair. In the case of (319), there is an interpretation in which the temperature went up from one point in a scale to the next point in the scale three times (e.g., from 37 to 38 , from 38 to 39 , and from 39 to 40 ). This interpretation is unavailable in (320). In combination with ka-, an absolute endpoint in the scale is required. This explains the oddness of (320). The only way to make sense of this example is by imposing an interpretation in which the temperature moved upward and then downward three times in likewise fashion, with the same upward endpoint in each case. (319) Onul i hwanca-uy cheyon-i sey pen Today this patient-Gen temperature-Nom three times noph-aci-ess-ta. high-CI-Past-Decl (320) ??Onul i hwanca-uy cheyon-i sey pen noph-a Today this patient-Gen temperature-Nom three times high-L ka-ss-ta. go-Past-Decl Because the endpoint in the case of -(e)ci is a relative endpoint, the construction is telic. In other words, no matter which point in the scale has been reached, that point is considered the telos. On the other hand, in the case of ka-, movement is toward an absolute telos. There is a potential telos, but there is no implication that it has been reached. While the -(e)ci construction is temporally bounded, the ka-construction is temporally unbounded at the R-time level. For these reasons, the -(e)ci construction, but not the ka-construction, is compatible with a measure modifier. (321) a.

Kang-i il mithe-(ka) kiph-eci-ess-ta. river-Nom one meter-(Nom) deep-CI-Past-Decl ‘The river deepened by 1 meter.’ b. *Kang-i il mithe-(ka) kiph-e ka-ss-ta. river-Nom one meter-(Nom) deep-L go-Past-Decl

Serial-Verb Construction and Verbs of Motion in Korean

113

(322) a.

Nal-i wancenhi etwuwe-ci-ess-ta. day-Nom completely dark-CI-Past-Decl ‘The day darkened completely.’ b. *Nal-i wancenhi etwuwe ka-ss-ta. day-Nom completely dark go-Past-Decl

The -(e)ci construction, but not the ka-construction, is compatible with the resultative/telic marker -e/a iss-. (323) a. Kang-i kiph-eci-e iss-ta. River-Nom deep-CI-Resul-Decl b. Nal-i etwuwe-ci-e iss-ta. day-Nom dark-CI-Resul-Decl (324) a. *Kang-i kiph-e ka-iss-ta. river-Nom deep-L go-Resul-Decl b. *Nal-i etwuwe ka-iss-ta. day-Nom dark go-Resul-Decl The -(e)ci construction, but not the ka-construction, is compatible with the telic temporal phrase. (325) a. *Kang-i han sikan man-ey kiph-e ka-ss-ta. river-Nom one hour in deep-L go-Past-Decl b. *Nal-i han sikan man-ey etwuwe ka-ss-ta. day-Nom one hour in dark go-Past-Decl (326) a. Kang-i han sikan man-ey kiph-eci-ess-ta. River-Nom one hour in deep-CI-Past-Decl ‘The river became deeper in an hour.’ b. Nal-i han sikan man-ey etwuwe-ci-ess-ta. day-Nom one hour in dark-CI-Past-Decl ‘The day became darker in an hour.’ Compare the above examples with the following ones, which contain an atelic temporal phrase: (327) a. Kang-i han sikan tongan kiph-e ka-ss-ta. river-Nom one hour for deep-L go-Past-Decl b. Nal-i han sikan tongan etwuwe ka-ss-ta. day-Nom one hour for dark go-Past-Decl (328) a. ??Kang-i han sikan tongan kiph-eci-ess-ta. river-Nom one hour for deep-CI-Past-Decl

114

Chapter 2

b. ??Nal-i han sikan tongan etwuwe-ci-ess-ta. day-Nom one hour for dark-CI-Past-Decl Based on the meaning distinctions discussed above, we put forth the following proposal: (329) In the context of a scalar verbal complement headed by an abstract comparative marker, V is spelled out as -(e)ci. In the absence of such a comparative marker, V is spelled out as ka-. Finally, we note that the -(e)ci construction can be embedded under the ka-construction. In such a case, the verbal phrase headed by -(e)ci functions as the path complement of ka. This is illustrated in (330) and (331). Their meaning can be paraphrased as follows: ‘Daylight moved toward a darker point in the scale’ and ‘John’s height moved toward a higher point in the scale.’ (330) Nal-i etwuwe-ci-e-ka-ss-ta. day-Nom dark-CI-L-go-Past-Decl Cf. *Nal-i etwuwe-ka-ci-ess-ta. day-Nom dark-go-CI-Past-Decl (331) John-i khi-ka khe-ci-e-ka-ss-ta. John-Nom height-Nom tall-CI-L-go-Past-Decl Cf. *John-i khi-ka khe-ka-ci-ess-ta. John-Nom height-Nom tall-go-CI-Past-Decl Note that the above sentences are atelic and therefore incompatible with a telic temporal modifier as well as with a measure modifier. (332) a. Nal-i han sikan tongan/*han sikan man-ey day-Nom one hour for/*one hour in etwuwe-ci-e-ka-ss-ta. dark-CI-L-go-Past-Decl b. John-i khi-ka il nyen tongan/*il nyen man-ey John-Nom height-Nom one year for/one year in khe-ci-e-ka-ss-ta. tall-CI-L-go-Past-Decl (333) a. Nal-i (*wancenhi) etwuwe-ci-e-ka-ss-ta. day-Nom (completely) dark-CI-L-go-Past-Decl b. John-i khi-ka (*wancenhi) khe-ci-e-ka-ss-ta. John-Nom height-Nom (completely) tall-CI-L-go-Past-Decl

Serial-Verb Construction and Verbs of Motion in Korean

115

(334) a. *Kang-i il mithe-(ka) kiph-eci-e-ka-ss-ta. river-Nom one meter-(Nom) deep-CI-L-go-Past-Decl b. *John-i khi-ka il inchi-(ka) khe-ci-e-ka-ss-ta. John-Nom height-Nom one inch-Nom tall-CI-L-go-Past-Decl The above facts indicate that (for some reason that remains to be understood) the modifiers must be adjoined to the higher verbal structure—that is, to the structure headed by ka- and not to the structure headed by -(e)ci. (335)

Appendix 2.2: On Distinguishing the ‘‘Prospective Possessor’’ Construction from the ‘‘Directed-Motion’’ and ‘‘Benefactive’’ Constructions In section 2.2.4, we mentioned an apparently contradictory fact regarding transitive contact verbs. In this appendix, we provide an analysis that resolves this apparent contradiction. We begin by restating the issue. The examples in (336) show that the transitive contact verbs cha- ‘kick’, mil‘push’, and kkul-/tangki- ‘pull’ do not imply directed motion. Therefore these verbs cannot take a path-denoting argument. On the other hand, these verbs can enter into an SVC with ka- and the construction headed by ka- licenses a path argument; see the examples in (337). (336) a. John-un ku kong-ul (*cengwen-ey) cha-ss-ta. John-Top the ball-Acc (garden-Loc) kick-Past-Decl ‘John kicked the ball to the garden.’

116

Chapter 2

b. John-un ku khathu-lul (*cengwen-ey) John-Top the cart-Acc (garden-Loc) ‘John pushed the cart to the garden.’ c. John-un ku khathu-lul (*cengwen-ey) John-Top the cart-Acc (garden-Loc) ‘John pulled the cart to the garden.’

mil-ess-ta. push-Past-Decl kkul-ess-ta. pull-Past-Decl

(337) a. John-un cengwen-ey kong-ul cha ka-ss-ta. John-Top garden-Loc ball-Acc kick go-Past-Decl ‘John goes to the garden, repeatedly kicking the ball.’ b. John-un cengwen-ey khathu-lul mil-e ka-ss-ta. John-Top garden-Loc cart-Acc push-L go-Past-Decl ‘John goes to the garden, repeatedly pushing the cart.’ c. John-un cengwen-ey khathu-lul kkul-e ka-ss-ta. John-Top garden-Loc cart-Acc pull-L go-Past-Decl ‘John goes to the garden, repeatedly pulling the cart.’ An apparent puzzle arises when the transitive contact verbs co-occur with an animate complement, Case marked with the Dative Case marker -eykey. The resulting sentences are well formed, without the presence of ka-; see (338). The latter sentences are also compatible with ka-; see (339). (338) a.

John-un Bill-eykey ku kong-ul cha-ss-ta. John-Top Bill-Dat the ball-Acc kick-Past-Decl ‘John kicked the ball to Bill.’ b. John-un Bill-eykey ku khathu-lul mil-ess-ta. John-Top Bill-Dat the cart-Acc push-Past-Decl ‘John pushed the cart to Bill.’ c. (?)John-un Bill-eykey ku khathu-lul kkul-ess-ta. John-Top Bill-Dat the cart-Acc pull-Past-Decl ‘John pulled the cart to Bill.’

(339) a. John-un Bill-eykey ku kong-ul cha ka-ss-ta. John-Top Bill-Dat the ball-Acc kick go-Past-Decl ‘John goes to Bill, repeatedly kicking the ball.’ b. John-un Bill-eykey ku khathu-lul mil-e ka-ss-ta. John-Top Bill-Dat the cart-Acc push-L go-Past-Decl ‘John goes to Bill, repeatedly pushing the cart.’ c. John-un Bill-eykey ku khathu-lul kkul-e ka-ss-ta. John-Top Bill-Dat the cart-Acc pull-L go-Past-Decl ‘John goes to Bill, repeatedly pulling the cart.’

Serial-Verb Construction and Verbs of Motion in Korean

117

Shall we then conclude that in the presence of a dative goal, ka- is optional? We do not think so. When we examine the sentences in (338) and (339) closely, we find that they have di¤erent meanings, while the sentences in (337) and (339) have comparable meanings. In the latter sentences, repeated events of ball-kicking, cart-pushing, and cart-pulling are involved. These meanings can be attributed to the fact that the event denoted by the directed-motion construction headed by ka- maps onto distinct subintervals at the aspectual temporal level (the R-level; see section 1.5.1) and the event denoted by the transitive contact verb modifies each of these subintervals. The di¤erences in meaning between (338) and (339) can be summarized as follows. First, as mentioned above, in the case of (339), there is an iterative reading (i.e., repeated events of ballkicking, cart-pushing, and so on). The iterative reading is unavailable in (338). This is as expected, since ka- is absent in these sentences. Second, in (339), the subject undergoes movement with the object. On the other hand, in (338), only the object undergoes movement. The third meaning di¤erence is particularly revealing because it suggests that the dative DP (marked by -eykey) in (338) is actually the argument of a silent applicative verb, which gives rise to a prospective possessor reading; see den Dikken 1995, Harley 2002, and Marantz 1993. Indeed, (338) entails that the object is intended for Bill. This is not the case in the sentences in (339); the subject is moving in the direction of Bill, while kicking the ball, pushing the cart, or pulling the cart. Therefore the object moves along with the subject, but the subject does not intend Bill as the prospective possessor of the object. Thus, (341) is not contradicted by the adjunct clause ‘without any intention to give a ball to Bill’. Compare with (340). (340) aBill-eykey kong-ul cwu-l uyto epsi John-un Bill-Dat ball-Acc give-Rel intention without John-Top Bill-eykey kong-ul cha-ss-ta. Bill-Dat ball-Acc kick-Past-Decl ‘John kicked Bill the ball (aalthough he did not have the intention of giving Bill the ball).’ (341) Bill-eykey kong-ul cwu-l uyto epsi John-un Bill-eykey Bill-Dat ball-Acc give-Rel intention without John-Top Bill-Dat kong-ul cha ka-ss-ta. ball-Acc kick go-Past-Decl ‘John goes to Bill, repeatedly kicking the ball, without having the intention of giving Bill the ball.’

118

Chapter 2

The above observation suggests that the dative argument in the examples in (338) is not a goal argument of the verb, but that it is intended as the prospective possessor argument licensed by an abstract verb (comparable to the dative argument in the English example John kicked Mary the ball ). Once we accept the presence of an abstract verb in (338), the illformedness of the examples in (336) is predicted because the DP marked by the locative -ey cannot function as a prospective possessor. We will refer to this abstract verb as low applicative (or V*). In the remainder of this appendix, we will develop the structural analysis of examples in (338). To do so, we compare it to other related constructions: (1) the cwu- construction (high applicative in the framework put forth by Pylkka¨nnen 2002), (2) an SVC in which the transitive contact verb modifies ka- (henceforth, contact verb þ ka- construction) (cf. (337) and (339)), and (3) a construction headed by the so-called goal verbs such as tenci‘throw’ (henceforth, the goal construction). To investigate the relative structural positions of the heads of these constructions, we will use two diagnostics: scrambling and adverb placement. The dative argument licensed by the low applicative V* in constructions with transitive contact verbs and the benefactive dative argument licensed by the high applicative cwu- occupy di¤erent structural positions. High applicatives introduce a benefactive argument—that is, an argument that benefits from the event denoted by little vP. On the other hand, the low applicative V* introduces a prospective possessor of the object involved in the event denoted by the lower VP (i.e., the Acc object). (342) [ ApplP DP [vP DP [VP  DP [VP DP V] V*]] v] cwu-] The cwu- construction has a striking resemblance to applicative constructions in Bantu languages, which introduce a benefactor in the applicative constructions (see Marantz 1993). The verbal morpheme cwu- merges with a vP and introduces a dative argument (DP-eykey) that is interpreted as a benefactor. Such -eykey marked DPs can appear in the sentence if and only if cwu- is present, as exemplified below.36 (343) a. John-i Mary-eykey kapang-ul tul-e *(cwu)-ess-ta. John-Nom Mary-Dat bag-Acc hold-L Ben-Past-Decl ‘John held the bag for Mary.’ b. John-i (*Mary-eykey) kapang-ul tul-ess-ta. John-Nom (Mary-Dat) bag-Acc hold-Past-Decl

Serial-Verb Construction and Verbs of Motion in Korean

119

(344) a. John-i Mary-eykey kheyikh-ul kwu-e *(cwu)-ess-ta. John-Nom Mary-Dat cake-Acc bake-L Ben-Past-Decl ‘John baked a cake for Mary.’ b. John-i (*Mary-eykey) kheyikh-ul kwu-ess-ta. John-Nom (Mary-Dat) cake-Acc bake-Past-Decl The Dative Case–marked argument in the above sentences is the benefactor of the event denoted by vP; it is not a prospective possessor. To illustrate the point, consider the example in (344). Mary is not necessarily the intended recipient of the cake; Mary can benefit in other ways, such as by John baking the cake in lieu of Mary so Mary can take a break. Given the fact that the construction is comparable to the Bantu high applicatives and given the obligatory presence of cwu- in the construction, we refer to it as the High cwu- construction. Further confirmation of the benefactive construal of the Dative Case–marked argument in this construction is provided by the fact that such an argument can be freely interchanged with the adjunct headed by -ulwihay, the Korean counterpart of English for. When the adjunct -ulwihay appears in the sentence, cwu- can be dropped because -ulwihay encodes a benefactive meaning. (345) John-i Mary-lulwihay kheyikh-ul kwu-e (cwu)-ess-ta. John-Nom Mary-for cake-Acc bake-L Ben-Past-Decl ‘John baked a cake for Mary.’ The scrambling and adverb placement tests provide evidence that the dative argument licensed by High cwu- and the dative argument licensed by the abstract V* are structurally distinct. In the latter case, the theme argument can scramble over the Dative Case–marked argument (i.e., a prospective possessor); see (346). On the other hand, in the cwu- construction, the theme argument cannot scramble over the Dative Case–marked argument; see (347). (346) a. John-i Mary-eykey kong-ul cha-ss-ta. John-Nom Mary-Dat ball-Acc kick-Past-Decl ‘John kicked Mary the ball.’ b. John-i kong-ul Mary-eykey cha-ss-ta. John-Nom ball-Acc Mary-Dat kick-Past-Decl ‘John kicked Mary the ball.’ (347) a.

John-i Mary-eykey kapang-ul tul-e cwu-ess-ta. John-Nom Mary-Dat bag-Acc hold-L Ben-Past-Decl ‘John held Mary the bag.’

120

Chapter 2

b. ??John-i kapang-ul Mary-eykey tul-e cwu-ess-ta. John-Nom bag-Acc Mary-Dat hold-L Ben-Past-Decl Given the assumption that low adverbs like the manner adverb ppalli ‘quickly’ and the instrumental adverb phokhu-lo ‘fork with’ demarcate the vP boundary (cf. Ko 2005), the relative positioning of these low adverbs in the constructions under discussion reveals the relative structural positions of the dative arguments and, indirectly, of the V that licenses them. The low-adverb placement test also shows that cwu- is higher in the structure than abstract V*. Neither the manner nor the instrumental adverbs can precede the -eykey marked argument in the High cwu- construction; see (348a) and (349a). On the other hand, they can precede the -eykey marked argument (i.e., the prospective possessor) in the construction with abstract V*; see (350a) and (351a). This contrast suggests that the -eykey marked argument in the cwu- construction is located above vP, while the -eykey marked argument in the V*construction is located below vP. (348) a. */??John-i ppalli Mary-eykey kapang-ul tul-e John-Nom quickly Mary-Dat bag-Acc hold-L cwu-ess-ta. Ben-Past-Decl b. John-i Mary-eykey ppalli kapang-ul tul-e cwu-ess-ta. c. John-i Mary-eykey kapang-ul ppalli tul-e cwu-ess-ta. (349) a. */??John-i tulaipe-lo Mary-eykey khemphywuthe-lul John-Nom screwdriver-with Mary-Dat computer-Acc kochi-e cwu-ess-ta. fix-L Ben-Past-Decl b. John-i Mary-eykey tulaipe-lo khemphywuthe-lul kochi-e cwu-ess-ta. c. John-i Mary-eykey khemphywuthe-lul tulaipe-lo kochi-e cwu-ess-ta. (350) a. John-i ppalli Mary-eykey kong-ul cha-ss-ta. John-Nom quickly Mary-Dat ball-Acc kick-Past-Decl b. John-i Mary-eykey ppalli kong-ul cha-ss-ta. c. John-i Mary-eykey kong-ul ppalli cha-ss-ta. ‘John quickly kicked Mary the ball.’ (351) a. John-i pal-lo Mary-eykey kong-ul cha-ss-ta. John-Nom foot-with Mary-Dat ball-Acc kick-Past-Decl

Serial-Verb Construction and Verbs of Motion in Korean

121

b. John-i Mary-eykey pal-lo kong-ul cha-ss-ta. c. John-i Mary-eykey kong-ul pal-lo cha-ss-ta. ‘John kicked Mary the ball with the foot.’ Consider next the directed-motion construction headed by ka-; see (337). With respect to scrambling, ka- patterns with Low V*. Indeed the Acc object may scramble over the locative; see (352). This indicates that ka- is lower than cwu-. (352) a. John-i cengwen-ey kong-ul cha ka-ss-ta. John-Nom garden-Loc ball-Acc kick go-Past-Decl ‘John ended up in the garden by repeatedly kicking the ball.’ b. John-i kong-ul cengwen-ey cha ka-ss-ta. John-Nom ball-Acc garden-Loc kick go-Past-Decl As for the manner adverb, its distribution depends on whether it modifies V1 (the subordinate V) or the matrix V (the motion). In the former case, the instrumental adverb must follow the locative argument; see (353). Indeed, the instrumental adverb pal-lo ‘with the foot’ can only modify the kicking event (and not the motion); therefore, it must be adjoined to the projection of V1 and follow the locative argument. On the other hand, in (354a) the manner adverb can precede the locative argument (it is adjoined to V2), but it is unambiguously interpreted as modifying the subject’s motion and not the kicking event. Compare with (354b–c), where the manner adverb follows the locative argument and it unambiguously modifies the kicking event. (353) a. ??John-i pal-lo cengwen-ey kong-ul cha ka-ss-ta. John-Nom foot-with garden-Loc ball-Acc kick go-Past-Decl b. John-i cengwen-ey pal-lo kong-ul cha ka-ss-ta. c. John-i cengwen-ey kong-ul pal-lo cha ka-ss-ta. (354) a. John-i ppalli cengwen-ey kong-ul cha ka-ss-ta. John-Nom quickly garden-Loc ball-Acc kick go-Past-Decl ‘John quickly went to the garden kicking the ball.’ b. John-i cengwen-ey ppalli kong-ul cha ka-ss-ta. ‘John went to the garden quickly kicking the ball.’ c. John-i cengwen-ey kong-ul ppalli cha ka-ss-ta. ‘John went to the garden quickly kicking the ball.’ Given that Korean is head final, the structurally higher V is rightmost. We have seen above that cwu- is structurally higher than ka-. Therefore,

122

Chapter 2

the prediction with respect to the order of morphemes is as follows. When the benefactive morpheme cwu- and ka- co-occur, combined with the transitive contact verb cha- ‘kick’, the order should be cha ka-cwu-, not cha-cwu ka. The prediction is borne out: (355) a. (?)John-i (Bill-eykey) cengwen-ey kong-ul cha John-Nom (Bill-Dat) garden-Loc ball-Acc kick ka-cwu-ess-ta. go-Ben-Past-Decl b. *John-i cengwen-ey kong-ul cha cwu ka-ss-ta. John-Nom garden-Loc ball-Acc kick Ben Past-Decl This is consistent with the SVC analysis we put forth in this chapter. The l-structure of cha- enters into an SVC with the directed-motion construction headed by ka-, giving rise to an analysis in which cha- modifies the directed-motion event (i.e., cha- is subordinate to ka-). The SVC merges with little v, giving rise to to a transitive verbal phrase, which in turn functions as complement of cwu-. The resulting structure is given below. (356)

Serial-Verb Construction and Verbs of Motion in Korean

123

To complete the picture, we compare the Low applicative construction (headed by abstract V*) and the goal construction (headed by the lexical V). Unlike the transitive contact verbs, goal verbs (like tenci- ‘throw’) entail directed motion. In other words, goal verbs, unlike contact verbs, merge with a goal argument, realized as a locative P (-ey) if the goal is inanimate and as a dative P (-eykey) if the goal is animate. The examples in (357) illustrate the point. The fact that locative -ey and dative -eykey can alternate suggests that both play the same role in the sentence. This, in turn, suggests that in (357a), DP-eykey is distinct from the prospective possessor argument licensed by V*. We argue that, like the -ey marked argument, the -eykey marked argument functions as the goal argument of the verb in this sentence. In this respect, verbs such as tenci- ‘throw’ are di¤erent from the transitive contact verbs—for example, (357b). (357) a. John-un cengwen-ey/Bill-eykey kong-ul tenci-ess-ta. John-Top garden-Loc/Bill-Dat ball-Acc throw-Past-Decl ‘John threw the ball to the garden/to Bill.’ b. John-un (*cengwen-ey)/Bill-eykey kong-ul cha-ss-ta. John-Top (garden-Loc)/Bill-Dat ball-Acc kick-Past-Decl ‘John kicked the ball to the garden/to Bill.’ The scrambling and adverb placement tests show that like the V*construction, the goal construction is also contained within the vP domain. On the one hand, the theme argument can scramble over the goal argument and, on the other hand, the manner and instrumental adverbs can precede the goal argument. (358) a. John-un cengwen-ey kong-ul tenci-ess-ta. John-Top garden-Loc ball-Acc throw-Past-Decl ‘John threw the ball to the garden.’ b. John-un Bill-eykey kong-ul tenci-ess-ta. John-Top Bill-Dat ball-Acc throw-Past-Decl ‘John threw the ball to Bill.’ (359) John-un kong-ul cengwen-ey tenci-ess-ta. John-Top ball-Acc garden-Loc throw-Past-Decl (360) a. John-un ppalli cengwen-ey kong-ul tenci-ess-ta. John-Top quickly garden-Loc ball-Acc throw-Past-Decl ‘John quickly threw the ball to the garden.’ b. John-un cengwen-ey ppalli kong-ul tenci-ess-ta. c. John-un cengwen-ey kong-ul ppalli tenci-ess-ta.

124

Chapter 2

(361) a. John-un mayn son-ulo cengwen-ey kong-ul tenci-ess-ta. John-Top bare hand-with garden-Loc ball-Acc throw-Past-Decl ‘John threw the ball to the garden with a bare hand.’ b. John-un cengwen-ey mayn son-ulo kong-ul tenci-ess-ta. c. John-un cengwen-ey kong-ul mayn son-ulo tenci-ess-ta. Nevertheless, the two constructions have di¤erent structures. Indeed, the -eykey marked arguments are licensed by di¤erent heads in the two constructions. The -eykey marked argument in the V*-construction is a prospective possessor argument licensed by an abstract applicative verb. On the other hand, the -eykey marked argument in the goal construction is contained within the l-structure of the verb (the lowest VP). Given the fact that V* is above the lower VP, an implication follows: the -eykey marked argument in the V*-construction is structurally higher than the -eykey marked argument in the goal construction. The generalizations discussed above are captured by the structure in (362). (362)

Serial-Verb Construction and Verbs of Motion in Korean

125

To recapitulate, in this appendix, we have addressed an apparently contradictory set of facts regarding transitive contact verbs. That is, while the transitive contact verbs cannot co-occur with a locative-marked complement, they do co-occur with a dative-marked argument. We resolved this contradiction by arguing that the dative-marked argument is a prospective possessor of an abstract applicative V and not a goal argument of the lexical verb. Indeed, transitive contact verbs do not license a goal argument; they must enter into an SVC with ka- in order to co-occur with a goal argument. Two clear predictions follow from the above claim. One immediate prediction (already mentioned above) is that in sentences with a transitive contact verb (without ka-), the dative-marked argument cannot alternate with the locative-marked argument. The second prediction is that in sentences with goal verbs such as tenci- ‘throw’, two dative-marked arguments should be able to co-occur: a higher dative argument, which functions as a prospective possessor, and a lower dative argument, which functions as the goal argument of the lexical verb.37 Alternatively, it is predicted that the dative-marked argument that functions as a prospective possessor and the locative-marked DP that functions as the goal argument of the verb should be able to co-occur in the same sentence; see the example in (363) and its associated structure in (364). There are speakers that accept examples such as (363). (See Miyagawa and Tsujioka 2004 for similar observations regarding the co-occurrences of two -ni marked DPs in Japanese.) But there seems to be idiolectal variation; most speakers with whom we consulted find such sentences awkward or degraded. Why this should be so remains to be explained.38 (363) John-un Bill-eykey kong-ul cengwen-ey tenci-ess-ta. John-Top Bill-Dat ball-Acc garden-Loc throw-Past-Decl ‘John threw Bill the ball to the garden.’

126

(364)

Chapter 2

Chapter 3 Motion Verbs in Germanic and Romance

In this chapter, we examine the manner-of-motion construction in Germanic and Romance in some detail. Our conclusion supports the oftenmade observation that the two are fundamentally di¤erent. In the first part of this chapter (section 3.1), we discuss the manner-of-motion construction in English and Dutch, as well as the cause-motion construction. While neither Germanic nor Romance languages are serial-verb languages, the use of manner-of-motion verbs in the directed-motion construction in Germanic is general and exceptionless. Building on insights due to Snyder 1995, 2001, and Beck and Snyder 2001, we argue that Germanic uses the Compound Rule (which is very productive and lexically unrestricted in this language family) to compose manner and directed motion. More precisely, we contend that manner-of-motion verbs in sentences with a directed-motion meaning are compounded with a covert light (aspectual) verb. In the second part of this chapter (section 3.2), we discuss Spanish and French, on the one hand, and Italian, on the other hand. Again building on the above-mentioned works by Snyder as well as Beck and Snyder, we argue that, because Romance lacks a productive and lexically unrestricted Compound Rule, it cannot use this rule to compose manner and directed motion. Yet, Italian appears to be able to compose manner with directed motion in some lexically restricted cases. We propose that Italian recruits a certain (independently established) auxiliary position in so-called restructuring contexts in order to compose manner and directed motion in some lexically restricted cases. We conclude with a summary and some final remarks regarding the nature of the dative preposition in Romance and Germanic (section 3.3).

128

3.1

Chapter 3

Germanic

This section is organized as follows. We first examine the dual behavior of manner-of-motion verbs in English (section 3.1.1) and Dutch (section 3.1.2). An analysis based on the Germanic Compound Rule is put forth to account for this dual behavior (section 3.1.3). We then examine the cause-motion examples in English and extend the compound-based analysis to these cases as well (section 3.1.4). 3.1.1 Manner-of-Motion Verbs in English As is well known, manner-of-motion verbs in English (and more generally, in Germanic languages) can take a dative directional complement (365), as well as other path-denoting PP complements (366). (365) a. b. c. d. e. f.

John danced to the kitchen. John swam to the boat. The bottle floated to the beach. John wobbled to the door. The soldiers marched to the park. The athletes ran to Paris.

(366) a. b. c. d.

The horse jumped over the fence. The bottle floated under the bridge. They danced out of the room. The horse galloped into the barn.

The above facts are remarkable given that manner-of-motion verbs do not generally entail directed motion (most notably, dance, wobble, and float). Manner-of-motion verbs are typically activity-denoting verbs. (367) a. b. c. d. e. f. g.

John danced (for an hour). John swam (for thirty minutes). The bottle floated (for a few seconds). John wobbled (for a minute). The soldiers marched (for two days). The athletes ran (for a while). The horse jumped/galloped (for a few minutes).

The phenomenon exemplified in (365) and (366) in English and other Germanic languages is exceptionless; it is a general process with no lexical restrictions. Postulating polysemy for each of these verbs would therefore miss a generalization.

Motion Verbs in Germanic and Romance

129

Nonetheless, there is some evidence (although not always solid) that there is an analysis of the English sentences in (365) in which the pathdenoting PP is the complement of a bounded directed-motion construction (368a), to be distinguished from another analysis in which the path-denoting PP is an adjunct to an unergative l-structure that denotes an unbounded activity (368b). As we will see in section 3.2, the evidence is more robust in Dutch, because this language provides us with an overt property to distinguish the two analyses, namely, choice of auxiliary. (368) a. [DP [V Ppath ]] (Complementation analysis) b. [ . . . [DP [v [V X]]] Ppath ] (Adjunct analysis) If the path-denoting PP functions as an adjunct, the verb is associated with an unergative structure and denotes an unbounded activity. Under this reading, the manner of motion is not necessarily coextensive with the path. The meanings associated with the sentences in (369) are akin to the ones in (370).1 (369) a. John ran for a while to the park (he walked the rest of the way). b. John walked for a while to the park (he ran the rest of the way). (370) a. On his way to the park, John ran for a while (he walked the rest of the way). b. On his way to the park, John walked for a while (he ran the rest of the way). Under the complementation analysis, the event is bounded and the manner of motion is necessarily coextensive with the path: (371) a. John danced (in)to the kitchen in a minute (ahe walked the rest of the way). b. John walked to the park in an hour (ahe ran the rest of the way). The hypothesis that the adjunct PP is structurally higher than the complement PP is supported by the following contrast. While the temporal phrase in (369) can precede the path-denoting PP, this is not possible in the case of (372).2 (372) a. ??John danced in a minute (in)to the kitchen. b. ??John walked in an hour to the park. There are other tests, such as extraction out of islands and VP-deletion, that are often used to distinguish complements from adjuncts. In the cases under discussion, we have found that these tests give rise to variable

130

Chapter 3

judgments. Consider the VP-deletion test. It is generally assumed that VP-deletion can strand an adjunct but not a complement. If that is the case, then we should expect the contrast below, where (373a) is judged as perfect and (373b) as marginal or less than perfect. Indeed, speakers that can use to in atelic situations do get the contrast (see note 1). (373) a.

John ran for a while to his house and Mary did so to her house (then they both got tired and walked the rest of the way). b. ??John ran to his house in twenty minutes and Mary did so to her house in thirty minutes.

Consider next the extraction test. It has been claimed that adjuncts cannot be extracted from weak islands, but complements can; see Rizzi 1990. Thus, in (374a), when can be interpreted as modifying the lower VP, while in (374b), it cannot. Given the surface string in (374b), when can only be interpreted as modifying the higher VP. This is due to the fact that extraction of adjuncts out of islands is impossible. Compare (374b) with long extraction of the direct object in (374c), which is somewhat acceptable to some speakers. Similarly, long extraction of the pathdenoting PP in (375b) is somewhat acceptable to such speakers, while (375c) is completely out. (374) a. Wheni do you think that Snow White will eat this apple ei ? b. *Wheni do you wonder which apple Snow White will eat ei ? c. ?Which applei do you wonder when Snow White will eat ei ? (375) a. To which park do you think John will/run walk ei in an hour? b. ?To which park do you wonder when John will run/walk ei in an hour? c. *To which park do you wonder when John will walk/run for a while ei (before hopping on a bus)? Cf. I wonder when John will walk/run for a while to the park (before hopping on a bus). A closer look at manner-of-motion verbs in English (and other languages) reveals that there are actually two subclasses: (376) a. run, walk, swim, gallop . . . b. wander, rove, wobble, sway, saunter . . . The first class of verbs, but not the second, can take a distance-denoting complement. We proposed in chapter 1 that those manner-of-motion verbs take a generic classifier that measures distance or interval, which

Motion Verbs in Germanic and Romance

131

we will refer to as the ‘‘distance classifier.’’ This distance classifier may be unspecified and incorporated into the verb, giving rise to the unbounded, activity-denoting unergative structure. Alternatively, it may be overt and specified (giving rise to a bounded process). Compare (377) with (378). (377)

John ran/walked/swam/galloped a certain distance/a mile.

(378) *John wandered/roved/wobbled/sauntered a certain distance/a mile. While we may think of ‘‘distance’’ as an abstract path, the fact that a verb can take ‘‘distance’’ as a complement does not entail that it can take a ‘‘physical’’ path as a complement—that is, it does not entail that such a verb will license a path-denoting PP as a complement. This will become clearer when we examine French and Spanish in section 3.2. We therefore conclude that the path-denoting PP in the examples in (371) is not a complement of the lexical verb at all. It must be the complement of an empty verb that heads a directed-motion construction, which in turn is modified by the manner verb. Still more striking is the behavior of the class of verbs in (376b), which in and of themselves cannot license a path complement of any kind; see (378). As we will see in the next section, Dutch clearly shows that the PP in examples like (379) has the status of a complement. Again, this suggests that the PP is a complement of an abstract head—that is, of an empty V that heads a directed-motion construction, which in turn is modified by the manner verb. The meanings obtained are akin to ‘The boy went to the lake in twenty minutes roving/wandering’ and to ‘The kid went to the door in three seconds wobbling’. (379) a. The boy roved/wandered to the lake in twenty minutes. b. The kid wobbled to the door in three seconds. To summarize, while a subset of the manner-of-motion verbs can take a distance-denoting complement, manner-of-motion verbs generally do not license a PP complement, a generalization that can be better appreciated in languages like Korean and Romance. This leads us to postulate an empty V that licenses the PP complement in sentences like (365) and (366); the lexical verb in turn modifies the empty V. 3.1.2 Manner-of-Motion Verbs in Dutch Like English, Dutch also has a productive and exceptionless usage of manner-of-motion verbs with a directed-motion reading. As shown by

132

Chapter 3

Hoekstra and Mulder 1990, Dutch clearly distinguishes event boundedness from event unboundedness in terms of auxiliary selection. More precisely, the past participle formed with zijn ‘be’ gives rise to a bounded meaning, while the past participle formed with hebben ‘have’ gives rise to an unbounded one. The examples below illustrate this point.3 While (380a) implies that Jan walked all the way to Groningen, there is no such implication in (380b) (he could have walked only part of the way). Similarly, (381a) implies that Jan swam all the way to the other side, while (381b) does not (he could have swum only part of the trajectory).4 (380) a. dat Jan naar Groningen gewandeld is. that Jan to Groningen walked is ‘Jan walked to Groningen.’ b. dat Jan naar Groningen gewandeld heeft. that Jan to Groningen walked has ‘On his way to Groningen, Jan walked.’ (381) a. dat Jan naar de overkant gezwommen is. that Jan to the other side swum is ‘Jan swam to the other side.’ b. dat Jan naar de overkant gezwommen heeft. that Jan to the other side swum has ‘On his way to the other side, Jan swam.’ In (380a) and (381a), the VP denotes directed motion (change of location), while in (380b) and (381b), it denotes an activity. Hoekstra and Mulder (1990) argue that the sentences in (380a) and (381a) have an unaccusative-like structure, in which the path-denoting PP has the status of a complement. The path is predicated of the subject, giving rise to an eventive meaning in which an individual moves in a certain manner to a certain location and this meaning can be measured by a telic temporal adverb, as shown in (382a). Hoekstra and Mulder furthermore argue that the sentences in (380b) and (381b) have an unergative structure, in which the path-denoting PP has the status of an adjunct and the VP denotes an unbounded activity. As expected, given the unbounded nature of the VP, it cannot be modified by a telic temporal phrase, as shown in (382b). (382) a.

dat Jan that Jan b. *dat Jan that Jan

in twee uur in two hour in twee uur in two hour

naar Groningen to Groningen naar Groningen to Groningen

gewandeld walked gewandeld walked

is. is heeft. has

Motion Verbs in Germanic and Romance

133

There is a distributional di¤erence between the PP in the unbounded construction and the PP in the bounded construction, which can be neatly accounted for in terms of the complement-adjunct distinction. Consider the perfective form in (383a), which asserts that Jan walked all the way to Groningen in two hours. The path-denoting PP in this case appears closer to the verb than the time adverbial. On the other hand, in the unbounded form in (383b), it is asserted that on his way to Groningen, Jan walked for two hours, with the implication that some other means was used during the rest of the trajectory. In this case, the time adverbial appears closer to the verb than the path-denoting PP; compare (383b) with (383c). (383) a.

dat that b. dat that c. ??dat that

Jan Jan Jan Jan Jan Jan

in twee uur naar Groningen is gewandeld. in two hours to Groningen is walked naar Groningen twee uur lang heeft gewandeld. to Groningen two hours long has walked twee uur lang naar Groningen heeft gewandeld. two hours long to Groningen has walked

As noted by Hoekstra and Mulder (1990), extraposition provides independent evidence that the PP is a complement in the bounded construction and an adjunct in the unbounded one.5 It is generally the case that adjuncts, but not complements, may be extraposed to the right of the verb in a V-final structure. The sentence in (384a), constructed with the auxiliary zijn, is bounded (it entails that Jan completed a circle around the fountain). In this case, the PP rond de fontein cannot be extraposed, as shown in (384b). The sentence in (385a), constructed with the auxiliary hebben, is unbounded (it does not entail that Jan completed a circle around the fountain). In this case, the PP rond de fontein, can be extraposed, as shown in (385b). (384) a.

dat that b. *dat that

(385) a. dat that b. dat that

Jan Jan Jan Jan

Jan Jan Jan Jan

rond de fontein gedanst/gelopen/?gerend is. around the fountain danced/walked/run is gedanst/gelopen/gerend is rond de fontein. danced/walked/run is around the fountain

rond de fontein gedanst/gelopen/gerend heeft. around the fountain danced/walked/run has gedanst/gelopen/gerend heeft rond de fontein. danced/walked/run has around the fountain

134

Chapter 3

The example in (386a) is unambiguously bounded, as indicated by the fact that the auxiliary hebben cannot be used. As expected, the PP cannot be extraposed, as shown in (386b). (386) a.

dat de vogel het raam uit gevlogen is/*heeft. that the bird the window out flown is/*has ‘that the bird has flown out the window.’ b. *dat de vogel gevlogen is het raam uit. that the bird flown is the window out

The paradigm in (384)–(385) is also obtained with the atelic PP richting de fontein ‘toward the fountain’, as illustrated below. Native speakers have di‰culty articulating the meaning di¤erence between (387a) and (388a), but it appears that (387a), unlike (388a), implies that the entire trajectory (whatever it might be) is accomplished dancing, walking, or running.6 (387) a. dat Jan richting de fontein gedanst/gelopen/?gerend is. that Jan toward the fountain danced/walked/run is b. dat Jan gedanst/gelopen/gerend *is richting de fontein. (388) a. dat Jan richting de fontein gedanst/gelopen/gerend heeft. that Jan toward the fountain danced/walked/run has b. dat Jan gedanst/gelopen/gerend heeft richting de fontein. Let us turn next to the Dutch counterparts of the English sentences in (379), which involve verbs that are incompatible with a path-denoting complement. Yet, the directional PP cannot appear to the right of the verb, which indicates that it has the status of a complement.7 Furthermore, in the perfect, such examples require the auxiliary zijn ‘be’, which indicates that they are unaccusative structures. (389) a.

Jan is richting het meer gezworven. John is toward the lake wandered/roved b. *Jan is gezworven richting het meer.

(390) a.

Jan is naar de deur gewaggeld. John is to the door wobbled b. *Jan is gewaggeld naar de deur.

To recapitulate, Dutch provides clear evidence that examples involving a manner-of-motion verb and a directional PP can be associated with an unaccusative structure, in which the PP has a complement status. As suggested earlier for English, we can account for these facts by analyzing the

Motion Verbs in Germanic and Romance

135

directional PP in such examples as a complement of an empty light V that heads the directed-motion construction. The lexical verb, which denotes the manner of motion, modifies the directed-motion construction. 3.1.3 The Analysis of ‘‘Manner B Directed Motion’’ in Germanic The question then arises as to how ‘‘directed motion’’ and ‘‘manner’’ are composed in English and Dutch. Unlike Korean, English and Dutch are not serial-verb languages.8 Yet, the semantics of the Germanic construction is very comparable to that of the Korean SVC that consists of a manner-of-motion intransitive verb followed by the light verb ka- ‘go’. This is shown by the fact that the two constructions are subject to the same restrictions as to which semantic type of lexical verbs can appear in them; see section 2.2. These are the manner-of-motion verbs discussed earlier and sound-of-emission verbs, as long as the sound can be interpreted as the result of the motion and as coextensive with the directed motion. This is illustrated by the paradigms below, discussed by Goldberg and Jackendo¤ 2004 and attributed to Levin and Rappaport 1990 (see also Rapapport and Levin 2001). Goldberg and Jackendo¤ note that while the sentences in (391) are acceptable, those in (392) are unacceptable for most speakers. They suggest that this is due to the fact that in the latter examples, the emitted sound fails to be understood as a result of the movement.9 (391) a. The trolley rumbled through the tunnel. b. The wagon creaked down the road. c. The bullets whistled past the house. (392) a. aThe car honked down the road. b. aThe dog barked out of the room. c. aBill whistled past the house. The oddness of examples like (393) follows as well. The verbs laugh, joke, and talk denote activities that are unrelated to the motion. (393) a. aThey laughed down the road. b. aThey joked out of the room. c. aThey talked past the house. The semantic constraint discussed above does not apply in the case of phrasal modification. In this case, the event denoted by the modifier and the event denoted by the matrix VP predicate, while temporally related, need not be intrinsically related. Compare (392) and (393) with the

136

Chapter 3

examples below, which involve phrasal modification that does not impose an intrinsic semantic relation between the modifier and the directed motion. (394) a. The car went down the road honking. b. The dog went out of the room barking. c. Bill went past the house whistling. (395) a. They went down the road laughing. b. They went out of the room joking. c. They went past the house talking. Obviously, there is something in common between Germanic and Korean. In both cases, the manner verb modifies the head of the directedmotion construction, giving rise to a tight semantic relation between the two. In Korean this is obtained by adjoining the l-structure of the manner verb to the head of the directed-motion construction via the Generalized Transformation (GT) below: (396) Merge a verbal l-structure with the head of another verbal l-structure. The above GT is not part of the grammar of Germanic languages. If it were, we should expect sentences like John ran go to the market. But Germanic has another mechanism that can achieve the same results. This is the Compound Rule below: (397) Merge two lexical categories of the same category type. Snyder (1995, 2001) and Beck and Snyder (2001) have argued that the existence of resultatives in Germanic (like the ones discussed here) is related to the fact that (398) Germanic has productive and compositional N-N compounding. N-N compounding is generated by the rule in (397); see Roeper, Snyder, and Hiramatsu 2002. While overt V-V compounding is virtually nonexistent in Germanic, rule (397) is co-opted in Germanic to compose ‘‘manner’’ and ‘‘directed motion.’’ This is achieved by merging a fully specified lexical item (such as dance, wobble, and run) with a category V with no P-features and no C-features: (399) a. [dance V] b. [wobble V] c. [run V]

Motion Verbs in Germanic and Romance

137

Such compounds can then function as the head of the directed-motion construction, as exemplified below. On the structure of directional PPs, see below. (400)

The verb go (like ka-) is not listed in the lexicon. It is the spell-out of V in a particular syntactic context: (401) V is spelled out as go (or come) when it functions as the head of a ‘‘bare’’ directed-motion construction. A ‘‘bare’’ directed-motion construction is one in which the structure that encodes directed motion is contained in a phase domain with no phonologically specified verbal head. The directed-motion construction is formed by merging V with a directional PP. The resulting structure is then merged with a DP specifier. The argument in the specifier position is interpreted as undergoing movement along the path denoted by the complex PP. (402)

We furthermore assume that (403) CP and the highest verbal phrase in the (extended) l-structure are phases and phases constitute the domain of spell-out.10 In (402), VP is a phase and V is the head of a ‘‘bare’’ directed-motion construction; therefore, V is spelled out as go. In (400), on the other hand, V fails to be spelled out because there is another verbal head in

138

Chapter 3

the l-structure that contains phonological material. In other words, (400) is not a ‘‘bare’’ directed-motion construction; the presence of the verbal modifier blocks the spell-out rule from applying. (Recall that in the SVC, the l-structures involved are spelled out before they undergo the Generalized Transformation that adjoins one to the other. Therefore, the head of the directed-motion construction is spelled out in Korean, namely as kaor o-, depending on the deixis specification on the complement P.) We discuss next some evidence in favor of the complex analysis of path-denoting PPs. Goldberg (1995) noticed that locative prepositions, such as within, inside, and outside, in combination with some manner-ofmotion verbs (such as run), can give rise to a directed-motion meaning. We have found English speakers for whom this is also true with the locative preposition in. Thus, while (404a) is unambiguously directional in meaning, (404b–d) are ambiguous between a locative and a directional meaning. (404) a. b. c. d.

John John John John

ran ran ran ran

into/out of the room. within arm’s length of the grenade. inside/outside the house. in the house.

Goldberg (1995) also notes that when the PPs in (404b–d) are topicalized, only the locative meaning is available. Compare (405a) with (405b–c). (405) a. Into the room, he ran, quick as lightning. (Directional meaning) b. Inside/outside the room he ran, quick as lightning. (*Directional reading) c. Within arm’s length of the grenade, he ran, quick as lightning. (*Directional reading) A ‘‘bare’’ directed-motion construction, the complement of which is by definition a path-denoting PP, can also license a directional meaning in the context of an overt locative preposition; see (406). But when such P is preposed, the outcome is ungrammatical; see (407b–c). These facts suggest that there is a covert directional P in such structures, which must be locally licensed. (406) a. John went into/out of the room. b. John went within arm’s length of the grenade, quick as lightning. c. John went inside the room/outside the room, quick as lightning.

Motion Verbs in Germanic and Romance

139

(407) a. Into/out of the room, he went, quick as lightning. b. *Within arm’s length of the grenade, he went, quick as lightning. c. *Inside the room/outside the room, he went, quick as lightning. The contrast between (404b–d) and (408b) furthermore shows that only the verbal form can license the directional reading, when combined with a locative preposition. The nominal form cannot license the directional meaning. (408) a. John’s running (in)to the house (Directional reading) b. John’s running in/inside/outside the house (*Directional reading) On the other hand, not all manner-of-motion verbs (when compounded with V) can license the empty directional P. Thus, while walk and run can give rise to a directed-motion meaning when combined with a locative preposition, dance and gallop do not:11 (409) a. John ran/walked in/inside the house. (Directional meaning) b. John danced in/inside the house. (*Directional meaning) c. The horse galloped in/inside the barn. (*Directional meaning) The above facts point to the presence of a (covert) path-denoting complementation analysis for the PPs in examples such as (404b–d) and (406b– c). In other words, such PPs are headed by an empty directional P. It has been proposed by several authors that path-denoting PPs are complex in nature; they contain a locative P embedded under a directional P (see Koopman 1997; Hale and Keyser 2002; Folli 2001). The complex nature of these PPs can be easily appreciated in cases like into and onto, where the locative has adjoined to the directional P. But in other cases, the preposition is covert, as is the case in (410b) and in (410c). The proposal is that when a locative P (like in/inside/outside) is associated with a directional meaning, it is due to the fact that it is embedded under another P.12 (410) a. [ P [in [to]] [ P (on) [the house]]] b. [ P to [ P e [the house]]] c. [ P e [ P in/inside/outside [ DP the house]]] The locative P in (410b) is locally identified by the directional P. On the other hand, the empty P in (411) is locally identified by a lexically restricted set of verbs, such as go, run, and walk.

140

Chapter 3

(411) [VP DP [V [ P e [ P in/inside/outside [ DP the house]]]]] The locality constraint on the empty directional P in the structure above is reminiscent of the locality constraint on an empty Comp in English. That is, an empty Comp is licensed only when it is ‘‘governed’’ by certain verbs (think, know, suspect, and so on) and not by others (whisper, shout, regret, deny). See Stowell 1981.13 (412) a. John thinks/knows/suspects [e [Mary will leave early]] b. *John thought/knew/suspected yesterday [e [Mary would leave early]] c. John regrets/denies/whispered [e [Mary left yesterday]] Likewise, we can state the locality constraint on empty Ps in terms of ‘‘government,’’ where we understand ‘‘government’’ as an extension of the notion of ‘‘sisterhood’’; see below. We assume the locality constraint applies at PF, where V is spelled out as go in the absence of another verbal head with phonological specifications. (413) a governs b if b or the projection of b is a sister to a. (414) An empty P is licensed i¤ it is governed by a lexically restricted verbal head or another lexical P at PF. It is also the case that in Dutch, as in English, only certain verbs can license a directional meaning with locative prepositions. In Dutch, postpositions are obligatorily directional. (415) a. Jan Jan b. Jan Jan c. Jan Jan

sprong de sloot in. (Directional only) jumped the ditch in liep/rende het bos in. (Directional only) walked/ran the woods in klom de berg op. (Directional only) climbed the mountain on

When they are prepositional, these same adpositions have a locative interpretation. As noted by den Dikken (2003), they may also have a directional reading, but only when governed by certain lexical verbs. Compare (416a–b) with (416c).14 The same analysis that we gave for the English examples in (404b–d) can be given for the Dutch examples in (416a–b) under the directional interpretation. They involve an empty directional P that must be governed by certain lexically restricted set of verbs, and these verbs are not necessarily the same in English and in Dutch; see (409a) and (416c).

Motion Verbs in Germanic and Romance

(416) a. Jan Jan b. Jan Jan c. Jan Jan

141

klom op de berg. (Locative or directional) climbed on the mountain sprong in de sloot. (Locative or directional) jumped in the ditch liep/rende in het bos. (Locative only) walked/ran in the woods

The question arises as to whether there is some fundamental grammatical property that distinguishes the English verbs run and walk (which can license an empty directional P) from verbs like dance and gallop (which cannot). Or is there any grammatical property that distinguishes the Dutch verbs in (416a–b) (which can license an empty directional P in prepositional structures) from those in (416c) (which cannot)? We do not think so. Possibly the verbs that can do so are the ones that are more frequently used in a directed-motion construction. In other words, which verbs can license an empty P and which verbs cannot might very well be a question of convention (perhaps grounded in frequency of usage). We note also that there are idiolectal variations in this respect (see note 11). To summarize, following insights originally due to Synder 1995, 2001, it was proposed that Germanic composes ‘‘manner’’ and ‘‘directed motion’’ via the Compound Rule in (397). In other words, this rule freely applies to nominal categories (giving rise to a very productive N-N compound vocabulary), but it does not apply freely to verbal categories. More specifically, it does not apply freely to phonologically specified verbal categories due to independent factors in the grammar of Germanic languages; see section 1.6.2. On the other hand, the Compound Rule can apply to a phonologically specified verbal lexical item (which expresses manner) and a phonologically unspecified verbal category, which when merged with a path-denoting PP can license a directed-motion meaning. In the following section, we will see that the same Compound Rule can be used to compose ‘‘causation’’ and ‘‘directed motion.’’ 3.1.4 The Cause-Motion Construction in Germanic In section 1.4, we have seen that the directed-motion construction can merge with little v and give rise to a causative construction. (417) [vP D [v [VP D [V V [Pdir ]]]]] There are many verbs that can ‘‘instantiate’’ such a construction. These are verbs that entail directed motion, such as send, give, receive, drop, and many others.15

142

Chapter 3

(418) [vP D [send v [VP D [V (send ) PP]]]] Yet there are many verbs that may appear in the cause-motion construction although they are not motion verbs or do not entail directed motion. Some examples from Goldberg (1995) are given below. (419) a. The professor talked us into a stupor. b. The critics laughed the play o¤ the stage. c. He sneezed the napkin o¤ the table. (420) a. Sam sawed/tore/hacked/ripped a piece o¤ the block. b. Sam rinsed/cleaned the soap out of her eyes. c. Sam stirred the paint thinner into the paint. As remarked by Goldberg (1995) and others, the lexical verbs in (419) are activity-denoting verbs. The object in these sentences is not an argument of the verb, as shown by the ungrammaticality of (421). While the sentences in (422) are grammatical, they have a meaning quite di¤erent from (420). The examples in (420) have a directed-motion meaning that is entirely lacking in the examples in (422). (421) a. *The professor talked us. b. *The critics laughed the play. c. *He sneezed the napkin. (422) a. Sam sawed/tore/hacked/ripped a piece. b. Sam rinsed/cleaned the soap. c. Sam stirred the paint thinner. Similarly, none of the verbs in (423) imply motion of the object, and therefore they should not be analyzed as ‘‘instantiating’’ a cause-motion construction. A causative event, by definition, entails a result; see section 1.4. (423) a. b. c. d.

Mary Mary Mary Mary

pushed the cart (to the barn). kicked the boy (out of the room). stared the boy (into a corner). scared the cat (out of the room).

Following the line of analysis discussed in the previous section, it may be assumed that the Compound Rule (397) can also be co-opted to generate such cause-motion examples. More precisely, in the cases discussed above, the lexical verb has been compounded with little v, generating compound forms such as the following:

Motion Verbs in Germanic and Romance

(424) a. b. c. d. e.

143

[laugh v] [sneeze v] [ push v] [kick v] [scare v]

These compounds are then merged with the directed-motion construction, giving rise to structures such as the ones in (425). The lexical verb is predicated of the DP in the Spec of little v, giving rise to a meaning akin (although not identical) to The critics caused the play to come o¤ the stage by laughing at it; Mary caused the boy to go into a corner by staring at him.16 (425) a. [vP the critics [[v laughed v] [VP the play [V [o¤ [(of ) the stage]]]]]] b. [vP Mary [[v stared v] [VP the boy [V [into [a corner]]]]]] Contacts verbs (such as kick, hit, smack, tap, jab) can also readily modify the cause-motion construction: (426) a. John kicked/hit/smacked the ball to left field. b. John tapped the ball to the pitcher. c. John jabbed the needle into my knee. As illustrated below, such sentences have a structure comparable to (425), where movement toward a goal is predicated of the object and the kicking/jabbing property is attributed to the causing event.17 (427) a. [John [v [kicked v] [the ball [V [to left field]]]]] b. [John [v [ jabbed v] [the needle [V [into my knee]]]]] Given the definition of spell-out domain given in (403), the embedded V in structures (425) and (427) will remain phonologically silent—that is, rule (401) will fail to apply in such cases. Indeed, definition (403) states that the verbal phase is the highest verbal projection in the (extended) l-structure. This means that little vP is the verbal phase and not the VP embedded under it. It is only in the absence of little v that VP counts as a phase. As expected, Dutch also has sentences of the type discussed above. The example in (428a) is from Hoekstra and Mulder (1990); the ones in (428b– c) were provided to us by Marcel den Dikken, personal communication. (428) a. dat Jan zijn tegenstanders op achterstand heeft gezwommen. that John his adversaries on arrears has swum ‘that Jan outswam his adversaries.’

144

Chapter 3

b. Jan schopte Marie de kamer out. Jan kicked Marie the room out ‘Jan kicked Marie out of the room.’ c. Het publiek lachte de acteurs van het toneel af. The audience laughed the actors from the stage o¤ ‘The audience laughed the actors o¤ the stage.’ We turn next to verbs of sound emission that cannot be interpreted as denoting the result of the motion and are therefore infelicitous in the change-of-location construction. See the examples in (392), which we repeat in (429). (429) a. aThe car honked down the road. b. aThe dog barked out of the room. c. aBill whistled past the house. As noted by Goldberg and Jackendo¤ (2004), the ill-formedness disappears in the way construction, as illustrated in (430). The reason is that in this case the sound-emission verb does not modify the lower V (the projection of which gives rise to the directed-motion meaning), but the higher, little v (the projection of which gives rise to the causative meaning).18 In e¤ect, in the examples in (430), the verb expresses the means by which the directed motion is brought about (as shown by the approximate paraphrase provided below each example). Other comparable examples are given in (431) (which involve sound-emission verbs and an animate object) and in (432) (which involve manner verbs and the ‘‘x-way’’ object). (430) a. The car honked its way down the road. Cf. The car made its way down the road honking. b. The dog barked its way out of the room. Cf. The dog made its way out of the room barking. c. Bill whistled his way past the house. Cf. Bill made his way past the house whistling. (431) a. The police car honked me to a stop. Cf. The police car made me come to a stop by honking (at me). b. The audience whistled the actors o¤ the stage. Cf. The audience made the actors go o¤ the stage by whistling (at them).

Motion Verbs in Germanic and Romance

145

(432) a. Bill belched/danced his way out of the restaurant. Cf. Bill made his way out of the restaurant by belching/ dancing. b. Bill elbowed his way through the crowd. Cf. Bill made his way through the crowd by elbowing. Consider next the following examples, in which the lexical verbs are manner-of-motion verbs that modify the motion undergone by the object. In these cases, the lexical verb is compounded with the lower V, as shown in (434). (433) a. The trainer jumped the lion through the loop. Cf. The lion jumped through the loop. b. The general marched the soldiers to camp. Cf. The soldiers marched to camp. c. The puppeteer danced the puppet across the stage. (From Folli and Harley 2004) Cf. The puppet danced across the stage. (434) a. [vP the trainer [v [VP the lion [[ jumped V] through the loop]]]] b. [vP the general [v [VP the soldiers [[marched V] to camp]]]] c. [vP the puppeteer [v [VP the puppet [[danced V] across the stage]]]] As has often been noted, when the cause-motion construction involves an agentive manner-of-motion verb and the object is animate, an ‘accompany’ or ‘go-along’ meaning is grafted onto the cause-motion meaning in many cases; see the examples in (435). And indeed, (435c) is acceptable to speakers only to the extent that they are willing to give it the ‘accompany’ or ‘go-along’ meaning. We consider it to be a secondary, conventionalized meaning that does not follow from the structure of the sentence. Indeed, if the object is inanimate, that bit of meaning disappears; see the examples in (436). We may assume that (435) and (436) have the same grammatical analysis, namely one comparable to (425) and (427), where the lexical manner-of-motion verb is predicated of the subject. When the object is animate, the manner of motion is attributed to the object as well by pragmatics (and not by the grammar). Yet, we believe that there is a reason for this grafting of a secondary meaning onto the construction in such cases. The structure that underlies all of these examples is the same one associated with the change-of-state causatives (e.g., John broke the vase) and, as is well known, this structure is generally

146

Chapter 3

associated with a direct-causation reading. This means that the causer directly acts on the causee to bring about the resulting event (but see below for a class of verbs that allows for a weakening of this condition). In (436), in which the causee is inanimate, the direct-causation requirement is readily met—that is, the causer is interpreted as carrying the causee. On the other hand, the animate object in examples such as (435) does not readily lend itself to the ‘carry’ meaning. This failure is ‘‘repaired’’ (so to speak) by grafting onto the construction the meaning of ‘accompany’ or ‘go along’.19 (435) a. John walked the guests to the door. b. John danced Mary across the ballroom. c. The instructor swam the kids across the lake. (436) a. Mary walked the package to the o‰ce. b. Mary swam the body across the lake. c. Mary danced the letter to the post o‰ce. The direct-causation meaning associated with the structure under discussion can also readily account for the type of contrasts shown below (discussed by Folli and Harley 2004). In e¤ect, the tide can be perceived as directly causing the log to move in a rolling manner, but the wind cannot directly cause the dog to go into the house. (437) a. The tide rolled the log up the beach. b. aThe wind walked the dog into the house. While all of the examples we have examined above involve a prototypical (‘‘strong’’) causative meaning, whereby the subject brings about the directed motion by directly acting on the object, there are other situations in which this is not the case. There are cases in which the subject may be said to have more of a facilitator role than a causative role; we refer to them as cases of ‘‘weak’’ causation. Compare the examples in (438) (which involve coercion) with the examples in (439) (which do not involve coercion). This ‘‘weak’’ causative meaning is licensed only by a certain semantic class of verbs (when these are adjoined to little v and modify the matrix ‘‘causing event’’). These are verbs that express persuasion, encouragement, urging, or instruction.20 (438) a. Sam frightened Bob out of the house. b. Sam coaxed him into the room. c. Sam lured him into the room. (Goldberg 1995)

Motion Verbs in Germanic and Romance

(439) a. b. c. d. e.

147

Sam persuaded Bob into his corner. Sam encouraged Bob up the mountain. Sam instructed him over the obstacles. ?Sam convinced Bob onto the other side. The crowd cheered the soldiers across the bridge.

In the previous section, we saw that a covert directional P can be licensed in the directed-motion construction when it is governed by a lexically restricted set of verbs. This is also the case in the cause-motion construction. This is illustrated by the example in (440), due to Goldberg 1995. The PP is ambiguous between a locative and a directional meaning, but only the latter gives rise to a cause-motion meaning—that is, ‘John made the clothes go into the bag by squeezing them’. If the PP is interpreted as a locative, then we obtain an interpretation in which John is inside the bag and he is squeezing the clothes. (440) John squeezed the clothes in the bag. As Goldberg notes, if the PP in the bag is topicalized, as in (441a), only the locative meaning is available. Compare that example with (441b), in which the topicalized PP is headed by an overt directional PP and the directed-motion meaning is available. (441) a. In the bag, John squeezed the clothes. b. Into the bag, John squeezed the clothes.

(Locative) (Directional)

The above contrast suggests once more that there is a covert preposition responsible for the path meaning involved in examples like (440); see the structure in (442). To account for the licensing of the covert P in such structure, we extend the definition of ‘‘government’’ given in (413) as in (443). The lexical verb squeeze will therefore license the empty P, to the extent that it governs the empty P at PF. This is indeed the case in example (440), which has the structure in (442). In this structure, the projection of the phonologically empty P is sister to V, the projection of which is governed by ‘‘squeeze v’’ at PF. This is not the case in (441a), in which the prepositional phrase headed by the phonologically empty is preposed to sentence initial position. (442) [vP John [[v squeezed v] [VP the clothes [V [ PP P [ PP in the bag]]]]]] (443) a governs b if b or the projection of b is a sister (i) to a or (ii) to d and d is the head of a projection sister to a. We have seen that in the directed-motion construction, not all lexical verbs can license a covert directional preposition. This is also true in the

148

Chapter 3

case of the cause-motion construction under discussion, as shown by the contrast below. The examples in (444a), (445a), and (446a) contain an overt directional preposition and they all give rise to a cause-motion reading. This is also the case in (446b) despite the absence of an overt directional preposition, but it is not the case in (444b) and (445b). The verbs kick and squeeze can license a covert directional preposition, but not the verbs scare and push. Note that kick and push are both contact verbs of the same semantic type. Yet kick, but not push, can license a covert directional P. As suggested earlier, the reason might very well be extragrammatical. Indeed, it might be due to the fact that kick is more frequently used in the cause-motion construction than push. (444) a. John scared the cat out of the room. b. John scared the cat outside the room. (445) a. John pushed the ball into the hole. b. John pushed the ball in the hole.

(Directional) (Locative only)

(Directional) (Locative only)

(446) a. John kicked the ball out of the room. b. John kicked the ball outside the room. directional)

(Directional) (Locative or

To summarize this section, we proposed that the Germanic Compound Rule can be co-opted to compose ‘‘cause’’ and ‘‘directed motion’’ in cases in which the lexical verb does not imply directed motion (i.e., cases in which the verb cannot ‘‘instantiate’’ directed motion). Two cases were identified: one in which the lexical verb modifies the ‘‘causing’’ event and another in which the lexical verb modifies the ‘‘directed-motion’’ event. It was suggested that in the latter case, the lexical verb merges with the phonologically unspecified V that heads the subordinate directed-motion event, while in the former case, the lexical verb is merged with little v, which heads the matrix ‘‘causing’’ event. See structures (434) and (425)/ (427) above. In either case, the embedded V remains phonologically unspecified. Indeed, the spell-out rule in (401), repeated in (447), fails to apply at the relevant phase domain, namely vP, because it contains a phonologically specified verbal head (which is either adjoined to the embedded V or to the matrix v). (447) V is spelled out as go (or come) when it functions as the head of a ‘‘bare’’ directed-motion construction. A ‘‘bare’’ directed-motion construction is one in which the structure that encodes directed motion is contained in a phase domain with no phonologically specified verbal head.

Motion Verbs in Germanic and Romance

149

3.1.5 The ‘‘go V ’’ Construction Given the assumption that go and come (in their motion meaning) are not listed as lexical items in the lexicon, but are the spell-out of the verbal head of a ‘‘bare’’ directed-motion construction, then the question arises as to the status of go (or come) in examples like those in (448). (448) a. Go see a movie. b. Come talk to me about your paper. When followed by the ‘‘bare’’ form of a verb, go/come are morphophonologically invariant; see Carden and Pesetsky 1977 as well as Jaeggli and Hyams 1993. (449) a. b. c. d.

*John goes talk to his advisor every day. *Mary comes talk to me whenever she has a problem. *I/you came talk to him yesterday. *He went eat at that restaurant yesterday.

While they are morphophonologically invariant, the following contrast suggests that they are specified for some features. In particular, they can be specified for features that are compatible with the invariant form come and go. (450) a. *He go(es) visit the dentist every year. b. I/you/we/they go visit the dentist every year. Jaeggli and Hyams have argued that these verbs are auxiliaries (i.e., functional elements) that are generated very low in the middle field, namely, immediately above the projection of the lexical verb. Thus, they can be preceded but not followed by adverbs such as seldom and often. (451) a. I seldom/often go talk to my advisor. b. I seldom/often come discuss this issue with you. c. *I go seldom/often talk to my advisor. d. *I come seldom/often discuss this issue with you. Invariant go/come can also be preceded by elements that occupy the T(ense) position, namely, modals, dummy do, and infinitival to: (452) a. I will go see this movie. b. She must/may come talk to you about this. c. I want to go see that movie. We would like to suggest though that go/come followed by a bare V are not auxiliaries; they are comparable to the go/come that takes a PP goal

150

Chapter 3

complement—that is, they are the spell-out of the verbal head in the directed-motion construction. In the examples in (448a) and (448b), the goal arguments are the VPs see a movie and talk to me about your paper. It is precisely for this reason that these forms cannot co-occur with a PP:21 (453) a. You will go (*to the kitchen) eat your supper (*to the kitchen). b. I will go (*to the library) read a book (*to the library). Evidence in favor of analyzing the VP as a telic, goal argument of go is provided by the contrast below, cited in Jaeggli and Hyams 1993 (they attribute it to Shopen 1971). The ill-formedness of (454b) is due to the fact that the second part of the sentence negates what the first part asserts, namely, that the goal or endpoint of the motion (buy vegetables every day) is attained. (454) a.

They go to buy vegetables every day, but there never are any vegetables. b. aThey go buy vegetables every day, but there never are any vegetables.

Jaeggli and Hyams note that invariant go/come requires an agentive interpretation. This is shown by the contrast in (455), as well as by the incompatibility of invariant go/come with stative verbs, shown in (456), and the lack of ambiguity in (457b). (455) a.

Big boulders roll down this hill every time there is an earthquake. b. *Big boulders come roll down this hill every time there is an earthquake.

(456) a. *Come know the answer to this problem. b. *Go be tall. (457) a. My children bother Mary. (Stative or eventive interpretation) b. My children go bother Mary. (Eventive interpretation) The above paradigms would seem to contradict what we independently know about unaccusative structures, which predicate directed motion of the argument in its DP specifier (which thus qualifies as primarily a theme and not an agent)—for example, the big boulders rolled down the hill. We think though that the above paradigm does not challenge a uniform

Motion Verbs in Germanic and Romance

151

analysis of go/come as directed motion. It is possible that a further requirement is imposed on the V complement, namely, that it denote a volitional event (i.e., an agentive embedded subject that is bound to the matrix subject). Cardinaletti and Giusti (2001) discuss a comparable construction in Marsala (spoken in Trapani, Sicily) and in Swedish. In neither language does the verb go/come impose an agentive interpretation. We can therefore assume that the ‘‘agentivity’’ requirement on the verbal complement of go/come is a quirk of the English construction (possibly due to its diachronic evolution). For the sake of parallelism with the PP complement cases, we propose that the structure with a ‘‘bare’’ V complement is as in (458), where the P that dominates the embedded V encodes the path. Such a structure is comparable to the one in (459). In both cases, the empty P is governed by go/come, thus complying with the PF licensing requirement in (414).22 (458) [VP DPi [ go/come [P [VP ei [roll down the hill]]]]] (459) [VP Theyi [ go/come [P [ PP ei [in [the room]]]]]] Cf. When they are summoned, they go/come in the room, quick as a lightning. Note that we can account for the ill-formedness of (460a) in the same way that we account for the ill-formedness of (460b). In both cases, the empty P fails to be governed by the verb at PF. (460) a. *In the room, they go/come quick as a lightning. b. *They said they will go eat their supper, and eat their supper they will go. Cf. They said they will go eat their supper, and go eat their supper they will. We can thus maintain the analysis proposed in the previous section, whereby go/come is the spell-out of the V that heads the directed-motion construction both in (458) and in (459). In (458), there are two VPs, an embedded and a matrix VP, each of which encodes an independent l-structure. Therefore, each VP constitutes a verbal phase and defines a distinct spell-out domain. The verb in the embedded VP is not ‘‘visible’’ at the point in which the matrix VP is spelled out. Therefore, the presence of the lower phonologically specified V does not block the application of rule (447) (¼(401)) to the higher V. The analysis proposed above has nothing to say about the invariant morphophonological status of go/ come when followed by a bare V; it is at present an idiomatic property

152

Chapter 3

of the construction. Possibly an explanation can be found in the diachronic development of the construction. 3.2

Romance

As we have seen in the first part of this chapter, English and Dutch have a rich variety of directed-motion and cause-motion sentences. Building on insights from Snyder 1995, 2001, and Beck and Snyder 2001, we put forth the hypothesis that these languages co-opt Compound Rule (397) to compose ‘‘manner’’ and ‘‘directed motion,’’ on the one hand, and ‘‘cause’’ and ‘‘directed motion,’’ on the other hand. The productive use of this rule in Germanic is evidenced by its open-ended possibilities to generate novel N-N compounds. This is not the case in Romance. Indeed, (461) In Romance, compounding is lexically determined and semantically frozen. Because Romance does not make productive use of Compound Rule (397), it cannot co-opt that rule to generate the variety of resultatives and cause-result examples found in Germanic. In e¤ect, as we will see in the second part of this chapter, in Romance, cases of directed motion and cause-motion are generally restricted to verbs that ‘‘instantiate’’ such constructions due to their lexical properties. This is more clearly so in Spanish and French, where dual behavior of verbs is rare (see section 3.2.1). The case of Italian is more intriguing because it has a somewhat larger and better established subset of manner-of-motion verbs that do exhibit dual behavior (Folli 2001). We put forth the novel hypothesis that Italian has co-opted the Aux(iliary) position for ‘‘restructured’’ motion verbs in order to express directed motion with manner verbs (section 3.2.2).23 3.2.1

Manner-of-Motion Verbs in Spanish and French

3.2.1.1 Spanish The oblique preposition a in Spanish is exclusively goal-denoting and it readily appears with the light verb ir ‘go’ (which is the spell-out of the V that heads the directed-motion construction), as well as with other verbs that entail directed motion. (462) a. Juan fue´ a Parı´s. ‘Juan went to Parı´s.’ b. Juan entro´ al salon. ‘Juan went into (entered) the livingroom.’

Motion Verbs in Germanic and Romance

153

c. Juan salio´ (de la casa) a la calle. ‘Juan went out (from the house) to the street.’ d. Juan volvio´/retorno´ (de Parı´s) a LA. ‘Juan returned (from Paris) to LA.’ e. La pelota se cayo´ al piso. ‘The ball fell to the floor.’ f. Juan subio´/bajo´ al tercer piso. ‘Juan went up/went down to the third floor.’ g. Juan se acerco´/se arrimo´ a Marı´a. ‘Juan came close to Maria.’ h. El tren llego´ a la estacio´n. ‘The train arrived at the station.’ i. El tren se dirigio´ a la estacio´n. ‘The train went in the direction of the station.’ The resultative nature of the structures underlying the above sentences is indicated by the fact that they can appear in the absolutive construction, headed by the past participle form of the verb. As shown by de Miguel 1992, only unaccusative verbs with a resultative meaning can appear in this construction. Some examples are provided below. (463) a. Una vez entrada Marı´a al salon, . . . once entered Marı´a to the room, . . . ‘Maria having entered the room, . . .’ b. Una vez subida Marı´a al tercer piso, . . . once went up Marı´a to the third floor ‘Maria having gone up to the third floor, . . .’ c. Una vez llegado el tren a la estacio´n, . . . once arrived the train to the station, . . . ‘The train having arrived at the station, . . .’ As noted in Morimoto 2001, among the inherent verbs of motion, some are obligatorily telic and some are not. (464) a. *Marı´a entro´ hacia el salo´n. Marı´a entered toward the room b. *El tren llego´ hacia la estacio´n. The train arrived toward the station c. *Juan se acerco´/se arrimo´ hacia Marı´a. Juan came close toward Marı´a (465) a. Marı´a subio´ (por el sendero) hacia la Catedral. Marı´a went up (on the narrow road) toward the Cathedral

154

Chapter 3

b. El The c. El The

barco bajo´ (por el rı´o) hacia el mar. boat went down (on the river) toward the sea tren se dirigio´ hacia la estacio´n. train headed toward the station

Nevertheless, the events in (465), like the ones in (464), are bounded at the event level of representation. Events that denote change of state or location are bounded in the sense that they contain a potential telos and are therefore analyzed as composed of distinct subevents, which makes them incompatible with gradable modifiers; see (466). This shows once more that the notion of E-boundedness is independent from actual telicity and T(emporal) boundedness (see section 1.5.2).24 (466) a. *Marı´a subio´ mucho/un poco/bastante hacia la Marı´a went up a lot/a little bit/quite a bit toward the Catedral. Cathedral b. *El barco bajo´ mucho/un poco/bastante hacia The boat went down a lot/a little/quite a bit toward el mar. the sea c. *El tren se dirigio´ mucho/un poco/bastante hacia The train headed a lot/a little bit/quite a bit toward la estacio´n. the station In Standard Spanish, the use of manner-of-motion verbs with the goaldenoting preposition a is practically nonexistent (among many others, cf. Aske 1989, Morimoto 2001, and Mateu 2002, but see below for some qualifications). (467) a. *Juan nado´ al barco. ‘Juan swam to the boat.’ b. *La botella floto´ a la playa. ‘The bottle floated to the beach.’ c. *Los soldados marcharon al campamento. ‘The soldiers marched to the camp.’ d. *El bebe´ gateo´ a la puerta. ‘The baby crawled to the door.’ e. *El caballo galopo´ al establo. ‘The horse galloped to the stable.’

Motion Verbs in Germanic and Romance

155

f. *Juan balio´ a la cocina. ‘Juan danced to the kitchen.’ g. *Juan (se) tambaleo´ a la puerta. ‘Juan wobbled to the door.’ The contrast between (467) and (462) shows, on the one hand, that a complementation analysis is not available for directional PPs with manner-of-motion verbs. On the other hand, it shows that the preposition a (although unambiguously goal-denoting in Spanish) cannot function as an adjunct either. Interestingly, a may be part of a directional adjunct when embedded under a preposition that denotes the initial point of the motion: (468) a. Juan camino´ de aquı´ a la escuela. ‘Juan walked from here to school.’ b. Juan nado´ de una orilla a la otra. ‘Juan swam from one shore to the other.’ c. El bebe´ gateo´ de aquı´ a la puerta. ‘The baby crawled from here to the door.’ d. La bailarina bailo´ de un lado al otro lado del escenario. ‘The dancer danced from one side to the other side of the stage.’ The adjunct analysis of the complex prepositions in (468) is further confirmed by the fact that they can receive an activity (unbounded) interpretation: (469) a. Camino´ (He) walked b. Camino´ (He) walked c. Camino´ (He) walked

un monto´n de aquı´ a la escuela. a lot from here to school un poquito/bastante de aquı´ a la escuela. a little bit/quite a bit from here to school media hora de aquı´ a la escuela. for half an hour from here to school

The claim then is that the complex PPs in (468) are outer adjuncts and do not a¤ect the aspectual properties of the inner predicate. The interpretation of such sentences is as in (470): (470) a. b. c. d.

On his way from here to school, Juan walked. On his way from one shore to the other, Juan swam. On his way from here to the door, the baby crawled. On her way from one side to the other side of the stage, the dancer danced.

156

Chapter 3

While nothing prevents a complex preposition from functioning as a complement (e.g., Maria fue´ de aquı´ a la escuela en autobu´s ‘Maria went from here to school on a bus’), there is evidence that the complex prepositions in (468) are unambiguously adjuncts. Indeed, if they could be analyzed as complements, they should give rise to a resultative interpretation and appear in the absolutive construction. This is not the case, as illustrated below. (471) a. b. c. d.

*Una vez caminado Juan de aquı´ a la escuela, . . . *Una vez nadado Juan de una orilla a la otra, . . . *Una vez gateado el bebe´ de aquı´ a la puerta, . . . *Una vez bailada la bailarina de un lado al otro lado del escenario, . . .

We have found some speakers that are willing to accept some of the manner-of-motion verbs (such as caminar ‘walk’, correr ‘run’, nadar ‘swim’) with the bare directional a-phrase.25 Nevertheless, none of these speakers accept the corresponding resultative forms: (472) a. *Una vez nadado el nin˜o a la orilla, . . . Once swam-3rd p.s.m. the boy to the shore, . . . b. *Una vez caminada Marı´a a la escuela, . . . Once walked-3rd p.s.f. Marı´a to school, . . . This leads us to believe that the speakers that accept some of the mannerof-motion verbs with the bare preposition a (such as caminar a la escuela ‘to walk to school’) are not actually analyzing the PP as a complement of the verb. We suggest that these speakers are in fact analyzing the PP as a complex adjunct, in which the P indicating the initial point of the path is a covert deictic element—for example, caminar (de aquı´/ahı´) a la escuela ‘to walk (from here/there) to school’. The prepositions de . . . a form a discontinuous complex P; they are semantically interdependent.26 The structure of such complex PPs would be as follows. The initial location functions as the specifier and the final location functions as the complement of the complex P ‘de a’. De then moves to a higher P position from where it can Case-mark D in the lower Spec position.27 (473) [ P de [aquı´ [ P (de) a [la escuela]]]] Spanish has another directional preposition (hasta), which can either introduce a spatial complement (474a) or a (reduced) temporal adjunct (474b). Compare (475a) with (475b). The ill-formedness of (475a) is due to the fact that ir is the spell-out of the V that heads the construction

Motion Verbs in Germanic and Romance

157

that expresses directed motion along a spatial dimension and it therefore requires a physical path-denoting complement. (474) a. Juan fue´ hasta Parı´s. ‘Juan went up to Paris.’ b. Juan camino´ (hasta Paris). ‘Juan walked (up to Paris).’ (475) a. *Juan fue´ hasta llegar a Parı´s. Juan went until arrive-inf. to Paris b. Juan camino´ hasta (llegar a) Parı´s. Juan walked until (arrive-inf. to) Paris ‘Juan walked until arriving in Paris.’ As a spatial, directional preposition, hasta may also be part of a complex preposition, embedded in a PP headed by de ‘from’, which introduces the initial point of the path. As such, it may function either as a complement (476a) or as an adjunct (476b). The latter denotes an unbounded activity; compare ‘On his way from Madrid to Paris, John walked’. (476) a. Juan fue´ de Madrid hasta Parı´s en coche. Juan went from Madrid up to Paris in a car ‘Juan drove from Madrid up to Paris.’ b. Juan camino´ de Madrid hasta Parı´s. ‘Juan walked from Madrid up to Paris.’ As expected, examples such as (474b) and (476b) cannot appear in the absolutive construction: (477) *Una vez caminada Marı´a (de Madrid) hasta Paris, . . . Once walked-3rd p.f. Marı´a (from Madrid) up to Paris, . . . Furthermore, as reported by Aske 1989, examples like (478a) are unacceptable.28 If adjunct hasta is actually a covert temporal adverb, then the ill-formedness of (478a) can be related to the ill-formedness of (478b). The unbounded activity caminar hasta (llegar) a Paris is incompatible with the telic temporal phrase (en treinta dı´as). (478) a. *Juan camino´ hasta Parı´s en treinta dı´as. ‘Juan walked to Paris in thirty days.’ b. *Juan camino´ hasta llegar a Parı´s en treinta dı´as. ‘Juan walked until arriving in Paris in thirty days.’ Interestingly, for some (but not all) speakers, it is easier to get the bounded interpretation in the case of complex PPs, with both the initial

158

Chapter 3

and the endpoints of the path specified, as in (479). How can we explain these facts? Certainly, the unergative/unbounded interpretation of the predicate cannot coexist with a telic temporal adverb; see (480). (479) Juan camino´ de Parı´s a/hasta Madrid en treinta dı´as. ‘Juan walked from Madrid (up) to Paris in thirty days.’ (480) Juan camino´ un monto´n de Parı´s a/hasta Madrid (*en treinta dı´as). ‘Juan walked a lot from Paris to Madrid (*in thirty days).’ We suggest that while (480) has an unergative structure and the PP functions as an outer adjunct, (479) is in fact a covert transitive. The PP in this case is the complement of a covert distance-denoting object; see (481). Indeed, as mentioned in section 1.4.1, a subclass of the manner-ofmotion verbs select a ‘‘distance’’ classifier as an object, which may remain covert and be incorporated into the verb or may be specified and realized overtly. (481) Juan camino´ el trecho de Parı´s a/hasta Madrid en treinta dı´as. ‘Juan walked the distance from Paris to Madrid in thirty days.’ Such transitive structures are indeed resultative in nature, as shown by the fact that they can appear in the absolutive construction:29 (482) Caminado el trecho de Parı´s a/hasta Madrid, . . . Walked-3rd p.s.m. the distance from Paris (up) to Madrid, . . . When the PP is preposed, the bounded reading is hard to obtain; see (483b) and the contrast between that example and (483c). This is due to the fact that a covert ‘‘distance’’ classifier cannot be licensed in topic position. Like the covert directional Ps that we discussed in sections 3.1.2 and 3.1.3, a covert classifier requires government by the licensing verb at PF. As we will see later, the same type of data is found in French. De Parı´s a/hasta Madrid, Juan camino´ sin parar. ‘From Paris to Madrid, Juan walked without stopping.’ b. ??De Parı´s a/hasta Madrid, Juan camino´ en treinta dı´as. ‘From Paris to Madrid, Juan walked in thirty days.’ c. El trecho de Parı´s a/hasta Madrid, Juan lo camino´ (en treinta dı´as). ‘The distance from Paris to Madrid, Juan walked it (in thirty days).’

(483) a.

Motion Verbs in Germanic and Romance

159

We note that those speakers who are willing to accept some of the manner-of-motion verbs (such as caminar ‘walk’, gatear ‘crawl’, nadar ‘swim’) with the bare directional a-phrase also accept (478a). This is as expected since these speakers allow for a covert deitic source-denoting preposition. In other words, we propose that the source of the abovementioned judgment is the transitive sentence below: (484) Juan camino´ [(el trecho) (de aquı´/ahı´) hasta Paris] en treinta dı´as. ‘Juan walked (the distance) (from here/there) up to Paris in thirty days.’ Further support for the above analysis is provided by the oddness of the sentences in (485a) and (486a). Morimoto notes that bailar ‘dance’ tends to be interpreted more as an internal bodily motion (like tambalearse ‘to wobble’ and balancearse ‘to sway’) than as an external one (e.g., caminar ‘walk’ and swim ‘nadar’). Verbs of internal bodily motion are odd in combination with a ‘distance’ classifer; see (485b–c) and (486b–c). Example (485a) does not find an acceptable source form in (485c) and (486a) does not find an acceptable source form in (486c). The only other possible source forms are ones with an elliptical temporal phrase, which are also somewhat odd: Juan bailo´ hasta (llegar a) la cocina ‘John danced until arriving at the kitchen’; Juan se tambaleo´ hasta (llegar a) la puerta ‘John wobbled until arriving at the door’. Interestingly, those speakers that accept (467a–e) reject (467f–g). This is indeed what our analysis predicts, since the only possible source for (467f ) and for (467g) are the ungrammatical forms in (485c) and (486c), respectively. (485) a. (??)Juan bailo´ hasta la cocina. ‘Juan danced up to the kitchen.’ b. *Juan bailo´ 3 metros. ‘Juan danced 3 meters.’ c. *Juan bailo´ el trecho (de aquı´) a/hasta la cocina. ‘Juan danced the distance (from here) to/up to the kitchen.’ (486) a. *Juan (se) tambaleo´ hasta la puerta. ‘Juan wobbled up to the door.’ b. *Juan (se) tambaleo´ 3 metros. ‘Juan wobbled 3 meters.’ c. *Juan se tambaleo´ el trecho (de aquı´) a/hasta la cocina. ‘Juan wobbled the distance (from here) to/up to the kitchen.’

160

Chapter 3

It was argued above that manner-of-motion verbs in Spanish are by and large unergatives, and as such, they unambiguously denote an activity. Nevertheless, there are a few verbs that express manner-of-motion and have unaccusative properties, such as rodar ‘to roll’ and deslizarse ‘to slide’. Indeed, the verbal structure headed by these verbs can have a resultative meaning and can therefore appear in the absolutive construction: (487) a. El barril rodo´ al pie´ de la colina/debajo de la mesa (en un segundo). ‘The barrel rolled to the foot of the hill/under the table (in a second).’ b. Rodado el barril al pie´ de la colina/debajo de Rolled-3rd p.s.f. the barrel to the foot of the hill/under la mesa, . . . the table, . . . ‘The ball having rolled to the foot of the hill/under the table, . . .’ (488) a. La moneda se deslizo´ dentro del agujero (en un segundo). ‘The coin slid inside of the hole (in a second).’ b. Una vez deslizada la moneda dentro del agujero, . . . Once slid-3rd p.s.f. the coin inside the hole, . . . ‘The coin having slid inside the hole, . . .’ Further evidence of their unaccusative nature comes from the fact that they have a lexical causative counterpart:30 (489) a. Pedro deslizo´ la moneda dentro del agujero. ‘Pedro slid the coin in the hole.’ b. ?Pedro rodo´ el barril al pie´ de la colina. ‘Pedro rolled the barrel to the foot of the hill.’ What distinguishes rodar and deslizar from other manner-of-motion verbs is that they readily lend themselves to a nonvolitional (nonagentive) interpretation. Possibly this is the factor that allows them to function as unaccusatives. This is also true in French (see below). In some Spanish idiolects, the verb saltar ‘jump’ seems to have a polysemous status: it may function either as an unergative, activity-denoting verb (490) or as an unaccusative verb that expresses directed motion (491a). Although the unaccusative form lacks a causative counterpart (like the achievement arrive), it seems that in this idiolect of Spanish, saltar can enter the absolutive construction, as shown in (491b).31

Motion Verbs in Germanic and Romance

161

(490) Juan salto´ de alegrı´a. Juna jumped from happiness (491) a. (*)El perro salto´ a la piscina en un segundo. The dog jumped into the pool in a second b. (*)Una vez saltado el perro a la piscina, . . . Once jumped the dog into the pool, . . . 3.2.1.2 French We now turn briefly to French manner-of-motion verbs. These have been discussed by Lamiroy (1983), who also points out the di¤erence between inherent verbs of motion and manner-of-motion verbs.32 The former but not the latter readily take a goal-denoting PP complement.33 Thus, the examples in (492) are considered marginal by this author and impossible by other native speakers.34 Compare with the intrinsic directed-motion verbs in (493).35 (492) a. ?/*Jean nage/rame a` la plage. ‘Jean swims/rows to the beach.’ Cf. Jean va a` la plage a` la nage/a` la rame. ‘Jean goes to the boat swimming/rowing.’ b. ?/*Jean galope a` la maison. ‘Jean gallops to the house.’ Cf. Jean va a` la maison au galop. ‘Jean goes to the house galloping.’ c. ?/*Jean valse a` la cuisine. ‘Jean waltzes to the kitchen.’ Cf. Jean va a` la cuisine en valsant. ‘Jean goes to the kitchen waltzing.’ d. ?/*Jean marche au magasin. ‘Jean walks to the shop.’ Cf. Jean va au magasin en marchant. ‘Jean goes to the shop walking.’ (493) a. Jean descende/monte au troisie`me e´tage. ‘Jean goes down/goes up to the third floor.’ b. Jean arrive a` la gare. ‘Jean arrives at the station.’ c. Jean se pre´cipite a` la gare. ‘Jean rushes to the station.’ d. Jean s’avance au bord du pre´cipice. ‘Jean moves to the edge of the abyss.’

162

Chapter 3

e. Jean recule au fond de la chambre. ‘Jean moves to the back of the room.’ f. Jean se retire au couvent. ‘Jean retires to the convent.’ The fact that manner-of-motion verbs do not exhibit the type of variable behavior that English manner-of-motion verbs do is further shown by the unambiguous interpretation of the locative clitic y when it co-occurs with manner-of-motion verbs. In the examples below, the clitic y unambiguously denotes the location of the activity denoted by the verb. (494) a. Jean Jean b. Jean Jean c. Jean Jean d. Jean Jean e. Jean Jean

y loc. y loc. y loc. y loc. y loc.

a nage´/rame´. Cl. swam/rowed a galope´. Cl. galloped a valse´. Cl. waltzed a marche´. Cl. walked a couru. Cl. ran

The verb sauˆter ‘jump’ is an exception. This verb appears to function ambiguously, either as an unaccusative (with a directed-motion meaning) or as an unergative (with an activity meaning); see (495a) and (495b). Thus, the clitic y, when it co-occurs with sauˆter as in (495c), can be interpreted as the location of the jumping activity or as the goal of the motion.36 (495) a. Pierre a sauˆte´ dans la picine. ‘Pierre jumps in(to) the pool.’ b. Pierre sauˆte dans la cours. ‘Pierre is jumping around in the yard.’ c. Pierre y a sauˆte´. (Ambiguous) Pierre loc. Cl. jumped Like Morimoto for Spanish, Boons (1987) has shown that in French we must distinguish manner-of-motion verbs that are compatible with a distance-denoting complement (as indicated by the possible co-occurrence with a measure phrase) from those that are not. As we have seen earlier, ramer, marcher, galoper, nager do not denote movement toward an endpoint, as shown by their incompatibility with a complement that denotes

Motion Verbs in Germanic and Romance

163

the endpoint of the path; see (492). Nevertheless, they can be combined with a distance-denoting complement. (As in the case of Spanish, we analyze these as hidden transitive structures, comparable to run a race.) (496) a. Pierre a rame´ 500 me`tres. ‘Pierre rowed (a distance of ) 500 meters.’ b. Pierre a marche´ 500 me`tres. ‘Pierre walked (a distance of ) 500 meters.’ c. Pierre a nage´ 500 me`tres. ‘Pierre swam (a distance of ) 500 meters.’ d. Pierre a galope´ (a distance of ) 500 me`tres. ‘Pierre galloped (a distance of ) 500 meters.’ On the other hand, verbs like vadrouiller ‘to rove’ and errer ‘to wander’ are not compatible with a measure complement, although they are compatible with a measure adjunct; see (497) and (498). This is also the case with manner-of-motion verbs like rouler ‘to sway’, tanguer ‘to swing’, and chanceler ‘to wobble’. Note that the type of meaning involved in (496) is very di¤erent from the type of meaning involved in (498). In the former cases, the entire distance of 500 meters was covered by rowing, walking, swimming, or galloping. In the latter cases, the measure phrase provides the spatial frame in which the activity of roving/wandering/swinging occurred. It does not imply that the entire distance was covered in that manner. (The same is true of their English counterparts.) (497) a. *Pierre a vadrouille´ 500 me`tres. *Pierre roved 500 meters. b. *Pierre a erre´ 500 me`tres. *Pierre wandered 500 meters. c. *Le bateau a tangue´ 500 me`tres. *The boat swung 500 meters. (498) a. Pierre a vadrouille´ sur (une distance de) 500 me`tres. ‘Pierre roved over a distance of 500 meters.’ b. Pierre a erre´ sur une distance de 500 me`tres. ‘Pierre wandered over a distance of 500 meters.’ c. Le bateau a tangue´ sur (une distance de) 500 me`tres. ‘The boat swung over a distance of 500 meters.’ Interestingly, Boons notes that predicates headed by the above-mentioned verbs in combination with jusqu’a` (the French counterpart of Spanish hasta) cannot be modified with the telic temporal phrase. See the French

164

Chapter 3

examples in (499) and compare them with their English counterparts, which are perfectly acceptable. (499) a. Pierre a vadruoille´/erre´ jusqu’a` l’e´tang (*en vingt minutes). ‘Pierre roved up to the pond (in twenty minutes).’ (Boons 1987) b. Le bateau a tangue´ jusqu’au port (*en vingt minutes). ‘The boat swung up to the dock (in twenty minutes).’ c. L’enfant a chancele´ jusqu’a` la porte (*en trois minutes). ‘The kid wobbled up to the door (in three minutes).’ We turn next to the verbs in (496), which are compatible with a distance-denoting complement, as shown by the fact that they can take a measure phrase complement. Native speakers find (500a) with a telictemporal phrase either marginal or unacceptable, in contrast with (500b).37 Note that, as in Spanish, the preposition a` can be embedded under a source-denoting preposition and the resulting PP denotes a path. Compare (501a) with (501b). (500) a. Pierre a marche´ jusqu’a` la maison (*en trente minutes). ‘Pierre walked up to the house (in thirty minutes).’ b. Pierre a marche´ du parque jusqu’a` la maison (en trente minutes). ‘Pierre walked from the park up to the park (in thirty minutes).’ (501) a. ?/*Pierre a marche´ a` la maison (en trente minutes). ‘Pierre walked to the house (in thirty minutes).’ b. Pierre a marche´ du parque a` la maison (en trente minutes). ‘Pierre walked from the park to the house (in thirty minutes).’ The counterpart of (500b) with a preposed PP gives rise to ungrammaticality. Thus, there is no contrast between the two examples in (502); they are equally ill-formed. Compare these with the forms in (503) and (504), which have an unbounded, manner-of-motion interpretation, typical of unergative structures. (502) a. *Jusqu’a` la maison, Pierre a marche´ en trente minutes. Cf. (500a) ‘Up to the house, Pierre walked in thirty minutes.’ b. *Du parque a`/jusqu’a` la maison, Pierre a marche´ en trente minutes. Cf. (500b)

Motion Verbs in Germanic and Romance

165

‘From the park (up) to the house, Pierre walked in thirty minutes.’ (503) a. Jusqu’a` la maison, Pierre a marche´ (au lieu de courir). ‘Up to the house, Pierre walked (rather than run).’ b. Du parque a`/jusqu’a` la maison, Pierre a marche´ (au lieu de courir). ‘From the park (up) to the house, Pierre walked (rather than run).’ (504) a. Jusqu’a` la maison, Pierre a beaucoup marche´. Up to the house, Pierre has a lot walked b. Du parque a`/jusqu’a` la maison, Pierre a beaucoup From the park (up) to the house, Pierre a lot marche´. walked As we did earlier for Spanish, we suggest that (500b) and (501b) are hidden transitives. More specifically, the complex PP (de . . . a`/jusqu’a`) can be analyzed as specifying a direct object with a covert classifier head with the meaning of ‘distance’ or ‘interval’. But unlike Spanish and English, the classifier head cannot be overtly realized in French with most manner-of-motion verbs. (We have no explanation for this crosslinguistic di¤erence.)38 (505) a. Pierre a marche´ (*la distance/le trajet) d’ici a`/jusqu’a` Pierre walked (the distance/the interval) from here to/up to la maison en trente minutes. the house in thirty minutes b. Pierre a galope´/trotte´ (*la distance/le trajet) d’ici Pierre galloped/trotted (the distance/the interval) from here a`/jusqu’a` la maison en trente minutes. to/up to the house in thirty minutes c. Pierre a nage´ (*la distance/le trajet) d’ici jusqu’au Pierre swam (the distance/the interval) from here to/up to the bateau en trente minutes. boat in thirty minutes We noted above that the path cannot be preposed under a bounded interpretation of the event; see (501b) and (502b) above, as well as (506a) and (506b) below. Once more, it is tempting to relate the ill-formedness of

166

Chapter 3

(502b) and (506b) to the presence of an empty direct object null classifier, as we proposed for Spanish in the previous subsection. As pointed out there, this covert classifier needs to be governed by the verb at PF in order to be licensed, a situation reminiscent of the null directional P in English (see section 3.1.1). In (506b), the preposed null object is not governed by the verb at PF. The above conjecture is supported by the contrast between (507a) and (507b). Unlike the manner-of-motion verbs in (505), the verb courir ‘run’ does allow for an overt ‘‘distance’’ classifier and, as expected, preposing is possible when the classifier is overt. Pierre a couru d’ici a`/jusqu’a` la maison en trente Pierre ran from here to/up to the house in thirty minutes. minutes b. *D’ici a`/jusqu’a` la maison, Pierre a couru en trente From here to/up to the house, Pierre ran in thirty minutes. minutes

(506) a.

(507) a. Pierre a couru la distance/le trajet d’ici a`/jusqu’a` la Pierre ran the distance from here to/up to the maison en trente minutes. house in thirty minutes b. Le trajet d’ici a`/jusqu’a` la maison, Pierre l’a couru The distance from here to/up to the house, Pierre it-ran en trente minutes. in thirty minutes Note that, due to locality conditions on extraction, the PP complement to the distance-denoting classifier cannot be extracted out of the DP, stranding the classifier head; see (508). This indicates that such an analysis is also impossible for the counterpart with a covert classifer head; see (506b). (508) *D’ici a`/jusqu’a` la maison, Pierre a couru le trajet en trente minutes. Finally, note that, like their Spanish counterparts, the nonvolitonal/ nonagentive French verbs rouler ‘to roll’ and glisser ‘to slide’ can readily take a PP directional complement. Given their nonagentive meaning, it is likely that the structure associated with such examples is indeed an unaccusative one.

Motion Verbs in Germanic and Romance

167

(509) a. La balle a roule´ dans le trou en deux seconds. The ball rolled in the hole in two seconds ‘The ball rolled into the hole in two seconds.’ b. La pie`ce de monnaie a glisse´ dans le trou en un second. The coin slid in the hole in a second ‘The coin slid into the hole in a second.’ (510) a. Dans le trou, la balle a roule´ en deux seconds. ‘Into the hole, the ball rolled in two seconds.’ b. Dans le trou, la pie`ce de monnaie a glisse´ en un second. ‘Into the hole, the coin slid in a second.’ 3.2.1.3 Summary We recapitulate briefly. We have examined above the behavior of the manner-of-motion verbs in Spanish and French. It was noted that except for some rare cases, Spanish and French mannerof-motion intransitive verbs do not give rise to a directed-motion reading, in particular in combination with the dative preposition. As for the cause-motion structure in (417), repeated in (511), it does exist in Romance, but it requires a verb that, by virtue of its lexical properties, can instantiate it. Some Spanish examples are given in (512).39 (511) [vP D [v [VP D [V V X dir ]]]] (512) a. Marı´a envio´ cartas a Parı´s. ‘Maria sent letters to Paris.’ b. Marı´a recibio´ cartas de Paris. ‘Maria received letters from Paris.’ c. Marı´a llevo´ cartas a Parı´s. ‘Maria took letters to Paris.’ d. Marı´a subio´/bajo´ la comida al tercer piso. ‘Maria took up/down the food to the third floor.’ e. Marı´a arrimo´/acerco´ la silla a la mesa. ‘Maria pulled up the chair to the table.’ f. Marı´a boto´ los papeles al basurero. ‘Maria threw the papers into the wastebasket.’ Romance does not have the rich variety of resultative examples that Germanic has. In particular, it lacks those forms that we have analyzed as involving modification of the cause event. The goal-denoting PP in (513) gives rise to ungrammaticality. The PPs in (514) are fine, but they can only give rise to a locative reading. Unlike their English counterparts, they cannot be associated with a directional meaning.40

168

(513) a. Marı´a Maria b. Marı´a Maria c. Marı´a Maria

Chapter 3

empujo´ la carreta (*al granero). pushed the wagon (to the barn) estiro´ la carreta (*al granero). pulled the wagon (to the barn) pateo´ la pelota (*al jardı´n). kicked the ball (to the garden)

(514) a. Marı´a empujo´/estiro´ la carreta en el granero. ‘Maria pushed/pulled the wagon inside the barn.’ b. Marı´a pateo´ la pelota afuera de la casa. ‘Maria kicked the ball outside the house.’ c. Marı´a asusto´ al gato afuera de la pieza. ‘Maria scared the cat out of the house.’ Recall that Germanic makes use of the Compound Rule (397) to generate a rich variety of directed-motion constructions. We have suggested that, while in Germanic the Compound Rule cannot generate verbal compounds that contain two phonologically specified Vs for reasons discussed in section 1.6.2, it can generate verbal compounds in which one of the Vs is phonologically unspecified, the intepretation of which derives from the construction itself. The latter option is not available in Romance because in Romance the Compound Rule is restricted to lexically specified items and its output is semantically frozen. Yet, as we will see in the next section, in Italian there is a semiproductive usage of manner-of-motion verbs in directed-motion constructions. While the phenomenon is by no means a general one as it is in Germanic, the semiproductive nature of the phenomenon in Italian needs to be accounted for. 3.2.2 Manner-of-Motion Verbs in Italian Folli (2001) identified three classes of motion verbs in Italian: (1) a class of verbs that imply directed motion and that are unambiguously unaccusative (i.e., they obligatorily take the essere auxiliary in the past perfect form and trigger agreeement between the participial verb and the subject), (2) a class of manner-of-motion verbs that denote an activity and are unambiguously unergative (i.e., they obligatorily take the avere auxiliary in the past perfect form and do not trigger participial agreement), and (3) a class of manner-of-motion verbs that can appear either in an unergative or in an unaccusative structure.41 (515) a. entrare (enter), uscire (exit), arrivare (arrive), atterrare (land), partire (depart), tornare (return), scappare (escape)

Motion Verbs in Germanic and Romance

169

b. galleggiare (float), camminare (walk), galoppare (gallop), danzare (dance), nuotare (swim), sciare (ski), passeggiare (walk around), vagabondare (wander) c. correre (run), rotolare (roll), rimbalzare (bounce), scivolare (glide, slide), gattonare (crawl), saltare ( jump), volare (fly), saltellare (hop) The examples in (516)–(517) illustrate class 2, and those in (518)–(519) illustrate class 3. (516) Maria ha camminato fino a casa. Maria has walked-3rd p.s.masc. fino a house ‘Maria has walked up to the house.’ (517) *Maria e` camminata ( fino) a casa. Maria has walked-3rd p.s.fem ( fino) a house ‘Maria has walked (up) to the house.’ (518) Maria e` corsa ( fino) a casa. Maria is run-3rd p.s.fem. ( fino) a house ‘Gianni has run (up) to the house.’ (519) a. Maria ha corso a Boston, alla maratona. Maria has run-3rd p.s.masc. in Boston, in the marathon ‘Maria ran in Boston, in the marathon.’ b. Maria ha corso fino a casa. Maria has run-3rd p.s.masc. fino a house ‘Maria has run up to the house.’ As in Dutch, the directed-motion construction with an unaccusative structure is bounded, while the directed-motion construction with an unergative structure is unbounded. (520) a. Maria e` corsa nel bosco (in un secondo). Maria is run-3rd p.s.fem. in the woods (in one second) ‘Maria ran into the woods (in one second).’ b. Maria e` corsa nel bosco (*per un’ora). Maria is run-3rd p.s.fem. in the woods (for an hour) ‘Maria ran into the woods (for an hour).’ (521) a. Maria ha corso nel bosco (per un’ora). Maria has run-3rd p.s.masc. in the woods (for an hour) ‘Maria ran in the woods (for an hour).’

170

Chapter 3

b. Maria ha corso nel bosco (*in un secondo). Maria has run-3rd p.s.masc. in the woods (in a second) ‘Maria ran in the woods (in a second).’ 3.2.2.1 Folli’s Analysis of Italian vs. English Folli’s analysis is based, on the one hand, on an event-based VP syntax (developed in Ramchand 2001, Butt and Ramchand 2001, and Folli and Ramchand 2001) and, on the other hand, on the semantic analysis of events put forth by Higginbotham 2000. The event-based VP syntax of Ramchand and colleagues is a refined version of Hale and Keyser’s l-syntax. One important di¤erence between the two is that the former postulates an abstract aspectual category termed Result v (Rv), which is present in accomplishment-denoting structures but not in activity-denoting structures. The intransitive and transitive accomplishments have the structures in (522) and (523), respectively. It is the telic nature of the path (represented as XP) that determines the presence of the abstract aspectual category RvP.42 The intransitive and transitive activities have the structures in (524) and (525), respectively. (522) [VP DP [V [ RvP DP [Rv XP]]]] (523) [vP DP [v [VP DP [V [ RvP DP [Rv XP]]]]]] (524) [VP DP [V . . . ]] (525) [vP DP [v [VP DP [V . . . ]]]] Following Higginbotham 2000, Folli assumes that predicates are specified with the Davidsonian e(vent)-variable, which represents the contribution of the predicate with respect to the temporal interpretation of the situation. Accomplishment predicates and directional PPs carry two e-variables, while activity predicates and locative PPs carry a single e-variable. Folli distinguishes two types of compositional analysis for generating the interpretation of directed motion: accomplishment adjunction and accomplishment creation. Following Higginbotham 2000, accomplishment adjunction and accomplishment creation are defined in terms of the operations of e-identification and e-projection, respectively. Folli claims that English uses only the operation of accomplishment adjunction to form a directional meaning with manner-of-motion verbs (which are activity-denoting verbs). To illustrate, we give below two of Folli’s examples: one with a directional preposition and one with a loca-

Motion Verbs in Germanic and Romance

171

tive preposition. In (526), the verb’s e-variable (e1) is identified with the first e-variable of the preposition into (represented as e1 ¼ e2). This e-variable (e1 ¼ e2), as well as the preposition’s second e-variable (e3), is projected onto the VP, giving rise to a directed-motion interpretation. In the resulting interpretation, e1(¼ e2) represents the initial point and e3 the endpoint of the motion. In (527), on the other hand, the verb’s e-variable is identified with the sole e-variable of the locative preposition along (e1 ¼ e2), giving rise to a unique temporal-point interpretation. (526) The boat floated into the cave.

(527) The boat floated along the canal.

While English is claimed to have recourse only to ‘‘accomplishment adjunction’’ to generate a complex eventive interpretation, Italian is claimed to use both ‘‘accomplishment adjunction’’ (which generates an unergative structure) and ‘‘accomplishment creation’’ (which generates an unaccusative structure). It is furthermore claimed that the di¤erence between English and Italian is related to a di¤erence between the preposi-

172

Chapter 3

tional structure of the two languages. According to Folli, in English, most prepositions are either purely directional, with a complex-eventive structure (e.g., to, into, onto), or ambiguous between directional or locative (e.g., under, over, below, behind, and so on), irrespective of their morphological complexity. On the other hand, in Italian, only morphologically complex prepositions have a complex-eventive structure—for example, fino a (until at), fino dietro (until behind), fino sopra (until over), fino sotto (until under). Monomorphemic prepositions, such as a (at), dietro (behind), sopra (over), sotto (under), in (in), dentro (inside), fuori (outside), are locatives with a simple-eventive structure, although as we will see below, in the right syntactic context, they can give rise to a directional meaning as well.43 Folli suggests that prepositions with a complex-eventive structure are like accomplishment predicates: they are associated with a structure that contains a Result Phrase (RvP). An example is given below. The first e-variable is provided by fino, while the second e-variable is provided by the Rv node, which is lexicalized by the locative preposition a (via movement). This is assumed to be crucial because Italian, unlike English, does not allow an Rv node to remain phonologically empty. (528)

Unlike Italian, in English an Rv node may remain empty, as is the case with the preposition to.

Motion Verbs in Germanic and Romance

173

(529)

Consequently, English can systematically form complex-eventive VP structures via accomplishment adjunction, with morphologically complex prepositions (526) or morphologically simple prepositions (527). Folli proposes that Italian can similarly form complex eventive VP structures via accomplishment adjunction, but only with morphologically complex prepositions. The sentences in (530), with a morphologically complex preposition, are thus assigned an analysis comparable to (526). (530) a. Maria ha corso fino a casa. Maria has run-3rd p.s.masc. fino a house ‘Maria has run to the house.’ b. Maria ha camminato fino a casa. Maria has walked-3rd p.s.masc. fino a house ‘Maria has walked to the house.’ Italian cannot form complex-eventive VP structures via accomplishment adjunction with morphologically simple prepositions because, as mentioned earlier, Italian does not allow an empty Rv node. As an adjunct, the morphologically simple preposition a is unambiguously interpreted as a locative: (531) Maria ha corso a Boston, alla maratona. Maria has run-3rd p.s.masc. in Boston, in the marathon ‘Maria ran in Boston, in the marathon.’ Italian therefore uses another operation, namely accomplishment creation, to form complex-eventive VP structures with morphologically simple prepositions. This is exemplified below. The verb provides the first e-variable (e1) and the Rv node provides the second e-variable. The latter e-identifies the e-variable introduced by the locative preposition a, namely

174

Chapter 3

e3. Therefore, e2 ¼ e3. Given that Italian does not allow an empty Rv, the locative preposition a must move to Rv to lexicalize it. Furthermore, as we will see shortly, the accomplishment-creation operation is restricted to a certain class of manner-of-motion verbs.44 (532) Maria e` corsa a casa. Maria is run-3rd p.s.fem. a house ‘Maria has run to the house.’

As mentioned earlier, Folli distinguishes three distinct classes of motion verbs in Italian, cited in (515) and repeated in (533). The verbs in (533a) denote directed motion by virtue of their lexical meaning; they obligatorily take a PP directional complement and are unambiguously associated with an unaccusative structure. The verbs in (533b) are activity-denoting verbs; they cannot take a directional PP complement and are unambiguously associated with an unergative structure. The third class of verbs given in (533c) are assumed to be polysemous. These can behave like the inherent directional verbs in (533a) (with an unaccusative structure) or like the activity-denoting verbs in (533b) (with an unergative structure). (533) a. entrare (enter), uscire (exit), arrivare (arrive), atterrare (land), partire (depart), tornare (return), scappare (escape) b. galleggiare (float), camminare (walk), galoppare (gallop), danzare (dance), nuotare (swim), sciare (ski), passeggiare (walk around), vagabondare (wander)

Motion Verbs in Germanic and Romance

175

c. correre (run), rotolare (roll), rimbalzare (bounce), scivolare (glide, slide), gattonare (crawl), saltare ( jump), volare (fly), saltellare (hop) According to Folli’s analysis, the manner-of-motion verbs in (533c), but not those in (533b), can create complex-eventive VPs via accomplishment creation—that is, they can head the structure in (532). Only these manner-of-motion verbs are lexically specified for an Rv complement. Indeed, given the point-locating nature of Italian a, the Rv node is essential in providing the other e-variable necessary to compose directed motion. Yet note that morphologically complex prepositions (which do encode a path meaning by virtue of the fact that they introduce an Rv node) are subject to the same lexical restrictions; see (534). Under Folli’s analysis, it is unclear why this should be so. Maria e` corsa fino a casa. Maria is run-3rd p.s.fem. fino a house ‘Maria has run to the house.’ b. *Maria e` camminata-3rd p.s.fem. fino a casa. Maria has walked fino a house ‘Maria has walked to the house.’

(534) a.

As we have seen above, in the analysis under discussion, the directional PPs that co-occur with a manner-of-motion verb in English are unambiguously considered to be adjuncts. In section 3.1.1, we have seen evidence that this is not the case. The complementation analysis is indeed available in both English and Dutch for directional PPs when in construal with manner-of-motion verbs. Furthermore, Folli’s analysis claims that the adjunct analysis, like the complement analysis, can generate an accomplishment event structure; see (526). We have already seen that the adjunct analysis in Germanic does not generate an accomplishment reading. The observation that an unergative remains aspectually unbounded even in the presence of an adjunct path-denoting PP is shown by the Dutch facts in (380)–(383), repeated below. Note in particular the contrasts in (537) and (538). The adjunct structures cannot be modified by a telic temporal adverb, which clearly shows that such structures are not bounded events (i.e., they are not accomplishments). (535) a. dat Jan naar Groningen gewandeld is. that Jan to Groningen walked is ‘Jan walked to Groningen.’

176

Chapter 3

b. dat Jan naar Groningen gewandeld heeft. that Jan to Groningen walked has ‘On his way to Groningen, Jan walked.’ (536) a. dat Jan naar de overkant gezwommen is. that Jan to the other-side swum is ‘Jan swam to the other side.’ b. dat Jan naar de overkant gezwommen heeft. ‘On his way to the other side, Jan swum has.’ (537) a.

dat Jan in twee uur naar Groningen gewandeld is. that Jan in two hour to Groningen walked is b. *dat Jan in twee uur naar Groningen gewandeld heeft.

(538) a.

dat Jan in twee uur naar Groningen is gewandeld. that Jan in two hours to Groningen is walked b. dat Jan naar Groningen twee uur lang heeft gewandeld. that Jan to Groningen two hours long has walked c. ??dat Jan twee uur lang naar Groningen heeft gewandeld.

We have also argued extensively that the adjunct analysis does not generate an accomplishment reading in French and Spanish, and there is evidence that this is the case in Italian, too. The sentences below, in which the PP headed by fino a is an adjunct, cannot be modified by the temporal in-phrase. This shows that the path in (539) is implied, but not grammaticality represented. We may attribute this to the claim that, as an adjunct, Italian fino a (like Spanish hasta and French jusqu’a`) is a reduced temporal phrase, as shown in (540). In these cases, the event denoted by the VP is clearly an activity. (539) a. Gianni ha corso fino a casa (*in un’ora). ‘Gianni has run to the house (in an hour).’ b. Gianni ha camminato fino alla spiaggia (*in un’ora). ‘Gianni has walked to the beach (in an hour).’ (540) a. Gianni ha corso (per un’ora) fino (ad arrivare) a casa. ‘Gianni has run (for an hour) until (arriving) at home.’ b. Gianni ha camminato (per un’ora) fino (ad arrivare) alla spiaggia. ‘Gianni has walked (for an hour) until (arriving) at the beach.’ The facts in (539), as well as the concomitant facts in Spanish and French discussed in the previous section and repeated in (541) and (542), are very

Motion Verbs in Germanic and Romance

177

important because they show that modifiers outside the VP cannot a¤ect the eventive properties of the VP. (541) a. Camino´/corrio´ un monto´n de aquı´ a la escuela. ‘He walked/ran a lot from here to school.’ b. Camino´/corrio´ un poquito/bastante de aquı´ a la escuela. ‘He walked/ran a little bit/quite a bit from here to school.’ c. Camino´/corrio´ media hora de aquı´ a la escuela. ‘He walked/ran half an hour from here to school.’ Cf. (469) (542) a. Il a beaucoup marche´ d’ici a` l’e´cole/jusqu’a` la e´cole. ‘He walked/ran a lot from here to school.’ b. Il a assez/un petit peu marche´ d’ici a`’l’e´cole/jusqu’a` l’e´cole. ‘He walked/ran a little bit/quite a bit from here to school.’ c. Il a marche´ pendant une demi heure d’ici a` l’e´cole/jusqu’a` l’e´cole. ‘He walked/ran half an hour from here to school.’ While one might attempt to argue that the auxiliary avere in Italian, hebben in Dutch, and to some extent avoir in French, are responsible for the imperfective nature of the verb, this cannot be argued to be the case in Spanish (since Spanish unergatives and unaccusatives both take the auxiliary haber). Furthermore, the above facts also hold in the future form of the verb, in Spanish as well as in French and Italian. (543) a. Caminara´ un poquito (de aquı´) hasta la escuela. (He) will walk a little bit (from here) up to the school b. Il marchera un petit peu (d’ici) jusqu’a` la e´cole. (He) will walk a little bit (from here) up to the school In (544), the temporal phrase gives rise to an ingressive meaning only (i.e., it specifies the time that it took for the event to get started); it does not measure the temporal span of the path. Under the telic interpretation, the sentences are ungrammatical. (But see section 3.2.1.1 for some qualifications regarding idiolectal variations.) (544) a. Caminara´ hasta la escuela (*en treinta minutos). b. Il marchera jusqu’a` l’e´cole (*en trente minutes). c. Camminera’ fino per la scuola (*en trenta minuti). ‘He will walk up to the school (in thirty minutes).’

178

Chapter 3

We conclude that the complementation analysis gives rise only to a bounded interpretation and the adjunct analysis gives rise only to an unbounded interpretation across languages. Nevertheless, there is another insight in Folli’s analysis that deserves further scrutiny, namely the status of the preposition to in English versus its Romance counterpart a.45 While the English to-phrase (as well as its Dutch counterpart) can function as an adjunct, the PP headed by the Romance preposition a cannot do so. Although Spanish a can never in and of itself function as a locative (unlike its Italian and French counterparts), it is possible that the Romance a is unambiguously a point-locating preposition and that in order to denote the endpoint of a path, it needs to be embedded under a P that denotes the initial point of a path. Yet, as we will see in section 3.3, this di¤erence between Romance and Germanic is not su‰cient to account for the di¤erences between the two sets of languages regarding the composition of manner and directed motion. 3.2.2.2 An Alternative Analysis of Italian: Recruiting the Auxiliary Position in ‘‘Restructuring Structures’’ to Compose Manner and Directed Motion The challenge that remains is to account for the subset of verbs that have a variable behavior in Italian—that is, those in (533c). (We note though that there are some idiolectal variations as to which verbs belong to this class.) The verbs correre ‘to run’, volare ‘to fly’, saltare ‘to jump’, saltellare ‘to hop’, and gattonare ‘to crawl’ are typically activity-denoting verbs, and as such they select the auxiliary avere ‘have’. (545) a. Maria ha corso nel parco. ‘Maria has run in the park.’ b. Maria ha saltato per due ore. ‘Maria has jumped for two hours.’ c. L’aereo ha volato per due ore. ‘The plane has flown for two hours.’ d. La rana ha saltellato per la stanza. ‘The frog has hopped around the room.’ e. La palla ha rimbalzato per dieci secondi. ‘The ball has bounced for ten seconds.’ f. Il bambino ha gattonato per la stanza. ‘The child has crawled around the room.’ As we saw earlier, when these verbs are combined with a goal-denoting PP, the resulting forms have the characteristics of unaccusatives—that is,

Motion Verbs in Germanic and Romance

179

they select the auxiliary essere ‘be’ to form the past perfect and they trigger participial agreement. (546) a. Maria e` corsa a casa. Maria is run-3rd p.s.fem. home b. Maria e` saltata nella piscina. Maria is jumped-3rd p.s.fem. into the pool c. Maria e` volata a Parigi. Maria is flown-3rd p.s.fem. to Paris d. La rana e` saltellata nella trappola. The frog is hopped-3rd p.s.fem. into the trap Another characteristic of unaccusatives is that they can appear in the absolutive construction (as we have seen in section 3.2.1.1 for Spanish). Compare the unaccusative arrivare with the unergative telefonare: (547) a. Maria e` arrivata a casa. Maria is arrived-3rd p.s.fem. home b. i. (Una volta) arrivata Maria a casa, . . . ii. (Una volta) arrivata a casa Maria, . . . (Once) arrived-3rd p.s.fem. home Maria, . . . (548) a. Maria ha telefonato a casa. Maria has telephoned home b. i. *(Una volta) telefonata Maria a casa, . . . ii. *(Una volta) telefonata a casa Maria, . . . (Once) phoned-3rd p.s.fem. home Maria, . . . The manner-of-motion verbs in (546) can also appear in the absolutive construction, although there is apparently a slight contrast between VSPP and VPPS word order. With these verbs, the VSPP order in the absolutive construction is less than perfect. Apparently, the presence of the adverb una volta ‘once’ facilitates the availability of the absolutive forms with these verbs. See (549)–(552) below. As mentioned earlier, for Folli, gattonare also belongs to the ambiguous class (directed-motion or activitydenoting). For our informants, the directed-motion reading is di‰cult to obtain with gattonare; see (553).46 This idiolectal variation (annotated in parentheses) could be due to frequency of usage. (549) a. ?Una volta corsa Maria a casa, . . . b. Una volta corsa a casa Maria, . . . Once ran-3rd p.s.fem. to the house Maria, . . .

180

Chapter 3

(550) a. ?Una volta saltata Maria nella piscina, . . . b. Una volta saltata nella piscina Maria, . . . Once jumped-3rd p.s.fem. into the pool Maria, . . . (551) a. ?Una volta volata Maria a Parigi, . . . b. Una volta volata a Parigi Maria, . . . Once flown-3rd p.s.fem. to Paris Maria, . . . (552) a. ?Una volta saltellata la rana nella trappola, . . . b. Una volta saltellata nella trappola la rana, . . . Once hopped-3rd p.s.fem. into the trap the frog, . . . (553) a. (?*)Il bambino e` gattonato sotto il tavolo. The child is crawled-3rd p.s.masc. under the table b. ?(*)Una volta gattonato il bambino sotto il tavolo, . . . c. (?*)Una volta gattonato sotto il tavolo il Once crawled-3rd p.s.masc. under the table the bambino, . . . child, . . . The manner-of-motion verbs discussed above are agentive in character; they require some volitionality on the part of the subject. These verbs contrast with manner-of-motion verbs scivolare ‘to slide’, rotolare ‘to roll’, and rimbalzare ‘to bounce’, which are not intrinsically agentive. The adverb una volta ‘once’ is not needed to facilitate the absolutive construction with these verbs and the VSPP order is completely natural; see (546a)–(552a) and (554b)–(556b). Perhaps these verbs are genuinely lexically ambiguous as proposed by Folli, although it deserves to be mentioned that some speakers prefer the unaccusative use of rotolare and scivolare. (Also, recall that the Spanish and French counterparts of these two verbs can denote directed motion; see section 3.2.1.) (554) a. La palla e` rotolata sotto il tavolo. The ball is rolled under the table b. Rotolata la palla sotto il tavolo, . . . c. Rotolata sotto il tavolo la palla, . . . Rolled under the table the ball, . . . (555) a. La moneta e` scivolata nel buco. The coin has slid in the hole b. Scivolata la moneta nel buco, . . . Scivolata nel buco la moneta, . . . Slid into the hole the coin, . . .

Motion Verbs in Germanic and Romance

181

(556) a. La palla e` rimbalzata addirittura sotto quel tavolo. The ball has bounced even under that table b. Rimbalzata la palla sotto quel tavolo, . . . Rimbalzata sotto quel tavolo la palla, . . . Bounced under that table the ball, . . . Following the compositional approach developed in the present work, we suggest that while rotolare, scivolare, and possibly rimbalzare, actually instantiate a directed-motion construction by virtue of their lexical meaning, the agentive verbs correre, volare, saltare, saltellare, and gattonare do not. These are unambiguously activity-denoting verbs. We therefore need an alternative account of the sentences in (549)–(553). We put forth the following hypothesis: (557) Italian recruits the auxiliary position designated for the motion ‘‘restructuring’’ verbs in order to compose ‘‘directed motion’’ and ‘‘manner’’ (in some lexically restricted cases). As is well known, Italian and Spanish (but not Modern French) have a class of modals (e.g., potere ‘can’, dovere ‘must’, volere ‘want’), aspectual verbs (e.g., cominciare ‘begin’, continuare ‘continue’, solere ‘used to’), and motion verbs (e.g., venire ‘come’, andare ‘go’, tornare ‘return’) that trigger a phenomenon known as restructuring. We will refer to these as the R-verbs (or VR ). Both in Italian and Spanish, these verbs give rise to clitic climbing (i.e., placement of the clitic on the R-verb instead of the lexical verb) and long object preposing in middle constructions. But even more remarkably, in Italian (but not in Spanish), these verbs trigger an ‘‘auxiliary switch,’’ as illustrated by the examples below (from Rizzi 1978). In a restructuring context (such as (558b)), the lowest lexical verb (venire) determines the choice of auxiliary. (558) a.

Maria ha dovuto venirci molte volte. Maria has modal come þ loc.Cl many times ‘Maria has had to come there many times.’ b. Maria c’e` dovuta venire molte volte. Maria loc.Cl þ is modal come many times c. *?Maria ci ha dovuto venire molte volte. Maria loc.Cl þ has modal come many times

We suggest that the choice of auxiliary is what makes such a recruitment process robust and gives a clear signal to the learner that a position

182

Chapter 3

designated for the motion R-verbs has been extended to a particular subset of manner-of-motion verbs. Cinque (2004) argues extensively for the functional or auxiliary status of R-verbs. (See also Strozer 1976; Picallo 1985, 1990; Rochette 1988, 1990.) Cinque develops a very fine-grained hierarchy of functional projections for the clause and locates di¤erent semantic subclasses of R-verbs within such a hierarchy. We will not dwell on that issue here, but will assume at least three main VR positions (some of which are recursive): VMod (for modals), VAsp (for the aspectuals), and VMT (for motion verbs). Hopefully, the relative order (or scope) of these three classes of R-verbs will follow from their semantics. (559) [VMod [VAsp [VMT [ . . . Vlex . . . ]]]] An example that illustrates the presence of all three types of R-verbs is given below (from Rizzi 1978): (560) Maria li avrebbe potuti Maria acc.Cl would-have-been able (Mod) stare per andare a prender lei stessa. to be on the point (Asp) of going (MT) to get herself ‘Maria would have been able to be on the point of going to get them herself.’ Returning to the unergative correre-class of verbs, these cannot license the presence of the auxiliary essere, nor can they license the presence of a goal-denoting argument. The presence of these elements indicates the presence of a directed-motion structure headed by an empty V. Where then is correre located in the structure? We suggest that correre (as well as other agentive manner verbs in that class) recruits the VMT position in (559). The structure of (546a), prior to undergoing Merge with T, will be as shown in (561). In that structure, correre modifies the directed-motion structure headed by an empty V. Furthermore, we make the assumption that the verbal head of the directed-motion construction remains phonologically unspecified in such a structure because in Italian, the verbal phase is extended to the VMT projection. Recall that go as well as its Italian counterpart andare are the spell-out of the verbal head of the directed-motion construction. This spell-out rule fails to apply when there is another verbal head within the phase that contains the directed-motion construction; see (447).

Motion Verbs in Germanic and Romance

183

(561)

Is there some evidence for the above analysis? More precisely, is there evidence that correre and other verbs in that class have the same syntactic status as the motion R-verbs when used in the directed-motion construction? The phenomenon of final -e deletion discussed by Cardinaletti and Shlonsky 2004 (C&S) seems to provide some evidence for the proposed analysis. C&S persuasively argue that there are two types of R-verbs: functional R-verbs (namely, the modals) and semifunctional R-verbs (aspectual, verbs of motion, and causatives). The latter have some properties of lexical verbs and some properties of functional verbs. One of the properties that distinguish functional R-verbs from purely lexical verbs is the phenomenon of final -e deletion in the infinitival form. C&S argue that the final -e in lexical verbs in the infinitive form (leggere) is the morphological reflex of a functional category F associated specifically with infinitives. Furthermore, they suggest that the enclitic that appears on the infinitive is associated with that F category. Consequently, when an enclitic is attached to the infinitive, the final -e is dropped; compare legger*(e) versus leggerli. See (562) below. In other words, the final -e and the enclitic are in complementary distribution because they are both associated with the same functional position. C&S argue that such final -e deletion is not to be confused with the phonological rule known as ‘‘troncamento’’ (which deletes a vowel before a consonant in certain prosodic domains). (562) a. Li ho voluti legger*(e) a Maria. (I) them-have wanted read to Maria

184

Chapter 3

‘I wanted to read them to Maria.’ (C&S, example (63)) b. Ho voluti leggerli a Maria. In the case of functional R-verbs (such as the modals volere, potere, or dovere), final -e deletion is obligatory; see (563). This follows from the fact that such verbs are not lexical and therefore lack an extended functional projection. (563) Lo vorrei poter(*e) fare andare a prendere a Maria. (I) it-would want be-able make go to fetch to Maria ‘I would like to be able to make Maria go fetch it.’ (C&S, example (57c)) On the other hand, the semilexical R-verbs (i.e., motion, aspectual, and causative verbs) optionally drop the final -e; see (564a–b). This is due to the fact that these verbs are optionally associated with the functional category F. Furthermore, a semilexical infinitive cannot drop its final -e if it has another semilexical infinitival to its left that has not undergone -e deletion; see (564c). This presumably follows from the fact that all verbs without an extended F category are merged above verbs that have an extended F category. (564) a.

Li ho voluti fare andare a prendere a (I) them-have wanted make go to fetch to Maria. ¼ C&S (64) Maria ‘I wanted to make Maria go fetch them.’ b. Li ho voluti far andar(e) a prendere a Maria. c. *Li ho voluti fare andar a prendere a Maria.

If the analysis we proposed for correre in (561) is correct, it predicts that this class of verbs should behave like the motion R-verb andare, namely, it should trigger optional final -e deletion in the same contexts that andare does. Strikingly, according to our informants, the prediction is borne out; see (565)–(568). Note in particular the contrasts between the second and third forms in the latter paradigm, which are comparable to the contrast between the second and third forms in (564). Recall that for our informants gattonare does not belong to the correre-class and, as expected, this verb fails to undergo final -e deletion; see (569).

Motion Verbs in Germanic and Romance

(565) a.

Li ho voluti fare correre a prendere a Maria. (I) them-have wanted make run to fetch to Maria ‘I have wanted to make Maria run fetch them.’ b. Li ho voluti far correr(e) a prendere a Maria. c. *Li ho voluti fare correr a prendere a Maria. d. Li ho fatti correr(e) a casa. (I) them-have made run home ‘I made them run home.’

(566) a.

Li ho voluti fare volare a prendere a Maria. (I) them-have wanted make fly to fetch to Maria ‘I have wanted to make Maria fly fetch them.’ b. Li ho voluti far volar(e) a prendere a Maria. c. *Li ho voluti fare volar a prendere a Maria. d. Li ho fatti volar(e) a casa. (I) them-have made fly home ‘I made them fly home.’

(567) a.

Li ho voluti fare saltare in braccio a te. (I) them-have wanted make jump onto your lap ‘I have wanted to make them jump onto your lap.’ b. Li ho voluti far saltar(e) in braccio a te. c. ?*Li ho voluti fare saltar in braccio a te. d. Li ho fatti saltar(e) in braccio a te. (I) them-have made jump onto your lap ‘I have made them jump onto your lap.’

(568) a.

L’ho lasciato saltellar(e) in una trappola. (I) them-have let hop into a trap ‘I have let them hop into a trap.’ b. Li ho voluti far saltellar(e) in una trappola. (I) them-have wanted make hop into a trap ‘I have wanted to make them hop into a trap.’ c. ?*Li ho voluti fare saltellar in una trappola.

(569) a. *?Il bambino e` gattonato anche sotto il tavolo. The child is crawled even under the table b. ?*L’ho lasciato gattonar anche sotto il tavolo. (I) them-have let crawl even under the table c. L’ho lasciato gattonare anche sotto il tavolo. ‘I have let them crawl even under the table.’

185

186

Chapter 3

More to the point, arrivare, being unambiguously a lexical verb, does not allow final -e deletion, any more than telefonare does; see (570)– (571). On the other hand, the possibility of final -e deletion with mannerof-motion verbs should vary with the interpretation. More precisely, we predict that in a ‘‘restructuring context’’ in which a manner-of-motion verb is associated with a directed-motion reading, final -e deletion should be possible, while in a ‘‘restructuring context’’ in which a manner-ofmotion verb is associated with an activity reading, final -e deletion should not be possible. According to our informants, the predicted contrasts, although subtle, do exist; see (572)–(575).47 (570) Gianni e` voluto arrivar*(e) a casa tardi. Gianni is wanted arrive home late ‘Gianni has wanted to arrive home late.’ (571) Gianni ha potuto telefonar*(e) a casa. Gianni has been-able phone home ‘Gianni has been able to phone home.’ (572) a. Gianni ha potuto correr*(e) a Boston, alla Gianni has been-able run in Boston, in the maratona. (Locational) marathon ‘Gianni has been able to run in Boston, in the marathon.’ b. Gianni e` potuto correr(e) a Boston. (Directional) Gianni is been-able run to Boston ‘Gianni has been able to run to Boston.’ (573) a. L’aereo ha potuto volar*(e) in una stanza per the toy-airplane has been-able fly in a room for dieci minuti, poi sono finite le pile. (Locational) ten minutes, then are finished the batteries ‘The toy airplane was able to fly in a room for ten minutes, then the batteries went dead.’ b. L’aereo e` potuto volar(e) in una the toy-airplane is been-able fly into a stanza. (Directional) room ‘The toy airplane was able to fly into a room.’ (574) a. Il bambino ha voluto saltar*(e) in una stanza per dieci The child has wanted jump in a room for ten

Motion Verbs in Germanic and Romance

187

minuti. (Locational) minutes ‘The child has wanted to jump in a room for ten minutes.’ b. Il bambino ti e` voluto saltar(e) in The child dat.cl. is wanted jump in braccio. (Directional) lap. ‘The child has wanted to jump onto your lap.’ (575) a. La rana ha potuto saltellar*(e) in una stanza per dieci The frog has been-able hop in a room for ten minuti. (Locational) minutes ‘The frog has been-able to hop in a room for ten minutes.’ b. La rana e` potuta saltellar(e) in una The frog is been-able hop into a vettura. (Directional) car ‘The frog has been able to hop into a car.’ Note that the contrast illustrated in (572)–(575) cannot be explained solely on the basis of the adjunct versus complement status of the PP following the targeted verb. Indeed, in (562), in (570), as well as in the example below, the PP is a complement. Nevertheless, final -e deletion is not possible. This clearly shows that, as argued by C&S, the functional status of the verb is a crucial factor in the licensing of final -e deletion. (576) Li ho voluti dare*(e) a Maria. (I) them-have wanted give to Maria ‘I wanted to give them to Maria.’ We therefore conclude that the verbs in the correre-class have the grammatical status of a functional (or semifunctional) category in their directed-motion use, as shown in (561). 3.3

Summary and Some Final Remarks

We have seen in this chapter that in Germanic any manner-of-motion verb (which is typically unergative) can be associated with an unaccusative structure with a bounded interpretation. The phenomenon is systematic and exceptionless. We have argued that those unergative verbs are

188

Chapter 3

unambiguously activity-denoting and that their unbounded property cannot be a¤ected by a path-denoting PP that is external to the VP. Although some manner-of-motion verbs in Romance (which are also typically unergative) can appear in an unaccusative (bounded) structure, the phenomenon is rare, more so in French and Spanish than in Italian. It has often been remarked that the dative preposition in Romance is a point-locating preposition (i.e., nondynamic), unlike the dative preposition in English (cf., e.g., Bergh 1948; Folli 2001).48 The question arises as to whether this di¤erence between the two sets of languages is su‰cient to account for the above-mentioned generalizations. If the dative preposition in Romance is indeed point-locating, then we have an account as to why it cannot give rise to a goal reading in an adjunct position. We may hypothesize that in order to give rise to a goal reading, the dative (point-locating) preposition must either combine with a verb of directed motion (ir . . . a in (577a)) or with a source-denoting preposition (de . . . a in (577b)). As we have seen in section 3.2.1.1, this complex preposition can function as the specifier of an abstract path classifier that denotes ‘distance’ or ‘interval’. This ‘‘distance’’ classifier is selected by verbs like bailar ‘dance’, nadar ‘swim’, caminar ‘walk’, and so on, but not by verbs like vagar ‘wander’/‘rove’ and tambalearse ‘sway’/ ‘wobble’. The ill-formedness of (577c) can then be attributed to the fact that vagar ‘wander’ is incompatible with any kind of path, whether it be an abstract distance-denoting complement (such as a mile) or a concrete, physical path (such as to the lake).49 Juan fue´ a Paris. Juan went point-loc. prep. Paris b. Juan bailo´ (el trecho) de aquı´ a la Juan danced (the distance) from here point loc. prep. the cocina. kitchen c. *Juan vago´ (el trecho) de aquı´ al lago. Juan wandered (the distance) from here point. loc. prep. lake

(577) a.

Korean locative -ey could be analyzed like the Romance locative a—that is, as an unambiguous point-locating preposition. On the other hand, in English, the preposition to can be analyzed as a directional P; therefore, it can give rise to a directed path reading in either complement or adjunct position. But this cannot be the end of the story.

Motion Verbs in Germanic and Romance

189

As we have seen in section 3.1, the wander-class of verbs in Germanic (although incompatible with a distance-denoting complement) can cooccur with a physical path-denoting complement; see (578) and (579). Whereas in English, the complement status of the directional P cannot be easily appreciated, we can observe it clearly in Dutch; see (580) and (581). The PP headed by naar in (581) cannot appear to the right of the verb, which shows that it has the status of a complement. Compare with the PP headed by richting in (580). (578) a. John walked/galloped/swam a mile. b. *John wobbled/roved/wandered a mile. (579) a. The boy roved/wandered (up)to the lake in twenty minutes. b. The kid wobbled to the door in three seconds. (580) a.

Jan is richting het meer gezworven. John is toward the lake wandered/roved b. *Jan is gezworven richting het meer.

(581) a.

Jan is naar de deur gewaggeld. John is to the door wobbled b. *Jan is gewaggeld naar de deur.

To conclude, although the point-locating status of the dative preposition in Romance can account for the fact that this preposition cannot give rise to a path reading in adjunct position, it appears that this factor cannot fully account for the di¤erences between Germanic and Romance. There must be some other property that distinguishes Romance and Germanic, namely, property P. In this chapter, we have endorsed the view that property P is the availability of productive, compositional compounding in the language (see Snyder 1995, 2001; Beck and Snyder 2001). More precisely, (582) In Germanic, compounding is productive and compositional, while in Romance compounding is lexically determined and semantically frozen (i.e., idiomatic). Because the Compound Rule is lexically unrestricted in Germanic, it can make use of it to compose manner and directed motion. Under such an analysis, the directional PP in the English and Dutch examples in (580) and (581) is not a complement of the lexical verb at all; it is a complement of an empty light verb. The lexical (manner-of-motion) verb, which is

190

Chapter 3

merged with the empty light verb (via the Compound Rule), is a modifier of the directed-motion structure headed by the empty V. The same analysis extends to a variety of cause-directed-motion examples in Germanic, in which the lexical verbs involved are not verbs of motion at all. Finally, we have seen that in Italian there are some genuine unergative (agentive) verbs that can appear in the directed-motion construction. We argued that Italian recruits the position for verbs of motion in ‘‘restructuring’’ structures (namely, a functional position) to compose manner and directed motion in some lexically restricted cases.

Notes

Chapter 1 1. For an excellent overview of di¤erent theories of the lexicon-syntax interface, see Rosen 1999. 2. Situations according to their temporal properties (known as lexical aspect) are generally classified into four main types (following Vendler 1957): a. States. Situations that involve no change throughout an interval (love, broken, angry). b. Activities. Situations that involve a process with transitions from one state to another, but lack a natural endpoint (run, laugh, drink juice, push a cart). c. Accomplishments. Situations that involve a process with transitions from one state to another toward a natural endpoint (break a glass, eat an apple, draw a circle, run a mile, run to the park). d. Achievements. Situations that lack a process; they are instantaneous transitions to a natural endpoint (reach the top, notice a problem). Alternatively, achievements can be analyzed as processes that involve a transition to an endpoint state from an immediately preceding state. Thus, achievements and accomplishments both have an endpoint, but achievements (unlike accomplishments) lack internal substates. 3. Presumably, Template Augmentation is not entirely free. For example, activity-denoting verbs rarely causativize in English. See chapter 3 in L&R 1995 and section 1.4.2 in this chapter for further discussion. 4. The underlined variable y in (16) and (18) is a constant participant, recoverable from the constant SWEEP, and not an event participant. 5. See also Van Hout 1996, which assumes that telicity is a feature on AgrO. The object in the Spec of this functional projection is interpreted as delimiter. 6. It is unclear how verbs such as go, arrive, depart, appear can be accommodated within Borer’s system, which has no place for a grammatical lexicon. Such verbs (even if small in number) suggest that not all verbs are born equal regarding their eventive properties. An approach that eliminates the lexicon from the grammar cannot accommodate the lexical contribution to the composition of phrasal syntax and to the aspectual meaning of the sentence.

192

Notes

7. Longobardi (2000) has argued that bare plurals in Romance can have a generic interpretation when they occupy a nonargument position. 8. This observation is also found in Bergh 1948. 9. Larson (1988) was the first to revive a decompositional analysis for causatives, but in his system, the verb actually selects the external argument, which is licensed in the Spec of the higher V. Note that ‘‘little v’’ is purely mnemonic; it has no intrinsic content. It is the verbal head that merges with a verbal l-structure and licenses an external argument. 10. As we will see in chapter 2, the Korean counterpart of English go/come is not categorially restricted because in this language the categorial distinction between A and V is neutralized. 11. There is another issue, namely, what is the relation between the extragrammatical conceptual properties associated with a lexical item and the grammatical l-structure that it instantiates? We will not be concerned with this important question here. 12. More precisely, F&P argue that the reason why movement out of a phasal category should occur through the edge of a phase is precisely because the linearization of words is computed at the phasal level, a property that follows immediately if indeed the phase constitutes the domain of spell-out. Ko (2005) provides further arguments (based on Korean scrambling) in favor of F&P’s view. We adopt here the general view put forth by F&P, namely, that the phase itself (rather than the complement of the head of the phase) constitutes the domain of spell-out. On the other hand, we do not adopt their definition of verbal phases. F&P propose that VP and not little vP counts as a phase (at least in some languages). We assume that the highest verbal projection in the (extended) l-structure functions as a phase. Thus, if little vP is present, that node (and not the VP embedded under it) counts as a phase. Thus, we also di¤er from Ko (2005), who assumes that both little vP and VP are phases. 13. The rule in (59) is subject to some semantic constraints. Consider the contrast the water became frozen versus *the glass became shattered. The di¤erence seems to be that the category-neutral root /shatter/ is lexically marked as [
durative]. Within the system sketched here, this means that if the category-neutral root inserted under A in (55b) is specified as [
durative], V cannot be spelled out as become (i.e., rule (59) is blocked). 14. It is well known that not all verbs of change of location actually transitivize (e.g., arrive, appear), but this may very well be an idiosyncratic property of these verbs (see Chierchia 2004), although curiously the verb arrive lacks a transitive form in many languages. The above conclusion receives some support from data in second-language acquisition reported in Cabrera and Zubizarreta 2003, 2005, and Cabrera 2005. This research shows that native English speakers learning Spanish and native Spanish speakers learning English overgeneralize causatives with unaccusatives (such as arrive/llegar and appear/aparecer) and not with unergatives (such as work/trabajar and laugh/reir), or they overgeneralize significantly more with unaccusatives than with unergatives.

Notes

193

C. Schutze (personal communication) has brought to our attention the example in (i), reportedly pronounced by someone in the control tower of Boston’s Logan Airport. But note the contrast between (i) and (ii)–(iii). This suggests that arrive in (i) has acquired a meaning close to land. (i)

We will arrive three planes every two.

(ii) *We arrived the children late. (iii) *We arrived the mail late. It could be that achievement verbs that denote instantaneous change of location do not lend themselves generally to direct causativization. Perhaps due to extragrammatical factors they are more appropriate in contexts of indirect causativization—for example, We made the children arrive late. Furthermore, it could be that the L2 learners that overgeneralize with arrive ignore the direct versus indirect distinction in the causativization process. 15. Note that if stative predicates are specified for the categorial features A and P and if little v does not merge with A and P (i.e., little v only merges with a V category), then it follows that stative predicates cannot be immediately embedded under a causative-denoting verbal projection. In other words, John fattened the cow would be loosely comparable to John caused the cow to become fat (and not to John caused the cow to be fat), and John put the book on the table would be loosely comparable to John caused the book to come to be on the table (and not to John caused the book to be on the table): (i) [John [v [the cow [V [(the cow) [a [fat]]]]]]] (ii) [John [v [the book [V [P [(the book) [on [the table]]]]]]]] Note that John sent the package to Paris does not entail that the package is in Paris; indeed, it can be followed by the phrase but the package never got to Paris. On the other hand, John put the book on the table does entail that the book is on the table. Our analysis captures the latter fact by establishing a direct predication relation between the DP the book and the PP on the table; see (ii) above. A similar relation between the DP the package and the PP to Paris is lacking in John sent the package to Paris; see (64). 16. Many authors have argued that causation is a relation between events rather than an argument-taking predicate, among them Dowty (1979), Wunderlich (1997), and McIntyre (2004). 17. There are a few apparent counterexamples in English to the generalization that unergative verbs do not undergo lexical causative formation. (i)

to work the students

(ii)

to burp the baby

(iii) to bleed the patient (iv) to sweat the cheese (v)

to run the machine

(vi) to jump the car (vii) to walk the dog

194

Notes

Levin and Rappaport (1995) argue that these are idiomatic cases. Examples (v)– (vii) are clearly figurative in meaning. Also, some of these examples, like (iv), are not acceptable to all speakers. Furthermore, verbs of bodily secretion (such as bleed and sweat) seem to oscillate between unergative and unaccusative behavior in some languages; see Sorace 2000, 2004. 18. Inner Aspect (related to the eventive or Aktionsart properties of the VP) is not to be confused with outer or grammatical Aspect (i.e., the perfect or the progressive). 19. In some languages (such as Korean), the relation between T and Asp and its associated Nom or Acc DP is licensed via AGREE (rather than via a Spec-head relation obtained via movement); see chapter 2. 20. Recall that directed motion can take place along a physical path (change of location) or along an abstract path or scale (change of state). 21. For discussion of serial-verb constructions in a variety of languages, see Bamgbose 1974; Sebba 1987; Baker 1989, 1991; Lefebvre 1991; Awoyale 1987, 1988; Manfredi 1991; De´chaine 1993; Collins 1997; Nishiyama 1998; also see the references cited in these sources. In addition, see Manfredi 2005 and Manfredi, forthcoming, for a di¤erent view from the one put forth by Baker and Stewart 1999. 22. Mateu 2002 postulates a GT comparable to (97a) to account for the composition of manner and directed motion in Germanic. We do not think that this is correct because Germanic languages are not of the SVC type. See section 1.6.4 for further discussion. 23. With Baker and Stewart, we assume that the operation Attract (which triggers movement of V to T) is blind to little v (which is phonologically unspecified). 24. If it is assumed that arguments are licensed outside the (extended) l-structure, one could argue that the head-initial versus head-final status of the VP is irrelevant in a language like Korean. What is relevant is the position of the inflectional categories—that is, Tense, Asp (in the Spec of which arguments are licensed), and Comp. Indeed, given the above assumption, it would be su‰cient that the inflectional categories be specified as head final to obtain the desired verb-final status of Korean. On the other hand, if arguments in a language like Korean do not move to the specifier of the licensing functional projection, but instead are licensed by virtue of entertaining an agreement relation with the head of the relevant functional projection via the operation Agree (cf. Chomsky 1999/2001), then the head-initial versus head-final status of the VP is crucial; see Ko 2005. In the present work, we adopt the latter view. Furthermore, we assume that V moves to T overtly in Korean to check its tense feature. 25. Baker and Stewart point out that the second object is like an E-type pronoun—that is, a definite description that picks up the referent introduced by ` zo´ bought few books and he read the few books that the first event; consider ‘O he bought’. Ideally, the interpretation of the second object will follow from the se-

Notes

195

mantic relation that the two events entertain in this type of SVC. Because the event denoted by the first V is a necessary condition for the event in the second V to take place, it should be the case that the worlds defined by the first event restrict the worlds defined by the second event. In the example under discussion, the worlds defined by the quantified event ‘‘few book-buying’’ restricts the worlds defined as events of ‘‘book-reading.’’ In section 2.1.2, we will see a Korean CSVC that involves a distributive marker on the first V that gives rise to an E-type interpretation of the object of V2. 26. Saramaccan RSVCs provide further evidence for the complex V analysis of this type of SVC. Veenstra 2004 reports that, in the Saramaccan RSVC, a phonological sandhi rule applies between V1 and V2, a rule that otherwise requires phonological adjacency between the two Vs. The analysis proposed by Baker and Stewart 1999 can capture this fact, assuming that the sandhi rule applies before deletion of the lower copy of V1. 27. V. Manfredi has pointed out to us that Ijaw, an unusual Niger-Congo language in that it is head final, also lacks an RSVC, as predicted. 28. The other interpretation is the consequential interpretation: ‘John swam and consequently ended up at the other side of the lake’. That interpretation is associated with the CSVC structure discussed earlier. 29. John ran/walked/swam in an hour can only be awkwardly associated with the meaning ‘It took an hour for John to begin running/walking/swimming’. The awkwardness comes from the fact that in the simple past, the Reference Time must coincide with the Event Time. See section 1.5. 30. Jackendo¤ 1992 argues against a causative interpretation of the wayconstruction and in favor of a simple directed-motion construction. But see Mateu 2005 and chapter 3 in favor of assigning a causative meaning to sentences like (123). 31. That (123) and (125) are not exact paraphrases of (122) and (124), respectively, is shown by the fact that the latter are more restricted than the former; see chapter 3. The same holds for the Korean counterparts to be discussed in chapter 2. 32. We note that N-N compounds in Korean, although productive, only have a fixed (idiomatic) meaning. Thus, the Korean counterpart of (132b) can only mean ‘a man that looks like a frog’. 33. There are some forms that appear to be counterexamples to that generalization, namely, make believe (e.g., John made believe that he was Santa Claus) and let go/fall/drop (e.g., John lets go of the rope), but these are rare and more likely generated by a low-level restructuring rule rather than via the Compound Rule. Such an analysis is supported by the fact that Germanic compounds are morphosyntactically head final and that is not the case in the above examples, where the first V (rather than the second V) is inflected. Other cases, such as stir-fried, slamdunked, drop-kicked, appear to be cases of backformation; compare to do a stirfry, to do a slam dunk, to do a dropkick. Note that these cases of backformation

196

Notes

actually follow the typical Germanic word order for compounds—that is, the rightmost constituent is the morphosyntactic head. Cases like helped run (as in John helped run the store) and helped solve (as in John helped solve the problem) should probably be analyzed as involving a verbal small-clause complement with a PRO embedded subject (as in the above-mentioned examples) or an overt embedded subject (John helped us run the store, John helped us solve the problem). We are grateful to Tom Roeper for providing many of these examples. 34. More precisely, the semantic rule that generates complex predicates can apply to discontinuous heads. For example, it can apply to a manner-of-motion verb and a directional preposition. 35. McIntyre (2004) has independently proposed similar structures, but making use of semantic-based categories, rather than syntactic categories. See section 1.6.4 for further discussion. 36. As mentioned in note 12, we assume Fox and Pesetsky’s version of spell-out domain. Within that view the phase itself constitutes the domain of spell-out, rather than the sister of the head of the phase (as originally proposed by Chomsky 1999/2001). 37. There are dialectal variations. We have found that some speakers do not like the use of gattonare ‘crawl’ to express directed motion; others do not like saltellare ‘hop’ in that usage either. 38. Note that the fact that VAux can be analyzed as the specifier of the verbal structure shows that it can be simultaneously analyzed as a head and as a phrase, a situation allowed by the theory of bare-phrase structure in which the distinction between the two is purely contextual. 39. The auxiliary switch is blocked when the preposition di ‘from’ intervenes between the R-verb and the lexical verb. See finire di V; tratare di V. 40. See Kayne 1993 for a proposal that attempts to relate participial agreement and choice of auxiliary. 41. The example below (from Rizzi 1978) shows that Auxpp can appear in a lower position (namely, between two R-verbs): (i) Maria ci potrebbe esser dovuta tornare. Maria loc.Cl. might be had to go back ‘Maria might have to go back there.’ 42. The VMod is clearly recursive, as shown by the example in the previous note. The habitual auxiliary solere/soler is often considered an aspectual verb. But we tend to think that it is a modal, and it can indeed appear above another modal verb: (i) Maria la suele querer visitar. Maria acc.Cl. usually wants to visit ‘Maria usually wants to visit her.’ 43. Also, Mateu assumes that the argument John is the subject of the PP complement, rather than of the eventive verb. Again, we ignore those minor details here.

Notes

197

Chapter 2 1. We put aside the case of covert coordination (briefly discussed in chapter 1), which we consider to be a creature of a di¤erent type. 2. The abbreviations used in the glosses are as follows: Nom: nominative case, Acc: accusative case, Gen: genitive case, L: linker, Neg: negation, Past: past tense, Pres: present tense, NL: nominalizer, Decl: declarative marker, Comp: complementizer, Conj: conjunction, Sub: subordination, Loc: locative, Top: topic marker, Contr: contrastive topic marker, Pl: plural marker, Hon: honorific marker, Caus: causative, Cl: classifier, Prog: progressive, Rel: relativizer, Resul: resultative marker, Ben: benefactive. 3. In accordance with vowel harmony, the choice between -e and -a is conditioned by the preceding vowel. Given the observation that the su‰x -e does not carry any semantic and structural content, Lee (1992) argues that the su‰x -e in the SVC is just a dummy linker. In the present work, we endorse her claim on the status of -e. In any case, it is highly unlikely that this su‰x is a conjunction marker since Korean has an independent conjunction that can be inserted between two verbs, giving rise to a di¤erent semantics. In this respect, we can treat V1 -e as equivalent to a bare verb in English. 4. Whitman (2000) argues that under the antisymmetric account, the preverbal negation is in Spec of NegP (like French pas), while the postverbal negation is in the head position of the NegP (like French ne). Although we maintain that part of his analysis of negation (with some alterations), we assume (contra Whitman) that Korean is head final. As is well known, the postverbal negation an is the head of NegP, as shown in (i). The morpheme -ci is a nominalizer, to which the V is adjoined. The dummy verb ha- is introduced under T to license the tense feature. (i) . . . [TP [NegP [[VP . . . V] ci ] an]] ha] The locus of preverbal negation an is Spec of Neg, as shown in (ii). The complements undergo scrambling and end up to the left of an. On the other hand, V moves to T (across empty Neg), as shown in (iii). Note that the intervening phonlogically empty Neg head does not block head movement; see Travis 1984 on the Head Movement Constraint. This suggests that head movement is a PF operation, since only intervening heads with phonological content block head movement. (ii) . . . [NegP an [[VP . . . V] Neg] T] (iii) . . . [NegP an [[VP . . . Vi ] Neg] Vi þT] 5. Note that Korean allows double negation, where one negates the propositional content and the other negates a constituent contained there. (i) John-i pap-ul an mek-cianh-ass-ta. John-Nom meal-Acc Neg eat-Neg-Past-Decl ‘It is not the case that John did not eat meal’ (Ahn 1991, 237). Similarly, (165b) is acceptable with the interpretation ‘It is not the case that John didn’t broil and eat the fish’, but it is unacceptable with the meaning indicated in the text.

198

Notes

6. Note though that NPI licensing across clausal boundaries in the context of propositional attitude verbs is definitely better than in the context of other types of verbs. See the contrast below: (i) ?John-un [amwuto kukos-ey ka-ss-ta-ko] mit-cianh-nun-ta. John-Top anyone there-Loc go-Past-Decl-Comp believe-Neg-Pres-Decl ‘John doesn’t believe that anyone went there.’ (ii) *John-un [amwuto kukos-ey ka-ss-ta-ko] malha-cianh-nun-ta. John-Top anyone there-at go-Past-Decl-Comp say-Neg-Pres-Decl ‘John doesn’t say that anyone went there.’ Cf. John-un [nwukwunka-ka kukos-ey ka-ss-ta-ko] John-Top someone-Nom there-at go-Past-Decl-Comp malha-cianh-ass-ta. say-Neg-Past-Decl ‘John didn’t say that someone went there.’ 7. NPI licensing also requires that the NPI be c-commanded by Neg at LF (postreconstruction of scrambled elements); Ahn 1991. Note that (176d) not only violates the clause-boundedness condition, but also the c-command requirement. 8. Stewart 1996 also discusses the role of the pause in the Covert Coordination construction in Edo. 9. In an earlier version of this book, we analyzed Korean SVCs as compounds. As pointed out by an anonymous reviewer, the arguments are not compelling. The impossibility of inserting adverbs between two serial verbs in Korean follows directly from the head-final status of the language. The impossibility of scrambling or topicalizing one of the two Vs (while stranding the other V) is probably due to some poorly understood constraints on SVCs. See the paradigm below. As pointed out by the above-mentioned reviewer, similar contraints apply to some of the SVCs in the West African languages, despite the fact that in such languages the two serial verbs are separated by an object; see De´chaine 1993. (i) a. John-i sayngsen-ul kwu-e mek-ess-ta. John-Nom fish-Acc broil-L eat-Past-Decl ‘John broiled and ate the fish.’ b. kwu-e mek-ess-ta John-i sayngsen-ul. broil-L eat-Past-Decl John-Nom fish-Acc c. sayngsen-ul kwu-e mek-ki-nun John-i ha-ss-ta. fish-Acc broil-L eat-NL-Top John-Nom do-Past-Decl ‘As for broiling and eating the fish, John did it.’ (ii) a. *kwu-e John-i sayngsen-ul (ha-e) mek-ess-ta. broil-L John-Nom fish-Acc (do-L) eat-Past-Decl b. *mek-ess-ta John-i sayngsen-ul kwu-e. eat-Past-Decl John-Nom fish-Acc broil-L c. *sayngsen-ul kwup-ki-nun John-i (ha-e) mek-ess-ta. fish-Acc broil-NL-Top John-Nom do-L eat-Past-Decl ‘As for broiling the fish, John did it and ate (the fish).’

Notes

199

10. Like the idiomatic V-V forms, idiomatic N-N forms disallow the appearance of delimiters between the two elements of the compound. (The examples below are from Sohn 2001, 245–246.) (i) a. b. c. d.

chel-say isul-pi mwul-kay nwun-mwul

(season-bird) (dew-rain) (water-dog) (eye-water)

(ii) a. b. c. d.

*chel-nun say *isul-nun pi *mwul-nun kay *nwun-nun mwul

‘migratory bird’ ‘drizzle’ ‘seal’ ‘tears’

11. Park 1994 provides the following evidence that -tul attaches to phrases. (Ko iss- is the progressive marker.) (i) Ai-tul-i pakk-eyse(-tul) sikkulepkey(-tul) nol-ko (-tul) iss-ess-ta. Child-Pl-Nom outside-at-(Pl) noisily-(Pl) play-KO-(Pl) BE-Past-Decl ‘Children are playing noisily outside.’ (ii) Ai-tul-i pakk-(*tul)-eyse sikkulepkey nol-(*tul)-ko Child-Pl-Nom outside-(*Pl)-at noisily play-(*Pl)-KO iss-(*tul)-ess-(*tul)-ta. BE-(*Pl)-Past-(*Pl)-Decl ‘Children are playing noisily outside.’ 12. The interpretation of the second object is akin to that of an E-type pronoun. See note 25 in chapter 1. 13. The locative -eyse in the context of ka- is associated with a source reading (movement away from an initial point). On the other hand, in combination with locative -ey, ka- gives rise to a movement-toward-a-goal meaning. We return to the latter form in section 2.2.1.2. 14. The fact that (i) is not an SVC is shown by the ill-formedness of (ii). Recall that NPI licensing is one of the properties that identifies the SVC as monoclausal and distinguishes it from the subordinate and coordinate structures; see section 2.1.1. We therefore conclude that the sentence in (i) contains an SC, with the subordinator -ese deleted. (i) John-i cengwen-ey ka (ese) ttwi-ess-ta. John-Nom garden-Loc go (Sub) run-Past-Decl ‘John went to the garden, and then he ran.’ (ii) *Amwuto cengwen-ey an ka ttwi-ess-ta. anyone garden-Loc Neg go run-Past-Decl 15. The following example seems to be a case where V1 expresses the purpose of the event denoted by V2. But there is a reason to believe that this is an idiomatic case. First, the purpose relation in the SVC is not productive. Second, ka- generally requires an obligatory goal argument; see section 2.2. Third, and most importantly, chaca ka- has a noncompositional meaning, which can be translated as ‘visit’.

200

Notes

(i) John-i (samwusil-ey) Mary-lul chaca ka-ss-ta. John-Nom (o‰ce-Loc) Mary-Acc look for go-Past-Decl ‘John went (to the o‰ce) to look for Mary.’ ‘John visited Mary (at the o‰ce).’ 16. While syntactic projections (generated via Merge) are not specified with Pfeatures, lexical items are. Indeed, we think that head movement is sensitive to whether the lexical item is specified with P-features or not. See note 4. As mentioned there, this suggests that head movement is a PF-operation. 17. Unfortunately, VP adverbs are not felicitous when they follow short-form negation (independently of the presence of the NPI). Therefore, we cannot insert cemcem ‘gradually’ immediately preceding V1 in the example in (221). This seems to be a general constraint on low VP adverbs. As noted in Hagstrom 1997 (who attributes the observation to Lee 1993), adverbs in Korean can be classified into two types. High adverbs can freely appear in various positions in the clause. On the other hand, low adverbs cannot. (i) (*cal) John-un (?*cal) sayngsenhoy-lul (cal) mek-nun-ta. (well) John-Top (well) sashimi-Acc (well) eat-Pres-Decl ‘John eats sashimi well.’ The low adverb may precede preverbal negation, but it has a focused status. (ii) John-un pap-ul (cal) an (?*cal) mek-nun-ta. John-Top meal-Acc (well) Neg (well) eat-Pres-Decl ‘John didn’t eat food well.’ (iii) John-i chelsa-lul (?cemcem) an (?*cemcem) kwupwuli-e kkunh-ess-ta. John-Nom wire-Acc (gradually) Neg (gradually) bend-L cut-Past-Decl The same applies to the example in (221); see below. (iv) Amwuto chelsa-lul (? cemcem) an kwupwuli-e kkunh-ess-ta. anyone wire-Acc (gradually) Neg bend-L cut-Neg-Past-Decl ‘No one bent and cut the wire.’ 18. There is another morpheme in Korean that can also spell out the directedmotion construction when the endstate is an abstract location, namely -(e)ci (comparable to English ‘become’). In appendix 2.1, we discuss the di¤erences and similarities between -(e)ci and ka-. 19. The progressive morpheme ko iss- seems to be compatible with some achievement verbs; compare (ii)–(iii) with (i). (i) *Han pyengsa-ka cwuk-ko iss-ta. one soldier-Nom die-Prog-Decl ‘One soldier is dying.’ (ii) John-i salaci-ko iss-ta. John-Nom disappear-Prog-Decl ‘John is disappearing.’ (iii) Yenkuk-i ttuthna-ko iss-ta. Performance-Nom end-Prog-Decl ‘The performance is ending.’

Notes

201

Given the fact that achievement verbs encode an instantaneous event and as such, they are nondurative, their compatibility with ko iss- is a puzzle. Yet, a careful examination of the data reveals that ko iss- combined with an achievement verb does not give rise to a progressive meaning. We conjecture that it gives rise to an incipient reading. In fact, some achievement verbs in English also co-occur with a progressive (e.g., John is reaching the top), but in these cases, ‘be ing’ encodes the meaning of ‘to be about to’. Some support for the above analysis is provided by the fact that the achievement verbs when combined with ko iss- are compatible with keuy ‘almost’. Ko iss-, in its progressive meaning, is not compatible with the adverb keuy. (iv) John-i (*keuy) nol-ass-ta. (Activity verbs) John-Nom (almost) play-Past-Decl ‘John (*almost) played.’ (v) John-i (*keuy) nol-ko iss-ta. John-Nom (almost) play-Prog-Decl ‘John is (*almost) playing.’ (vi) John-i cip-ul (keuy) ci-ess-ta. (Accomplishment verbs) John-Nom house-Acc (almost) build-Past-Decl ‘John (almost) built a house.’ (vii) John-i cip-ul (*keuy) cis-ko iss-ta. John-Nom house-Acc almost build-Prog-Decl ‘John is (*almost) building a house.’ (viii) Sengcang-i (keuy) ttuthna-ss-ta. growth-Nom (almost) end-Past-Decl ‘The growth (almost) ended.’

(Achievement verbs)

(ix) Sengcang-i (keuy) ttuthna-ko iss-ta. growth-Nom (almost) end-Prog-Decl ‘The growth is (almost) about to end.’ There is another puzzle regarding the ko iss- construction in Korean. Verbs such as ‘have’, ‘know’, ‘understand’, and ‘love’ are typical stative predicates in English and therefore cannot co-occur with the progressive. On the other hand, the Korean counterparts of these verbs—kaci- ‘have’, al- ‘know’, ihayha- ‘understand’, and salangha- ‘love’—are compatible with ko iss-. (x) John-i ku kong-ul kaci-ko iss-ta. John-Nom the ball-Acc have-Prog-Decl (xi) John-i ku mwuncey-lul ihayha-ko iss-ta. John-Nom the problem-Acc understand-Prog-Decl (xii) John-i ku sasil-ul al-ko iss-ta. John-Nom the truth-Acc know-Prog-Decl Interestingly, ko iss-, when it co-occurs with the verbs listed above, does not express a process in progress. Rather, it expresses the resulting state. In this respect, when ko iss- co-occurs with emotive and cognitive verbs, its function is similar to the resultative marker -e/a iss-. Given their compatibility with ko iss-, Ahn (1995)

202

Notes

claims that the above Korean verbs are not stative. Ogihara (1998) claims that the Japanse counterparts are not stative verbs either. His argument is based on the observation that Japanese counterparts of Korean al- and salangha- can co-occur with -te iru, which is known to be incompatible with stative predicates. 20. Note that if the singular subjects in (228a) and (228b) are replaced by a plural subject, the output sentences are compatible with the adverb cemcem. Again, this is just as expected since predication between a predicate and a plural argument gives rise to a plurality of events that can be mapped onto a series of subintervals at the aspectual temporal level; see section 1.5.1. (i) Pyengsa-tul-i cemcem cwuk-ess-ta. soldier-Pl-Nom gradually die-Past-Decl ‘Soldiers died one after the other.’ (ii) Pyengsa-tul-i cemcem salaci-ess-ta. Soldier-Pl-Nom gradually disappear-Past-Decl ‘Soldiers disappeared one after the other.’ Note that the example in (i) above is compatible with the progressive marker ko iss-. Compare (iii) below with (i) in note 19. (iii) Manhun pyengsa-tul-i cencayngthe-eyse cwuk-ko iss-ta. many soldier-Pl-Nom battlefield-Loc die-Prog-Decl ‘Many soldiers are dying in the battlefield.’ 21. Based on the bound-morpheme status of the intransitive path verbs tul-/na-, the argument that tul-e ka-ta ((move) into-L go-Decl) and na-ka-ta ((move) out of-go-Decl) are lexical compounds has been put forth; see Suh 2000. We disagree with the premise that compounds containing bound morphemes must necessarily be formed in the lexicon. 22. The following examples are incompatible with ka-, which indicates that these predicates are not compatible with an interpretation that involves subintervals. (i) Wuli-nun Seoul-yek-ey(se) nayli-e (*ka)-ss-ta. we-Top Seoul-station-Loc (move) down-L (*go)-Past-Decl ‘We got o¤ at Seoul station.’ (ii) John-i mal-ey olla-(*ka)-ss-ta. John-Nom horse-Loc (move) up-go-Past-Decl ‘John climbed on the horse.’ 23. A caveat is in order here regarding the aspectual property of intransitive path verbs when they refer to directed movement along an abstract path or scale. In such cases, the resulting predicates can be modified by cemcem ‘gradually’. (i) Kakyek-i cemcem oll-ass-ta. prices-Nom gradually (move) up-Past-Decl (ii) Onto-ka cemcem oll-ass-ta. temparature-Nom gradually (move) up-Past-Decl This suggests that the [
durative] feature of the predicate in the examples discussed in the text is not determined solely by the verb; it is determined compositionally by the verb and its locative argument.

Notes

203

24. -ey has a locative function with the copular verb iss- ‘be’; see (i) below. It is morphologically related to the form -eyse (which contains a fossilized morpheme -se), which also functions as a locative marker. Yet, unlike -ey, -eyse occurs with activity verbs like nol- ‘play’; see (ii) below. While -ey has the status of a complement, -eyse introduces an adjunct (cf. Lee 1999). When -eyse appears with verbs of directed motion such as ka- ‘go’ and o- ‘come’, -eyse gives rise to a source meaning; see (iii) below (cf. Sohn 2001, 334–335). (i) John-i cip-ey iss-ta. John-Nom house-Loc be-Decl ‘John is at home.’ (ii) John-un hakkyo-eyse nol-ass-ta. John-Top school-Loc play-Past-Decl ‘John played at school.’ (iii) John-i hakkyo-eyse o-ass-ta. John-Nom school-Loc come-Past-Decl ‘John came from school.’ 25. On the adjunct status of Korean -lo, see Chae 2000; Lee, Nam, and Kang 1998; and Lee 2003. We assume that -kkaci can also function as an adjunct. On the other hand, -kkaci can also function as the complement of a path-denoting direct object, as illustrated in (i). This is comparable to the Spanish case in (ii), to be discussed in chapter 3. (i) John-i cengwen-kkaci (osolkil-ul) kel- / talli-ess-ta. John-Nom garden-up to (little path-Acc) walk- / run-Past-Decl (ii) Juan corrio´/camino´ (el trayecto) hasta la casa. ‘Juan ran/walked (the path) up to the house.’ 26. The Linear Correspondence Axiom or LCA (proposed by Kayne 1994), or some version of it (see Chomsky 1994, 1995), derives the generalization that Spec must be to the left of the head. The LCA also requires that languages be universally head initial. In this chapter, we have assumed the classical analysis of Korean as head final. Either we must reanalyze Korean as head initial and motivate the necessary leftward movements that yield the right order, or we must come up with a mixed system that imposes a leftward Spec position but allows for headfinal structures. We will not attempt either exercise here, but we will assume that the second option is in the right direction. 27. The fact that the contact object is aimed in a certain direction probably follows from the fact that the body (or relevant body part) of the subject is aimed in a certain direction. The question arises whether the grammar should represent such a notion, for example by having an implicit body-part instrumental modifying the subject. This would be similar to Hale and Keyser’s proposal for verbs like pour, whereby the l-structure contains an implicit manner adverb that is subject oriented. We leave this question open for future research. 28. An anonymous reviewer reports that she or he does not like the examples in (270). We have consulted five Korean native speakers and they all judged these examples to be acceptable.

204

Notes

29. The example in (277c) is fine under a CSVC reading, namely, in a situation in which finishing the reading assignment is a necessary condition for John to be able to go to school. 30. The following examples are not counterexamples to the observation that Korean lacks RSVCs. These are cases of subordinate structures with a deleted subordinator. That they are not SVCs is shown by the fact that short-form negation cannot scope over the sequence of Vs and therefore cannot license the NPI in subject position; see (ib), (iib), (iiib), and (ivb). (i) a. (?)Ai-tul-i ttangpatak-ulo cwul-ul tangki-(ese) nemeci-ess-ta. child-Pl-Nom ground-toward rope-Acc pull-(Sub) fall-Past-Decl ‘The children pulled the rope and fell to the ground.’ b. *Amwuto ttangpatak-ulo cwul-ul an tangki-(ese) nemeci-ess-ta. anyone ground-toward rope-Acc Neg pull-(Sub) fall-Past-Decl Intended meaning: ‘No one pulled the rope and fell to the ground.’ (ii) a.

Kyengchal-tul-i kangto-lul olaystongan chwukyekhay-(ese) policeman-Pl-Nom robber-Acc for a long time chase-(Sub) phikonha-ss-ta. tired-Past-Decl ‘The policemen chased a robber for a long time and became tired.’ b. *Amwuto kangto-lul olaystongan an chwukyekhay-(ese) anyone robber-Acc for a long time Neg chase-(Sub) phikonha-ss-ta. tired-Past-Decl Intended meaning: ‘No one chased a robber for a long time and became tired.’

(iii) a.

John-i namwu-ey meli-lul pwuticchi-(ese) ssuleci-ess-ta. John-Nom tree-Loc head-Acc hit-L-(Sub) collapse-Past-Decl ‘John hit his head against a tree and collapsed.’ b. *Amwuto namwu-ey meli-lul an pwuticchi-(ese) ssuleci-ess-ta. anyone tree-Loc head-Acc Neg hit-(Sub) collapse-Past-Decl ‘No one hit his head against a tree and collapsed.’

(iv) a.

John-i namwu-ey meli-lul pwuticchi-(ese) kicelha-ss-ta. John-Nom tree-Loc head-Acc hit-(Sub) faint-Past-Decl ‘John hit his head against a tree and fainted.’ b. *Amwuto namwu-ey meli-lul an pwuticchi-(ese) kicelha-ss-ta. anyone tree-Loc head-Acc Neg hit-(Sub) faint-Past-Decl ‘No one hit his head against a tree and fainted.’

31. As mentioned earlier, cha- ‘kick’ modifies the motion event and not the causative event in (292). On the other hand, in the English sentence John kicked the ball to the garden, kick does modify the causative event. We discuss this type of English sentence in chapter 3. 32. Intransitive path verbs are comparable to particles in English: go up, go down, go out. On the other hand, English lacks transitive particles, with a function com-

Notes

205

parable to the transitive path verbs in Korean. But note the use of the particle in examples like ‘they will up the prices’. 33. Kim (1995) put forth a di¤erent claim for the interpretation of ka- compounded with a transitive path verb. This author argues that ka- compounded with the transitive path verb does not entail movement of the subject although it entails movement of the object. The speakers we have consulted disagree with that judgment. 34. Note that while mil- and kkul- are morphologically simple, kwulli- is morphologically complex. It is made up of two morphemes: the unaccusative kwulu- and the causative morpheme -i. 35. As was mentioned in section 2.2.1.1, activity verbs lack an endpoint and therefore cannot function as a path-denoting argument of ka-. This is also true for -(e)ci. (i) a. *Aki-ka ca-ci-ess-ta. baby-Nom sleep-CI-Past-Decl b. *John-i ilhay-ci-ess-ta. John-Nom work-CI-Past-Decl c. *John-i nol-aci-ess-ta. John-Nom play-CI-Past-Decl 36. Sentence (344a) is ambiguous. On the one hand, it can have an SVC reading, ‘John baked a cake and gave it to Mary’, in which case cwu- keeps its literal meaning (namely, ‘give’). On the other hand, cwu- can have a benefactive reading, ‘John baked a cake for Mary.’ The following sentence also illustrates an ambiguity between the benefactive reading and the SVC reading. (i) John-i Mary-eykey ssuleyki-lul cwuw-e cwu-ess-ta. John-Nom Mary-Dat trash-Acc pick up-L give-Past-Decl ‘John picked up and gave the trash to Mary.’ ‘John picked up the trash on behalf of Mary (for Mary).’ There are cases where no ambiguity arises. When the first V does not give rise to a transferable object, cwu- can only be associated with the benefactive construal, as illustrated by the example below. (ii) John-i Mary-eykey nolay-lul pwul-e *(cwu)-ess-ta. John-Nom Mary-Dat song-Acc sing-L Ben-Past-Decl ‘John sang a song for Mary.’ 37. In theory, double dative-marked arguments are predicted to be possible. But they are not. We assume that double dative-marked arguments are ruled out by the morphological constraint banning two occurrences of Dative Case. 38. Another fact that needs to be explained is why in Korean the verbs of obtaining ( find, buy) and verbs of creation (draw, bake) cannot combine with the abstract V* (unlike English)—for example, (344). Possibly, this is due to the competition with the cwu- construction (which encodes the wider notion of benefaction and which subsumes the notion of prospective possessor). We will address this issue in future work.

206

Notes

Chapter 3 1. English, unlike Dutch, makes a distinction between to and toward. Although the preposition to is preferably used as telic, the atelic meaning is available, as shown by the fact that John went to Paris does not necessarily imply that John reached Paris. On the other hand, the preposition toward is unambiguously atelic. The judgments reported in (369) are valid only for speakers comfortable using to in atelic situations. 2. Similar judgments are reported by Folli and Harley 2004. 3. Inchoatives in Dutch also select the auxiliary zijn and are perfective: (i) De vaas is/*heeft gebroken. The vase is/*has broken (ii) De temperatur is/*heeft gestegen/gedaald. The temperature is/*has increased/decreased 4. We thank Marcel den Dikken for providing many of the examples in this section and for spelling out the meaning di¤erences between the forms constructed with zijn and the forms constructed with hebben. 5. See also den Dikken 1995. 6. Note that (387a) illustrates once more the distinction between event boundedness (defined at the E-level) and temporal boundedness (defined at the R-level); see section 1.5.1. In (387a) the event is an accomplishment, although it is temporally unbounded (due to the atelic nature of the PP). Indeed, den Dikken (p.c.) reports that when modified by an atelic temporal phrase, (387a) results in an awkward sentence. 7. It seems that (389a) is somewhat awkward, but much better than its variant with heeft ‘has’. 8. The following cases are not SVCs (see Carden and Pesetsky 1977; Jaeggli and Hyams 1993; Cardinaletti and Giusti 2001). See section 3.1.5. (i) I go and buy bread every day. (ii) I go buy bread every day. 9. Goldberg and Jackendo¤ (2004) note the contrast between (392) and (i)–(iii) below. In these cases, the verb modifies the ‘‘cause,’’ rather than the ‘‘movement.’’ We return to these examples in section 3.1.4. (i)

The car honked its way down the road.

(ii) The dog barked its way out of the room. (iii) Bill whistled his way past the house. 10. See note 12 in chapter 1. 11. There are some idiolectal variations as to which verbs may license the directed-motion reading in combination with a locative preposition. Some speakers find it di‰cult with stumble, saunter, jog, amble, drift, trudge, stagger, limp, and stroll, but easier with crawl, slither, dart, slide, and climb. Other speakers do not find any di¤erence between the two sets of verbs.

Notes

207

12. As has often been noted, the preposition under gives rise both to a locative and a directional reading, but in the latter case, its complement does not specify the endpoint of the path. Instead, it provides a reference point with respect to which the endpoint of the path is construed; see section 1.5.2. (i) a. The boat floated to/toward the bridge (the bridge is the endpoint of the path). b. The boat floated under the bridge (the other side of the bridge, rather than the bridge, is the endpoint of the path). It could be that in (ib), under the bridge is the specifier of a covert particle with a meaning comparable to through. This is suggested by the Dutch counterparts of (ib), and other similar examples, which are overtly circumpositional. In (ii), it is the postpositional part of the circumposition, through, that conveys that the boat ended up on the other side of the bridge. (ii) De boot dreef onder de brug door. The boat floated under the bridge through The example in (iii) is explicit in saying that the airplane ended up on the other side of the city; its variant without heen would be compatible with that interpretation but it also accommodates a reading in which the airplane is flying around over the city, in circles. (iii) Het vliegtuig vloog over de stad heen. the airplane flew over the city PRT This is also the case in (iv), though here the e¤ect of heen is less definitive: both versions are compatible with a ‘walking around in circles’ reading (although heen is not easily construed that way). But to express that John actually exited the building after having walked through it, one would still prefer to use the circumposition with heen. (iv) Jan liep door het gebouw (heen). Jan walked through the building (through) We are grateful to Marcel den Dikken for providing the Dutch data to us. 13. We thank an anonymous reviewer for reminding us of this phenomenon. 14. When governed by a noun, in is also unambiguously locative (whether it is pre- or postpositional): (i) a. de weg the road b. de weg the road

in het bos (Locative only) in the woods het bos in (Locative only) the woods in

15. It is plausible to analyze send as a causative form of go, receive as a causative form of come, and drop as a causative form of fall. 16. Levin and Rappaport (1995) and Rappaport and Levin (1998) have noted the contrast below: (i) a. John broke the eggs into a bowl. b. aJohn broke the dishes o¤ the table.

208

Notes

As pointed out by an anonymous reviewer, the contrast might be pragmatic in nature. It makes sense to get the inside of the eggs into a bowl by breaking the eggs, but it does not make sense to get the dishes o¤ the table by breaking them. 17. Compare with the Korean SSVCs that contain contact verbs, where the movement is predicated of the subject; see section 2.2.4. 18. Mateu (2005) also analyzes such examples as being associated with a causative structure. 19. Apparently this is also the case in Dutch, so in this language the ‘accompany’ or ‘go-along’ meaning is the preferred one as well. 20. The examples in (439a–d) are due to an anonymous reviewer and (439e) to H. Borer (who attributes it to S. Rothstein). (439e) has a noncausative analysis as well. 21. Lamiroy (1983) has proposed the same type of analysis for French aller/courir acheter le journal and its Spanish counterpart ir/correr a buscar el diario. See note 32. 22. Cardinaletti and Giusti (2001) note that in Marsala, the counterpart of go/ come is followed by a linker a that is diachronically related to the conjunction. Nevertheless, we may assume that synchronically a functions as a path-denoting preposition. (i) Vaju a pigghiu u pani. (I)go-1s to fetch-1s the bread ‘I go to fetch bread.’ In Swedish, the linker och still has the morphophonological form of a conjuction. Yet it is possible that at present it functions semantically as a path-denoting element. (ii) Jag ga˚r och go¨r mig en grogg. I go-pres and make-pres myself a grog ‘I go to make myself a grog.’ One might conjecture that the English ‘‘go-V’’ form (go eat your supper) also evolved from the ‘‘go and V’’ form (go and eat your supper), although the latter form is still in use and the two have distinct properties. 23. While we continue to use the term restructuring verbs (introduced by Rizzi 1978), we will not assume a ‘‘restructuring’’ process to account for the phenomenon in question. See section 3.2.2 for futher details. 24. Morimoto 2001 also notes that the notions of E-boundedness and telicity must be pulled apart. See also Demonte 2003. 25. In fact, Aske 1989 reports that the verb caminar (walk) is able to take a dative directional PP in the dialect described by this author. (i) Mi ejercicio consiste en caminar a la biblioteca dos veces al dı´a. ‘My exercise consists in walking to the library twice a day.’ (Aske 1989) 26. De is di¤erent from English from. The latter corresponds more closely to desde.

Notes

209

27. On other types of Spanish complex prepositions, see Bosque 1996. 28. More precisely, Aske’s examples are the following: (i) *Juan camino´ hasta la cima en una hora. ‘Juan walked up to the summit in an hour.’ 29. While the following are perhaps less natural than (482), they are (to our ears) acceptable forms: (i) Una vez galopado ese trecho, el jockey debera´ volver al punto de partida. ‘Once galloped that distance, the jockey should return to the starting point.’ (ii) Una vez nadado ese trecho, el nadador debera´ volver al punto de partida. ‘Once swam that distance, the swimmer should return to the starting point.’ 30. Modification by a degree adverb is possible; see the examples below. But in these examples, the degree adverbs mucho and un poco/poquito, like the measure phrase tres metros, are interpreted as modifying the extent of the path rather than the event itself. See Le llevo´ un minuto al barril rodar un poquito/tres metros ‘It took a minute for the barrel to roll a little bit/three meters’. In this respect, these verbs seem to be similar to scalar verbs like increase/decrease—for example, the temperature increased/decreased a lot/a bit in a minute. (i) El barril rodo´ mucho/un poco/tres metros (en un minuto). ‘The barrel rolled a lot/a little bit/three meters (in one minute).’ (ii) La moneda se deslizo´ mucho/un poco/tres centı´metros (en un minuto). ‘The coin slid a lot/a little bit/three centimeters (in one minute).’ 31. Note that the verb correr, as in (i) below, has been bleached of its manner component; it means ‘to go quickly’. (i) Corrio´ al hospital (en taxi). (I) run to the hospital (in a taxi) 32. Lamiroy’s main focus is the study of bare infinitival complements with manner verbs, such as courir acheter le journal ‘to run buy the newspaper’, and intrinsic directed-motion verbs, such as aller acheter le journal ‘to go buy the newspaper’. Lamiroy argues that in both cases the VP complement functions as the goal or telos of the motion. Compare with *marcher acheter le journal ‘to walk buy the newspaper’; *nager chercher le ballon ‘to swim find the ball’. 33. Unlike Italian, auxiliary selection in French does not correlate perfectly with unaccusativity. While the auxiliary eˆtre ‘be’ can only appear with unaccusative verbs and verbs with a reflexive clitic, the auxiliary avoir ‘have’ appears with unergatives and with many unaccusatives (alternating anticausatives, in particular). 34. Other manner-of-motion verbs cited by Larmiroy are claudiquer (to limp), clopiner (to hobble), bondir (to leap), crawler (to crawl), danser (to dance), grimper (to climb), pagayer (to paddle), patiner (to skate), pe´daler (to pedal), plonger (to dive), ramer (to row), ramper (to slither), rouler (to roll), sauˆter (to jump), sautiller (to hop), skier (to ski), sprinter (to sprint), trotter (to trot), trottiner (to jog along), voler (to fly).

210

Notes

35. Other intrinsic directed-motion verbs are partir (to leave), passer (to pass), rentrer (to go in), sortir (to go out), venir (to come), as well as (s’en) fuir (to run away or flee), s’expatrier (to expatriate oneself ), se glisser (to slide or glide), se faufiler (to inch or edge one’s way into), s’e´chaper (to escape), s’esquiver (to slide or sneak away), se pre´cipiter (to plunge into; to rush over or forward), foncer (to charge at or into), s’elancer (to rush forward; to dash toward), se ruer (to tear or hurl into). 36. For some (but not all) speakers, courir ‘run’ seems to exhibit variable behavior (like sauˆter ‘jump’): (i) a. Jean Jean b. Jean Jean

court a` la maison en cinq minutes. runs to the house in five minutes y court en cinq minutes. loc. cl. runs in five minutes

(Unaccusative)

(ii) Jean court dans le parque pendant cinq minutes. Jean runs in the park during five minutes

(Unergative)

37. Boons gives (i) as well formed. We therefore assume that he would equally accept (520a). We have observed similar variability among Spanish speakers; see section 3.2.1.1 and note 38. (i) Max a rame´ jusqu’au rocher en vingt minutes. Max rowed up to the rock in twenty minutes 38. We propose to extend the analysis mentioned in the text to account for French speakers like Boons, who accept examples like (i) in note 37. That is, for such speakers, the structural representation of such sentences contains a complex PP with a covert deictic source (de . . . jusqu’a` ‘from up to’), which specifies a silent object classifier that denotes ‘distance’; see (i) below and (505) in the text. (i) Mar a rame´ (d’ici/la`-bas) jusqu’au rocher en vingt minutes. Max rowed (from here/there) up to the rock in twenty minutes 39. Note that Spanish and other Romance languages have a productive set of prefixed denominal verbs with a causative meaning. These are verbs that are morphologically derived from nouns via a‰xation of the prefix de. This prefix is associated with the causative meaning ‘make go out’ or ‘remove from’, and the noun to which it is a‰xed can be analyzed as the incorporated figure-denoting object (along the lines of Hale and Keyser 2002). (i) a. des þ huesar/grasar la carne (remove the bone/fat from the meat) b. des þ aguar la ban˜adera (remove the water from the bathtub) c. des þ lechar al ternero (remove the mother’s milk from the calf ) d. de þ sangrar al animal (remove the blood from the animal) e. *de þ gasar al bebe´ (remove the gas from the baby) f. *de þ zapatar al bebe´ (remove the shoes from the baby) It is tempting to analyze such cases in terms of ‘‘incorporation,’’ in which the object undergoing movement, as well as the directional preposition, are incorporated into the verb, as schematized below:

Notes

211

(ii) [ . . . [vP dei - v [VP huesj -ar]] [(huesoj ) [V [(dei ) la carne]]]] de prefixation to nouns is lexically restricted; it combines with certain nouns, such as hueso ‘bone’, agua ‘water’, leche ‘milk’, and sangre ‘blood’ (ia–d), but not with others, such as gas ‘gas’ and zapato ‘shoe’ (ie–f ). On the other hand, de prefixation productively attaches to verbs that have been formed via prefixation of en to a noun. The prefix en can be analyzed as the morphological realization of the incorporated preposition and the noun as the incorporated figure-denoting object. See Mateu 2002 for an analysis along similar lines. (iii) a. b. c. d. e. f.

de(s) þ en þ sillar el caballo (remove the saddle from the horse) de(s) þ em þ botellar el vino (remove the wine from the bottle) de(s) þ em þ paquetar el regalo (remove the gift from the package) de(s) þ en þ vainar el cuchillo (remove the knife from its sheath) de(s) þ en þ capuchar al prisionero (remove the hood from the prisoner) de(s) þ en þ mascarar al culpable (remove the mask from the culprit)

It is to be noted though that verbs like em-botellar (loc. prefix þ bottle) and des-em-botellar (neg. pref. loc. pref. þ bottle) do not have the same syntactic properties. (iv) a.

Em botellamos el vino en una botella de vidrio. (We) loc. pref þ bottled the wine in a glass bottle b. Des em botellamos el vino (*de una (We) neg. prefix-loc. pref. þ bottle the wine (*out of a botella de vidrio). glass bottle)

The above paradigm raises doubts as to whether the prefixed verbs in (i) and (iii) should be derived from a syntactic structure like the one in (ii). 40. There is some interesting crosslinguistic variation in the class of ‘push’ verbs. In French there are two di¤erent verbs pousser ‘push’ and there are two di¤erent verbs tirer ‘pull’. One of them actually entails movement of the object, which is realized as a direct object. Its complementation is comparable to mettre ‘put’, with the di¤erence that the end location is obligatorily expressed with mettre, but can be left unspecified in the case of pousser/tirer: (i) Pierre a pousse´ le chariot (dans le garage). ‘Pierre pushed the cart into the garage.’ ‘Pierre pushed the cart toward an unspecified location (while he was in the garage).’ (ii) Pierre a tire´ le chariot (dans le garage). ‘Pierre pulled the cart into the garage.’ ‘Pierre pulled the cart toward an unspecified location (while he was in the garage).’ Compare the French examples in (iii) with their English counterparts in (iv): (iii) Pierre a pousse´/tire´ le chariot (amais le chariot n’a pas bouge´). (iv) Pierre pushed/pulled the cart (but the cart did not move).

212

Notes

The other use of these verbs is incompatible with directed movement of the object. The object on which force is exerted is realized as a PP. Comparable forms also exist in English. (v) Pierre a pousse´ sur le chariot (*vers le garage). Pierre pushed on the cart (*toward the garage) (vi) Pierre a tire´ sur le chariot (*vers le garage). Pierre pulled on the cart (*toward the garage) The Spanish counterparts, on the other hand, are ill-formed: (vii) *Pedro empujo´ sobre el carro (hacia el garage). Pedro pushed on the cart (toward the garage) (viii) *Pedro estiro´ del carro (hacia nosotros). Pedro pulled on the cart (toward us) To summarize, in English and Spanish, the transitive forms of push/pull do not entail movement, but they are compatible with it. In French the transitive form does entail movement, but of the ‘come-to-be Loc’ type, not of the ‘come/go to Loc’ type; see *Pierre a pousse´ le chariot au garage ‘Pierre pushed the cart to the garage’. Unlike English push/pull, French pousser/tirer might actually constitute a case of polysemy. This class of verbs deserves further crosslinguistic research. 41. We have found that there is some idiolectal variation with respect to class 3. See below for further discussion. 42. For a critique of the latter point, see note 44. 43. The latter set of prepositions contrasts with the preposition lungo ‘along/ beside’, which can never give rise to a directional meaning, irrespective of the context in which it appears. Indeed, with the preposition lungo, it is impossible to construct the meaning ‘Maria ran until she was beside the beach’. (i) *Maria e` corsa lungo la spiaggia. ‘Maria ran along the beach’ (intended meaning). (ii) Maria ha corso lungo la spiaggia. ‘Maria ran beside the beach.’ 44. Following Ramchand 2001, Folli assumes that the telic nature of the path determines the presence of the abstract aspectual category RvP. In our discussion of Dutch, we have seen that this is not the case; see (387) and (388) in section 3.1.2. An event can be bounded even if the preposition is atelic. This can also be appreciated in Italian with an example like (i) below, which contains the atelic preposition verso ‘toward’. The PP specifies the direction of the path (i.e., it provides a reference point with respect to which the movement is defined), but it does not specify the endpoint of the motion. Since the endpoint is not specified, the presence of the temporal phrase in un secondo in (ii) can only be interpreted as specifying the time it takes for the running event to get started (rather than the length of the running event). Nevertheless, the construction has the properties of a bounded construction, as indicated by the choice of auxiliary and the presence of participle agreement.

Notes

213

(i) Maria e` corsa *(verso la spiaggia). Maria is run-3rd p.s.fem. *(toward the beach) (ii) Maria e` corsa verso la spiaggia in un secondo. Maria is run-3rd p.s.fem. toward the beach in one second ‘It took Maria one second to start running toward the beach.’ Note furthermore that the preposition verso (like other directional prepositions) can give rise to a bounded construction only with a lexically restricted class of verbs—that is, the class of verbs in (533c). See the contrast below. (iii) Maria e` corsa verso la spiaggia. Maria is run toward the beach ‘Maria has run toward the beach.’ (iv) *Maria e` camminata verso la spiaggia. Maria is walked toward the beach ‘Maria has walked toward the beach.’ 45. For a comparison between directional prepositions in Germanic (in particular, Swedish) and Romance (in particular, French), see Bergh 1948. 46. One of our informants found the directional use of saltellare somewhat odd as well. We will ignore these finer distinctions here. 47. A. Cardinaletti has brought to our attention the contrast between (i) and (ii) on the one hand, and the contrast between (iii) and (574) in the text. These contrasts show that the PP with a point-locating meaning has the status of an adjunct. The dative clitic in (i) and (iii) cannot bind into the PP adjunct because it does not c-command the PP adjunct. (i) *Il bambino ti ha voluto saltare in braccio per dieci the child dat. Cl þhas wanted jump on lap for ten minuti. (Locational) minutes ‘The child has wanted to jump on your lap for ten minutes.’ (ii) L’aereo e` potuto volar(e) in una stanza. the toy plane is been-able fly into a room ‘The toy airplane has been able to fly into a room.’

(Directional)

(iii) *Il bambino ti ha voluto saltare in braccio per dieci The child dat. Cl þhas wanted jump on lap for ten minuti. (Locational) minutes ‘The child has wanted to jump on your lap for ten minutes.’ 48. Note though that the Spanish dative preposition, unlike its Italian and French counterpart, cannot function as a locative. It can only function as goal-denoting, although only in complement position (see section 3.2.1.1). 49. Talmy (1985) has suggested that in Romance path ‘‘conflates’’ with motion. Under Talmy’s view, one might speculate that it is precisely because path ‘‘conflates’’ with motion in Romance that the dative preposition in this language family functions as a point-locating preposition. But as we argue below in the text, this hypothesis fails to account for the entire range of facts.

References

Ahn, H.-D. 1991. Light Verbs, VP-Movement, Negation and Clausal Architecture in Korean and English. Doctoral dissertation, University of Wisconsin at Madison. Ahn, Y. 1995. The Aspectual and Temporal System of Korean from the Perspective of the Two-Component Theory of Aspect. Doctoral dissertation, University of Texas at Austin. Aske, J. 1989. Path predicates in English and Spanish: A closer look. In Proceedings of the Berkeley Linguistics Society, vol. 15, 1–14. Awoyale, Y. 1987. Perspectives on verb serialization. In Victor Manfredi, ed., Niger-Congo Syntax and Semantics 1, 3–36. Papers from the April 1987 NigerCongo Syntax and Semantics Workshop, Boston University. Awoyale, Y. 1988. Complex Predicates and Verb Serialization. Lexicon Project Working Papers 28. Unpublished manuscript, Department of Linguistics and Philosophy, MIT. Baker, M. 1989. Object sharing and projection in serial verb construction. Linguistic Inquiry 20 (4): 513–553. Baker, M. 1991. On the relation of serialization to verb extensions. In Claire Lefebvre, ed., Serial Verbs: Grammatical, Comparative, and Cognitive Approaches, 79–102. Amsterdam: John Benjamins. Baker, M., and O. T. Stewart. 1999. On Double-Headedness and the Anatomy of the Clause. Unpublished manuscript, Rutgers University. Bamgbose, A. 1974. On serial verbs and verbal status. Journal of West African Languages 9:17–48. Beck, S., and W. Snyder. 2001. Complex predicates and goal PPs: Evidence for a semantic parameter. In L. Domı´nguez and A. Johansen, eds., Proceedings of the 25th Boston Conference on Language Development, vol. 1, 114–122. Somerville, MA: Cascadilla Press. Bergh, L. 1948. Moyen d’Exprimer en Franc¸ais l’Ide´e de Direction. Doctoral dissertation, University of Go¨teborg, Sweden.

216

References

Boons, J. P. 1987. La notion se´mantique de de´placement dans une classification syntaxique des verbes locatifs. Langue Franc¸aise 76:5–40. Borer, H. 1994. The projection of arguments. In E. Benedicto and J. Runner, eds., University of Massachusetts Occasional Papers in Linguistics Functional Projections 17: 19–47. Borer, H. 2005. Structuring Sense, Volume 2: The Normal Course of Events. Oxford: Oxford University Press. Bosque, I. 1996. Preposicio´n tras preposicio´n. In Homenaje a Ramo´n Trujillo, 13– 145. Tenerife: Montesinos. Butt, M., and G. Ramchand. 2001. Complex aspectual structure in Hindi/Urdu. Oxford Working Papers in Linguistics. Oxford: Oxford University. Cabrera, M. 2005. The Acquisition of Causative Structures in English and Spanish as Second Languages. Doctoral dissertation, University of Southern California. Cabrera, M., and M. L. Zubizarreta. 2003. On the acquisition of Spanish causative structures by L1 speakers of English. In J. M. Liceras, H. Zobl, and H. Goodluck, eds., Proceedings of the 2002 Generative Approaches to Second Language Acquisition (GASLA-6): L2 Links, 24–33. Somerville, MA: Cascadilla Press. Cabrera, M., and M. L. Zubizarreta. 2005. The role of the L1 in the overgeneralization of causatives in L2 English and L2 Spanish. In Selected Proceedings of the 33rd Linguistic Symposium on Romance Languages (LSRL 33). Amsterdam: John Benjamins. Carden, G., and D. Pesetsky. 1977. Double-verb constructions, markedness, and a fake coordination. Chicago Linguistic Society 13:82–92. Cardinaletti, A., and G. Giusti. 2001. ‘‘Semi-lexical’’ motion verbs in Romance and Germanic. In N. Corver and H. van Riemsdijk, eds., Semi-lexical Categories: On the Function of Content Words and Content of Function Words, 371–414. Berlin: Mouton de Gruyter. Cardinaletti, A., and U. Shlonsky. 2004. Clitic positions and restructuring in Italian. Linguistic Inquiry 35:519–557. Carstairs-McCarthy, A. 1999. The Origins of Complex Language: An Inquiry into the Evolutionary Beginnings of Sentences, Syllables, and Truth. Oxford: Oxford University Press. Carstens, V. 2001. Antisymmetry and word order in serial constructions. Language 78:3–50. Chae, H.-R. 1999. A definition and classification of locomotion verbs. Studies in Modern Grammar 15:79–100. The Society of Modern Grammar. (In Korean.) Chae, H.-R. 2000. Complements vs. Adjuncts. Studies in Modern Grammar 19:69–85. The Society of Modern Grammar. Chierchia, G. 2004. A semantics for unaccusatives and its syntactic consequences. In Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert, eds., The Unaccusativity Puzzle, 22–59. Oxford: Oxford University Press.

References

217

Choe, H.-S. 1988. Restructuring Parameters and Complex Predicates: A Transformational Approach. Doctoral dissertation, MIT. Choi, S., and M. Bowerman. 1991. Learning to express motion events in English and Korean: The influence of language-specific lexicalization patterns. Cognition 41:83–121. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1991. Some notes on economy of derivation and representation. In R. Freidin, ed., Principles and Parameters in Comparative Grammar. Cambridge, MA: MIT Press. Chomsky, N. 1993. A Minimalist Program for linguistic theory. In K. Hale and S. J. Keyser, The View from Building 20, 1–52. Cambridge, MA: MIT Press. Chomsky, N. 1994. Bare phrase structure. MIT Occasional Papers in Linguistics 5. Department of Linguistics and Philosophy, MIT. (Reprinted in G. Webelhuth, ed., Government and Binding Theory and the Minimalist Program, 383–439. Oxford: Blackwell, 1995.) Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. 1999. Derivation by phase. MIT Occasional Papers in Linguistics 18. (Also in M. Kenstowicz, ed., Ken Hale: A Life in Language. Cambridge, MA: MIT Press, 2001.) Chung, T. 1993. Argument Structure and Serial Verbs in Korean. Doctoral dissertation, University of Texas at Austin. Cinque, G. 2004. ‘‘Restructuring’’ and functional structure. In A. Belletti, ed., Structures and Beyond, vol. 3, 132–191. Oxford Studies in Comparative Syntax. Oxford: Oxford University Press. Clark, E. V. 1987. The principle of contrast: A constraint on language acquisition. In B. MacWhinney, ed., Mechanisms of Language Acquisition. Hillside, NJ: Erlbaum. Collins, C. 1993. Topics in Ewe Syntax. Doctoral dissertation, MIT. Collins, C. 1997. Argument sharing in serial verb construction. Linguistic Inquiry 28:461–497. De´chaine, R.-M. 1993. Predicates across Categories: Towards a Category-Neutral Syntax. Doctoral dissertation, University of Massachusetts at Amherst. Demonte, V. 2003. Que´ es sinta´ctico y que´ es le´xico en la interfaz entre sintaxis y le´xico-sema´tica: Hipo´tesis y perspectivas. Paper presented at the VI Coloquio Internacional de Linguı´stica Hispa´nica, University of Leipzig, October. den Dikken, M. 1995. Particles. On the Syntax of Verb-Particle, Triadic, and Causative Constructions. Oxford: Oxford University Press. den Dikken, M. 2003. On the Syntax of Locative and Directional Adpositional Phrases. Unpublished manuscript, CUNY, New York. Dowty, D. R. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel.

218

References

Dowty, D. R. 1991. Thematic proto-rules and argument selection. Language 67:547–619. Fillmore, C. J. 1988. The mechanism of ‘‘Construction Grammar.’’ BLS 14:35– 55. Folli, R. 2001. Constructing Telicity in English and in Italian. Doctoral dissertation, Oxford University. Folli, R., and H. Harley. 2004. Event-path homomorphisms and the accompanied-action reading in motion causatives. Paper presented at WECOL, USC, Los Angeles. To appear in the Proceedings. Folli, R., and G. Ramchand. 2001. Getting results: Motion verbs in Italian and Scottish Gaelic. In K. Megerdoomian and L. Barel, eds., WCCFL 20 Proceedings, 192–205. Somerville, MA: Cascadilla Press. Folli, R., and G. Ramchand. Forthcoming. Prepositions and results in Italian and English: An analysis from event decomposition. In H. Verkyul, H. van Hout, and H. de Swartz, eds., Perspectives on Aspect, 1–20. Dordrecht: Kluwer Academic Publishers. Fong, V., and C. Poulin. 1997. Verb Classes and Argument Structure in French and English. Unpublished manuscript, Stanford University. Fox, D., and D. Pesetsky. 2005. Cyclic linearization of syntactic structure. Theoretical Linguistics: Object Shifts, ed. Katalin E´. Kiss, 31 (1–2): 1–46. Freeze, R. 1992. Existentials and other locatives. Language 68:553–595. Goldberg, A. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Goldberg, A., and R. Jackendo¤. 2004. The English resultative as a family of constructions. Language 80:532–568. Gruber, J. S. 1965. Studies in Lexical Relations. Doctoral dissertation, MIT. Hagstrom, P. 1997. Scope interactions and phrasal movement in Korean negation. In S. Kuno et al., eds., Harvard Studies in Korean Linguistics VII. Cambridge, MA: Department of Linguistics, Harvard University. Hale, K., and S. J. Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In K. Hale and S. J. Keyser, The View from Building 20, 53–110. Cambridge, MA: MIT Press. Hale, K., and S. J. Keyser. 2002. Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT Press. Harley, H. 2002. Possession and the double object construction. Yearbook of Linguistic Variation II, 29–68. Amsterdam: John Benjamins. Hay, J., C. Kennedy, and B. Levin. 1999. Scalar structure underlies telicity in degree achievements. In T. Matthews and D. Strolovitch, eds., SALT 9, 127–144. Ithaca, NY: CLC Publications. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Doctoral dissertation, University of Massachusetts at Amherst. (Published by Garland Press, New York, 1988.)

References

219

Higginbotham, J. 2000. Accomplishments. Unpublished manuscript, Oxford University. Hoekstra, T., and J. Mulder. 1990. Unergatives as copular verbs: Locational and existential predication. Linguistic Review 7:1–79. Im, H.-P. 1975. Bwuceongpep e wa sangtaycinswuluy ko (Infinitives e and Stative ko). Nonmwuncip 8, Kwukmintayhak. (In Korean.) Im, S.-C. 2001. Typological Patterns of Motion Verbs in Korean. Doctoral dissertation, State University of New York at Bu¤alo. Jackendo¤, R. 1983. Semantics and Cognition. Cambridge, MA: MIT Press. Jackendo¤, R. 1990. Semantic Structure. Cambridge, MA: MIT Press. Jackendo¤, R. 1992. Babe Ruth homered his way into the hearts of America. In T. Stowell and R. Oehrle, eds., Syntax and Semantics, vol. 26, 155–178. New York: Academic Press. Jackendo¤, R. 1996. The proper treatment of measuring out, telicity, and perhaps even quantification in English. Natural Language and Linguistic Theory 14:305– 354. Jaeggli, O., and N. Hyams. 1993. On the independence and interdependence of syntactic and morphological properties: English aspectual come and go. Natural Language and Linguistic Theory 11:313–346. Jo, I.-H. 1990. Multi-verb constructions in Korean. In B. D. Joseph and A. M. Swuicky, eds., When Verbs Collide: Papers from the 1990 Ohio State MiniConference on Serial Verbs, 265–287. Columbus: Department of Linguistics, Ohio State University. Kang, S.-Y. 1991. Parametric Head-Licensing in Syntax. Doctoral dissertation, Sogang University, Seoul. Kayne, R. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 47:3–31. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kim, Y.-J. 1995. Verb lexicalization patterns in Korean and some issues of language acquisition. Language Research 31 (3): 501–543. Ko, H. 2005. Syntactic Edges and Linearization. Doctoral dissertation, MIT. Koopman, H. 1997. Prepositions, Postpositions, Circumpositions and Particles: The Structure of Dutch PPs. Unpublished manuscript, UCLA. Kratzer, A. 1996. Severing the external argument from its verb. In J. Rooryck and L. Zaring, eds., Phrase Structure and the Lexicon, 109–137. Dordrecht: Kluwer. Krifka, M. 1989. Nominal reference, temporal constitution and quantification in event semantics. In R. Bartsch, J. van Benthem, and P. van Emde Boas, eds., Semantics and Contextual Expression, 75–115. Dordrecht: Foris. Krifka, M. 1992. Thematic relations as links between nominal reference and temporal constitution. In I. A. Sag and A. Szabolcsi, eds., Lexical Matters, 29–54. Stanford, CA: CSLI.

220

References

Lamiroy, B. 1983. Les Verbes de Mouvement en Franc¸ais et en Espagnol. Linguisticae Investigationes Supplementa. Amsterdam: John Benjamins and Leuven: Leuven University Press. Larson, R. 1988. On the double object construction. Linguistic Inquiry 19:335– 391. Larson, R. 1991. Some issues in verb serialization. In C. Lefebvre, ed., Serial Verbs: Grammatical, Comparative and Cognitive Approaches. Amsterdam: John Benjamins. Lee, C. 1999. A Generative Lexicon Approach to ‘‘-ey’’ and ‘‘-eyse’’ in Korean. In H. Lee et al., eds., To Where Modern Linguistics Goes. Seoul: Hanshin Publishing Company. (In Korean.) Lee, C. 2003. Change of Location: With Reference to Change of State, Creation/ Removal, Degree and Telicity. Unpublished manuscript, Seoul National University. Lee, C., S. Nam, and B.-M. Kang. 1998. A generative approach to the lexical semantics of Korean predicates. Korean Journal of Cognitive Science 9 (3): 19–33. Seoul: Korean Society for Cognitive Science. Lee, J.-H. 1993. Postverbal adverbs and verb movement in Korean. In P. Clancy, ed., Japanese/Korean Linguistics 2. Stanford, CA: SLA/CSLI. Lee, S. 1992. The Syntax and Semantics of Serial Verb Constructions. Doctoral dissertation, University of Washington. Lee, Y. 2003. Two kinds of structural relationships in SVCs. In G. Iverson and Sang-Cheol Ahn, eds., Explorations in Korean Language and Linguistics, 443– 458. Seoul: Hankook Publishing Company. Lefebvre, C. 1991. Take serial constructions in Fon. In C. Lefebvre, ed., Serial Verbs: Grammatical, Comparative, and Cognitive Approaches, 37–79. Amsterdam: John Benjamins. Levin, B., and M. Rappaport. 1990. Wiping the state clean: A lexical semantic exploration. Cognition 41:123–155. Levin, B., and M. Rappaport. 1995. Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge, MA: MIT Press. Li, Y. 1993. Structural head and aspectuality. Language 69:480–504. Longobardi, G. 2000. Postverbal subjects and the mapping hypothesis. Linguistic Inquiry 31:691–702. Lord, C. 1974. Causative constructions in Yoruba. Studies in African Linguistics Supplement 5:195–204. Manfredi, V. 1991. A´gbo` and E´hugbo`: I`gbo Linguistic Consciousness, Its Origins and Limits. Doctoral dissertation, Harvard University. Manfredi, V. 2005. Aspect versus the Serialization Parameter. Unpublished manuscript, Boston University.

References

221

Manfredi, V. Forthcoming. Tense parameters and serial verbs. In E. Aboh and J. Essegbey, eds., Studies in the Syntax of Kwa: A Generative Perspective. Berlin: Mouton de Gruyter. Marantz, A. 1984. On the Nature of Grammatical Relations. Cambridge, MA: MIT Press. Marantz, A. 1993. Implications of asymmetries in double object constructions. In Sam A. Mchombo, ed., Theoretical Aspects of Bantu Grammar 1, 113–151. Stanford, CA: CSLI. Marantz, A. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. In A. Dimitriadis, L. Siegel, C. Surek-Clark, and A. Williams, eds., Proceedings of the 21st Penn Linguistics Colloquium. UPenn Working Papers in Linguistics 4:201–225. Mateu, J. 2002. Argument Structure: Relational Construal at the Syntax Interface. Doctoral dissertation, Universitat Auto`noma de Barcelona. Mateu, J. 2005. Arguing our way to the Direct Object Restriction on English resultatives. Journal of Comparative Germanic Linguistics 8:55–82. Mateu, J., and G. Rigau. 2002. A minimalist account of conflation processes. In A. Alexiadou, ed., Theoretical Approaches to Universals. Amsterdam: John Benjamins. McIntyre, A. 2004. Event paths, conflation, argument structure, and VP shells. Linguistics 42 (3): 523–571. Megerdoomian, K. 2002. Beyond Words and Phrases: A Unified Theory of Predicate Composition. Doctoral dissertation, University of Southern California. Miguel, E. de. 1992. El Aspecto en la sintaxis del espan˜ol: Perfectividad e impersonalidad. Madrid: Ediciones de la Universidad Auto´noma de Madrid. Miyagawa, S., and T. Tsujioka. 2004. Argument structure and ditransitive verbs in Japanese. Journal of East Asian Linguistics 13:1–38. Morimoto, Y. 2001. Los Verbos de Movimiento. Madrid: Visor Libros. Muysken, P. 1988. Parameters for serial verbs. Paper presented at the First NigerCongo Syntax and Semantics Workshop, Boston University. Nishiyama, K. 1998. V-V compounds as serialization. Journal of East Asian Linguistics 7:175–217. Ogihara, T. 1998. The ambiguity of the -te iru form in Japanese. Journal of East Asian Linguistics 7 (2): 87–120. Park, M. 1994. A Morpho-Syntactic Study of Korean Verbal Inflection. Doctoral dissertation, University of Connecticut. Picallo, C. 1985. Opaque Domains. Doctoral dissertation, CUNY, New York. Picallo, C. 1990. Modal verbs in Catalan. Natural Language and Linguistic Theory 8:285–312. Pinker, S. 1989. Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, MA: MIT Press.

222

References

Pylkka¨nnen, L. 2002. Introducing Arguments. Doctoral dissertation, MIT. Ramchand, G. 1995. Aspect and Predication: The Semantics of Argument Structure. Unpublished manuscript, Oxford University. Ramchand, G. 2001. L-Syntax, Selection, and Semantics. Unpublished manuscript, Oxford University. Ramchand, G. 2002. First Phase Syntax. Unpublished manuscript, Oxford University. Rappaport, M., and B. Levin. 1998. Building verb meanings. In M. Butt and W. Geuder, eds., The Projection of Arguments: Lexical and Compositional Factors, 97–134. Stanford, CA: CSLI. Rappaport, M., and B. Levin. 2001. An event structure account of English resultatives. Language 77:766–797. Riemsdijk, H. C. van. 2002. The unbearable lightness of going: The projection parameter as a pure parameter governing the distribution of elliptic motion verbs in Germanic. Journal of Comparative Germanic Linguistics 5:143–196. Riemsdijk, H. C. van. Forthcoming. The morphology of nothingness. In S. Haraguchi, O. Fujimura, B. Palek, and M. Watanabe, eds., Proceedings of the Linguistics and Phonetics Conference (LP2002), Meikai University, Shin-Urayasu, Japan, September 2002. Prague: Karolinum Press. Ritter, E., and S. Rosen. 1998. Delimiting events in syntax. In M. Butt and W. Geuder, eds., The Projection of Arguments, 135–164. Stanford: CSLI. Ritter, E., and S. Rosen. 2000. Event structure and ergativity. In C. Tenny and J. Pustejovsky, eds., Events as Grammatical Objects, 187–238. Stanford: CSLI. Rizzi, L. 1978. A restructuring rule in Italian syntax. In S. J. Keyser, ed., Recent Transformational Studies in European Languages, 113–158. Cambridge, MA: MIT Press. Rizzi, L. 1990. Relativized Minimality. Cambridge, MA: MIT Press. Rochette, A. 1988. Semantic and Syntactic Aspects of Romance Sentential Complementation. Doctoral dissertation, MIT. Rochette, A. 1990. On the restructuring classes of verbs in Romance. In A. Di Sciullo and A. Rochette, eds., Binding in Romance: Essays in Honour of Judith McA’Nulty, 96–128. Ottawa: Special Publication of the Canadian Linguistic Association. Roeper, T., W. Snyder, and K. Hiramatsu. 2002. Learnability in a minimalist framework: Root compounds, merger, and the syntax-morphology interface. In I. Lasser, ed., The Process of Language Acquisition. Frankfurt: Peter Lang Verlag. Rosen, S. 1999. The syntactic representation of linguistic events. Glot International 4 (2): 3–11. Sebba, M. 1987. The Syntax of Serial Verbs. Amsterdam: John Benjamins. ¨ zyu¨rek. 2004. Children creating core properties Senghas, A., S. Kita, and A. O of language: Evidence from an emerging sign language in Nicaragua. Science 305:1779–1782.

References

223

Shopen, T. 1971. Caught in the act. In Papers from the Seventh Regional Meeting of the Chicago Linguistic Society, 254–263. Chicago: Chicago Linguistic Society. Snyder, W. 1995. Language Acquisition and Language Variation: The Role of Morphology. Doctoral dissertation, MIT. Snyder, W. 2001. On the nature of syntactic variation: Evidence from complex predicates and complex word formation. Language 77:324–342. Sohn, H.-M. 1976. Semantics of compound verbs in Korean. Linguistics Journal of Korea 1:142–150. Sohn, H.-M. 2001. The Korean Language. Cambridge: Cambridge University Press. Song, S.-C. 1976. On an abbreviation phenomenon in Korean. Linguistics Journal of Korea 1. Sorace, A. 2000. Gradients in auxiliary selection with intransitive verbs. Language 76:859–890. Sorace, A. 2004. Gradience at the lexicon-syntax interface: Evidence from auxiliary selection and implications for unaccusativity. In A. Alexiadou, E. Anagnostopoulou, and M. Everaert, eds., The Unaccusativity Puzzle, 243–268. Oxford Studies in Theoretical Linguistics. Oxford: Oxford University Press. Spencer, A., and M. Zaretskaya. 1998. Verb prefixation in Russian as lexical subordination. Linguistics 36:1–39. Stewart, O. T. 1996. Adverb placement and the structure of the serial verb construction. Proceedings of NELS 26, 409–423. Cambridge, MA: MIT/Harvard GLSA. Stewart, O. T. 1998. The Serial Verb Construction Parameter. Doctoral dissertation, McGill University. Stowell, T. 1981. Origins of Phrase Structure. Doctoral dissertation, MIT. Strozer, J. 1976. Clitics in Spanish. Doctoral dissertation, UCLA. Suh, C.-S. 1996. Kwuke Mwunpep (Korean Grammar). Seoul: Hanyang University Press. (In Korean.) Suh, Y. 2000. A Study of the Auxiliary Verb Construction and Verb Serialization in Korean. Doctoral dissertation, University of Washington. Sung, K.-C. 1972. Emi -ko wa -e ey Tayhaye (On the su‰xes -ko and e). Kuke Kyoyuk 17–20, Hapbyongho. (In Korean.) Talmy, L. 1985. Lexicalization patterns: Semantic structure in lexical forms. In Timothy Shopen, ed., Language Typology and Semantic Description, Vol. 3: Grammatical Categories and the Lexicon, 36–149. Cambridge: Cambridge University Press. Tenny, C. L. 1987. Grammaticalizing Aspect and A¤ectedness. Doctoral dissertation, MIT. Tenny, C. L. 1994. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht: Kluwer.

224

References

Tenny, C. L. 1995. How motion verbs are special: The interaction of semantic and pragmatic information in aspectual verb meanings. Pragmatics and Cognition 3:31–37. Travis, L. 1984. Parameters and E¤ects of Word Order Variation. Doctoral dissertation, MIT. Travis, L. 2000. Event structure in syntax. In C. Tenny and J. Pustejovsky, eds., Events as Grammatical Objects, 145–185. Stanford, CA: CSLI. Van Hout, A. 1996. Event Semantics in the Lexicon-Syntax Interface: Verb Frame Alternations in Dutch and Their Acquisition. Doctoral dissertation, Tilburg University. Veenstra, T. 2004. Unaccusativity in Saramaccan: The syntax of resultatives. In A. Alexiadou, E. Anagnostopoulou, and M. Everaert, eds., The Unaccusativity Puzzle, 269–287. Oxford Studies in Theoretical Linguistics. Oxford: Oxford University Press. Vendler, Z. 1957. Verbs and times. Philosophical Review 56:143–160. Verkuyl, H. 1993. A Theory of Aspectuality: The Interaction between Temporal and Atemporal Structure. Cambridge: Cambridge University Press. Whitman, J. 2000. Preverbal elements in Korean and Japanese. In G. Cinque and R. Kayne, eds., Handbook of Comparative Syntax. New York: Oxford University Press. Wumbrand, S. 2004. Two types of restructuring: Lexical vs. functional. Lingua 114:991–1014. Wunderlich, D. 1997. Cause and the structure of verbs. Linguistic Inquiry 28:27– 68. Yi, E.-Y. 1997. DE- and RE-constructing Coordination in Korean. Doctoral dissertation, Cornell University. Zagona, K. 2004. Viewpoint Aspect, Telicity, and Boundedness. Unpublished manuscript, Department of Linguistics, University of Washington.

Index

Aske, J., 48, 154, 157 Aux-selection in Italian, 51–52, 168–169, 178–182. See also Restructuring Baker, M., 32–33, 36–38, 73, 74, 92, 97– 98 Beck, S., 45, 127, 136, 152, 189 Bergh, L., 48, 188 Boundedness, 25–28 di¤erence between ka- and -(e)ci construction in temporal boundedness, 112 event boundedness, 25, 154 mismatch between the duration of Reference time and Event time, 26, 87 temporal boundedness, 25–26 Boundedness of directed-motion structure in Dutch and English, 41–42, 129, 131, 132–134 in Korean, 88 in Romance, 49, 154, 169–170 Boons, J.-P., 48, 162 Borer, H., 2, 8, 9, 10–11, 24 Bowerman, M., 83 Butt, M., 170 Carden, G., 149 Cardinaletti, A., 53, 151, 183–184 Carstairs-McCarthy, A., 54 Carstens, V., 36, 108 Case assignment, 15 Case licensing, 24 Category-neutral root, 21 Causation, 146 Causatives, 4–5 causative with unaccusatives, 23 causatives with unergatives, 23 lexical causatives, 23, 25 structure of causatives, 22 Cause-directed motion, 42–44, 141, 167

causative counterpart of ka- ‘go’ in Korean, 99 in Dutch, 44, 143–144 in English, 42–44, 46, 141–143, 144, 145– 146, 147, 148 ‘go-along’ meaning, 145–146 in Korean, 99 in Romance, 49, 167–168 verbs instantiating cause-motion construction, 141 Central coincidence, 17 Chae, H.-R., 84 Choi, S., 83 Chomsky, N., 15, 20, 21, 24, 31 Chung, T., 64 Cinque, G., 51, 53, 182 Clark, E.-V., 54 Classifier covert classifier, 165–166 distance classifier, 131, 158 Collins, C., 32, 98 Composition of manner and directed motion, 45, 54, 127, 134–135, 136 syntactic parameter, 55 Compositional N-N compounding, 44–45 in Germanic, 45, 136 in Romance, 45 Compound Rule, 45, 48, 127, 136, 141, 142, 148, 152, 168, 189 Conflation, 14, 55, 56 Construction, 1, 43 constructional approach, 8, 10, 44 functional-based constructional approach, 8, 10 instantiation of, 44 lexical-based constructional approach, 11 modification of, 44 morphological signature of, 20, 78, 114, 138

226 Construction (cont.) spell-out rule, 137, 148, 149 syntactic approach, 8 Contact verbs, 93 contact verb þ ka-construction, 118 contact verbs in SVC, 94, 105 in English, 46, 142, 143 in Korean, 93, 95, 116, 125 lack of causative meaning in Korean, 96 lack of directed-motion interpretation, 115 as modifier of the cause-motion construction, 143 unergative property of, 93–94 Coordinate structure with -ko in Korean, 60–61, 62–63 biclausal properties of, 60–61, 62–63 comparison with SVC in Korean, 68–69 Covert coordination, 33 Cwu- construction in Korean, 118–120 Davidsonian e(vent)-variable, 170, 171, 172 accomplishment adjunction, 170–171, 173 accomplishment creation, 170, 173 De´chaine, R.-M., 8, 32, 33, 36 de Miguel, E., 153 den Dikken, M., 16, 18, 117, 140 Directed-motion structure, 13, 137, 42–44 directed-motion verbs, 24, 29 in Germanic, 41–48 in Italian, 52 in Romance, 48–54 Directional PP, 134–135, 137 covert directional PP, 138, 147, 148 locality constraint on covert directional PP, 138–140, 147 Double dative-marked arguments, 125 Dowty, D.-R., 28 Durative feature, 30, 102 Edo, 32 Event types, 25 Experiencer verbs, 18, 19 Features C-features, 20, 31–32, 36, 45, 72, 92, 136 FF-features, 20, 31–32, 34, 36, 72, 74–75, 92 P-features, 20, 31–32, 36, 45, 72, 74, 92, 136 Fillmore, C. J., 1 Final -e deletion, 183 functional R-verbs, 183–184 semifunctional R-verbs, 183–184 Folli, R., 2, 7, 8, 11–12, 16, 48, 49, 139, 145, 146, 152, 168, 170–175, 188

Index Fong, V., 7 Fox, D., 21 Freeze, R., 18 Full decompositional approach, 17 General rule of syntactic composition, 15 Generalized Transformation (GT) Rule, 34, 72, 92, 136 Giusti, G., 151 Goal construction, 118, 123–124 Goldberg, A., 1, 15, 20, 42–44, 89, 135, 138, 142, 147 Government, 140 extended government, 147 Gruber, J.-S., 15 Hale, K., 1, 2, 8, 11, 13–17, 23, 54, 55, 139, 170 Harley, H., 2, 117, 145, 146 Hay, J., 28 Head-final language, 39 Headness, 36, 50, 74 morphosyntactic head, 36, 40, 50, 74, 92 semantic head, 36, 50, 92, 93 Heim, I., 11 Higginbotham, J., 170 Hiramatsu, K., 45, 136 Hoekstra, T., 2, 132, 133, 143 Hyams, N., 149, 150 Igbo, 3 Im, H.-P., 63 Im, S.-C., 79, 83 Inclusiveness Condition, 31 Interpretation of A, 18 Interpretation of Dutch postposition, 140 Jackendo¤, R., 15, 42–44, 89, 135, 144 Jaeggli, O., 149, 150 Kayne, R., 18, 108 Kennedy, C., 28 Keyser, J., 1, 2, 8, 11, 13–17, 23, 54, 55, 139, 170 Kita, S., 3 Ko, H., 74 Koopman, H., 16, 139 Kratzer, A., 15 Krifka, M., 15 Lamiroy, B., 48, 161 Larson, R., 13, 33 l-structures, 13, 20, 21 extended l-structures, 22, 35 Lee, C., 84 Lee, S., 61, 62, 69

Index Lee, Y., 62 Levin, B., 1, 4, 6–7, 10, 13, 28, 135 Lexical approach, 4 Argument Realization Condition, 6 Canonical Realization Rule (CRR), 5 Event-structure template, 4–5 linking Rule, 6 projectionist approach, 10 Subevent Identification Condition, 5 template Augmentation, 7, 10, 12 Light verb, 45 (e)ci-, 30, 109–115 empty light verb, 47, 53–54, 56 go/come, 20, 44, 46 invariant go/come, 149–150 ka-, 30, 40, 78–81, 91 Linear Correspondence Axiom (LCA), 108 Li, Y., 37, 39, 68, 109 Locative PPs lexical restriction on licensing directional reading with, 139, 140–141 Lord, C., 97 Low applicative construction, 123–124 applicative verb, 95, 117, 118 prospective possessor, 95, 117, 118 McIntyre, A., 55, 56 Manner-of-motion verbs, 2, 130 the lexical approach, 7–8 unaccusative properties of, 129, 132 unergative properties of, 128, 132 Manner-of-motion verbs in Germanic, 11, 128–135, 187 in Dutch, 2, 131–135, 175–176 Manner-of-motion verbs in Korean, 40, 84–90 unergative property of, 84–85 Manner-of-motion verbs in Romance, 48– 54, 152–155, 188 in French, 161–167 in Italian, 3, 13, 49 in Spanish, 152, 154–155 unaccusative properties of Spanish manner-of-motion verbs, 160–161 three classes in Italian, 168–170, 174–175 Marantz, A., 14, 20, 117, 118 Mateu, J., 2, 31, 42, 48, 55–56, 154 Measure verbs, 18, 19 Megerdoomian, K., 24 Merge, 31, 45–47, 56 Minimality Condition, 36 Miyagawa, S., 125 Morimoto, Y., 48, 153, 154, 159 Mulder, J., 2, 132, 133, 143 Muysken, P., 37, 68

227 Negation in Korean, 58 Nicaraguan Sign Language (NSL), 3 Nishiyama, K., 33, 39 Path, 78, 84 absolute endpoint, 111, 112 abstract path, 78–81, 110 physical path, 81 relative endpoint, 111, 112 Path-denoting PP, 16, 138–139 adjunct analysis (unbounded interpretation), 129, 132, 178 complementation analysis (bounded interpretation), 129, 132, 178 path-denoting adjunct PPs vs. complement PPs, 129–130, 133 Path verbs in Korean, 81–84, 101–108 aspectual property of causative path verbs, 102 aspectual property of intransitive path verbs, 83 Pesetsky, D., 21, 149 Phases, 21, 46, 137, 143 phasal extension, 53, 181–182 Picallo, C., 51, 182 Pinker, S., 1, 54, 111 Polysemy, 13, 128 Possessor verbs, 18, 19 Poulin, C., 7 Preposition, 12 morphologically complex preposition, 12, 16 (see also Path-denoting PP) morphologically simple preposition, 12 Pylkka¨nnen, L., 118 Ramchand, G., 2, 8, 11, 170 Rappaport, H., 1, 4, 6–7, 10, 13, 135 Restructuring, 51, 181 auxiliary switch, 51, 181 Dutch modals, 53–54 restructuring verbs (R-verbs) in Italian, 51, 51–52, 181–182 Rigau G., 2, 42, 48 Ritter, E., 2, 8, 9–10, 24 Rizzi, L., 51, 130, 181, 182 Rochette, A., 51, 182 Roper, T., 45, 136 Rosen, S., 2, 8, 9–10, 24 Senghas, A., 3 Serial Verb Construction (SVC), 3, 8, 32– 41, 57–64, 75–77, 92, 96, 100, 105, 108 consequential SVC (CSVC), 33, 36, 75–76 lack of the RSVC in Korean, 97, 98–99, 109 (see also Resultative SVC) monoclausal property of, 61–62

228 Serial Verb Construction (SVC) (cont.) noncompound properties of Korean SVCs, 64–67 contrastive markers, -nun, -man, -to, 64– 66 distributive (plural) marker -tul, 66–67 resultative SVC (RSVC), 33, 37–40 semantic relation within CSVC, 68, 69–70, 75, 108 semantic relation within SSVC, 41, 70–71, 76–77, 108 simultaneous SVC (SSVC), 33, 40–41, 76– 77, 97 Sholonsky, U., 53, 183–184 Shopen, T., 150 Snyder, W., 44–45, 47, 127, 136, 141, 152, 189 Sohn, H.-M., 63, 99 Song, S.-C., 69 Sorace, A., 51 Sound-of-emission verbs, 89, 135, 144 Spell-out, domain of, 21, 46, 91, 137 Spencer, A., 42, 47 S-syntax, 13 Stewart, O.-T., 32–33, 37–38, 73, 74, 92, 97, 98 Stowell, T., 140 Strozer, J., 51, 182 Subject-to-subject raising, 19 Subordinate structure with -se in Korean, 59–60 biclausal properties of, 59–60 comparison with SVC in Korean, 71 covert subordination, 63–64 Suh, C.-S., 71 Sung, K.-C., 69 Talmy, R., 48 Telicity, 25, 28 Temporal Iconicity Condition (TIC), 37, 39, 68–69, 72, 98, 99, 108, 109 Tenny, C., 14, 15, 23 The way construction, 43, 144–145 Travis, L., 24 Tsujioka, T., 125 Unaccusative verb, 13, 15–20 l-structure of, 22 Unergative verb, 13–15 l-structure of, 14, 22, 24–25 Uniqueness principle, 111 van Riemsdijk, H.-C., 53, 56 Vendler, Z., 25, 28 Wumbrand, S., 53

Index Yi, E.-Y., 61, 62, 64 Zagona, K., 25–26 Zarestskaya, M., 42, 47