Constructions: A Construction Grammar Approach to Argument Structure

C

o n s t r u c t io n s

Cognitive Theory o f L an gu age and Culture A series edited by G illcs Fauconnier. G eorge Lakoff, and Eve Sweetser

Co n st r u c t io n s A Construction Grammar Approach to Argument Structure

Adele E. Goldberg

The University of Chicago Press Chicago and London

A dele E. G oldberg is assistant professor o f linguistics at the University or California, San Diego.

The University of Chicago Press, Chicago 60637 The University o f Chicago Press, Ltd., London K 1995 by The University o f Chicago All rights reserved. Published 1995 Printed in the United Stales of America

04 03 02 01 00 99 98 97 96 95

5 4 3 2 1

ISBN (cloth): 0-226-30085-4 ISBN (paper): 0-226-30086-2

Library o f Congress Cataloging- in-Publication Data Goldberg, Adele E. Constructions : a construction grammar approach to argument structure / Adele E. Goldberg. p. cm. — (Cognitive theory of language and culture) Originally presented as the author's thesis (Ph.D.)— University of California, 1992. Includes bibliographical references (p.) and index. ISBN 0-226-30085-4 (cloth). — ISBN 0-226-30086-2 (pbk.) I. Grammar, Comparative and general— Syntax. 2. Semantics. 3. Generative grammar. 1. Title. II. Series. P29I.G 65 1995 4 15— dc20 94-20705 CIP

© The paper used in this publication meets (he minimum requirements of the American National Standard for Information Scicnccs— Pcrmancncc of Paper for Printed Library M aterials, ANSI Z39.48-1984. This book is printed on acid-free paper.

ToAli

Contents

Acknow ledgm ents t

ix

Introduction

1

1 .1 T h e C o n c e p t o f C o n s tru c tio n s

I

1.2 A B rie f In tro d u c tio n to C o n s tru c tio n G ra m m a r

6

1.3 A n A ltern ativ e A cco u n t: L e x ic o se m a n tic R u le s

7

1.4 A d v an tag es o f th e C o n s tru c tio n A c c o u n t

9

1.5 T ra d itio n a l M o tiv a tio n s for L exical R u les

2

21

The Interaction between Verbs and Constructions 2.1 F ra m e se m a n tic s

2 .2 T h e N atu re o f V erb M e an in g

27

2 .3 T h e N a tu re o f C o n s tru c tio n a l M e a n in g

31

2.4 T h e In teg ratio n o f V erb a n d C o n s tru c tio n

43

2.5 P o ssib le R e la tio n s b etw een V erbs an d C o n s tru c tio n s 2 .6 C o n c lu sio n

3

59

66

Relations am on g Constructions

67

3.1 R elev an t P sy ch o lo g ica l P rin c ip le s o f L a n g u a g e O rg a n iz a tio n 3.2 M o tiv atio n

3 .4 R e la tin g P a rtic u la r C o n s tm c tio n s 3 .5 M u ltip le In h e rita n c e

72

81

97

3.6 In h e rita n c e w ith in C o n s tru c tio n s 3.7 C o n c lu sio n

98

99

O n Linking

10 1

4.1 T ra n s fo rm a tio n a l A p p ro a c h e s o f A rg u m e n t S tru c tu re 4 .2 G e n e ra liz a tio n s a c ro ss C o n s tru c tio n s 4 .3 C o n c lu s io n

5

101

108

1 19

Partial Productivity 5.1 In tro d u c tio n

120

120

5.2 In d irect N eg ativ e E v id e n c e 5 .3 C irc u m sc rib in g V erb C la sse s 5.4 E x c e p tio n s

122 125

129

5.5 A cco u n tin g fo r the E x c e p tio n s: A U sa g e -b a se d A c c o u n t 5.6 C o n c lu sio n

67

69

3.3 R e p re se n tin g M o tiv atio n by In h e rita n c e

4

24

25

133

139

v ii

viii

6

Contents

The English Ditransitive Construction 6.1 Introduction

6.2 The Existence of the Construction 6.3 The Semantics 6.4 Conclusion

7

151

7.3 The Various Interpretations 7.4 Semantic Constraints

153

161

164

7.5 The Load/Spray Alternation 7.6 Conclusion

175

179

The English Resultative Construction 8.1 Introduction

8.3 Middle Formation 8.4 Other Accounts

182

183 185

8.5 The Existence of a Resultative Construction

188

8.6 Constraints on the Resultative Construction

193

198

The Way Construction 9.1 Introduction

199

199

9.2 The Existence of the Construction

199

9.3 The Semantics of the Way Construction 9.4 Semantic Constraints 9.6 Relation to Resultatives 9.7 Conclusion

214

215

217

Conclusion

2 19

10.1 Other Constructional Approaches 10.2 Summary

202

212

9.5 The Lexical Complex Predicate Approach

10

180

180

8.2 The Status of the Postverbal N P

8.7 Conclusion

152

152

7.2 The Existence of the Construction

9

141

142

The English Caused-Motion Construction 7.1 Introduction

8

14 I

141

219

224

Notes

229

Bibliography

243

Index

261

Acknowledgments

This book grew out o f my Ph.D. thesis (G oldberg 1992b), which was completed at the University o f California, Berkeley. An enorm ous debt is owed to my advisor, George Lakoff, for his w isdom , enthusiasm , and encourage­ ment, his ever-ready exam ple and counterexam ple, and for sharing his time and his deep insights with incredible generosity. I'd like to thank Charles Fillmore for instilling in me a deep respect for the com plexities o f the data, and for sharing his wisdom. His enduring insights have profoundly influenced this work in innum erable ways. I ’m also grateful for his spearheading the developm ent of the theory o f C onstruction Gram m ar, on which the present work is based. Work in Construction G ram m ar includes, for exam ple, Fillm ore, Kay and O 'C onnor’s analysis o f the let alone and the more, the m errier constructions (1988), Brugman’s analysis o f have constructions (1988), K ay’s work on even (1990), the “ W hat, me w orry?” construction o f Lam brecht (1990). and Sweetser’s analysis o f modal verbs (1990). C onstruction G ram m ar is also developed in R llm ore (1985b, 1987. 1988. 1990), Fillm ore & Kay (1993), Filip (1993), Jurafsky (1992), Koenig (1993), and M ichaelis (1993). The present work owes its greatest debts to Lakoff's in-depth study of there constructions (1984) and to Fillmore (1987), who suggested that the m eaning o f an expression is arrived at by the superim position o f the meanings o f open class words with the m ean­ ings o f gramm atical elements. I’m grateful to Dan Slobin for his encouragem ent and guidance, and for providing a reality check on the plausibility of psychological claim s. In the final stages o f w riting my dissertation, I was fortunate enough to work closely with Annie Zaenen. 1 am imm ensely grateful for her advice, her many leads to relevant literature, and for our many interesting and helpful discussions, which have deeply influenced my work. Other members of the Berkeley faculty contributed in countless ways to my education. Eve Sweetser tirelessly read and offered valuable com m ents on many papers; Paul Kay provided much helpful input, and was consistently w ill­ ing to lend an ear and a critical eye; R obert W ilensky offered many helpful discussions and some wonderful data. Len Talmy was always w illing to discuss all manner o f ideas. Visitors Don Form an, Knud Lam brecht, M inoko Nakau, IX

x

A c k n o w le d gm e n ts

Frederika Van der Leek, and Robert Van Valin offered different perspectives and very helpful discussions. I’d like to offer personal thanks to C laudia Brugman. M ichele Emanatian, Hana Filip, Jean-Pierre Koenig, and Laura M ichaelis for support o f every kind, including countless enlightening discussions on topics related to alm ost every aspect o f this m onograph. I’m also grateful to Jess Gropen, Beth Levin. Steve Pinker, and Ray Jackendoff, for theirow n inspirational work and for their help­ ful feedback and discussion. D uring the w riting and rewriting o f this m anuscript, I was able to spend a good deal of tim e at Berkeley, Xerox Palo Alto Research Center, Stanford, and the University o f California. San Diego, so there are many people to thank for very helpful suggestions and discussions, including Farrell Ackerm an, Joan Bresnan, Tony Davis, Jane Espenson, G illes Fauconnier. Joe Grady, Marti H earst, Kyoko Hirose, Rolf Johnson, Dan Jurafsky. Suzanne Kemmer, Yuki Kuroda, Ron Langacker, M aarten Lem mens, John M oore, Terry Regier, Hadar Shem-Tov, Eve Clark, C leo Condoravdi, Mark Gaw ron, Jess Gropen, Geoff N unberg, Ivan Sag, Tom Wasow, Ali Yazdani, and Sandro Zucchi. Several UCSD students carefully read the m anuscript and made very helpful sugges­ tions, particularly Kathleen A hrens, M ichael Israel, and Bill Morris. For help preparing the manuscript I would like to thank Kathleen Ahrens, Bill Byrne, and N itya Sethuram an. Finally, for editorial assistance I thank G eoff Huck and Karen Peterson, and for the m ost careful, w ell-informed copy-editing 1 could have imagined, I thank C hristine Bartels. E xcerpts o f this book first appeared, in different form, as articles or book chapters. I thank the publishers for perm ission to include revised material from: “ The Inherent Sem antics o f A rgum ent Structure: The Case o f the English D itransitive C onstruction,” Cognitive Linguistics 3 ( l) : 3 7 - 7 4 , 1992; “ A Se­ m antic Account o f Resultatives,” Linguistic A nalysis 2 1 :6 6 - 9 6 , 1991; “ It Can’t Go Down the Chim ney Up: Paths and the English Resultative,” BLS 17; “ M aking O ne’s Way Through the D ata,” in A. Alsina, J. Bresnan, and P. Sells (eds.), Com plex Predicates, CSLI Publications, forthcoming; “ Another Look at Some Learnability Paradoxes,” in Proceedings o f the 25th A nnual Stanford Child Language Research Forum. CSLI Publications. For providing financial support, com fortable offices, and stimulating envi­ ronm ents, I’d like to thank the Sloan Foundation, who funded the Cognitive Science Institute at Berkeley, the International Com puter Science Institute (1CSI), the C enter for the Study o f Language and Information (CSLI), and Xerox PARC. A crucial debt is owed to my family: my mom, Ann Goldberg, for being a voice o f reason on topics related to this monograph and on all others; my sib­

A c k n o w le d g m e n ts

xi

lings, Ken G oldberg and Elena G oldberg Man, and my grandparents. Harry and Birdie Goldberg and Rose W allach, for their consistent loving support, and just for being themselves. I am also deeply grateful to the memory of my father, Melvin Goldberg, for his unparalleled courage, curiosity, and com passion. Finally, I am im m ensely grateful to Ali Yazdani, for always being there, even though there have been many miles between us. This book is dedicated to him.

1 Introduction

i. i

Th e C o n c e p t

of

C o n s t r u c t io n s

What is it children learn when they learn to speak a language? W hat is the nature of verb m eaning and what is its relation to sentential m eaning? How and 10 what extent are novel utterances based on previously learned utterances? Those questions are addressed here through a study o f basic sentence typc.s— the “ simple sentences” of traditional gram m arians. A central thesis of this work is that basic sentences of English are instances o f constructions— form -m eaning correspondences that exist independently o f particular verbs. Fhat is. it is argued that constructions them selves carry meaning, indepen­ dently of the words in the sentence. The notion construction has a tim e-honored place in linguistics. Traditional gramm arians have inevitably found it useful to refer to properties of particular constructions. The existence of constructions in the gram m ar was taken to be a self-evident fact that required little com m cnt. In the early stages o f transfor­ mational gram m ar (Chom sky 1957, 1965), constructions retained their central role, construction-specific rules and constraints being the norm. In the past two decades, however, the pretheoretical notion o f construction has com e under attack. Syntactic constructions have been claim ed to be epiphenom enal, arising solely from the interaction of general principles (C hom sky 1981, 1992); the rejection o f constructions in favor o f such general principles is often assumed now to be the only way to capture generalizations across patterns. At the same time, the rising tide of interest in sem antic and pragm atic prop­ erties has led to a renewed focus on the idiosyncratic properties of particular sentence patterns (cf. Levin 1993, for exam ple). In order to reconcile the theo­ retical desire for construction-independent principles with the em pirical neces­ sity of recognizing pattern-specific properties, all such idiosyncratic properties have been attributed to individual lexical items, lexical entries being the last refuge of the idiosyncratic. There is no question that a large am ount of inform ation is contributed by individual lexical items (cf. chapters 2 and 5). However, in this work it is ar­ gued that an entirely lexically-based, or bottom -up, approach fails to account for the full range of English data. Particular sem antic structures together with their associated formal expression m ust be recognized as constructions inde­ pendent o f the lexical items which instantiate them. i

2

Chapter O n e

T his m onograph thus represents an effort to bring constructions back to their rightful place on center stage by arguing that they should be recognized as theoretical entities. Single-clause patterns hold a special interest because these cases clearly lie at the heart o f any theory o f grammar. If it can be shown that constructions are essential to a description o f the domain o f simple clauses, then it must be recognized that constructions are crucial to the description o f lan­ guage. Chapters 3 and 4 argue that em pirical generalizations across construc­ tions can in fact naturally be captured within a construction-based framework. A nother goal of this monograph is to explicate the semantics associated with particular clausal patterns. T he semantic properties to be discussed must be accounted for by any framework, regardless of where the semantics is encoded or w hat o n e’s assum ptions about the lexicon and syntax are. It has long been recognized that differences in com plem ent configuration arc often associated with differences in meaning. For example, the dilransitive re­ quires that its goal argum ent be anim ate, while the same is not true o f para­ phrases with to: (1)

a. b.

I brought Pat a glass o f water, (ditransitive) I brought a glass of water to Pat.

(2)

a. *1 brought the table a glass o f water, (ditransitive) b. I brought a glass o f water to the tabic. (Partee 1965:60)

Fillm ore (1968, fn. 49) noted that sentences such as the following differ in meaning: (3)

a. b.

Bees are sw arm ing in the garden, The garden is swarm ing with bees.

(3b) suggests that the w hole garden is full o f bees, whereas (3a) could involve bees in only a part o f the garden. A nderson (1971) observed that the follow ing sentences also differ in meaning: (4)

a. b.

I loaded the hay onto the truck, I loaded the truck with the hay.

W hile (4b) implies that the truck is entirely filled with hay (or at least relevantly affected), no such im plication exists in (4a). W orks by G reen, Oehrle, Bolinger, Borkin, and W ierzbickaand by Interpre­ tive Sem anticists such as Chom sky, Partee, and Jackendoff have drawn atten­ tion to system atic differences in m eaning between sentences with the same lexical items in slightly different constructions.' Borkin (1974), for exam ple, provides the following contrast:

Introduction

(5)

3

a. When 1 looked in the files, 1 found that she was M exican. b. ?W hen I looked in the files I found her to be M exican. c. *W hen I looked in the files I found her M exican.

Borkin argues that the pattern in (5c) is only possible with verbs o f proposition when the proposition expressed is considered to be a m atter of judgm ent, as opposed to a m atter o f fact. The pattern in (5b) prefers but does not require the proposition to express judgm ents, and the full clausal form with thatcom plem entizer in (5a) freely allows matters o f judgm ent or fact. W ierebicka (1988) contrasts (6a) and (6b): (6)

a. b.

I am afraid to cross the road. 1 am afraid of crossing the road.

Only in (6a) is the speaker presumed to have som e intention o f crossing the road. This difference in interpretation is argued to account for why (7a) is in­ felicitous unless the falling is interpreted as som ehow volitionally intended:2 (7)

a. #1 am afraid to fall down. b. I am afraid o f falling down.

Similar observations o f subtle differences in m eaning led B olinger to con­ clude: “ A difference in syntactic form always spells a difference in m eaning” (1968:127). T he same hypothesis— which we may term the Principle of No Synonymy of G ram m atical Form s— has been form ulated by Givon (1985), K irsn er(1985),L angacker(l985), Clark (1987), and W ierzbicka(1988). It will be adopted here as a w orking hypothesis.3

In this monograph, 1 explore the idea that argum ent structure construc­ tions are a special subclass o f constructions that provides the basic means of clausal expression in a language/ Exam ples o f English argum ent structure co n ­ structions to be discussed here include the following: 1. Ditransitive

X

2. Caused Motion

X c a u s e s Y to

m ove

3. Resultative

X c a u s e s Y to

becom e

4. Intrans. Motion

X

causes

m oves

Y to

Y

r e c e iv e

Z

Z

Z

Subj V Obj Obj2 Pat faxed Bill the letter. Sub V Obj Obi Pat sneezed the napkin off the table. Subj V Obj Xcomp She kissed him unconscious. Subj V Obi The fly buzzed into the room.

4

C hapter O n e

5. Conativc

X

d ir e c t s a c t io n

at Y

Subj V Oblu, Sam kicked at Bill.

On a constructional approach to argum ent structure, systematic differences in m eaning between the same verb in different constructions are attributed di­ rectly to the particular constructions. We will see that if we consider various constructions on their own term s, interesting generalizations and subtle seman­ tic constraints em erge. Several constructions can be shown to be associated with a fam ily o f distinct but related senses, much like the polysemy recognized in lexical items. Moreover, these constructions them selves are shown to be interrelated. The analysis I am going to propose draws on research in Construction G ram ­ m ar (cf. Fillm ore 1985b, 1987, 1988, 1990; Fillmore & Kay 1993; Lakoff 1987; Fillmore, Kay & O ’C onnor 1988; Brugman 1988; Kay 1990; Lambrecht 1990, 1994; G oldberg L991a, 1992a; M ichaelis 1993; Koenig 1993; Filip 1993). A ccording to Construction Grammar, a distinct construction is defined to exist if one or more of its properties are not strictly predictable from knowl­ edge o f other constructions existing in the gram m ar:5 C is a c o n s t r u c t io n iff*, C is a form -m eaning pair < F„ S ,> such that some aspect of F, or some aspect of S, is not strictly predict­ able from C ’s component parts or from other previously established constructions. C onstructions are taken to be the basic units o f language. Phrasal patterns are considered constructions if som ething about their form or m eaning is not strictly predictable from the properties o f their com ponent parts or from other constructions.6 T hat is, a construction is posited in the gram m ar if it can be shown that its m eaning and/or its form is not com positionally derived from other constructions existing in the language (cf. section 1.2). In addition, ex­ panding the pretheoretical notion o f construction somewhat, morphemes are clear instances o f constructions in that they are pairings o f meaning and form that are not predictable from anything else (Saussure I9 I6 ).7 It is a conse­ quence o f this definition that the lexicon is not neatly differentiated from the rest o f grammar. Constructions can be understood to correspond to the “ listem es” o f DiSciullo and W illiam s (1987)— that is, the entities o f gram m ar that m ust be listed. However, our view o f the collection o f listem es is radically different from theirs. They state categorically: “ If conceived of as the set of listemes, the lexicon is incredibly boring by its very nature. It contains objects o f no single specifiable type (words, VPs, m orphem es, perhaps intonational patterns.

Introduction

b

and so on), and those objects that it does contain are there because they fail to conform to interesting laws. The lexicon is like a prison— it contains only the lawless, and the only thing that its inm ates have in com m on is law lessness” (p. 3). This view o f the lexicon, or w hat might be better term ed the constructicon, following Jurafsky (1992), is rejected in the present work. The collection of constructions is not assum ed to consist o f an unstructured set of independent entities, but instead it is taken to constitute a highly structured lattice o f inter­ related information. The reb u o n s between constructions are discussed in ch ap ­ ters 3 and 4. A basic axiom that is adopted is: know ledge o f language is knowledge. Many o f the findings of the follow ing chapters are thus expected, particularly that linguistic constructions display prototype structure and form networks o f associations. Hierarchies o f inheritance and sem antic networks, long found useful lor organizing other sorts o f knowledge, are adopted for explicating our linguistic knowledge (cf. Q uillian 1968; Bobrow & W inograd 1977; Fahlman

VST)9, 'WWensV^ \9?>6, Norv'ig &. LaVoff

lurafsVy \99T).

On the basis of research on language acquisition by Clark (1978), Slobin (1985), and Bowerman (1989), together with the findings presented here, it is hypothesized that Simple clause constructions are associated directly with semantic structures which reflect scenes basic to human experience.’ In particular, constructions involving basic argum ent structure are show n to be associated with dynam ic sccnes: experientially grounded gestalts, such as that of someone volitionally transferring som ething to som eone else, som eone causing som ething to move or change state, som eone experiencing som ething, som ething moving, and so on. It is proposed that the basic clau.sc types of a language form an interrelated network, with sem antic structures paired with particular forms in as general a way as possible.

♦ This book is structured as follows. The rest o f this chapter presents ar­ guments for adopting a constructional approach to argum ent structure. Chapter 2 analyzes the nature of verb m eaning, the nature of constructional meaning, and the relation between the two. C hapter 3 suggests an account of how to capture relations am ong constructions and generalizations across co n ­ structions; an inheritance hierarchy o f constructions is posited, and the inheri­ tance links themselves are treated as objects in the system. In chapter 4, the idea of a monostratal theory is defended, and the way linking generalizations are to be captured within a constructional approach is discussed. C hapter 5

6

C hap ter O n e

presents an account o f the partial productivity o f constructions; this work adapts insights from Pinker (1989) to a system w ithout lexical rules. Chapters 6 - 9 involve more specific analyses o f several English construc­ tions: the ditransitive construction (e.g., C hris fa x e d her the news), the “ caused-m otion" construction (e.g., Sally sneezed the napkin o ff the table), the resultative construction (e.g., Sam talked him self hoarse), and the way con­ struction (e.g., B ob elbow ed his way through the crowd). Specific arguments for the existence o f each o f these constructions are given in those chapters. 1.2

A B r ie f I n t r o d u c t i o n

to

C o n s t r u c t io n G

rammar

The basic tenet o f C onstruction G ram m ar as developed in Fillm ore & Kay 1993, Fillm ore, Kay & O ’C onnor 1988, Lakoff 1987, Brugman 1988, Lam brecht 1994, is that traditional constructions— i.e., form -m eaning corre­ spondences— arc the basic units o f language. T heorists w orking within this theory share an interest in characterizing the entire class o f structures that make up language, not only the structures that are defined to be part o f “ core gram m ar." This interest stem s from the belief that fundam ental insights can be gained from considering such non-core cases, in that the theoretical m achinery that accounts for non-core cases can be used to account for core cases. In addition, much o f actual corpus data involves such non-core cases. C onstruction G ram m arians also share an interest in accounting for the conditions under which a given construction can be used felicitously, since this is taken to be part o f speakers’ com petence or know ledge of lan­ guage; from this interest stems the conviction that subtle semantic and prag­ m atic factors are crucial to understanding the constraints on gramm atical constructions. These tenets, which in m any respects hearken back to G enerative Semantics (e.g. L akoff 1965, 1970a,b, 1971, 1972,1976; Lakoff & Ross 1976; Langacker 1969; Postal 1971;D ow ty 1972; Keenan 1972; M cCawley 1973,1976) are also shared by the theory o f C ognitive G ram m ar (Langacker 1987a, 1991), the framework im plicit in much o f W ierzbicka’s work (e.g., W ierzbicka 1988), and by many functionalist approaches to gram m ar (e.g., B olinger 1968; DeLancey 1991; Giv6n 1979a,b: Haim an 1985a; Foley & Van Valin 1984). Work in G en­ eralized Phrase Structure G ram m ar (GPSG ) and in Head-Driven Phrase Struc­ ture G ram m ar (HPSG ) (G azdar et al. 1985; Pollard & Sag 1987, 1994) also em phasizes the central role o f the sign in gramm ar. In many ways, aspects o f the proposals made here are also com patible with recent work by Levin (1985), Levin & Rapoport (1988), Pinker (1989) and Jackendoff (1990a). Some sim i­ larities and differences are discussed below. Ow ing in part to the fact that Construction G ram m ar has grown largely out

'n troau ccion

7

of work on frame semantics (Fillm ore 1975, 1977b, 1982, 1985a) and an experientially based approach to language (Lakoff 1977, 1987), the approach to semantics that is adopted by the theory is one that crucially recognizes the importance of speaker-centered “construals” of situations in the sense of Langacker (1987a, 1991). This approach to sem antics is discussed in chapter 2.

♦ In Construction Grammar, no strict division is assum ed between the lexi­ con and syntax. Lexical constructions and syntactic constructions differ in in­ ternal complexity, and also in the extent to which phonological form is specified, but both lexical and syntactic constructions are essentially the same type of declaratively represented data structure: both pair form with meaning. It is not the case, however, that in rejecting a strict division. Construction Gram m ar denies the existence o f any distinctly m orphological or syntactic con­ straints (or constructions). Rather, it is claim ed that there are basic com m onali­ ties between the two types o f constructions, and moreover, that there are cases, such as v erb -p article com binations, that blur the boundary. A nother notion rejected by Construction G ram m ar is that o f a strict divi­ sion between semantics and pragm atics. Inform ation about focused constitu­ ents, topicality, and register is represented in constructions alongside semantic information. Construction G ram m ar is generative in the sense that it tries to account for the infinite num ber of expressions that are allowed by the gram m ar w hile at­ tempting to account for the fact that an infinite num ber of other expressions are ruled out or disallowed. C onstruction G ram m ar is not transform ational. No underlying syntactic or sem antic forms are posited. Instead, C onstruction G ram m ar is a monostratal theory o f gram m ar like many other current theories, including Lexical Functional G ram m ar (LFG ) (Bresnan 1982), Role and Ref­ erence G ram m ar (Foley & Van Valin 1984). GPSG (G azd aret al. 1985), HPSG (P o llard & S ag 1987, 1994), and C ognitive G ram m ar (Langacker 1987a, 1991). The rationale for this and some consequences are discussed in chapter 4. It is perhaps easiest to explore the constructional approach by first contrast­ ing it with the relevantly sim ilar proposal described in the follow ing section. 1.3

A

n

A

l t e r n a t iv e

A

c c o u n t : L e x ic o s e m a n t ic

R ules

The recognition o f subtle sem antic differences between related syntactic (subcategorization) frames has been growing, and there has also been increas­ ing focus on the fact that there appears to be a strong correlation between the meanings of verbs and the syntactic fram es they can occur in, leading many researchers to speculate that in any given language the syntactic subcategori­

8

Chapter O n e

zation frames o f a verb may be uniquely predictable from the verb's lexical sem antics (e.g.. Levin 1985; Chom sky 1986; C arter 1988; Levin & Rapoport 1988; Rappaport & Levin 1988; Pinker 1989; Gropen et al. 1989). T he following factors have led these theorists to postulate lexical rules which are designed to operate on the sem antic structures o f lexical items: ( I) overt com plem ent structure appears to be predictable by general linking rules that map sem antic structure onto syntactic form, and (2) the same verb stem often occurs with more than one com plem ent configuration. For exam ple. Pinker (1989) proposes that the prepositional/ditransitive al­ ternation (the “ dative" alternation) results from a semantic rule rather than being the product o f a syntactic transform ation. Specifically, he suggests that productive use o f the ditransitivc syntax is the result o f a lexicosem antic rule w hich takes as input a verb with the sem antics ‘X

causes

Y to

go to

Z ’ and

produces the sem antic structure X c a u s e s Z to h a v f Y \ The double object syntax, he argues, is then predictable from near-universal linking rules m ap­ ping the argum ents o f a verb with the m eaning X c a u s e s Z to h a v e Y ' into the ditransitive form . In this wav. Pinker argues that the dative rule produces a “ conceptual gestalt shift,” — that it is, in effect. a semantic operation on lexical structure (cf. also Gropen et al. 1989). T he general approach can be outlined as follows: la. The syntactic com plem ent configuration o f a clause is taken to be uniquely predictable from the sem antic representation o f the matrix verb. The m apping from sem antic representations to particular com plem ent configurations is perform ed via universal, or near-universal, linking rules. lb . Different syntactic com plem ent configurations therefore reflect differ­ ences in the sem antic representations of the main verb. 2. Different sem antic representations o f a particular verb stein, i.e., different verb senses, are related by generative lexical rules which take as input a verb with a particular semantics and yield as output a verb with a differ­ ent semantics. 3. D ifferences in sem antics are not necessarily truth-functional differences, but may represent a different construal of the situation being described: that is, the relevant sem antics is speaker-based. These principles are detailed m ost explicitly in Pinker 1989. but are also shared by Levin 1985, Levin & Rapoport 1988, and Gropen ct al. 1989. By postulating rules that operate on sem antic structure, as opposed to rules or transform ations that arc purely or prim arily syntactic, these theories manage to incorporate im portant insights. As was discussed above, different construc­ tions are typically, possibly always, accom panied by slightly different sem an­ tic interpretations: these sem antic differences are respected as soon as the

Introduction

9

forms are learned (Bowerman 1982; Gropen et al. 1989). By postulating sem antics-changing rules, as opposed to syntactic rules with additional sem an­ tic constraints, such theories capture the insight that changes in com plem ent configurations arc crucially semantic. Regularities in the syntax are captured by linking rules m apping the semantic structure to surface form. To a large degree, as will become apparent below, the lexical rule approach is directly com parable to the approach being proposed here. They share the em phasis on sem antic differences am ong different com plem ent configurations. The strongest differences between the two approaches stem from the increased focus o f the present approach on the nature o f the relation between verb and construction (the lexical rule approach represents this relation only im plicitly in the iatem ent o f the rule itself). By recognizing constructions and verbs to be nterrelated but independent, the nature o f constructional meaning, the prin­ ciples that relate verb and construction, and the relations am ong constructions are brought to the foreground. These topics are the focus o f m uch of the pres­ ent work. In addition, on the present approach it is not necessary to posit an additional verb sense for each new syntactic configuration in which the verb appears. Several general reasons to prefer the constructional approach to the lexical rule approach just described are detailed in the following section. S pe­ cific arguments for the existence o f each construction analyzed in chapters 6 - 9 are provided in those chapters. 1.4

A

dvantages of the

C o n s t r u c t io n A

ccount

i a. i Implausible Verb Senses Are Avoided The constructional approach avoids the problem o f positing im plausible verb senses to account for exam ples such as ihe following: (8) He sneezed the napkin off the table. (9) She baked him a cake. ( 10) Dan talked him self blue in the face. In none of these cases docs the verb intuitively require the direct object com ­ plement. To account for (8), for exam ple, a lexicosem antic theory would have to say that sneeze, a parade exam ple ol an intransitive verb, actually has a threeargument sense. ‘X c a u s e s Y to m o v e Z by sneezing’. To account for (9), such a theory would need to claim that there exists a special sense o f bake that has three arguments: an agent, a theme, and an intended recipient. This in effect argues that bake has a sense which involves som ething like ‘X c m 's i

Y to

have

in t e n d s

to

Z \ To account for (10), the theory would need to postulate a

special sense o f talk, 'X

causes

Y to

bhcom e

Z by talking’.

If additional senses w ere involved, then it would follow that each o f these

10

C hapter O n e

verbs is am biguous between its basic sense and iis sense in the syntactic pattern above. Therefore we would expect that there would be some language that dif­ ferentiates between the two senses by having two independent (unrelated) verb stems. For exam ple, alongside the equivalent o f the English word sneeze we might expect to find another stem move

say, m oop— that meant ‘X

causus

Y to

Z by sneezing’. However, to my know ledge there is no language that has

distinct verb stems for any o f the m eanings represented by exam ples (8 -1 0 ). On a constructional approach, we can understand aspects of the final inter­ pretation involving caused m otion, intended transfer, or caused result to be contributed by the respecti ve constructions. That is, we can understand skeletal constructions to be capable o f contributing argum ents. For example, we can define the ditransitive construction to be associated directly with agent, patient, and recipient roles, and then associate the class o f verbs o f creation with the ditransitive construction. We do not need to stipulate a specific sense o f bake unique to this construction. In general, we can understand the direct objects found in the above exam ples to be licensed not directly as argum ents o f the verbs but by the particular constructions. This idea is discussed in more detail in chapter 2. O ther exam ples where it is im plausible to attribute the com plem ent configu­ ration and the resulting interpretation directly to the main verb include the following: (11)

“ Despite the President’s efforts to cajole or frighten his nine million sub­ jects into line . . . ” (N ew York Times, 29 May 1993)

(12)

“ My father fro w n ed away the com plim ent and the insult." (Stephen M cCauley, Easy Way Out, 1993)

(13)

“ Sharon was exactly the sort o f person w ho’d intimidate him into a panic.” (Stephen M cCauley, Easy Way Out, 1993)

(14)

“ 1 cannot inhabit his mind nor even imagine my way through the dark labyrinth of its distortion.” (Oxford University Press corpus)

(15)

Pauline sm iled her thanks. (Levin & Rapoport 1988)

(16) The truck rum bled down the street. (Levin & Rappaport Hovav 1990b) T he suggestion being made here is to account for these cases, in which the w hole is not built up from the lexical items in a straightforward way, by pos­ tulating a construction that is itself associated with meaning.

i .4.2

Circularity Is Avoided A nother im portant advantage of the construction-based approach is that

it avoids a certain circularity o f analysis resulting from the widespread claim in current linguistic theories that syntax is a projection o f lexical requirements.

Introduction

I I

This claim is explicit in the Projection Principle o f G overnm ent and Binding Theory (GB) (Chomsky 1981), the Bijection Principle of Lexical Functional Gram m ar (Bresnan 1982), and in all current accounts w hich attem pt to predict overt syntax from sem antic roles or theta role arrays. In all of these fram e­ works, it is the verb w hich is taken to be of central im portance. T hat is, it is assumed that the verb determ ines how many and w hich kinds o f com plem ents will co-occur with it. In this way, the verb is analogized to the predicate of formal logic, w hich has an inherent num ber o f distinct argum ents. The verb is taken to be an n-place relation “ w aiting” for the exactly correct type and num ­ ber of arguments. But note, now, that an ordinary verb such as kick can appear with at least eight distinct argum ent structures: 1. Pal kicked the wall. 2. Pat kicked Bob black and blue. 3. Pat kicked the football into the stadium. 4. Pat kicked at the football. 5. Pat kicked his foot against the chair. 6. Pat kicked Bob the football. 7. The horse kicks. 8. Pat kicked his way out o f the operating room. Theories which assum e that the verb directly determ ines particular com ple­ ment configurations are forced to claim that kick is a binary relation with agent and patient argum ents and therefore occurs with transitive syntax, except in Pat kicked Bob the football, in w hich it is a ternary relation with agent, recipi­ ent, and patient argum ents and therefore occurs in the ditransitive construction, and in Pat kicked the fo o tb a ll into the stadium, where kick is again ternary, but now with agent, theme, and goal argum ents, and must “ therefore” occur with the direct object and prepositional com plem ents; and so on. Thus both the evi­ dence for the claim that kick has a particular «-argum ent sense and the expla­ nation for kick having the corresponding com plem ent configuration com e from the fact that kick can occur overtly with a particular n-com plem ent construc­ tion. That is, it is claim ed that kick has an n-argum ent sense on the basis of the fact that kick occurs with n com plem ents; it is sim ultaneously argued that kick occurs with n com plem ents because it has an /(-argument sense. This is where the circularity arises. A constructional approach to argum ent structure allow s us to avoid the cir­ cularity o f arguing that a verb is an n-ary predicate and “ therefore” has n com plem ents when and only when it has n com plem ents. Instead, the ternary relation, for exam ple, is directly associated with the skeletal ditransitive con­ struction. The verb, on the other hand, is associated with one or a few basic senses w hich must be integrated into the m eaning o f the construction. Under

12

C hapter O n e

what conditions this is possible is the subject o f the following chapte:. Instead o f positing a new sense every time a new syntactic configuration is encountered and then using that sense to explain the existence o f the syntactic configuration, a constructional approach requires that the issue o f the interaction between verb m eaning and constructional m eaning be addressed. 1 .4.3

Semantic Parsimony Levin (1985) suggests that evidence for different verb senses does exist.

For exam ple, she argues that “ there is evidence that when the verb slide is found in the double object c o n stru c tio n ,. . . its sense is not the purely physical transfer sense o f slide but rather a transfer o f possession sense” (p. 35). She cites the fact that “ the goal argum ent o f a change of possession verb must denote an entity capable o f ow nership, but the goal argum ent o f a change of location verb need not," as illustrated by her exam ples (17a, b). (17) a. b.

She slid Susan/*the door the present, She slid the present to Susan/to the door.

Thus two distinct senses of slide would be posited to account for the contrast in (17). O ne sense o f slide, ‘slid e,’, would constrain its goal to be animate, while the other, ‘slide2\ would have no such constraint. The two different syntactic realizations are claim ed to follow from universal or near-universal linking patterns m apping sem antic argum ent structures to overt com plem ent configurations. The linking rules would be sensitive to the fact that ‘slid e,’ requires its goal to be anim ate, as follows: ‘slid e,’: < ag t, pat, g o alan.....‘s lid e /: < ag t, pat, go al> I Linking Rules J, She slid Susan the present. She slid the present to Susan. However, general linking rules do not insure that ‘slid e,’ will only occur with the ditransitive construction, as is desired. Verbs which uncontroversially lexi­ cally constrain their goals to be anim ate— such as give or hand— can be used with both syntactic patterns: hand: < ag t, pal, g o a l ....:ilr> ( ~ 'slid e ,’ )

./ Joe handed his m other a letter.

\ Joe handed a letter to his mother.

T hat is, we would need to stipulate that ‘slid e ,’ may only occur with the ditran­ sitive construction. Instead o f positing both an additional sense of slide and a stipulation that this sense can only occur in the ditransitive construction.

Introduction

13

we can attribute the constraint that the goal m ust be animate directly to the construction. Still, it might be argued that ' s l i d e i s not actually constrained to appear ditransitively, and that it is this sense which (just like give and hand) appears in expressions such as (13): (18) She slid the present to Susan. (The reason we might assum e that (18) involves an unconstrained sense o f slide is that She slid the present to the door is also acceptable.) This does not alleviate the problem , however; we still need to insure that the ditransitive con­ struction can only occur with ‘slid e,’. T hat is, instead o f needing to stipulate that ‘slid e,’ can only appear ditransitively, we would now need to posit a con­ straint on the construction that perm its it to only occur with verbs which co n ­ strain their goals to be animate. But with this constraint in place, there is no need to posit an additional verb sense. More generally, I concur with Levin that the sem antics o f (and constraints on) the full expressions are different w henever a verb occurs in a different con­ struction. But these differences need not be attributed to different verb senses; they are more parsim oniously attributed to the constructions themselves.

t .4.4

Compositionality Is Preserved

A construction is posited in the gram m ar if and only if som ething about its form, meaning, or use is not strictly predictable from other aspects o f the grammar, including previously established constructions. In order to under­ stand this principle, we must first consider the notion o f compositionality. Frege is generally acknow ledged to have originally form ulated the idea that sem antics need be com positional: the m eaning of every expression in a lan­ guage must be a function of the m eanings o f its im m ediate constituents and the syntactic rule used to com bine them. M ontague stated the analogous condition that there must be a hom om or­ phism from syntax to semantics: that is, there m ust be a structure-preserving m apping from syntax to semantics. Letting a be a function from syntax to semantics, ’ + a rule of syntactic com position, and ‘ + wm recipient, experiencer > instrum ent > patient, them e > location, source, goal The roles expressed by the hierarchy are argum ent roles, or role types in the sense of Dowty 1986. That is, they are more general than the verb-specific participant roles. Since participant roles are typically instances o f one o f these roles, the hierarchy serves to define a p artial ordering o f all roles. For exam ple, the “ hitter” role is higher on the hierarchy than the “ hittee" role. But the fact that the ordering is partial means that not all roles are ordered with respect to each other. Passive applies only to verbs which are associated with two or more roles, one o f which is higher than the others. Passive

role,

role2

(ro lej)

I

Deprofile r o le ,> role„ n * l

_________________

|

1

Figure 2.12

Cutting. The term “cutting” is intended to invoke the notion o f a direc­ tor cutting one o f the participants out o f the picture. Stative constructions in

58

Chapter Tw o

Bantu (M chom bo 1992), impersonal passive constructions in German, and the middle construction in English serve to cut a profiled participant. The differ­ ence between a shaded participant role and a cut participant role is that the latter cannot be expressed. For exam ple, the agent role is cut in the English middle construction: (54) *This bread cuts easily by Sarah. R ole m erging. Reflexive constructions, for instance in Romance, serve to m erge one participant role with another. The merged participant roles are fused with a single argum ent role, and are then linked with a single gram m ati­ cal function.15 Null com plem ents. Fillm ore (1986) distinguishes two distinct ways in which verbs may lexically specify that a certain participant role can fail to be expressed. In the first type o f case, the unexpressed role receives an indefinite interpretation; the referent’s identity is either unknown or irrelevant. These are indefinite null com plem ents. For exam ple, the objects o f eat and drink are not expressed in (53), and their referents’ identities— that is, w hat was eaten or dru n k — are irrelevant.16 (55) After the operation to clear her esophagus, Pat ate and drank all evening. The unexpressed source role in the following is similar: (56) Chris drove across the country. W hile it is entailed that Chris drove from som ew here, the identity of the source need not be recoverable by either speaker or hearer; it is left indefinite. A sim ilar case involving an unexpressed path argum ent is given in (57): (57) She ran for two hours. Since the unexpressed role in each o f these exam ples has no special prom i­ nence and is nonsalient, these are clear cases o f nonprofiled roles. That is, the food and drink participants o f eat and drink, respectively, are participant roles but are not lexically profiled. T he same is true o f the source (and goal and path) roles o f drive and run. The second type o f unexpressed com plem ent discussed by Fillmore is differ­ ent: the referent’s identity in this case m ust be recoverable from context. This is the definite null com plem ent. Exam ples o f this type include the following (the square brackets are used to indicate where the absent role would normally be expressed):

The Interaction betw een Verbs a n d Constructions

(58) a. b. c.

59

Chris blamed Pat [ ]. Lee found out [ ]. Jo won [ ]!

Only in contexts in which both speaker and hearer can be expected to be able to recover the unexpressed argum ents are these cases Felicitous; it is in this sense that they are definite null com plem ents. Since the contextual constraint ensures that the participant role in question is accessed and salient (in order to be identified), the definite null com plem ent is considered profiled. Fillmore provides a test to distinguish the two types o f unexpressed roles. He notes that while it is perfectly acceptable for a speaker to adm it ignorance of the identity of a m issing indefinite argum ent, it sounds odd for a speaker to admit ignorance of a m issing definite com plem ent: (59) a. b.

He found out! #1 wonder what he found out. (definite null com plem ent) He was running. I w onder where he was running to. (indefinite null com plem ent)

Fillmore observes that in English, w hether a verb allows an argum ent to be un­ expressed with a definite interpretation is a lexical specification. T his assum p­ tion is necessary in order to account for distinctions o f the follow ing k in d :17 (60) (Why did you marry her?) Because M other insisted/*required/*dem anded. (1 9 8 6 :9 8 ) Only insist allows a definite null com plem ent; the closely related require and dem and do not. Ai the same time, many other languages, including Japanese, Korean, and Hungarian, allow definite null argum ents freely. In these lan­ guages, often only the verb is overtly expressed: all o f the verb’s participants may receive a definite interpretation in context. Below, profiled definite om is­ sible participant roles will be represented by the role nam e in boldface su r­ rounded by square brackets: [role]. To sum m arize, there are several ways in which profiled participant roles can be accounted for w ithout being overtly expressed. The verb may occur in a construction which specifically shades, cuts, or merges a certain role or. in languages like English, the verb may lexically designate that a particular role may be unexpressed if it receives a definite interpretation. 2.5

P o s s ib l e R e l a t io n s

betw een

V

erbs a n d

C o n s t r u c t io n s

On a constructional approach to argum ent structure, in w hich the sem anics of the verb classes and the sem antics o f the constructions are integrated to

60

Chapter T w o

yield the sem antics o f particular expressions, the question arises as to what range o f verb classes can be associated with a given construction. C ould any verb class in principle be conventionally associated with a par­ ticular construction?18 For exam ple, if we accept that the ditransitive construc­ tion is directly associated with a particular semantics, roughly, ‘X r e c e iv

)

causes

Y to

Z ', then why would it not be possible in principle for, say verbs of

mood like sadden, anger, regret to be used with the ditransitive construction as in (61) to imply the resulting em otional state? (61) *Joe angered Bob the pink slip. ( “ Joe gave Bob a pink slip, causing Bob to become angry.’’) O bviously we w ant to rule out such a possibility. In order to circum scribe the possible types o f verb classes that can be as­ sociated with particular constructions, we need to exam ine more closely the types o f relations that the verb’s sem antics may bear to the sem antics of the construction. Com m only, the event type designated by the verb is an instance o f the more general event type designated by the construction. For exam ple, consider the use o f hand in (62): (62) She handed him the ball. H and lexically designates a type o f transfer event; at the same time, transfer is the sem antics associated with the ditransitive construction. A nother example of this kind is put, used as in (63): (63) She put the phone on the desk. Put lexically designates a type of caused-m otion event, and caused motion is of course the sem antics associated with the caused-m otion construction. O ther system atic relations between verbs and constructional meanings have been discussed under the heading o f “ conflation patterns” (Talmy 1985a). In our term s, conflation patterns correspond to m ism atches between the semantics o f the verb and the sem antics designated by the construction. The mismatches can be o f several types. As had been implicit in much o f the generative sem antics literature (e.g., Lakoff 1965; M cCawlcy 1973) and has more recently been recognized by Talmy (1985), Levin and Rapoport (1988), and Jackendoff (1990a), verbs w hich do not directly denote the m eaning associated with the construction of­ ten denote the m eans by w hich the action is perform ed. T his is the relation that verbs of ballistic motion bear to the m eaning o f the ditransitive construction. For exam ple, in (64) kicking is the means by which transfer is effected.

The Interaction betw een Verbs a n d Constructions

61

(64) Joe kicked Bob the ball. ( “Joe caused Bob to receive the ball by kicking it.” ) In the case of causative constructions, the verb designates the result associ­ ated with the construction. The construction supplies an agent argum ent which does not fuse with any of the participant roles associated with the verb. For example, consider the Chichewa causative m orphem e its in (65) (from A lsina & M chom bo 1990): (65) N'ungu

i-na-phik its-a

maungu

kwtf kidzldzi.

9 porcupine 9s-ps-cook-CAUSE-fv 6 p u m p k in s to

1 ow l.

'T he porcupine had the pum pkins cooked by the ow l.' Alsina (1993) analyzes this morphem e as having the following sem antic representation: (66) CAUSE < a g t pat PRED < . . . » The causative morpheme is thus a construction, into which the verb’s sem antics (represented by PRED) integrates. This m oiphological construction is quite analogous semantically to the lexically unfilled English constructions that have been discussed so far. T he verb stem and the causative m orphem e must integrate, just as the English verb must integrate into the various English constructions.

The Causal Relation Hypothesis Croft (1991) proposes a general constraint on possible conflation pat­ terns. He suggests that “ individual lexical items appear to denote only causally linked events” (p. 160) (see also M atsum oto 1991 for discussion o f the cen ­ trality of causality in this respect). To illustrate his point, Croft cites the follow ­ ing exam ple adapted from Talmy (1985a): (67) The boat sailed into the cave. He argues that the sailing m anner and the implication o f motion can only be conflated if the activity o f sailing causes the motion. That is, the follow ing is unacceptable: (68) *The boat burned into the cave. Example (68) cannot mean that the boat entered the cave while b urning.1'’ C roft’s claim can be restated in term s of the present account in the followng way:

62

Chapter T w o

Causal Relation Hypothesis: The meaning designated by the verb and the meaning designated by the construction must be integrated via a (temporally contiguous) causal relationship. Evidence supporting C roft's claim com es from the distribution of verbs of sound em ission w ith constructions that designate motion. Such verbs can be used freely when the sound is a result o f the motion and occurs sim ultaneously with the motion: (69) a. b. c. d. e.

The The The The The

wooden-legged man clum ped into the room. train screeched into the station. fly buzzed out o f the window. truck rumbled down the street. (Levin & Rappaport 1990b) elevator creaked up three flights.

For instance, the clum ping noise of (69a) is a result o f the man’s moving. For m ost speakers verbs o f sound em ission cannot be used for coincidentally co­ occurring (or characteristic) sounds, where no causal relationship is involved: (70) a. b. c. d.

*The *The *The *The

bird chirped out o f the cage. dog barked into the room. rooster crowed out of the barn. man laughed out o f the room.

However, Croft's claim is not sufficient to account for all cases. This brings us to the following section.

Violations of the Causal Relation Hypothesis T here are several types o f violations of the Causal Relation Hypothesis that are allowed by particular constructions. The construction exemplified by (71) allow s verbs which designate events not causally related, at least to a lim ­ ited extent (cf. chapter 9). ( 7 1) She kicked her way out of the room. For exam ple, the follow ing exam ples from the O xford University Press corpus involve only the m anner o f m otion, not the m eans o f motion (cf. Levin & R apoport 1988; Jackendoff 1990a):10 (72) a. b. c.

“ ‘I knitted my way across the A tlantic,’ he reveals.” “ . . . w ithout a party to go to, he nods and winks his way through the set cram m ed with seaside sing-alongs.” . . [anyone] watching would have thought he was scowling his way along the fiction shelves in pursuit o f a book.”

The Interaction be tw e en Verbs a n d Constructions

63

Interestingly, the way construction tends to be used with pure m anner verbs only when the manner is particularly salient and em phasized. This is reflected in the fact that, not uncom monly, m anner cases involve two or three conjoined verbs, as in exam ple (72b). Returning to verbs o f sound em ission again, it seem s that they can m argin­ ally be used in the motion construction when the verbs do not designate a sound resulting from the motion. In particular, if the sound is the m eans of identifying the path o f motion, the expressions seem at least m arginally acceptable: (73) a. ?The police car scream ed down the street, b. ?Thc train whistled into the station. The conative construction exemplified by (74) also perm its exceptions to the Causal Relations Hypothesis: (74) a. b.

Ethel struck at Fred, Ethel shot at Fred.

In this case the verb designates the intended result o f the act denoted by the construction. The sem antics of the construction can be represented roughly as ‘X

d ir e c t s a c t io n

at

Y ’. T hat is, Ethel does not necessarily strike Fred, but

striking him is the intended result o f the directed action. The construction can be represented as follows: Conative Construction

Sem

D1RECT-ACTION-AT


+ contact

I Syn

V Figure 2.13

The fact that a verb that is related to the construction by the intended-result relation m ust be (+ motion, + contact] serves to allow verbs such as shoot, hit, kick, and cut, w hile correctly ruling out verbs such as *m ove (no contact) and *touch (no motion) (Guerssel et al. 1985; Laughren 1988). This constraint is captured by restricting the class o f verbs which can instantiate PRED when the R-rclation is one of intended result.

64

Chapter T w o

T his representation allow s us to assim ilate expressions such as (74a, b) above to other related expressions, for instance those in (75): (75) a. b-

Fred looked at Ethel, Ethel aimed at Fred.

Look and aim are not [ + motion, + contact] verbs,21 and yet they bear an obvi­ ous sim ilarity to the cases above. They differ from these earlier cases in that now the verb’s sem antics is an instance o f the semantics o f the construction. T hat is, ‘look' and ‘aim ’ are instances of ‘D IR ECT-A C TIO N -A T’. For ex­ am ple, aim fuses with the conative construction as follows: Composite Structure: Conalive + aim

DIRECT-ACTION-AT


1 target 1

>

OBL..J,-.

Figure 2.14 T he m eaning o f the construction rem ains constant, regardless o f whether the verb designates an instance or the caused result; it is the relation between the m eaning o f the verb and the m eaning o f the construction— the R -relation— which is different. Particular R-relations m ust be able to refer to classes o f verbs in order to capture the [ + motion, + contact] constraint. The conalive construction can be represented as follows: Conative Construction

Sem

DIRKCT-ACTION-AT


i' O BL-a, ••

The Inter.icnon betw een Verbs a n d Constructions

65

Verbs may also code particular preconditions associated with the sem antics of the construction. For exam ple, creation verbs designate an act of creation, which is a precondition for transfer. C onsider (76): (76) Sally baked Harry a cake. This sentence does not entail that the baking itself was causally related to the transfer. The baking does not cause the transfer, and the transfer does not cause the baking. However, the creation o f the cake is a necessary precondition of the transfer. An im portant question is, why should these relations be privileged? W hy should means, preconditions, and to a lesser extent, the m anner involved in an event be more likely candidates for use in a construction w hich im plies the entire event than, say, the mood o f one o f the participants? This deeper question is difficult to answer, but if we consider certain verbs’ inherent sem antics to bear a m etonym ic relationship to the sem antics of the construction, we may find a partial explanation. The sem antics associated with the construction defines a sem antic frame, and the verb m ust inherently desig­ nate a particular salient aspect o f that frame.

The Fusion of Roles M atsumoto (1991) notes that when two verbs are com bined to form a com plex motion predicate in Japanese, they must share at least one role. He labels this constraint the Shared Participant Condition. In our term s, this con­ straint can be translated into the claim that at least one participant role and argument role m ust be fused; thus not all o f the argum ent roles can be contrib­ uted by the construction.

Summary of the Relations between Verb Semantics and Construction Semantics Let e c be the event type designated by the construction, and e , the event type designated by the verb. I. e , must be related to e c in one o f the following ways: A. e y may be a subtype o f ec B. e v may designate the means o f e c C. e , may designate the result of et D. e v may designate a precondition o f e c F.. To a very limited extent, e v may designate the m anner of e c, the means o f identifying ec, or the intended result of ec II. e c a n d e , must share at least one participant (M atsum oto 199I).22

66

C hapter T w o

D o all o f the possible relations in (I) have equal status? Clearly not. That e v may be a subtype o f e c is prototypical and universal. The possibility that c. may code the means o f cc seems to be a language-specific parameter: English, D utch, and C hinese allow this relation; Rom ance. Semitic, and Polynesian languages apparently do not (Talmy 1985a). O ther relations, for exam ple that e w may designate the precondition, manner, or result o f e t , are construction specific. The result o f integrating the verb with the construction must be an event type (E) that is itself construable as a single event. That is, only a single event can be expressed by a single clausc. Some of the constraints on exactly what this entails are discussed in chapters 7 and 8. 2.6

C o n c l u s io n

In this chapter, I have attem pted to argue for som e o f the basic claims underlying this m onograph, and have laid out some o f the m achinery needed to make these claim s precise. Follow ing the discussion in chaplcr 1, where it was argued that constructional m eaning exists independently o f verb meaning, the type o f sem antics associated with verbs and constructions has been discusscd in more detail. Verbs and other lexical item s have been argued to be associated with rich fram e-sem antic knowledge. Basic sentence-level constructions, or argument structure constructions, have been argued to designate scencs which are in som e sense basic to human experience (cf. also Fillm ore 1968, Langacker 1991). That is, it is claim ed that the set o f basic clause types o f a language are used to encode general event types such as those denoting that someone did som ething to som eone, som ething moved, som eone caused som ething to change stale, som eone experienced som ething, som eone possessed something, and so forth. Evidence for the idea that these event types have a privileged status com es from certain language acquisition facts noticed by Clark (1978), Slobin (1985), and Bow erm an (1989). in addition it has been argued that these basic senses are extended in various ways so that particular syntactic frames are associated with a family of related m eanings. This idea has been explicitly contrasted with the idea that the se­ m antics associated with a construction is ultimately generalized, or that it is abstracted to a single m ore general sense. Finally, constraints on the types o f potential relations between verbs and constructions have been suggested, extending observations by Talmy (1985a), C roft (1991), and M atsum oto (1991).

3 Relations among Constructions

The repertoire o f constructions is not an unstructured set. There are sys­ tematic generalizations across constructions. In this chapter, several organi­ zational principles are discussed and applied to the constructions analyzed in this work. It is argued that constructions form a network and are linked by inheritance relations w hich motivate many o f the properties o f particular co n ­ structions. T he inheritance network lets us capture generalizations across con­ structions w hile at the sam e lim e allow ing for subregularities and exceptions. Before explicating the nature of the relations between the constructions we have looked at, it is im portant to describe the general psychological principles of language organization that will be assumed. 3.1

R e l e v a n t P s y c h o l o g ic a l P r in c ip l e s

of

La n g u a g e O

r g a n iz a t io n

Each o f the following principles is stated in term s o f constructions, since constructions are the basic units in our system. All o f these principles have direct analogues in various functionalist frameworks. I. The Principle o f M axim ized M otivation: If construction A is related to construction B syntactically, then the system o f construction A is m oti­ vated to the degree that it is related to construction B sem antically (cf. Haiman 1985a; Lakoff 1987). Such m otivation is maximized. II. The Principle o f No Synonymy: If two constructions are syntactically distinct, they musi be sem antically or pragm atically distinct (cf. Bolinger 1968; Haim an 1985a; Clark 1987; M acW hinney 1989). Pragm atic aspects of constructions involve particulars o f inform ation structure, in­ cluding topic and focus, and additionally stylistic aspects o f the co n ­ struction such as register (cf. discussion in section 1.1). Corollary A: If two constructions are syntactically distinct and S(em antically)-synonym ous. then they must not be P(ragm atically)-synonym ous. Corollary B: If two constructions are syntactically distinct and P-synonym ous, then they must not be S-synonym ous. III. The Principle o f M axim ized Expressive Power: The inventory o f con­ structions is m axim ized for com m unicative purposes. IV. The Principle o f M axim ized Economy: The num ber of distinct construc-

67

68

C hapter Three

tions is m axim ized as much as possible, given Principle III (Haiman 1985a). In support o f these principles, consider the analogy Haiman (1985a) pro­ poses between the form of a language and a diagram such as map or a musical score. Haiman suggests that w hile a map dcpicts geography and a musical score depicts a melody, language depicts our construal of reality. There are several relevant facts about diagram s. Haiman notes that in an ideal diagram , every point should correspond to some point in the reality being depicted. He refers to this property as isomorphism, which also seems to imply that every point in the geography or in the melody corresponds to a unique point on the map or musical score, respectively. Moreover, every relation be­ tween two points on a diagram should correspond to a relationship between points in reality. This second property is referred to as motivation. Haiman notes that w orking against strict adherence to these two properties is the fact that diagram s are designed to sim plify: they only need to represent, not repro­ duce. For exam ple, a map does not show all o f the details o f the territory being represented and a musical score diagram does not uniquely determ ine the way the music is to be played. Moreover, diagram s often contain certain distortions: G reenland is represented loo largely in most maps, and a low note in treble clef notation is higher than a high note in bass clef. However, the general principles o f isom orphism and motivation arc ob­ served to a large degree: each point o f a map corresponds roughly to one point in the world, each representation o f a musical note corresponds to only one pilch most o f the lime. Also, Ihe distance between two points on the map is generally greater when the corresponding distance in the world is greater; within the sam e clef, higher notes are higher than lower notes. T he analogy to natural language runs as follows. T he principle of iso­ m orphism covers two aspects. On the one hand, differences in form imply d if­ ferences in m eaning (or pragm atics), as dem anded by the Principle of No Synonym y (cf. section 1.1). Haiman attributes this principle to H um boldt, Vendryes, O gden, and R ichards, and it has been echoed more recently by. among others, B olinger 1968, Clark 1987, and M acW hinney 1989. Conversely, a dif­ ference in m eaning or pragm atics should lead to a difference in form, in accor­ dance with w hat we have called above the Principle o f M aximized Expressive Power. Noting a need for sim plification, H aim an allows for derivations from iso­ m orphism . He suggests that deviations from this rule in natural languages oc­ cur in cases o f polysem y and homonymy, but that such exceptions can be attributed to a general need for sim plification, just as in the case o f diagrams. This observation is captured by the Principle of M aximized Economy.

Relations a m o n g Constructions

69

Therefore, while the Principle o f M axim ized Econom y w orks lo constrain the multitude of constructions, the Principle o f M axim ized Expressive Power works in the opposite direction, creating the tendency for more distinct form s; that is, a maximally expressive system w ould have a distinct label for every distinct item in the user’s world. These two principles mutually constrain each other. With one possible exception, all of the functional principles listed above are widely assumed and are sufficiently intuitive so that a more extended defense o f them is not attem pted here. The one principle which is som ew hat less widely adopted within linguistics is the Principle of M axim ized M otivation. Since this principle plays an important role in the discussion of the relations between constructions, it is w orthwhile discussing it in m ore detail. 3.2

M

o t iv a t io n

The term “ m otivation” was introduced into linguistics by Saussure. In the Cours he provides the exam ple o f d ix -n e u f'nineteen’, noting that w hile the parts of this word are arbitrary signs, the com plex taken as a w hole is m oti­ vated. It is clear that it is not predictable that d ix-n eu f should take the form it does. A unique m orphem e could have been introduced to signify the concept ‘nineteen’, or n eu f-d ixcould have been used. Still, there is an obvious sense in which the term is not arbitrary. Motivation in this sense lies between predictability and arbitrariness. In an intuitive sense, it often constitutes explanation. If a (som ew hat hapless) French child were to ask, “ Why is this many [pointing to nineteen things] referred to by ' d ix -n e u f'T ' a natural response would be to point out that n e u f means ‘n in e’ and dix means ‘ten’ and that nine plus teen is nineteen. Haiman argues that making generalizations and sim plifications is a neces­ sary function of language, because it would be im possible in our finite world, with our finite m em ories, to have distinct nam es for the infinite num ber of actual distinctions in the world. Rather than recognizing an infinity o f sounds and concepts, human language recognizes a finite inventory of phonem es and morphemes. In order to reveal the im portance of motivation in gram m ar, H ai­ man recounts J. L. B orges’s tale o f “ Funes the M em orious" (1962). B orges’s hero, Funes. has undergone an accident that has left him with a perfect memory. Funes can rem em ber "the outlines o f the foam raised by an oar in the Rio Negro the night before the Q uebracho uprising.” Since he has a perfect m em ­ ory, he devises his own language, in which every sense experience and every concept he recognizes is given a separate name: “ It bothered him that a dog at 3:14 (seen from the side) should have the sam e name as the dog at 3:15 (seen from the front)." Funes scorns the use of m nemonic classification: “ In place of

70

Chapter Three

7,013, he would say M axim o Perez; in place o f 7,014, he would say. The rail­ road; . . . in place o f five hundred, he would say nine." By rejecting principles of organization, Funes's language is not motivated. Every difference is a com plete difference; there is no motivation to code gen­ eralizations and sim ilarities. It is adm ittedly often not predictable which gen­ eralizations or sim ilarities a language will encode; however, unless the ne­ cessity o f motivation in a gram m ar is recognized, we cannot account for the fact that F unes’s language is an inconceivable human language. L angacker (1987a) has also stressed the im portance o f a notion between predictability and arbitrariness. He notes that our inability to predict what pat­ tern a language uses does not entail that the choice has no semantic basis. For exam ple, he observes that w hile the fact that scissors, pants, glasses, and bin­ oculars have the form o f plurals is not predictable from their designations, it is nonetheless motivated by the bipartite character o f the type o f object the words designate (1987a: 47). L akoff (1987) suggests a precise definition for the term “ motivation” in gramm ar. A given construction is m otivated to the degree that its structure is inherited from other constructions in the language. On Lakoff's (1987) account o f f/iere-constructions, the “ based-on” relation is o f central importance. It is said to be an asym m etric inheritance relation, so that if construction A is based on construction B, then A inherits all o f B's properties that do not specifically conflict with its own specifications. Lakoff suggests that the more the proper­ ties o f a given category are redundant, the more it is motivated and the better it fits into the system as a whole. An optim al system is a system that maximizes m otivation. There may be many optim al gram m ars since motivation can be m axim ized in many ways. R esearchers in child language acquisition are also arguing against the idea of a strict dichotom y between predictability and arbitrariness. M ore and more they are advocating learning m echanism s in which there is no sharp division between obligatory rules and probabilistic tendencies (e.g.. Bates & MucW hinney 1987; M acW hinney 1989. 1991; Pinker 1987). Evidence that a relation in form aids in the acquisition of concepts which are related in m eaning com es from studies o f children’s learning o f taxonomic re­ lations. G elm an, W ilcox, and C lark (1989) have shown that children learn the nam es o f subordinate term s more easily when those term s are com pounded with basic level term s that the child already knows. For exam ple, children were m ore likely to learn the nam e for a new type o f car when it was called a fep ca r than when it was sim ply called a fep. This finding is not obvious, since it would seem on the face of it that a child would have to learn more in learning the com pound term than in learning the uncom pounded novel term. However,

Relancns amona Consrnjaions

7I

when motivation is taken into account as an aide in learning, the findings can be seen to be natural. Children learn new term s for concepts which are related to other, already familiar concepts more easily w hen the new term s are syste­ matically related to the term s for the fam iliar concepts.1 A recognition o f the importance o f m otivation-like reasoning is grow ing in the field of Artificial Intelligence. Abduction, or reasoning to the best ex p la­ nation. has been argued to be useful in attem pts to model human inferences (W ilensky 1982). Typically one must know the outcom e in order to perform abduction, which distinguishes it from deduction. In critical respects, the seek­ ing out o f linguistic motivation can be understood to be abductive inferencing applied to language learning, whereas predictability corresponds to the result o f applying deductive reasoning. That is, abductive reasoning involves afterthe-fact inferencing to determ ine why a given sequence o f events should have occurred as it did. The given sequence o f events is not, however, a priori pre­ dictable. Similarly, while speakers cannot predict w hether or to what extent two related concepts will be related formally, it is claim ed that they nonetheless search for such relations in order to “ m ake sense o f” the input forms, fitting the new forms into the network of interrelated constructions that constitutes their knowledge ol language. This idea has been suggested by, for exam ple, Bates and M acW hinney (1987), who propose that relations between forms, meanings, and fo rm -m e an in g pairs are (unconsciously) observed and pon­ dered in their own right. If W ilensky is right in arguing that people seek out abductive explanations—that is, m otivation— in trying to account for se­ quences o f events, then this would give us reason to suspect that speakers m ight unconsciously apply the same principles in trying to acquire language. Connectionist representations also make no sharp division between what is predictable and what is arbitrary, instead allow ing there to be correlations o f varying strength (cf. Rum elhart & M cClelland 1986). Individual correlations can be interpreted as m otivating factors: they may influence the system in a certain direction but they are not in isolation predictive. Such m odels have begun to be applied to linguistic phenom ena. In these system s violable con­ straints— constraints which add to the naturalness (or unnaturalness, if framed negatively) o f a given expression— are o f central im portance. This ideas has given rise to constraint optim alization theories in phonology (Smolensky 1986; Legendre, M iyata & Smolensky 1990; Prince & Sm olensky 1991; G oldsm ith 1993). More generally, in connectionist networks, item s o f new inform ation are more easily incorporated when analyzed as variations on known inform ation; new patterns are autom atically assim ilated to old patterns as much as possible. Optimization in such systems therefore produces m otivated structures.

72

Chapter Three

Incorporating motivation into the gram m ar captures a fundamental structur­ alist insight w hich has been overlooked by most formal linguistic theories. This insight is that elem ents in a system influence each other even when they do not literally interact. Evidence for this kind o f influence is abundant in the domain o f phonology, for exam ple in the phenom ena o f analogic extension and resto­ ration, back form ations, push and drag chains, paradigm atic leveling, and in the very fact that, to a striking degree, sound change is regular. These phe­ nom ena attest to the fact that speakers (unconsciously) seek out regularities and patterns, and tend to impose regularities and patterns when these are not readily available. T he idea o f explicitly linking constructions that are related in various ways is in accordance with w hat is known about the lexicon. Current research over­ w helm ingly rejects the idea that the lexicon is sim ply a list of unrelated facts or com pletely independent pieces of knowledge. Instead, memory in general, and the lexicon in particular, have been shown to involve a richly intercon­ nected web o f inform ation. Various psycholinguistic prim ing experim ents have shown that form and m eaning relations between lexical items are cognitively real (e.g., M eyer & Schvaneveldt 1971; Ratcliff & M cKoon 1978; Anderson 1984). 3.3

R e p r e s e n t in g M

o t iv a t io n b y I n h e r it a n c e

To capture relations of m otivation, asym m etric inheritance links are pos­ ited between constructions which are related both sem antically and syntacti­ cally. T hat is, construction A motivates construction B iff B inherits from A. Inheritance allow s us to capture the fact that two constructions may be in some ways the same and in other ways distinct. The idea o f using inheritance as a method of capturing generalizations origi­ nated in com puter science, as a way to represent data structures in as general a form as possible (cf. Fahlman 1979;Touretzky 1986). Inheritance has since been found to be useful in many program m ing and know ledge representation sys­ tems, including FRL, KRL, KL-ONE, KODIAK, SM ALLTALK, FLAVORS, L O OPS, ADA, and object-oriented LISP. By postulating abstraction hierar­ chies in which lower levels inherit inform ation from higher levels, information is stored efficiently and made easily modifiable. Use of the concept of inheritance is also currently growing as a way to cap­ ture linguistic generalizations, for exam ple in work by Bobrow and Webber (1980), Hudson (1984), Lakoff (1984), Flickinger, Pollard, and Wasow (1985), W ilensky (1986), Pollard and Sag (1987), Jurafsky (1992), Thom ason (1992). Davis (1993). T he follow ing inheritance system draw s on aspects o f cach of these theories.

Relations a m o n g Constructions

73

Following Lakoff 1984, W ilensky 1986, and Jurafsky 1992, the data struc­ tures in our system are constructions. Constructions are specified as to which other, more abstract constructions they inherit from, or equivalently— to use the term inology o f W ilensky 1986— w hich other constructions they are d o m i­ nated by.

3.3. i

General Properties of Inheritance Notation An inheritance relation between two constructions C, and C : such that

C. inherits from C, will be represented as follows:

C 2 inherits from C, C | dominates Ci C | motivates C j I = inheritance link Figure 3 .1

inherited inform ation will be represented in italics; that is, all inform ation which is shared between the dom inating and dom inated node is italicized in the dominated construction. As before, profiled inform ation is written in boldface.

Multiple Inheritance Is Allowed In accord with all o f the linguistic applications o f inheritance cited above, multiple inheritance paths are allowed. T hat is, inheritance system s may re­ semble tree diagram s if each child has only one parent, but in the general case they are “ tangled” and can be represented as ‘D irected Acyclic G raphs’ (DAGs). This allow s a given construction in the hierarchy to inherit from more than one dom inant construction.

Normal M ode Inheritance Following Flickinger, Pollard and Wasow (1985) the norm al m ode o f inheritance is distinguished from the com plete mode. The normal m ode is d e­ signed to allow for subregularities and exceptions, and is the only type to be used here. In the normal mode, inform ation is inherited from dom inant nodes transitively as long as that inform ation does not conflict with inform ation specified by nodes lower in the inheritance hierarchy. Lakoff (1984), in his analysis of r/iere-constructions, refers to this type of inheritance as "in h eri­

74

Chapter Three

tance with overrides" (cf. also Zadrozny & M anaster-R am er 1993). Normal inheritance is simply a way o f stating partial generalizations. The com plete mode o f inheritance, which is not exploited here, is designed to capture purely taxonom ic relations and constraints. In the com plete mode, all inform ation specific to every node which directly or indirectly dominates a given node is inherited Inform ation from one node may not conflict with that of a dom inant node w ithout resulting in ill-form edness. This is the type o f inheritance norm ally assum ed in unification-based gram m ars (e.g., Kay 1984; Fillm ore & Kay 1993).

Real Copies; Full-Entry Representations Fahlman (1979) distinguishes real copying from virtual copying o f in­ form ation. In real copying, dom inated constructions contain all the information that the dom inating constructions do: each construction is fully specified, but is redundant to the degree that inform ation is inherited from (i.e., shared with) dom inating constructions. This is the type o f inheritance em ployed here. Jurafsky (1992) likens this type o f inheritance to the ‘'luJI-entry" theory o f redun­ dancy rules, as opposed to the "im poverished-entry” theory (cf. Jackendoff 1975). Thus the inheritance m echanism o f our system is not an on-line process, but rather a static relation defined by shared inform ation (cf. Jackendoff 1975; A ronoff 1976; Bresnan 1978, 1982; Hudson 1984; Lakoff 1984; and Pollard & Sag 1987 for related m echanism s). In virtual copying, on the other hand, dom inated constructions are only par­ tially specified: inherited inform ation is only stored with the dom inating con­ struction. U nder this m echanism , inferences are com puted by searching up the inheritance tree to determ ine the full specifications of a given construction. This type of inheritance is not exploited here. A llow ing each construction to be fully specified would seem to be an inef­ ficient way to store inform ation; however, this inefficiency is not necessary, depending on the particular im plem entation adopted. A connectionist system can capture the redundancy w ithout inefficiency by allow ing inherited infor­ mation to be shared inform ation; that is, instead o f stating the specifications twice, aspects o f the patterns that are inherited are shared by two overlapping patterns. Similarly, in a sym bolic system, it is possible to avoid fully specifying particular inform ation twice by allow ing particular specifications within con­ structions to have pointers to other inform ation.

3.a.2

Inheritance Links as Objects So far we have not said how inheritance links make explicit the particular

types of relations that may hold am ong elem ents of constructions. That is, in­

Relations a m o n g Constructions

75

heritance links capture the fact that all nonconflicting inform ation between two related constructions is shared. However, we have not said anything about how to distinguish am ong various different types o f inheritance relations. In order to make explicit the specific ways that constructions may be related, we will adopt another idea from com puter science, that o f object-oriented d e­ sign.1 In particular, the inheritance links themselves will be treated as objects in our system (cf. also W ilensky 1991). Like constructions, they are assumed to have internal structure and 10 be related hierarchically. Links can be of sev­ eral types, with various subtypes each. This idea is useful because various kinds of relations am ong constructions recur in the gram m ar; in order to ca p ­ ture these generalizations, it is useful to be able to explicitly notate inheritance links as being of specific types. Moreover, as discussed below, by treating links as objects we are able to represent the fact that extensions may be created productively. Four m ajor types of inheritance links are distinguished: polysem y links, metaphorical extension links, subpart links, and instance links.3

Polysemy |IP) Links Polysemy links capture the nature o f the sem antic relations between a particular sense of a construction and any extensions from this sense. The syn­ tactic specifications o f the central sense are inherited by the extensions; there­ fore we do not need to state the syntactic realization for each extension— such specifications are inherited from the dom inating construction. T he same gen­ eral type o f link is posited to capture m orphological polysemy. Each particular extension is related by a particular type of I P-link. For e x ­ ample, in chapter 2 it was argued that the ditransitive syntactic pattern is asso­ ciated with a family of related senses, rather than a single abstract sense. The following pattern of polysemy was observed: 1. 'X

causes

Y to

r e c e iv e

Z ' (central sense)

Example: Joe gave Sally the ball. 2. Conditions of satisfaction imply 'X

causes

Y to

r e c e iv e

Z’

Example: Joe promised Bob a car. 3. ‘X

ENABLES

Y to

R EC EIVE

Z’

Example: Joe perm itted Chris an apple. 4. ‘X

causes

Y not to

r e c e iv e

Z’

Example: Joe refused Bob a cookie. 5. *X IN T E N D S to C A USE Y 10 R E C E IV E 7 ' Example: Joe baked Bob a cake. 6. ‘X

ac

is to

cause

Y to

r e c e iv e

Z at some future point in tim e’

Example: Joe bequeathed Bob a fortune.

76

C hapter Three

The caused-m otion construction has a strikingly sim ilar pattern o f polysemy: 1. ‘X c a u s e s Y to m o v e Z ’ (central sense) Exam ple: Pat pushed the piano into the room. 2. Conditions of satisfaction imply ‘X c a u s e s Y to

move

Z'

Exam ple: Pat ordered him into the room. 3. ‘X

enables

Y to

move

Z’

Exam ple: Pat allowed Chris into the room. 4. ‘X

causes

Y not to

m o v e from

Z’

Exam ple: Pal locked C hris into the room. 5. ‘X h e l p s Y to m o v e 7.' Exam ple: Pat assisted Chris into the room. In both cases, several o f the extensions involve the type o f family of related causal relations discussed by Talmy (1976, 1985a, 1985b) under the rubric of “ force dynam ics.” In particular, enablem ent, resistance, and aiding are con­ cepts force-dynam ically related to causation, which is a central com ponent of the central senses o f the two constructions. Each o f these concepts involves two entities w hich are construed as interacting via transm ission o f energy ei­ ther in the sam e or in opposing directions (cf. also Jackendoff 1990a for discussion). Extensions 2, 3, and 4 o f the two constructions are quite analogous. The particular verbs involved are different, but the relations between the central sense o f transfer or caused motion and the entailm ents of these extensions is the same. Jackendoff’s (1990a) analysis o f the infinitive (or “eq u i") pattern indicates that it, too, has a rem arkably sim ilar pattern o f interpretations. At the same tim e, the full patterns o f polysem y in the two constructions analyzed above are not identical. For exam ple, while the caused-m otion con­ struction can be used to entail ‘X h e l p s Y to m o v e Z ’, no such interpretation is possible for the ditransitivc construction: ( I) *She helped him the prize. (Intended meaning: She helped him to get the prize.) Therefore the patterns o f polysem y m ust in general be learned for each indi­ vidual construction. Each o f the extensions constitutes a m inim ally different construction, moti­ vated by the central sense; that is, each sense can be represented by a construc­ tion that is m inim ally different from that o f the central sense. The semantic relations are captured by particular U-links, and all inform ation about syntactic specifications is inherited from the central sense. For exam ple, the fifth extension o f the ditransitive, sometim es called the

Relations a m o n g C onstructions

77

"benefactive” construction, can be represented thus, with inform ation that is inherited from the central sense italicized: Diransitive Consluclion

Sem CAUSE-RECEIVE 1 1 PRED

Syn




1




SL'BJ

V

OBJ

OBJ,

lP: intended causation Benefactive-Ditransitive Construction

Sem

INTEND C A U SE-R EC EIV E < precondition PRED

Syn

V

agt

rec

|

I

pat

>

1 >


DO Abs > Erg D O > 10 PO > SO 3

The principles governing English word order and case marking facts can now be stated quite simply: All and only term s (i.e., Subj, DO, 10, PO, SO) are not marked with prepositions; the word order must be: Subj-V -PO /D O -SO /lO nonterms. Dryer goes on to argue that this fact— the sim plification o f word order and case marking descriptions given the existence o f the categories PO and S O — is evidence for his A ntidative analysis. But in order for this fact to substantiate his claim that prepositional paraphrases with to are derived from ditransitive expressions. Dryer must assum e ( I ) that the two form s necessarily share a level o f representation, and (2) that derivations cannot create gram m atical relations (therefore, if PO and SO exist, they must be base generated). If we do not assume ( I), with its im plication that either the ditransitive or the prepositional paraphrase is derived, there is no reason why the two form s cannot both be base generated. T hat is, the argument as to w hether the categories PO and SO exist in English has no bearing on the question w hether an alternation account is warranted or not; D ryer provides no independent evidence for the deriva­ tional analysis.6 Not all transform ational or derivational accounts rely crucially on an under­ lying shared representation between pairs of expressions that share a rough semantic equivalence (e.g., Perlm utter 1978; Perlm utter & Postal 1983b; Aissen 1983; Farrell 1991). However, there are other reasons to avoid a transfor­ mational relationship between related constructions if possible. Bowerman (1982) and Gropen et al. (1989) show that in child language ac­ quisition, semantic restrictions are operative as soon as certain constructions are produced, there being no period o f unconstrained overgeneralization on the

10 6

C hap ter l-our

basis o f a purely syntactic relation.7 For exam ple, Gropcn et al. (1989) show that the sem antic restriction that the recipient o f a ditransitive must be animate is operative as soon as the ditransitive syntax is produced. Thus none of the following possible types o f overextensions were ever uttered by any of the children they observed: (9) *Amy took Chicago Interstate 94. (Amy took Interstate 94 to Chicago.) (10) * Betty threw the tree the box. (Betty threw the box to the tree.) (11) * Alex put his head a gun. (Alex put a gun to his head.) (12) *Babs took fun a trip. (Babs took a trip for fun.) (Gropen et al. 1989:218) As G ropen et al. note, this calls into question the idea that the dative rule is fundam entally a syntactic operation; there is no clear reason why a syntactic operation would be instantaneously constrained by an arbitrary semantic con­ dition. Moreover, since an unconstrained rule would be easier to leam and rep­ resent and would provide more expressive power (Pinker 1989), it is not clear why the sem antic constraint on this putative syntactic rule is not ignored by new generations o f speakers. A nother problem with approaches that rely on transform ations is that they posit an often unw arranted asym m etry between two constructions that are thought to be related. In the case o f the ditransitive. H e gave the book to her is usually supposed to be more basic than He gave her the book (contra Dryer 1986). A typical reason given is that the verbs w hich allow ditransitives are a proper subset o f those that allow prepositional paraphrases. However, this is not actually so: refuse and deny do not have paraphrases with to or for, and neither do many m etaphorical expressions. For example: (13) a. She gave me a headache/a kiss/an idea. b. *She gave a headache/a kiss/an idea to me. M oreover, O ehrle (1976) has argued that there is no principled way to distin­ guish those cases which have prepositional paraphrases from those that do not. D evelopm ental data (Gropen et al. 1989) show s that the ditransitive and prepositional paraphrases occur at roughly the same time in children s speech, with neither construction reliably preceding the other, so that evidence for an asym m etry cannot be grounded in evidence from children’s acquisition o f the forms. C onsider also the English locative alternation. In general, the pattern asso­ ciated with (14) is supposed lo b e more basic than that associated with (15) (cf. Channon 1980; Perlm utter & Postal 1983a).

O n U n k in g

107

(14) He loaded hay onto the wagon. (15) He loaded the wagon with hay. However, when different verbs are exam ined, this claim o f asym m etry is not clearly warranted. So, although stack and plaster allow both argum ent struc­ tures, there is no intuition that the onto variant is more basic than the with variant. That is, the follow ing appear to have equal status in term s o f being basic or unmarked: (16) a. b.

He stacked the shelves with boxes, He sacked boxes onto the shelves.

(17) a. b.

He plastered the wall wtih posters, He plastered posters onto the wall.

M oreover, adorn, blanket, block, cover, dam, enrich, fill, dirty, titter, smother, soil, trim, endow, garnish, imbue, pave, riddle, saturate— to name a few— only occur with the with variant. In fact, in a detailed study o f locative verbs, Rappaport and Levin (1985) found that out o f 142 verbs studied, only 34 alternated, with exceptions existing in both directions (Pinker 1989). On a constructional approach, we need not assum e an asym m etrical relation­ ship between two constructions that are found to be related. We can describe instances o f partial overlap o f syntax, sem antics, or pragm atics as such, without necessarily assuming that one of the constructions involved is basic, the other derived. For example, we can state that the sem antics associated with the ditransitive construction is related to the sem antics o f the paraphrase with to; we do not need to assum e the prim acy o f one over the other. And we can describe similarly the relations between paraphrases with to and other instances o f the caused-motion construction, for instance between the following (a) and (b) expressions: (18) a. b.

Ethel brought the w rench to Fred. Ethel brought the w rench toward Fred.

(19) a. b.

Ethel threw the ball to Harry. Ethel threw the ball over Harry.

These cases are discussed in chapter 7. To summ arize, accounts of argum ent structure which relate one construction to another by a syntactic transform ation that derives one from the other have several drawbacks: I. Expressions lhat are claim ed to share a level o f representation are not fully synonymous. This raises the following problems:

108

Chapter Four

(a) W hich aspects o f sem antics are relevant to determ ining semantic equivalence, and thus a shared level o f representation, has never been adequately detailed. (b) In many cases the only m otivation (often implicit) for proposing a derivational relationship in the first place is semantic synonymy. 2. The sem antic distinctions are learned as early as the forms themselves, w hich casts doubt on the idea that the transform ations are basically or prim arily syntactic. 3. Such accounts postulate an asym m etry between the two forms in ques­ tion. However: (a) T here are typically lexical item s that only have the output form o f a putatively optional transform ation. (b) The two forms are often learned at roughly the same age (with nei­ ther one reliably preceding the other). W hile transform ational accounts explicitly represent semantic relations am ong constructions, the constructional approach, as we saw in the previous chapter, takes a different view. On the constructional approach, semantic sim i­ larities that do not coincide with formal sim ilarities are captured implicitly, because o f a relation between the specified sem antics, but are not explicitly notated in the gramm ar. The intuition is that the existence o f a given form with a particular m eaning in no way m otivates the existence o f a different form with a closely related meaning. Therefore, inheritance links are not posited between constructions that are not related formally. Only relations involving both form and m eaning (or som etim es ju st form; cf. Fillm ore & Kay 1993) are explicitly represented by positing inheritance links. An apparent benefit to transform ational approaches is that they allow the relationship between underlying form and m eaning to be stated in a straight­ forward. and often transparent, way." The question arises, how arc crosslinguistic generalizations about the relationship o f sem antic representation to over/ syntactic expression to be captured within a constructional approach'7 T his is the subject o f the following section. 4.2

G e n e r a liz a t io n s a c r o s s C o n s t r u c t io n s

G eneralizations across constructions concerning word order facts, casem arking properties, and links between sem antics and gram m atical relations can all be captured by stating these generalizations at a sufficiently high node in an inheritance hierarchy o f constructions. Thus, such generalizations are inherited through dom inated constructions, unless a particular construction prevents such inheritance by having a conflicting specification. The following is an overview of all the relations discussed in the previous chapter:

O n L in k in g

10 9

□

Leaving the polysemy and m etaphorical extensions out o f the diagram , we can represent som e o f the more general relations am ong constructions as follows:

Subj.-Pred. Construction

F igure 4 .2

110

C hap ter Four

The fact that English is an SVO language can be captured by specifying a word order constraint on the top node o f the diagram, at the level of the su b jectpredicate construction. Certain constructions further down the inheritance hierarchy, such as the topical ization construction or the locative there construc­ tion (not shown), can override the word order constraint with constructionspecific constraints. Thus generalizations about word order can be captured while at the same tim e other constructions with exceptional word orders are perm itted. Subregularities arc expressed sim ilarly by stating a generalization at a node that is interm ediate on the hierarchy. It should be stressed that if a generalization is construction-specific this does not entail that it is not part o f a recurring pattern crosslinguistically. We know that many languages have constructions closely analogous to, for instance, the English transitive, ditransitive, loeative, and topicalization constructions. It is quite possible that there is a universal inventory o f possible argument structure constructions relating form and meaning, and that particular languages make use o f a particular subset o f this inventory. A long with many other theories o f thematic roles, ours makes no assumption that them atic roles are prim itives (cf, Jackendoff 1972, 1983, 1987; Foley & Van Valin 1984; Rappaport and Levin 1988; Gropen et al. 1989; Pinker 1989; Van Valin 1990b). Instead, roles are taken to be slots in relational semantic structures. A rgum ent roles are defined to be slots in the semantic representation of particular constructions and participant rotes are defined to be slots in the rich sem antic representation o f predicates (cf. chapter 2). The linking o f se­ m antics to syntactic expression is claim ed to be generally determined within constructions, that is, at the level o f argum ent roles. At the same time, excep­ tional linking patterns may be stated as part o f particular lexical entries (cf. Fillm ore & Kay 1993).

4.2. i

Empirical W eaknesses of Construction-Independent U n k in g Rules It m ight be tem pting to think that individual constructions are not the

right level at which to capture generalizations about syntactic expression, and that instead very general linking rules m apping particular roles onto gram m ati­ cal relations or syntactic configurations should be a priori preferable. Such general linking theories have been proposed for some time. For ex­ am ple, Fillm ore (1968) suggested that subject selection was determ ined with reference to a fixed them atic role hierarchy; the highest available role on the hierarchy would be m apped onto the subject. More recent attem pts to relate argum ent structure and overt syntactic form in a general way can be found in, for exam ple, Foley & Van Valin 1984, C arter 1988. Pinker 1989. and Rappa-

O n L in k in g

I I I

port & Levin 1988. Such linking theories are motivated by the fact that there are intra- and inter-language generalizations about the kinds o f com plem ents particular predicates have. The attem pt, then, is based on the fact that clearly, syntactic form is not related in an arbitrary way to the sem antics o f predicates. In this section, 1 first review evidence that in a m onostratal account, conslruction-specific linking rules are required— that it is not possible to state all linking generalizations in a construction-independent way: certain m appings o f semantic arguments to gram m atical forms are only relevant to particular co n ­ structions (cf. also Koenig 1993). The two argum ent roles discussed are those of recipient and theme. Recipients In English, recipient argum ents (or the first argum ent o f an abstract predicate HAVE or RECEIV E) can be linked to three different gram m atical relations. W hich gram m atical relation is actually expressed depends on the construction at hand. For exam ple, in the ditransitive construction recipient arguments are expressed as objects: (20) Sam gave M ary a cake. Subj V Obj O bj2 They also appear in oblique phrases in the transfer-caused motion con­ struction:9 (21) Sam gave the piece o f land to his son. Subj V Obj Obi Recipient argum ents are also som etim es expressed as subjects: (22) Sam received/got/acquired a package. Subj V O bj, What we have then is the situation diagram m ed below. The sam e generally defined argument occurs overtly in different syntactic positions (bearing differ­ ent gramm atical relations). W hich syntactic position is actualized is determ ined by the construction, not by the them atic role in isolation. Recipient

/N

Obj | Subj Obi Figure 4 .3

Thus the mapping from sem antics to gram m atical relations is not determ ined by a function that is based solely on the them atic role to be expressed. Instead, we find cases wherein the syntactic expression is construction specific.

1 12

C h a p re r F o u r

Themes C onsider what the “ them e" argum ent would be mapped onto in a con­ struction-independent account if the “ them e” is defined to be an argument which undergoes a change of slate or location. A specific attem pt at such an argum ent has been form ulated within the monostratal linking theory o f LFG by L. Levin (1987), A lsina and M chom bo (1990), Bresnan and Kanerva (1989), Bresnan and Moshi (1989), Bresnan and Zacncn (1990), and Ackerman (1990). This theory is chosen here for discussion because to my knowledge it is the m ost detailed attem pt at a linking theory within a monostratal framework. In this theory, gram m atical relations are predicted from the argument structure o f particular predicates. A rgum ent structures are represented by argument (theta) role arrays, although there is no strong assumption that the argument roles are prim itives instead o f being derived from a richer decompositional sem antics. In fact, most proponents o f this theory suppose that the roles are shorthand for different argum ent places in some logical decomposition in the style made fam iliar by G enerative Sem antics (see Jackendoff 1972, 1983. 1987; Foley & Van Valin 1984; Rappaport and Levin 1988; Gropen et al. 1989; and Pinker 1989 for argum ents that thematic roles are not primitive). Tw o abstract features are postulated, [r] and [o], which categorize four types o f gram m atical relations: SUBJ [ —r, - o ] O BL0 [+ r, - o ] OBJ [ —r, -fo] O B J0 [ + r, + o ] (O BJ0 denotes the second object o f dilransitives.) The feature [r] stands for (sem antically) restricted; [o] stands for objective or objeci-like. Thematic roles are assigned features in two ways. On the one hand, they have an intrinsic classification (IC), w hich is said to be based on their inherent sem antic prop­ erties. A first approxim ation o f this basic classification is given below. Intrinsic Classification ( first approxim ation): • theme roles: [ —r] • all other roles: [ - o j On the other hand, roles receive a default assignment: the highest theta role on the proposed hierarchy receives a [ —r] feature as a default, the rest receive [ + r] (Bresnan & M oshi 1989; Alsina & M chom bo 1990). The hierarchy that is adopted is: agent > beneficiary > goal > instrum ent > theme > location A lsina and M chom bo (1990) propose that applicatives in Chichewa are formed

O n I in k in g

I 13

by a lexical rule which adds a “ dependent” argum ent to the argum ent structure o f the matrix verb; (23) illustrates this for co o k:w (23) cooku < a g t p a t> => cook, < a g t

p a t>

The “ dependent” -subscript on the theta role ‘0 ' is intended to capture a se­ mantic property that is claim ed to be loosely correlated with affectedness. O ther LFG accounts have described this sem antic property as “ applied” (B res­ nan & Moshi 1989), “ patient-like” (Bresnan 1990)yjr “ affected” (Ackerm an 1990). Unfortunately, this attribute is not fully explicated in any o f these analy­ ses. For the sake of consistency, 1 will refer to it as “ dependent” throughout. The rule in (23) is analogous to the sem antics-changing lexical rules pro­ posed in Levin & Rapoport 1988 and Pinker 1989. Several difficulties stem ­ ming from the semantic claims inherent in this type of approach have been discussed in chapter I. The prim ary focus within LFG , however, is on the link­ ing between lexical sem antics and surface syntax; it is this aspect o f the ap­ proach which is considered here. Alsina and M chom bo (1990) state the intrinsic classification assignm ent for the applicative construction in general term s:11 (24) W hen there is a theme and a(nother) dependent argum ent, then one will receive ( —r] and the other will receive [+ o ] (in languages that allow only one direct object, like Chichew a and English). Dependent recipient roles have a special status, since when present, they must occur directly after the verb, can be expressed as an object-m arkeron the verb, and can be passivized, whereas the co-occurring theme argum ent cannot (M orolong & Hyman 1977; Alsina & M chom bo 1990). In these ways the recipient dependent argum ent is direct-object-like, and like other direct objects may ap ­ pear as subject in passives. These facts are accounted for if a dependent recipi­ ent is necessarily either a surface SUBJ or OBJ. A lsina and M chom bo therefore propose that dependent recipient roles must receive a [ —r] classification, which distinguishes SUBJ and OBJ relations from other relations. We thus have the following revised list o f intrinsic classifications now: Intrinsic Classification (revised): • them e/patient roles: [ —r] or [+ o ] • “dependent” recipient role. [ —r] ■ other dependent roles: [ - r] or [+ o ] • all other roles: [ - o ] A nother exam ple o f a sem antics-changing lexical rule is proposed by A ck­ erman (1990) in order to account for the locative alternation involving verbs such as “ spray” and “ load” in Hungarian (exam ples are A ckerm an’s):

I 14

(25)

C h a p t e r (-our

a paraszt

( r a = )ra k ta

aszen at

a s z e k e rre

the p easan t (onto)loaded-3sg/D F.F the hay-ACC the w agon-suB L ‘T he p easan t Ioaded0 the hay o n to the w ag o n .’ (2 6 )

a paraszt

m eg = rakta

aszek eret

szendval

the p easan t PERF-loaded-3sg/DEF the w a g o n -A rc hay-lNSTR ‘T he p easan t load ed , the w agon with hay.’ T he locative argum ent o f meg = rakta in (26) is a dependent argument, which gets the intrinsic classification [ - r ] just as dependent arguments do on Alsina and M chom bo's account. Thus we have the following: (2 7 )

“ load0” < a g t loc them e>

“ lo ad ,” < a g t lo c ^ ,,^ ,., them e>

T he with variant given in (2 6 ) seems to fit squarely within the domain o f gen­ eralization in (24), which would predict that the them e argum ent should receive a [+ o ] intrinsic classification, ultim ately resulting in its being linked to an OBJ« relation like the theme argum ent in applicative constructions. However, as (2 6 ) shows, the theme argum ent in the presence o f another dependent argu­ ment is not linked to OBJ„ but rather mapped to OBL. In order to avoid the conclusion that locatives should be expressed in a way directly parallel to applicatives, A ckcrm an stipulates that the them e argum ent does not get the intrin­ sic classification [+ o ], but instead receives [+ r] as intrinsic classification and then [ - o] as a default.12 For languages that have ditransitive (or applicative) expressions and both form s o f the locative alternation, for exam ple, English and Chichew a,l, we now need to postulate the following intrinsic classifications: Intrinsic Classification (final formulation): 1. If there is another dependent role which is a recipient, thcme/patient role: [+ o ] 2. If there is another dependent role which is a locative, theme/patient role: [+ r] 3. If th ere is no o th e r d e p e n d e n t ro le, th e m e /p a tie n t: [ — r] 4. dependent recipient role: [ - r] 5. all other roles: [ - o ] T hese rules are in part language specific, since not all languages have all of the relevant constructions. More crucially for our purposes, observe the three feature-assigning rules ( 1 - 3 ) involving the theme role. Each of these rules is sen­ sitive to other roles present in the argum ent structure. This context sensitivity is expected on a constructional approach to linking, bur may have been precisely w hat the LFG linking theory had specifically wanted to avoid (M chom bo, personal com m unication).

O n L in k in g

I 15

Taking the A lsina & M chombo and A ckerm an analyses together, we find that the theme can receive the intrinsic classification [ - r ] , [ + o], or l + r] (de­ pending on what other roles are assigned), that is, all but one possible assign­ ment ( [ - o ] ) . It becom es difficult to see w hat is supposed to be m eant by “ intrinsic classification.” T hat is, w hat is called “ intrinsic classification" is assigned not on the basis o f intrinsic properties o f the argum ent, but rather on the basis o f properties o f other, co-occurring arguments. Moreover, as A cker­ man ( 1992, note 21) observes, this assignm ent o f features still does not account for the expression of themes as subjects in the intransitive m otion construction. Adding to this difficulty o f too w ide a range of possible intrinsic classifica­ tions for a given role is the fact that these classifications are not assigned on the basis o f independent evidence. W hat determ ines which o f the possible clas­ sifications is actually assigned is w hat construction is supposed to be predicted. In this way, the classifications becom e circular: the recipient is dependent, and thus [ - r] as opposed to [ - o], just in case it is supposed to be the OBJ in the ditransitive construction. Since the notion “ dependent” is never adequately defined, no independent criterion for the assignm ent o f dependent status is o f­ fered. In short, since the linking theory can by its nature capture the necessary facts— because the abstract features [r] and [o] are all that is needed to code the gramm atical relations— unless independent criteria for assigning gram ­ matical relation features are found, the formalism only serves to code the syn­ tactic structure that is supposed to be predicted. The general point is that the linking o f the theme role to an overt gram m ati­ cal relation crucially depends on w hat other argum ents are present. The same point can be made equally well with English data. The theme argum ent, that is, the entity whose motion or location is at issue, can occur as subject in the intransitive motion construction: (28) The boy ran home. As object in sim ple transitives: (29) Pat moved the bat. As oblique in a version of the locative alternation: (30) Pat loaded the truck with hay. And as the second object in ditransitives: (31) Pat threw Chris the ball. W hich gram m atical relation is actually realized depends on which construction is expressed.

I 16

4 .2.2

C hap ter Fo u r

Capturing Linking Generalizations with Constructions O nce we recognize the large num ber of construction-specific linking

rules, the question arises w hether there are any more general linking rules. In this section, D ow ty’s (1991) account o f linking generalizations is discussed, and it is shown that his observations can be adapted to the present framework quite naturally. Dowty (1991), follow ing the spirit o f Foley and Van Valin (1984), suggests two general m acro-role types: Proto-Agent and Proto-Patient (cf. Foley & Van Valin’s “ A ctor” and “ U ndergoer” ). These roles are defined as prototype con­ cepts, m uch like the prototypes discussed by Rosch and her colleagues (e.g., Rosch and M ervis 1975). The lists o f Proto-Agent and Proto-Patient properties draw in part on the em pirical findings o f Keenan (1976, 1984), who detailed many properties associated with Subject and O bject crosslinguistically. Proto-A gent properties: 1. volitional involvem ent in the event or state 2. sentience (and/or perception) 3. causing an event or change o f state in another participant 4. m ovem ent (relative to the position of another participant) 5. exists independently o f the event named by the verb Proto-Patient properties: 1. undergoes change o f state 2. increm ental theme 3. causally affected by another participant 4. stationary relative to m ovem ent o f another participant 5. does not exist independently o f the event, or not at all. G iven these definitions o f Proto-Patient and Proto-Agent properties, Dowty proposes the following principle: Argument Selection Principle: In predicates with grammatical subject and object, the argument for which the predicate entails the greatest number of Proto-Agent properties will be lexicalized as the subject; the argument for which the predicate entails the greatest number of Proto-Patient properties will be lexicalizcd as the direct object. Proto-roles reflect higher-order generalizations about lexical meanings. There­ fore, the principles above allow for a small num ber o f lexical exceptions (e.g., undergo, sustain, tolerate, receive, inherit).14

O n L in k in g

I 17

In syntactically ergative languages, for exam ple Dyirbal (Dixon 1972), cer­ tain Mayan languages including Mann (England 1983) and Q uiche (Treschsel 1982), the subject in transitive clauses is the argum ent with more Proto-Patient properties. For these languages Dowty reverses the syntactic association. He notes, however, that “what we do not find, even in split ergat/vity, is ‘random ’ alignment from one verb to another, e.g. ‘build’ with A gent absolutive but ‘kill’ with Patient absolutive” (p. 582). It Lurns out that D ow ty’s linking generalizations are naturally accounted for in the present framework. Notice the dom ain of application o f D ow ty’s prin­ ciple: “ In predicates with gram m atical subject and object. . . .” Clearly the principle is only relevant if the transitive construction is involved. Therefore, the generalization that if there is a SUBJ and an OBJ, then the role that is more agent-like, the “ pioto-agent,” is linked with SUBJ and the “ proto-patient” role is linked with OBJ, can be captured by specifying these linking rules within a skeletal transitive construction, and allow ing other constructions to inherit from this construction.

Transitive Construction

Sem (

Syn

proto-agent

proto-patient

I

I

SUBJ

OBJ

Figure 4 .4

In syntactically ergative languages, the transitive construction has the re­ verse linking, so that SUBJ is linked with the proto-patient role and OBJ is linked with the proto-agent role; these linkings are then inherited by other con­ structions as long as those constructions' specifications do not conflict. Just like D ow ty’s proposal, the constructional account allow s for a limited number of lexical exceptions. Exceptions are cases which do not inherit from (i.e., are not motivated by) the transitive construction, or cases which inherit only the form but override the meaning o f the construction. The num ber of exceptions is limited because such non-m otivated (or less well m otivated) cases exist at a cost to the overall system, in accord with the Principle o f M axi­ mized M otivation (cf. chapter 3). Thus the inheritance hierarchy allow s us to capture the relevant generalizations w hile at the same tim e allow ing for a lim ­ ited num ber o f lexicalized exceptions.

118

4.2.3

C h a p te r F o u r

Som e Speculations on the Semantics of the Transitive Construction

By being posited as a unitary structure with sem antics consisting merely o f two abstract “ proto-roles,” the transitive construction in figure 4.4 is repre­ sented as if it had a single, very general abstract meaning. It may well turn out that this construction is more like the others discussed in this monograph in that it may be m ore felicitous to assign a fam ily o f related m eanings lo it, with the prototypical “ transitive scene” (cf. L akoff 1977; H opper & Thompson 1980; Rice 1987a,b) being the central sense. In this case, the central sense would be quite specific, being that o f a volitional actor affecting an inanimate patient— a causative event.'5 Extensions from the prototype would license a w ider range o f transitive expressions. O ne indication that m ultiple senses may be involved stem s from the fact that there exist clusters o f cases which are not instances o f the general semantic tem plate given in figure 4.4. For exam ple, alongside the “exceptional” receive and inherit, we find the closely related verbs have, own, acquire, get. The ex­ istence o f so many cases m ilitates against the idea that they are in fact excep­ tional. On a m ultiple-sense view, it is possible to posit additional senses o f the transitive construction, for exam ple a sense ‘h a v e (X, Y )’ (cf. Pinker 1989). A nother benefit o f the m ultiple-sense view is that it is able to capture the generalization, discussed in chapter 2, that crosslinguistically language learn­ ers apply transitive m arkers first to expressions designating prototypical tran­ sitive scenes, before extending them to less prototypically transitive ex ­ pressions (Slobin 1985). Languages differ in how and to what extent the transitive construction is extended to express nonprototypical sem antically transitive scenes. In recog­ nizing distinct senses, we might be able to develop a model in which we would be able to isolate the true crosslinguistic generalizations. In other words, lan­ guages differ in how they express noncanonical transitive predicates, such as predicates o f possession ( “ have” , “ ow n” ), but languages express semantically canonical transitive predicates by m eans o f a transitive construction."’ Lan­ guage-specific idiosyncrasies would arise, according to this view, in just how languages extend their inventory o f gram m atical argum ent structure construc­ tions to cover expressive requirem ents. Linking generalizations apply to the basic senses o f constructions. A ddi­ tional senses o f the transitive construction, related to it by polysemy links (cf. chapter 3), would inherit the linking specifications o f the construction they are dom inated by. Therefore, once we posit extensions o f the basic transitive con-

O n U n k in g

I 19

siruction, the fact that these extensions have the sam e syntactic expression would follow. 4.3

C o n c l u s io n

To summ arize, this chapter has presented general argum ents for a m on­ ostratal approach to the relation between overt syntactic expressions and se­ mantic representations. It has been suggested that the degree o f regularity in the relation between sem antic role types and overt syntactic expression is sometimes exaggerated, and that many linking generalizations are construction specific. The cross-constructional generalizations that do exist are naturally captured in the present fram ework by slating the relevant regularities at a high node in the hierarchy o f constructions; subregularities or m inor patterns are captured by stating the respective generalizations at interm ediate nodes. Ex­ ceptions are allowed to exist, but only at a cost to the overall system.

5 Partial Productivity

If you invent a verb, say greem, which refers to an intended act of communication by speech and describes the physical characteristics of the act (say a loud, hoarse quality), then you know . . . it will be possible to greem, to greem for someone to get you a glass of water, to greem to your sister about the price of doughnuts, to greem “ Ecch" at your enemies, to have your greem frighten the baby, to greem to me that my examples are absurd, and to give a greem when you see the explanation. Arnold Zwicky (1971) 5. i

In t r o d u c t io n

It has been a long-standing puzzle that many constructions are used som ew hat productively (as implied by the above quotation),' yet resist full pro­ ductivity. This chapter addresses the issue of partial productivity for the most part by exam ining the ditransitive construction as an example. In section 5.5.2 other constructions, which can be seen to be either more productive or less productive, are considered. The ditransitive construction can be used som ew hat productively; that is, the construction can be extended to new and hypothetical verb forms (e.g., Wasow 1981). For exam ple, the new lexical item fa x can be used ditransitivcly as in (1): (1)

Joe faxed Bob the report.

Also, hypothetical lexical item s are readily adapted to the ditransitive syntax. As M arantz (1984) notes, if we define a new verb, shin, to mean “ to kick with the shin,” it is quite natural for us to allow this new verb to be used ditransi­ tivcly, as in (2): (2)

Joe shinned his team m ate the ball. (p. 177)

Experim ental evidence confirm s the fact that speakers extend constructional patterns for use with novel verbs (Pinker, Lebeaux & Frost 1987; Pinker 1989; M aratsoset al. 1987; Gropen et al. 1989, 1991; Braine ct al. 1990). At the same time, the ditransitive pattern is not com pletely productive within any generally defined class of verbs. Seem ingly closely related words show

120

Partial Productivity

12 1

distinct differences as to w hether they allow ditransitive syntax. T he following contrasts exist in many dialects: (3)

a. Joe gave the earthquake relief fund $5. b. *Joe donated the earthquake relief fund $5

(4)

a. Joe told Mary a story. b. *Joe whispered Mary a story.

(5)

a. Joe baked M ary a cake. b. *Joe iced Mary a cake.

Brown and Hanlon (1970) have argued that children are neither corrected nor miscomprehended more often when they speak ungram m atically, so that they have no recourse to "negative evidence” that could allow them to either un­ learn or avoid learning the above type o f ungram m atical sentences (cf. Braine 1971; Baker 1979). The standard solution to the no-negative-evidence problem in the case of vocabulary learning is to assum e that there is indirect negative evidence in the form of attested input, assum ing a principle that synonym y is avoided (cf. d is­ cussion in chapter 3). That is, a child may overgeneralize the past tense co n ­ struction to produce coined as the past tense of come, but upon learning that came is synonymous, the child will expunge corned from her vocabulary, since she will assume that the language does not have two term s corned and cam e which are synonymous. Such indirect evidence is not forthcom ing in an o bvi­ ous way in the case of alternative syntactic patterns. It is not likely that the child simply expunges (6) upon hearing (7) because many verbs (e.g., give) do occur in both forms. (6) *H e w hispered the woman the news. (7)

He whispered the news to the woman.

Moreover, as noted above, experim ental evidence shows that children do not learn how to use syntactic patterns entirely conservatively, that is to say, solely on the basis o f the input. If properly prim ed, they are w illing to extend their use of verbs to previously unheard but related patterns. An apparent paradox arises then, since if speakers have a productive m echa­ nism that allows them to extend the use of the ditransitive syntax to new and novel verbs, it is not clear w hat prevents them from overgeneralizing to pro­ duce the above ill-formed exam ples (3 b -5 b ). This paradox is often sidestepped in linguistic theories. Thus, w hether relation-changing lexical rules are intended to be purely redundant generaliza­

122

C h a p r e r F iv e

tions over stored item s in a fixed lexicon, or rather generative rules which produce new forms productively, is often not made entirely clear. Jackendoff (1975), for example, states that his lexical rules are intended only to account for existing regularities (both m orphological and sem antic) within the lexicon. These rules are represented by two-way arrows which encode the sym m etric relation “ is lexically related to.” This aspect o f Jackendoff’s ac­ count is crucial, since he argues explicitly against L akoff’s (1965) proposal that productive rules generate “ hypothetical lexical entries.” However, Jackendoff also suggests that “ after a redundancy rule is learned, it can be used generatively, producing a class o f partially specified possible lexical entries” (p. 668). Bresnan (1982) also attem pts to find a middle ground between nonproduc­ tive and fully productive rules. W hile the lexical rules o f LFG are explicitly conceived o f as “ redundancy rules,” the m etaphor o f a lexically changing pro­ cess is pervasive. The following is Bresnan's early description o f the passive lexical rule: Passivization in English Functional change: M orphological change:

(SUBJ) —» 0 / (BY OBJ) (OBJ -> (SUBJ) V —» V|Par1|

The use o f single-headed arrows and the word “change” indicate that the rule is a generative relation-c/iangm g rule. In fact, the notion o f a "redundancy rule" itself is slightly oxym oronic, since a redundant statem ent o f regularity is not in any normal sense rule-like. In the rem a in d ero f this chapter, a resolution o f the paradox o f partial produc­ tivity is suggested, involving two types o f learning mechanisms. The first is a certain type o f indirect negative evidence; the second m echanism , presumably working in tandem with the first, draws largely on work by Pinker (1989) and Levin (1993) and the related experim ental evidence o f Gropen et al. (1989). 5.2

In d ir e c t N

e g a t iv e

E v id e n c e

1 do not attem pt to survey the full range o f efforts to suggest that some type o f indirect negative evidence is available here (see Bowcrman 1988 and Pinker 1989 for detailed discussion o f the problem and critiques o f many pos­ sible solutions), but there is one possibility (raised in Pinker 1981, 1984, and then rejected in Pinker 1989) that deserves further study. Since we have assum ed that no two constructions are entirely synonymous both sem antically and pragm atically (cf. chapter 3), it should be possible to find contexts in which a given construction is the most preferred. If the pre­

Partial P ro d u c tivity

12 3

ferred form is not used, then the child is able to tentatively infer that that form is disallowed. The inference would have to be tentative, since it is unrealistic to expect speakers to system atically use the most felicitous form in all contexts. However, if the situation repeats itself several tim es, the ch ild ’s tentative hy­ pothesis may becom e a fairly strong conviction. In this way, children would have the opportunity to unlearn certain ovcrgeneralizations. A simple case that may illustrate this is lexical and periphrastic causatives. It is well known that lexical causatives are used for cases o f direct causation, whereas periphrastic causatives may be used for indirect causation (cf. 7.4.2). Therefore, for exam ple, after seeing a m agician m ake a bird disappear, the child may expect to hear a lexical causative as in (8), given that the causation is direct: (8) *The m agician disappeared the bird. Instead, however, the child may hear a periphrastic causative: (9)

Look! The magician m ade the bird disappear.

The child may now tentatively hypothesize that the lexical causative is un­ available. That is, since the causation is direct, the lexical causative would be preferable if it were an option. As another example, consider a child's strategy in determ ining w hether a given verb can occur in the ditransitive construction. As noted by ErteschikShir (1979), and discussed in section 3.4.2 above, the ditransitive and its prepo­ sitional paraphrase with to differ in the inform ation structure o f the clause. In particular, the ditransitive construction requires that the recipient argum ent be nonfocused (or “ non-dom inant” in E rteschik-Shir’s term inology) and the transferred entity be focused ( “dom inant” ). Prepositional paraphrases prefer the opposite information structure: the recipient tends to be focused, the trans­ ferred entity nonfocused. Both o f these generalizations are motivated by the fact that focused information tends to com e at the end o f the nuclear clause. If the recipient is nonfocused and the transferred entity is focused, we find the ditransitive more acceptable than the prepositional paraphrase: (10) a. b.

Sally gave him a brand-new red Volkswagen. > Sally gave a brand-new red Volkswagen to him.

If the recipient argum ent is focused and the transferred entity nonfocused, we find the reverse situation: (11) a. b

Sally gave that to a charm ing young man. > Sally gave a charm ing young man that.

124

C h a p t e r Five

W hen using verbs which can occur in both constructions, speakers are free to exploit the difference in pragm atic structure. There is, in fact, evidence that children are sensitive to these pragm atic factors (Gropen et al. 1989). Indirect evidence would arise, then, from situations in which the discourse context m atches a certain form but the speaker nevertheless uses a less felici­ tous form. For exam ple, speakers use the prepositional form for w hisper even when the inform ation to be conveyed more closely matches the information structure o f the ditransitive construction. Thus, if a child hears (12) instead o f (13), when the latter might be expected given the fact that the news is the focused inform ation, the child will infer that the ditransitive form is not a pos­ sibility for whisper. (12) Sally w hispered som e terrible news to him. ( 13) *Sally w hispered him some terrible news. Pinker raises this possibility in several places (Pinker 1981, 1984:400). However, he ultim ately rejects the idea that this m echanism could be sufficient for learning to disallow particular form s, for two reasons (Pinker 1989:16). The first objection he raises is that children’s sensitivity to discourse contexts is statistical, not absolute. T hat is, children do not treat discourse effects as a determ inant factor in choosing alternate argum ent structures; they are more likely to use the argum ent structure with the better-suited pragm atics, but they do not always do so (G ropen et al. 1989). However, the very fact that children are more likely to use the construction with better-suited pragm atics is suffi­ cient to show that they do have an im plicit know ledge o f the inform ation struc­ ture and are able to attend to it. For instance, it is possible that a child wouldn't notice the first time that whisper was used with the focus on the transferred entity, or the second time. But eventually, the child would presum ably notice; at that tim e she would be able to use the input evidence to form the hypothesis that whisper cannot be used in the ditransitive form. Even if we strengthen P inker’s first objection and assume that adult's sensi­ tivity to discourse contexts is also only statistical and not absolute, we do not underm ine this strategy. That is, even if we concede that neither adults nor children can be assum ed to always use the most felicitous form, this strategy is not ruled out. All that is required is that the child be capable o f recognizing a statistical correlation in the input data (K apur 1993). T he second objection Pinker raises stem s from his assum ption that other focusing devices such as pronouns, cleft constructions, and contrastive stress can be used to override the default differences in inform ation structure between alternative argum ent structures. On that view, adult speakers would be able to com pensate for using less-preferred argum ent structures by overlaying these

Partial P rod u c tivity

125

less-preferred argument structures with various focusing devices (e.g., pro­ nouns, focus constructions, and stress), thus altering the inform ation structure encoded by particular argument structures as a default. Assuming that speakers make use o f these strategies, the input would be for the most part optim al, and children would have no reason to infer that the speaker in a given situation would have used a different argum ent structure if he could have. However, this suggestion is ultimately not persuasive, since focusing devices are not able to alter or override the inform ation structure of the clause but instead are required to obey the independently existing inform ation structure o f the clause. For exam ple, pronouns are preferred in nonfocus positions: (14) a. She gave it to a woman. > b. *She gave a woman it. (15) a. She gave her a brand-new house. > b. #She gave a brand-new house to her. Similarly, focus has been argued to only pick out argum ents that are in focusable positions as defined for a given construction. For exam ple, as m entioned above. Erteschik-Shir (1979) has argued that the recipient argum ent of the d i­ transitive construction is not available as focus because the construction re­ quires that argument to be nonfocused (or “ non-dom inant” ): (16) a.??W ho did you give the book? b.??lt was M ary you gave the book. c.??W as it Mary you gave the book? Finally, stress is also more felicitous on argum ents that are in focus position, thereby generally em phasizing the inform ation structure rather than overriding it. Thus exam ple (17a) is more felicitous than (17b): (17) a. She gave that to A W O M A N SHE JU ST MET. > b. #She gave A W OM AN SHE JU ST M ET that. Therefore, focusing devices might well be lim ited to giving the child addi­ tional evidence for the inform ation structure o f the clause, rather than serving to dilute other evidence by providing ways for the adult speaker to circum vent the inform ation structure associated with a particular argum ent structure. Since two constructions generally differ either sem antically or pragm atically, the hy­ pothesis that indirect negative evidence is inferred from hearing a verb in a less-than-optim al construction deserves further study. 5.3

C ir c u m s c r ib in g V

erb

C lasses

Pinker (1989), arguing against any negative evidence (direct or indirect), ultimately provides a different, com pelling resolution o f the paradox of partial

12 6

C h a p t e r hive

productivity. A broad-range rule is proposed to capture the necessary condi­ tions for a verb’s occurrence in additional syntactic frames. In the case o f the ditransitive, Pinker posits a broad-range rule that slates in effect that a “ pro­ spective possessor” must be involved— that is, the first object referent must be understood to be a prospective possessor. This general rule does not provide sufficient conditions, however, there being many verbs that can be understood to involve a prospective possessor w hich do not allow ditransitive syntax (e.g., donate, contribute, pull, shout, choose, credit, say). D rawing on work by G reen (1974) (and Levin 1985 and Rappaport & Levin (1985) for the locative alternation), Pinker suggests that sufficient conditions are determ ined by a set o f narrow-range rules w hich classify verbs into nar­ rowly defined sem antic classes. The specific classes that Pinker proposes are the follow ing (cf. also Gropen et al. 1989; Levin 1993); 1. Verbs that inherently signify acts o f giving: e.g., give, pass, hand, sell, trade, lend, s e n r , fe e d 2. Verbs o f instantaneous causation of ballistic motion: e.g., throw, toss, flip, slap, poke, fling, shoot, blast 3. Verbs o f sending: e.g., send, mail, ship 4. Verbs o f continuous causation o f accom panied motion in a deictically specified direction: bring, take 5. Verbs o f future having (involving a com m itm ent that a person will have som ething at som e later point): e.g., offer, prom ise, bequeath, leave, refer, forw ard, allocate, guarantee, allot, assign, advance, award, reserve, grant 6. Verbs o f com m unicated message: e.g., tell, show, ask, teach, pose, write, spin, read, quote, cite 7. Verbs o f instrum ent o f com m unication: e.g., radio, e-mail, tele­ graph, wire, telephone, netmail, fa x 8. Verbs o f creation: e.g., bake, make, build, cook, sew, knit, toss (when a salad results),/!* (when a meal results), p o u r (when a drink results) 9. Verbs o f obtaining: e.g., get, buy, find, steal, order, win, earn, grab It may seem that if we adm it the possibility o f indirect negative evidence as suggested above, there is no need to adopt Pinker’s suggestion that narrowly defined sem antic classes also play a role in the acquisition o f argument struc­ ture. However, circum scribing narrowly defined classes o f verbs to be associ­ ated with a particular construction will allow us to account for extremely low -frequency or novel non-alternating verbs (since the assumed notion o f in­ direct negative evidence presupposes hearing the verb in a non-optimal con­ struction on several occasions).3 For exam ple, taking Z w icky’s exam ple of a

Partial P ro d u c tivity

127

novel verb greem, defined as a m anner-of-speaking verb referring to speech o f a loud, hoarse quality, speakers presumably know that they cannot say (18): (18) *He greemed her the news. This knowledge cannot be attributed to any kind of indirect negative evidence, because the verb is novel; speakers would not have had a chance to unlearn or avoid learning it in this use. O ther situations in w hich the type of indirect nega­ tive evidence suggested above would not be an aid to acquisition m ight include cases in which the construction in question is so low-frequency that the child can never with any modicum o f confidence expect its occurrence, and cases in which there is no construction which is closely enough related sem antically to the target construction so that the child would be able to infer that the speaker would have used the target construction if possible.3 Moreover, the generaliza­ tions Pinker describes are real; it is necessary to account for the fact that verbs which are used in particular argum ent structures do often fall into sim ilarity clusters (cf. Levin 1993). There is no reason not to believe that children exploit multiple sources of evidence for learning argum ent structure; it is suggested here that they m ake use o f narrowly defined verb classes as well as appealing to some degree o f indirect negative evidence as described above. Before continuing with Pinker's argum ent, we might make several small com ments on the particular set o f subclasses he proposes, listed above. The fifth subclass, “ verbs o f future having," can be seen to conflate three distinct subclasses. Some of the verbs are used in expressions which imply that the subject argument actually acts to cause the first object argum ent to receive the second object argum ent at som e later point in tim e (e.g., bequeath, leave, fo r ­ ward, allocate, assign). In other cases, only if the conditions o f satisfaction (Searle 1983) associated with the act denoted by the predicate hold does the subject argum ent cause the first object argum ent to receive the second object argum ent at som e later point in tim e (e.g., prom ise, guarantee, owe). Finally, some verbs are used in expressions w hich imply that the subject argum ent only enables the first object argum ent to receive the second object argum ent (e.g., permit, allow) (cf. discussion in section 3.3.2). The sixth class, “ verbs of com m unicated m essage,” should be understood to include verbs whose inherent sem antics involves a com m unicative act, in order to distinguish this class from sim ilar verbs such as say. assert, claim, and doubt which might be described as verbs o f propositional attitude. U nderstood in this way, several o f the verbs listed by Pinker seem to be misclassified; for example, pose and spin do not obviously fall into the class o f “ verbs o f co m ­ municated m essage,” and accordingly (at least in my dialect) are not readily dativizable:

12 8

C h a p t e r F ive

(19) a.?*Bill posed him a problem . b.?*Bill spun her a fairy tale. Both this class and the seventh class, “ verbs o f instrum ent o f com m unication,” should be classified as m etaphorical classes since they are based on a system­ atic m etaphor that involves understanding com m unicated inform ation as being linguistically packaged and exchanged between interlocutors (Reddy 1979). Finally, at least one additional subclass should be added to the list, namely, verbs o f refusal such as refuse, deny. Expressions involving these verbs, like (20a, b), imply that the subject argum ent refuses to causc the first object argu­ ment to receive the second object argument. (20) a. b.

Bill refused Joe a raise. The com m ittee denied him a prom otion.

In any case, we need only accept the spirit of Pinker’s analysis— that there is a need to identify narrowly defined sem antic subclasses— in order to accept his conclusion that this type o f narrow circum scription allow s us to capture the fact that subclasses o f verbs w hich refer to the same kinds o f general events as the ones listed, but do not fall into any o f the above particular classes, fail to dativize. His exam ples o f such nondativizing classes are as follows: 1. Verbs o f fulfilling (X gives som ething to Y that Y deserves, needs, or is w orthy of): * / presented him the award; * I credited him the discovery. * B ill entrusted/trusted him the sacred chalice; * / supplied them a bag o f gro­ ceries. (I would also include in this class concede, furnish, and donate.] 2. Verbs o f continuous causation o f accom panied motion in some m an­ ner: *1 pulled/carried/pushed/schlepped/lifted/low ered/hauledJohn the box. 3. Verbs o f m anner o f speaking: * John shouted/scream ed/m urm ured/ whispered/yodeled B ill the news. 4. Verbs o f proposition and propositional attitudes: *1 said/asserted/ questioned/claim ed/doubted her something. 5. Verbs o f choosing: * / chose/picked/selected/favored/indicatedher a dress. Gropen el al. (1989) provide experim ental evidence to show that speakers are sensitive to certain m orphophonological constraints. In particular, verbs with particular m orphem es such as per-, con-, -mil, -sume and polysyllabic verbs with non-initial stress are disallow ed from participation in the ditransitive construction. These constraints largely coincide with distinctions between Latinate and native vocabulary, and between specialized and more basic vocabu­ lary; however, we clearly would not w ant to ascribe recourse to etymological

Partial P ro d u c tivity

12 9

information to children, and the experim ents in support o f these particular constraints controlled for sem antic inform ation. Therefore, the constraints are stated in term s o f morphophonology. They are used to explain the following: (21) Chris b o u g h t/* p u rch ased /* o b tain ed /* co llected h im some food. (22) Jan to ld /* ex p la in e d /* re p o rted /* a n n o u n ce d C h ris a story. However, the constraints do not apply to every narrowly defined class o f verbs. Verbs of future having, in particular, are not subject to them: (23) Chris assigned/allotted/guaranteed/bequeathed him the tickets. The class o f instrum ent-of-com m unication verbs and the class o f creation verbs also include verbs which are exceptions to the m orphophonological constraints: (24) Chris e-m ailed/radioed/arpanetted him a message. (25) Chris xeroxed/therm ofaxed/nroffed him a copy. Gropen et al. suggest that each o f the verbs in (2 4 -2 5 ) is classified, indepen­ dently o f the morphological criteria, as a special kind o f com plex stem having a noun or name as its root. They cite evidence that tacit know ledge of a w ord’s stem being from another syntactic category allow s it to be treated specially with respect to morphological processes (cf. Pinker & Prince 1988). To account for these cases, we can state the generalization that a verb from any class which is understood to have a noun or nam e as its root is not constrained by the morphophonological constraints. The narrowly defined subclasses o f verbs together with the m orphophono­ logical constraints provide a high degree o f predictive power. A new or non­ sense verb which falls into one o f the recognized narrow classes o f verbs and which, if applicable, obeys the m orphophonological constraints is autom ati­ cally licensed to be used ditransitively (but see the next section). Verbs in co n ­ flict with these requirem ents are ruled out. This circum scription o f narrow domains in which the ditransitive is productive goes a long way toward ac­ counting for the apparent paradox that Pinker set out to resolve: that the ditran­ sitive syntax can be extended to new and novel verbs, but at the sam e tim e is not available to all verbs o f any broadly defined class. 5.4

E x c e p t io n s

The above generalizations are com pelling, and in fact every researcher who has studied the sem antics o f the ditransitive construction in any detail has found it necessary to classify verbs which occur in the construction as belong­ ing to narrowly defined subclasses as a descriptive device (cf. G reen 1974;

13 0

C h a p te r Five

Oehrle 1976; W ierzbicka 1986). Still, there are various kinds of exceptions to the generalizations ju st described. First, there are a couple o f members in some subclasses which do occur, yet the subclasses are not fully productive. Second, there is at tim es a certain degree of variability in judgm ents for verbs which are supposedly within the same narrowly defined class. Finally, there are excep­ tional verbs such as envy and forgive w hich do occur in the ditransitive con­ struction although they do not entail the relevant semantics. Each of these cases is discussed in turn. In section 5.3, an interpretation of the nature o f the verb classes is suggested which can naturally account for all o f these seem ingly problem atic phenom ena.

5.4.1

Unproductive Subclasses The small classes of verbs of perm ission (permit, allow) and verbs of

refusal ( refuse, deny) are unique in not form ing productive subclasses: (26) Sally perm itted/allow ed/* let/*enabled Bob a kiss. (27) Sally refu sed /denied/*prevented/*disallow ed/*fo rb ad eh im a kiss. These classes actually have a slightly different status in the theory proposed by Pinker (1989), because the verbs in these classes do not alternate with prepo­ sitional paraphrases. Thus on Pinker’s account, these sem antically related verbs are not eligible to undergo the lexical rule. However, since we are not postulat­ ing a lexical rule, we cannot appeal to the sam e solution. We need another way to account for their lack o f productivity. 5 .4.2

Differences in Judgm ent w ithin Classes

An expected source o f idiosyncrasy stems from the fact that the deter­ mination o f the narrowly defined class w hich a given verb belongs in is not always entirely clear-cut. For exam ple, I have suggested that bequeath falls into the dativizing class o f verbs o f future having, along with leave, forw ard, allocate, etc. However, it seem s that on sem antic grounds it might be equally plausible to instead classify bequeath in the nondativizing class of verbs of fulfilling (X gives som ething to Y that Y deserves, needs, or is worthy of), along with present, credit, entrust, donate, etc. Because o f these two classifi­ cation possibilities, we would expect bequeath in fact to dativize in some dia­ lects and not to dativize in others. In general, in the case of verbs that on the basis o f their m eaning may fall into one o f two classes, one which can appear dilransilively and one which cannot, we would expect to find some dialectal variation as to w hether these verbs can be used ditransitively. A nother source of lexical idiosyncrasy is evidenced by the fact that speak­

Partial P ro d u c tivity

13 1

ers occasionally report different degrees o f gram m atically even am ong verbs which are uncontroversially within the sam e narrow -range class. For exam ple, throw and blast both fall within the class o f “ verbs o f instantaneous causation of ballistic motion” (Pinker 1989; cf. above), yet (28) is decidedly better for many speakers than (29): (28) She threw him a cannonball. (29) *She blasted him a cannonball. Similarly, (30) is judged to be more gram m atical than (31), although both d e­ sign and create should fail to dativize because o f the verbs’ non-initial stress. (30) ?Sally designed him a sculpture. ( 3 1)??Sally created him a sculpture. These facts are not obviously accounted for on Pinker’s proposal, since accord­ ing to his theory the productive rule should operate blindly within narrowly defined classes; there is no reason to expect som e instances to be judged more acceptable than others.4

5.4.3 Positive Exceptions There are a few ditransitive expressions that do not entail any associated transfer. Some uses o f ask can be fit into the pattern described above if they are interpreted as instances of the m etaphor involving understanding inform ation as traveling from speaker to hearer (cf. Reddy 1979). For exam ple, (32) can be understood to mean that Amy caused Sam to “ receive" a question: (32) Amy asked Sam a question. However, other uses o f ask are clearly exceptional; consider (33): (33) Amy asked Sam his nam e/his birthday/his m arital status. This type o f exam ple clearly does not imply that Sam potentially receives his name, his birthday, or his marital status. G rim shaw (1979) discusses these “concealed questions" at some length. She argues that noun phrases such as those above, which are questions semantically, can appear as argum ents o f any verb which subcategorizes for an NP in that position and which selects for a question com plem ent. Thus exam ple (33) is motivated by factors that are independent of the ditransitive construction, resulting in a case o f “ targetstm cture conspiracy” in the sense o f G reen (1973).5 Forgive and especially envy, as used in (34) and (35), respectively, are also exceptional:

13 2

C h a p t e r F ive

(34) He forgave her her sins. (35) He envied the prince his fortune. The subjects in these cases arc not causal, and no reception is involved. How­ ever, these predicates have illum inating semantic histories. Forgive and envy historically had senses that were closely related to give. Forgive used to mean “ to give or g ran t” (OED). Envy used to mean “ to give grudgingly” or “ to refuse to give a thing to” (OED). This of coursc is not evidence that fo rg ive or envy are part o f the synchronic sem antic pattern outlined above. But the his­ torical facts do suggest that these predicates were at least at one time associated with this sort of pattern. Correspondingly, these facts also suggest that a con­ struction can occasionally be frozen without continuing reference to the origi­ nal semantics. However, it seem s reasonable that syntactic change should tend toward pat­ terns that are more transparent to the speaker. If the construction with the se­ m antics outlined here is psychologically real, then it would be natural for odd cases o f ditransitives involving fo rg ive and envy to drop out of use. I myself find archaic-sounding scntcnces involving fo rg ive and envy much more ac­ ceptable than m odern-sounding sentences; for example: (36) a. She forgave him his sins. b.?*S he forgave him his goof. (37) a. She envied him his vast fortune. b.?*S he envied him his extensive stock portfolio. And in fact, other speakers are even less accepting o f these constructions. In attem pting to explain the idea o f positive exceptions to a class o f undergraduate cognitive science students, I wrote sentence (36a) and (37a) on the board. In response an audible groan arose from the class. W hen asked what was wrong, the students said they didn’t find those sentences acceptable (this judgm ent was held by more than half o f them). Thus it seems indeed that envy and forgive are dropping out o f the language (at least am ong speakers under twenty-one), ju st as we would expect if the sem antics associated with the ditransitive pattern were synchronically real. N onetheless, envy and fo rg ive have been exceptions for som e time, and have been learned as such by generations o f speakers. Thus an adequate account of gram m ar must allow for some degree o f lexical idiosyncrasy, despite the pow­ erful effect o f sem antic m otivations (cf. also Lakoff 1965; Fillmore 1977b; Rosen 1984; M ithun 1991; Dowty 1991). N ote that these cases are unproble­ matic from the point o f view o f learning, since the child has positive evidence that the verbs in question are used in the ditransitive construction, and can therefore learn them on an instance-by-instance basis as idioms.

Partial P rod uctivity

5.s

A

c c o u n t in g f o r t h e

E x c e p t i o n s .-A U

s a g e -b a s e d

A

133

ccount

5.5. i Productivity Defined by Verb Clusters In the preceding sections, we have seen that even after em bracing the idea of narrowly defined verb classes to account for the partial productivity of the ditransitive construction, there rem ains a residue o f lexical idiosyncrasy. There are small subclasses which are not productive, varying degrees of ac­ ceptability within seem ingly productive subclasses, and positive exceptions to the semantic generalizations, such as envy and forgive. This idiosyncrasy is in fact expected if one considers certain experim ental findings. Gropen et al. (1989) suggest that speakers “ tend to be conservative” in their use of lexical items. Specifically, they show that people tend to use lexical items in the sam e constructions in which they have heard those items used by others, but that they can, if properly prim ed, extend the uses to new patterns.6 This phenom ena would be im possible if people did not store in memory the specific syntactic patterns that a word is heard used with (see also Bybee 1985 and Langacker 1987a for particular usage-based models o f gram ­ mar). This being the case, a certain degree o f lexical idiosyncrasy is to be expected. However, the existence o f som e degree of lexical idiosyncrasy should not be taken as counterevidence against the existence o f narrowly defined sem antic subclasses o f verbs that occur in the ditransitive construction. Although the exact form ulation of these classes has differed, their existence as such has been recognized by every researcher w ho has looked in any detail at the verbs oc­ curring in this construction. And, as has ju st been discussed (and is spelled out in more detail in Pinker 1989), the existence o f such classes helps to explain the phenom ena o f partial productivity. These two facts— that there are narrowly defined productive verb classes, and that at the same time wc find scattered positive exceptions and varying degrees o f acceptability within these narrowly defined classes— can be rec­ onciled by recognizing verb classes to be im plicitly represented speakerinternally as generalizations over learned instances. Because m em ory is associative, sim ilar verbs used in the same constructions are classified together by general categorization processes. Therefore the claim is that speakers at­ tempt to categorize learned instances. Narrowly defined verb classes, then, are implicitly represented as clusters of semantically related verbs known to occur with a given construction. New or previously unclassified verb forms are attracted to existing clusters on the basis o f sim ilarity to existing cases. However, judgm ents o f sim ilarity are notori­ ously variable across speakers and contexts, and two activities can alm ost al­ ways be said to be sim ilar in som e respect. Therefore, in order to adequately

13 4

C h a p te r Five

defend the idea that the use o f new and novel senses is determ ined by similarity to existing cases, one must be able to define the sim ilarity metric which is to be used as the basis o f com parison. On the present account, the characteriza­ tions o f the verb classes them selves can be viewed as providing a sim ilarity m etric. For exam ple, if one o f the verb classes associated with the ditransitive is “ verbs o f ballistic m otion,” then we can consider shin to be relevantly like kick in that it is a verb o f ballistic motion. The determ ination o f w hich verb classes arc relevant, or alternatively, what features o f sim ilarity are important, requires em pirical crosslinguistic study, and I do not claim to provide an account here (but cf. Pinker 1989 and Levin 1993 for discussion and suggestions, and M ufw ene 1978 for an early discus­ sion on this subject). Only by looking at which distinctions are made crosslinguistically can we determ ine w hat the scm antically (or m orphophonologically) relevant aspects o f verb m eaning arc that determ ine the basis of the clustering into subclasses. On this view, frequency is expccted to affect the classification o f new verbs. Tw o kinds o f frequency inform ation need to be distinguished. On the one hand there is token frequency, which refers to the num ber o f times a given instance (e.g., a particular word) is used in a particular construction; on the other hand there is type frequency, which refers to the num ber o f distinct words that occur in a particular construction. M acW hinney (1978) and Bybee (1985) have ar­ gued that it is the type frequency o f a particular process (or a particular con­ struction) that plays a crucial role in determ ining how likely it is that the process may be extended to new forms: the higher the type frequency, the higher the productivity. To see the relevance o f the type/token frequency distinction for productivity, consider the following exam ple cited by Bybee (1 9 8 5 :1 3 2 -1 3 3 ). She notes that G uillaum e (1927) docum ented the fact that French-speaking children most frequently overgeneralize the use o f first-conjugation suffixes with verbs of other conjugations. He also observed the num ber o f verbs o f each conjugation used spontaneously in children’s speech. Bybee cites the following table, which shows the num ber o f occurrences o f each conjugation class and the num ber of verbs used from each class:

Conjugation Class

Number o f Uses

Number o f Verbs

First ( ch a n ter)

1.060(36.2% )

12A (76.07c)

Second (fin ir)

173 (6%)

Third (vendre)

1,706 (57.8%)

10(6.1% ) 29(17.9% )

Partial P ro d u c tivity

13b

Although more than half o f the num ber o f tokens ( “ uses” ) o f verbs were of the third conjugation, the num ber of different verbs that occurred in this class was much sm aller than the num ber that occurred in the first conjugation. Correspondingly, the first conjugation was seen to be much more productively used. The proposal to implicitly represent verb classes as sim ilarity clusters can perhaps be made more clear by the follow ing rough-and-ready representation (which does not take morphophonological sim ilarity into account):

o

forgive

o

envy

Figure 5.1

Each circle represents a lexical entry; the entries are projected onto two dimensions, with sem antically closer verbs being represented by physically closer circles. By way of dem onstration, one or more instances within a given cluster have been labeled. T he circles representing bake and cook, for exam ple, are close together to indicate their being in the same narrowly defined class.7 Type frequency can be discerned by considering the num ber o f circles in a given cluster. Clusters containing m ore circles are m ore likely to be productive. Subclasses with only two m em bers such as the verbs o f refusal (deny, refuse) are expected not to be productive, bccause of their low type frequency. The idea that verbs are represented this way in an associative m em ory is of course inspired by recent connectionist representations. However, the diagram need not be construed us necessarily presupposing a connectionist model o f memory; all that is needed is an associative memory (e.g., as proposed in work in the dom ain of morphology o f Pinker and Prince (1991)). This view o f the way new verbs are attracted to learned instances makes

136

C h a p t e r f iv e

several predictions. First, it predicts that subclasses with few members will not contain enough instances to create a sim ilarity class, and so will not be produc­ tive. Secondly, it predicts the possibility of differences in judgm ents within sim ilarity classes. Such differences will result from (1) the degree o f similarity between the case being judged and other cases within the subclass, and (2) the relative type frequency that the relevant cluster displays. It is not necessary (or possible) to exhaustively list all the verbs that can potentially occur in a given construction. Novel cases are analogized to previously learned cases on the basis o f their sim ilarity to these fam iliar cases and the type frequency o f these fam iliar cases. O ccasional positive exceptions (such as envy and forg ive for the ditransitive construction) are tolerated because speakers simply associate the words with the constructions idiosyncratically. There is no danger o f productive extensions from these outliers because they, like subclasses o f fewer than two members, do not constitute a cluster, and therefore do not attract novel cases. T he representation in figure 5 .1 entails that the knowledge that certain verbs are used in a particular construction is part o f a speaker’s com petence. How­ ever, it is not necessary that each new entry be stored as an additional member o f a cluster, throughout the speaker's life. It is possible that once a critical num ber o f instances in a particular cluster is learned— insuring that novel in­ stances that fall into the class will be included— new cases are no longer stored in memory since they would provide only entirely redundant information. It is also possible that learned instances are not necessarily stored as discrete, clearly individuated cases; rather, the edges o f learned instances that form a cluster may blend into each other, delim iting an area in semantic space without specifically retaining each individual instance within. 5 .5.2

Varying Degrees o f Productivity The ditransitive provides a good exam ple of a construction with associ­

ated verb classes. T he degree o f productivity o f other constructions can be seen to form a cline between those constructions which are not fully productive even within narrowly defined verb classes and those which approach full produc­ tivity as long as general constraints are obeyed. An exam ple of the first case, that o f very lim ited productivity (at least in som e dialects), is the resultative construction. As discussed in chapter 8, there is a large degree o f idiosyncrasy as to w hich verbs can occur with which resultatives, Notice the following contrasts: (38) a. She shot him dead. > b.??She blasted him dead.

Partial P rod u c tivity

13 7

(39) a. She cried herself to sleep. > b. She cried herself asleep. > c.??She wept herself to sleep/asleep. (40) a.??He ate him self asleep. b.??He cried him self sick. At the same time, it is clear that resultatives are not entirely idiom atic and do occasionally occur productively. For example: (41) a. b. c.

“ I cried m yself well-nigh blind." (Tennyson, “ G randm other X ” (1884); cited by V isser 1963) “ Drive your engine clean" (M obil ad; cited by Rappaport Hovav & Levin 1991) “She could wonder herself crazy over the human eyebrow.” (R. L. Stevenson, “ Virginibus Puerisque” (1881))

The particular factors which underlie the limited productivity o f this construc­ tion must include semantic factors of the type outlined in chapter 8. In addition, morphophonological factors, such as the ones Gropen et al. (1989) found to be relevant in the case o f the ditransitive, and the token frequency of the analogi­ cal source may need to be taken into account. T he role o f each of these factors remains an issue for further research. An exam ple at the opposite end of the continuum is the way construction discussed in chapter 9. This construction appears to be alm ost entirely produc­ tive. The following exam ples com e from the O xford U niversity Press corpus: (42) a. b. c. d. e.

“ But he consum m ately ad-libbed his way through a largely secret press m eeting.” “ . . . nasty gossip about me now sludging its way through the intes­ tines of the society I know . . . ” ". . . their custom ers snorted and injected their way to oblivion and som etim es died on the stairs.” “ . . . [they] hoped they too could massage their way to keeping power." “ Lord King craftily joked and blustered his way out o f trouble al the m eeting."

As discussed in chapter 9, the few non-occurring verbs (such as the vanilla motion verbs go, walk, m ove) can be accounted for by general semantic co n ­ straints on the construction. Interestingly, the token frequency of this construc­ tion is low, with one exam ple occurring approxim ately every 4 0 ,0 0 0 - 56,000 words in the Lund Corpus o f conversational texts and the Wall Street Journal.8 This provides support for the idea that productivity has little to do with token frequency and more with type frequency.

13 8

C h a p t e r Five

The range o f differing productivity illustrated by these exam ples is exactly what wc would expect given the usage-based model o f gram m ar described above. T hat is, learned and thus stored resultative cases are few and only dot the sem antic landscape; little or no clustering o f exam ples is attested. Therefore novel extensions sound novel and are not fully idiom atic, unlike such exten­ sions o f the ditransitive cases as fa x or M arantz’s novel verb shin. At the same time, attested way construction exam ples seem to span the spec­ trum o f sem antic space, given the general constraints imposed by the construc­ tion. Since the construction has such a high type frequency o f attested verbs, novel verbs are freely used in it.

5.5.3

Productive Links As discusscd in chapter 3, relations between constructions, represented

by various types o f inheritance links, are also objects in our system. Different instances o f a given link occur with different type frequencies, ju st as different instances o f a particular construction occur with different type frequencies. For exam ple, the causative-inchoative relation, which is represented by a kind o f subsum ption inheritance link, occurs between the caused-m otion and intransitive motion constructions, the resultative and the intransitive resultative constructions, and the sim ple causative and simple inchoative constructions. Therefore this particular link would be said to have a type frequency o f at least three. Som e o f the polysem ous extensions we have seen occur in both the ditran­ sitive construction and the caused-m otion construction. The type frequency of each o f these polysem y links is increased with every construction which is extended in the same way. Bccause productivity is directly correlated with type frequency, the higher the type frequency, the more likely a particular inheri­ tance link will exist between pairs o f new constructions that are relevantly sim i­ lar to the pairs o f existing constructions which the inheritance link already relates. In the lim iting case, a link will apply fully productively, yielding exten­ sions every time a novel construction is encountered, as long as that construc­ tion satisfies the particular sem antic characteristics o f the existing instances. In this case, the link between the two constructions is quite analogous to a rule, in that the existence o f one form can be used to predict the existence o f the other form. For exam ple, the passive construction, discussed briefly at the end of chapter 2, is instantiated by many different particular versions, each corre­ sponding to an active construction with the relevant sem antics (the active con­ struction m ust have at least two argum ents, with one being higher on the role hierarchy than the other). Because the type o f link between active and passive

Partial P ro d u c tivity

13 9

constructions occurs between so many different active and passive pairs, it has an extremely high type frequency. Therefore the passive link is, in effect, rulelike in its application. 5.6

C o n c lu s io n

The account proposed here to explain the partial productivity o f co n ­ structions involves two types o f learning m echanism s. The first is a type o f indirect negative evidence, based on the hypothesis that every construction contrasts with every other construction. Therefore, upon hearing a verb in a construction that must be considered non-optimal given the current context, the learner tentatively hypothesizes that the verb cannot occur in w hat would be the preferred construction. The reasoning is roughly, “ If that construction could have been used, I guess it would have been used; therefore maybe it can’t be used.” Upon w itnessing the verb in a non-optim al construction, given the context, a num ber o f times, the learner’s hypothesis that the verb cannot occur in the optim al construction is strengthened. This strategy was first proposed by Pinker (1 9 8 1); his later rejection o f this strategy (Pinker 1989) was argued here to have been unwarranted. The second learning m echanism , presum ably working in tandem with the first, draws largely on recent work by Pinker (1989) and the related experim en­ tal evidence o f Gropen et al. (1989). Specifically, the need to circum scribe narrowly defined sem antic subclasses characterized by local productivity is acknowledged. The account proposed here differs som ew hat from Pinker’s and G ropen’s in that on the present account, the narrowly defined subclasses are understood to be clusters defined by sem antic and m orphophonological sim ilarity that are conventionally associated with the construction, as opposed to subclasses that are conventionally allowed to undergo a lexical rule. Moreover, on the account presented here, the verb classes are interpreted as im plicit generalizations over learned instances in order to account for small nonproductive subclasses, dif­ ferences in judgm ents even within narrowly defined classes, and the existence of positive exceptions such as envy and forgive. Viewing verb classes as clus­ ters of cases in an associative memory also allows us to assim ilate other con­ structions which involve markedly more or less productivity. In particular, it was shown that we might actually expect the fact that the resultative co nstruc­ tion is productive only to a limited degree, w hereas the way construction is almost fully productive. It may seem that by allowing the know ledge o f whether a verb is used in a particular construction to be stored, we underm ine the existence o f the con­ struction as an independent entity. That is, if we need to posit the fact that kick

140

C h a p r e r F iv e

can be used with the ditransitive construction as a separate piece o f gram ­ matical knowledge, why not instead posit a new sense o f kick, along the lines suggested by sem antics-changing lexical rule accounts (cf. the discussion in chapter 4)? The reason for postulating constructions is analogous to the reason why other researchers have wanted to postulate a lexical rule: in order to capture generalizations across instances. Moreover, it is claim ed here that what is stored is the know ledge that a particular verb with its inherent meaning can be used in a particular construction. This is equivalent to saying that the com posite fused structure involving both verb and construction is stored in memory. By recognizing the stored entity to be a com posite structure, we gain the benefits described in chapter 1 over a lexical rule account. For exam ple, we avoid im­ plausible verb senses such as “ to cause to receive by kicking." It is the com ­ posite structure o f verb and construction that has this meaning. We also allow other syntactic processes to refer to the inherent lexical sem antics of the verb. Thus we do not lose the inform ation conveyed by the verb, because the verb is not changed into a new verb with a different sense.

6 The English Ditransitive Construction

6. i

In t r o d u c t io n

The dilransiiive construction has already been discussed with respect to its polysem y (chapter 2) and its partial productivity (chapter 5). In this chapter, I concentrate on particular sem antic constraints and m etaphorical extensions o f the construction. Highly specific sem antic constraints are associated directly with the ditransitive argum ent structure, revealing a m ore specific sem antic structure than is generally acknowledged. In particular, the central sense is argued to involve transfer betw een a volitional agent and a w illing recipient. Several system atic metaphors are identified and associated with the construc­ tion, show ing that expressions such as M ary gave Joe a kiss and M a ry ’s b e­ havior gave John an idea, w hich arc often assum ed to be idiosyncratic, are instances o f a large and productive class o f expressions that are based on sys­ tematic metaphors. Before getting to those constraints, however, evidence that a construction is indeed required in this case is reviewed. 6.2

T h e E x is t e n c e

o f the

C o n s t r u c t io n

Following the program laid out in previous chapters, we need to show that aspects o f the syntax or sem antics o f dilransiiive expressions are not pre­ dictable from other constructions existing in the gramm ar. First, to see that the construction contributes sem antics not attributable to the lexical item s in­ volved, consider the verb bake when used ditransilively: (1)

Sally baked her sisler a cake.

This expression can only mean that Sally baked a cake with the intention o f giving it to her sister. It cannot mean that Sally baked the cake so that her sister w ouldn't have to bake it; nor can it mean that Sally baked the cake as a d em ­ onstration o f cake-baking, or that she baked a cake for herself because her sister wanted her to have one. Unless we associate the “ intended transfer” aspect of m eaning to the construction, we are forced to say that bake itself m eans som e­ thing like ‘X intends to cause Y to receive Z by baking.’ This “ transfer sen se” o f bake would be posited only to avoid attributing aspects o f the sem antics to the construction. The positing o f such ad hoc verb senses w hich only occur in a particular construction was argued against extensively in previous chapters. 141

14 2

C h a p t e r Six

In addition, as was noticed by Partee (1 965:60) and Green (1974:103), the goal argum ent o f ditransitives must be anim ate— that is, it must be a recipient: (2)

a.

She brought the b o ard er/* th e border a package, (cited by Gropen et al. (1989), attributed to J. Bresnan)

As has been argued in chapters 1 and 4, this sem antic constraint is most parsi­ m oniously attributed to the construction. D itransitive expressions arc syntactically unique in allow ing two nonpredicativc noun phrases to occur directly after the verb; the fact that English will allow such a configuration is not predictable from other constructions in the language. In addition, this is the only construction which links the recipient role with the OBJ gram m atical function. The construction was represented in figure 2.4, repeated here as figure 6.1. Dilransiu've Construction

Sem

CAUSE-RECEIVE




|r PRKD


He caused the dog to move into the bathroom ) (11) Joe hit the ball across the field. (—» He caused the ball to move across the field) As has been noticed by Aske (1989), it is also the case that many verbs do not necessarily code motion independently o f this construction. A ske provides the following contrast, and notes that it is im plausible to posit a distinct m otion sense for the predicate squeeze: ( 12) a.

Frank squeezed the ball.

(-b The ball necessarily moves.) b.

Frank squeezed the ball through the crack. ( - » The ball necessarily moves.)

Also, as has been noted by G reen (1973), Randall (1983), and H oekstra (1988), many transitive verbs which can occur in this construction do not bear the same

154

Chapter Seven

sem antic relation to theirdirect object as they do in simple transitive sentences. C onsider (13a—c): (13) a. b. c.

Sam saw ed/tore/hacked/ripped a piece off the block. Sam rinsed/cleaned the soap out of her eyes. Sam stirred the paint thinner into the paint.

These do not entail (1 4 a-c): (14) a. b. c.

Sam saw ed/tore/hacked/ripped a piece. Sam w ashed/rinsed/cleaned the soap. Sam m ixed/stirred the paint thinner.

Verbs can som etim es appear in this construction that do not independently li­ cense direct object com plem ents at all: (15) The audience laughed the poor guy off o f the stage. (16) Frank sneezed the napkin off the table. (17) In the last Star Trek episode, there was a woman who could think people into a different galaxy. The m ove to postulate novel causative motion senses for each of these verbs, thereby positing the m eaning o f the w hole in the m eaning o f the parts by stipu­ lation, has been argued against in chapter 1. Several other proposals to account for the caused-m otion interpretation by com positionally deriving the meaning from the com bination o f verb and preposition are critically discussed below.

♦ Several authors have proposed accounts that avoid positing rampant lexi­ cal polysem y to account for the caused-m otion construction and the related resultative construction (see chapter 8). Gawron (1985, 1986), Pustejovsky (1991a), Rappaport Hovav and Levin (1991), and H oekstra (1992), for ex ­ am ple, have argued that the m eanings o f resultatives and/or caused-motion ex­ pressions such as those in exam ples ( 1 - 6 ) above do not require positing additional verb senses. T hese authors argue that the m eaning of the entire sen­ tence can be com positionally derived from com posing the m eanings o f the con­ stituent parts. A general problem with com positional accounts such as these stems from the far-reaching conclusion drawn from the fact that we may be able to prag­ m atically infer the m eaning of the construction. If one knows that a construc­ tion has a particular form, then it is som etim es the case that one may reasonably infer that it has the particular interpretation that it has. However, it is fallacious to argue that because we may be able to pragm atically infer the m eaning of a

The English Caused-Monon Construction

155

construction, its existence is therefore predictable rather than conventionalized. Such reasoning is based solely on a model o f interpretation, yet we also m ust account for production. For instance, w hile we may be able to infer that ex ­ pressions of the form exem plified in ( 1 - 6 ) have the sem antics they do, we cannot predict that constructions o f this type will exist. In fact, Talmy (1985a) points out that this pattern of expression is not available as a productive form in Romance, Sem itic, or Polynesian language families, although it does occur in Chinese as well as in Dutch and English.' M akkai's (1972) distinction between “ idioms o f encoding” and "idiom s of decoding” can be used to make the point. Decoding idioms are idiom s which a listener would be unable to confidently interpret without having learned the idiom separately. Encoding idiom s are idiom s whose m eaning may be infer­ able; all the same, without having heard the idiom a speaker would have no way of knowing that it was a conventional way o f saying w hat it says. Fly by night, by and large are exam ples o f decoding idioms; serial killer, sofa bed are examples o f encoding idioms. By referring to both kinds o f terms as idioms, M akkai makes the point that neither kind is predictable from general pragmatic principles. Still, it is worth exam ining the com positional accounts m entioned above on their own terms. Gawron (1985, 1986) argues that caused-m otion expressions consist o f two predicates— a verb and a preposition —and that both o f these retain their nor­ mal meanings. The relation between the two predicates, if not determ ined in­ dependently by the verb’s sem antics, is said to be pragm atically inferable from the possible relations that can hold in general between predicates in a single clause. A new way for com plem ents to be sem antically conjoined with verbs is introduced to account for some expressions sim ilar to those in exam ples ( I -7 ): “ c o -p r e d ic a tio n The verb and the preposition act as co-predicators, sharing one argum ent and com bining sem antically in pragm atically inferable ways. For example, John broke the hammer against the vase is analyzed roughly as ‘Break(John.the-ham m er). Against(lhe-ham m er, the-vase)’. The preposi­ tion against is claim ed to be responsible for the interpretation that the ham m er com es into forceful contact with the vase.: Pustejovksy’s (1991a) account is relevantly similar. He suggests that the verbs involved in caused-m otion or resultative expressions are lexically tran­ sitive process verbs that are com bined with independent PPs (or, in the case o f resultatives, APs). The PP is associated with its own event structure: that o f a state. The com position process-plus-state is claim ed to inevitably yield a transition (accom plishm ent) interpretation, namely that of a caused motion or caused change o f state. Pustejovsky claim s that the PP or AP is an adjunct. However, the syntactic

I 56

Chapter Seven

attachm ent o f the PP or adjectivc is not that o f an adjunct. This is true regard­ less of w hether the PP is sister to the verb as claim ed here or is part o f a small clause that is sister to the verb. M oreover, it is not possible to attribute adjunct status to the result phrase and claim that the causative interpretation is inferred, because there exist cases w hich do not receive a causative interpretation. These are the w ell-know n depictive predicates as well as standard PP adjuncts.’ (18) D epictives a. T he w itch-hunters burned her alive. b. Sam passed Bob the towel wet. (19) Adjunct PPs a. Lisa slept under the bridge. b. Joe played in the house. A problem for both G aw ron’s and Pustcjovsky’s approach is the existence o f caused-m otion expressions that involve predicates w hich cannot occur tran­ sitively at all. For exam ple, we cannot account for (16) repeated here as (20), in term s of ‘Snee7.e(Fred,the-napkin),O ff(the-napkin,thc-iable)\ because the first predication— ‘Sneeze(Fred,the-napkin)’— is nonsensical (cf. also ex­ am ples (15, 17)). (20) Fred sneezed the napkin off the table. T his tyi>c o f exam ple is not discussed by Pustejovsky or Gawron. Presumably, both accounts would require a three-argum ent sense for verbs such as sneeze, thereby resorting to the kind o f polysem y they had otherw ise sought to avoid. H oekstra’s (1992) account actually focuses on the type o f resultative found in exam ples (13), (20), and those below: (21) a. b.

Fred mixed the paint thinner into the paint, Fthel w ashed the soap out o f her eyes.

N otice that in exam ples (21a, b) the verb does not bear its normal relation to the direct object com plem ent. H oekstra analyzes such resultatives as involving an intransitive process verb com bined with a small clause stative predicate.4 This is sim ilar to a proposal m ade by Rappaport Hovav and Levin (1991) to account for this type o f resultative. They also propose that the verb is lexically intransitive— in particular, that it is unergative— and that it is com bined with an independent small clause com plem ent.5 Both H oekstra's and Rappaport Hovav and Levin’s analyses require some way to join the main verb and the small clausc. They cannot simply say that the verb selects for (or theta-m arks) the small clause com plem ent, since this

The English Caused-Motion Construction

IS 7

would imply an additional sense for each verb, som ething they wish to avoid. For exam ple, they wish to avoid positing a new sense o f drink that has an agent and a state argum ent to account Tor (22): (22) She drank him under the table. Rappaport Hovav and Levin suggest that the verb case-m arks the NP of the small clause. But case m arking norm ally applies to NPs inside small clause

arguments o f the verb. Since Rappaport Hovav and Levin do not wish to claim that the sm all clause is an argum ent, they do not, as far as I can tell, explain what licenses the occurrence o f the sm all clause itself. To account for the resuluitive interpretation, Rappaport Hovav and Levin suggest that because the small clausc XP is attached at the lowest bar level w ithin the VP, it has to be semantically integrated into the “ core eventuality" nam ed by the verb. H ow ­ ever. norm ally only lexically-subcategorized-for com plem ents are attached at the lowest bar level within the VP; therefore the sem antic generalization hinges on a syntactic constraint that is left unexplained. Hockstra proposes a new way to syntactically com pose the verb and small clause. To do this, he needs to rely on rather nonstandard and often seem ingly unmotivated assum ptions. A critical assum ption o f his is that points o f time associated with verbs are theta-m arked. In the case o f an activity predicate like drink, a final tim e point /„ may be theta-m arked through binding the “e-position" o f the small clause w hich denotes a state.® The resultative inter­ pretation is claim ed to be determ ined by the way in which the sm all clause is licensed, namely, through the final point in the event structure o f the m atrix verb (p. 162). However, H oekstra claim s that the resultative can only occur with atelic activity verbs;7 a priori it would seem that there is no “ final point” o f time associated with atelic predicates. H oekstra docs not make d e a r how or why it is that such a final point is available. To sum m arize, there is a general problem with attem pts to account for caused-m otion and resultative expressions by means o f a sim ple concatenation of two independently existing predicates, the sem antic interpretation being ar­ rived at by general pragm atic principles. Such analyses do not account for the fact that such concatenation is allowed in the language in the first place. Unless we treat the secondary predicate as an argum ent or an adjunct, there are no preexisting means by which to concatenate the two predicates. T reating the secondary predicate as an argum ent o f the m atrix verb is tantam ount to creating an extended sense o f the verb, a m ove that all o f these accounts wish to avoid and which has been argued against in chapter 1. Treating the secondary pred i­ cate as an adjunct K not a viable solution either, because its syntactic position

I 58

Chapter Seven

(w hether sister to the verb or part o f a small clause) is not that o f an adjunct; moreover, the sem antic interpretation is different than that of other adjunct sec­ ondary predicates such as depictive predicates and PP adjuncts. H oeksira’s suggestion o f a new way to join verb and secondary predicate is not well m o­ tivated, and Gawron's proposal requires additional verb senses for some cases and non-argum ent PP phrases for others. In the latter case, the burden of the change in m eaning falls onto the preposition. This raises the question whether the preposition in itself could be held responsible for the caused-motion se­ m antics (cf. also Aske 1989). A m ajor problem with attributing the burden o f the sem antic interpretation o f caused motion to the preposition is that many prepositions which appear in this construction favor a locative interpretation: (23) Fred stuffed the papers in the envelope. (24) Sam pushed him within arm ’s length o f the grenade. (25) Sam shoved him outside the room. It might be suggested that these prepositions are system atically am biguous in English, being able to receive either a locative or a directional interpretation. However, such a proposal fails to account for the intuition that prepositions such as inside, in, outside, and within do not intuitively code motion on either use. M ore to the point, these term s are not am biguous in all contexts. For ex ­ am ple, when fronted they can only receive a locative interpretation: (26) a. Into the room he ran, quick as lightning. b. *Inside the room he ran, quick as lightning, (on the directional reading that he ran into the room) c. *W ithin the room he ran, quick as lightning, (on the directional reading) An account relying on an am biguity o f these term s would need to specify in exactly w hich contexts the am biguity could arise. In response to the possible suggestion that we might attribute the motion interpretation to either the verb or the preposition, but that one or the other m ust lexically specify m otion, consider exam ples such as the following: (27) Sam squeezed the rubber ball inside the jar. (28) Sam urged Bill outside o f the house. In these eases, neither the verbs squeeze or urge nor the prepositions inside or

outside independently code motion.

1he English Caused-Motion Construction

159

Therefore, since the causal interpretation cannot be system atically attributed to either the verb or the preposition or their com bination, and since system ati­ cally attributing the motion interpretation to the preposition requires that seem ­ ingly locative prepositions such as within are said to be am biguous although they are dem onstrably not am biguous in other contexts, we attribute the causedmotion interpretation to a construction w hich com bines the verb and direc­ tional preposition yielding a particular, conventionalized interpretation. The possibility o f allow ing “ basically” locative, nondirectional PP's in this construction raises a question for our account, since we have specified that the construction must contain a PP coding a directional phrase. However, we noted that terms which are intuitively locative cannot receive a directional interpre­ tation in all contexts. W hat needs to be recognized to account for these cases is a particular process o f accommodation (cf. Talmy 1977; C arter 1988) or

coercion (M oens & Steedman 1988; Croft 1991; Pustejovsky 1991b; Sag & Pollard 1991) by which the construction is able to coerce the locative term into a directional reading. On the view taken here, coercion is not a purely pragm atic process; rather, it is only licensed by particular constructions in the language. T hat is, coercion is only possible when a construction requires a particular interpretation that is not independently coded by particular lexical items. To the extent that the o c­ curring lexical items can be coerced by the construction into having a different but related interpretation, the entire expression will be judged grammatical." On this view, the locative term s are not independently am biguous, but instead are capable o f being coerced by particular constructions into having the related directional meaning. In the case at hand, we can understand the locative term s to be coerced into having a directional m eaning by the caused-m otion construc­ tion itself. In order for coercion to be possible, there needs to be a relationship between the inherent m eaning o f the lexical items and the coerced interpretation. Clearly it is not possible for ju st any lexical item to be coerced into receiving a direc­ tional interpretation. The relationship between the m eaning of the locative term and the directional interpretation it receives is one o f endpoint focus (Brngm an 1988). That is, the location encoded by the locative phrase is interpreted to be the endpoint o f a path to that location. To summ arize, because attem pts to attribute the m eanings o f entire expres­ sions of caused motion to the m eanings o f individual lexical item s fall short of accounting for the data in a natural way, a construction is posited in the gram ­ mar. The construction can be represented as follows;

160

Cnnpu?r Seven

Cauied-M olion Construction

Sem

CAUSE-MOVE

1

PRED

Syn

< cause

1


>

OBJ

Figure 7.!

A distinct but related construction must be posited to account for intransitive motion cases; this construction can also add a motion interpretation to verbs that do not lexically code m otion;’ (29) The bottle floated into the cave. (Talmy 1985a) T he intransitive m otion construction is related to the caused-m otion construc­ tion as follows: Caused-Moiion Construction

Sem

CAUSE-M OVE PRED

< cause

(hem e > >


>


«d-class: a mass o f a size, shape, or type defined by the intended use o f a container (and not purely by its geom etry) is put into the container, enabling it to accom plish its function: load, pack (of suitcases), stock (of s h e lv e s). . .'5 Verbs o f the slather-class require all three participant roles to be expressed. Notice, for exam ple, that one can say either (118a) or ( 1 18b): ( 1 18) a. b.

Sam slathered shaving cream onto his face, Sam slathered his face with shaving cream.

None o f the follow ing exam ples, with one role unexpressed, are allowed: ( 119) a. *Sam slathered shaving cream. b. *Sam slathered his face. c. *Shaving cream slathered onto his face. This is represented by profiling all three roles: slather (slatherer, thick-mass, target)

Slather is com patible with both the caused-m otion construction and the causative-plus-vw7/i-adjunctconstructions in the follow ing way. Both construc­ tions allow all three roles to be expressed, so there is no problem satisfying the constraint that profiled roles are obligatory. Since there are three profiled par-

The English Caused Motion Construction

177

licipants, one may be fused wilh a nonprofiled argum ent role, in accord with the Principle o f Correspondence discussed in section 2.4.2. The verb’s participant roles are fused with each of the construction’s argu­ ment roles in accord with the Principle of Sem antic C oherence, also discussed in section 2.4.2, as follows. T he verb’s participant roles fuse with the causedmotion construction’s argum ent roles in that the slatherer can be construed as a cause, thick-mass as a type o f theme since it undergoes a change of location, and the target as a type o f goal-path. Slather is com patible with the causativeplus- wj'f/i-adjunct since the target can be construed as a type o f patient, in that the entity which is slathered on can be construed as totally affected. The withphrase is obligatory even though it is an adjunct, because the profiled status of the verb’s thick-mass role requires that the role be expressed. Verbs o f the heap- and cram-classes are similar. Thus both full variants of the alternation are acceptable with heap: ( 120) a. b.

Pat heaped mash potatoes onto her plate, Pat heaped her plate with mash potatoes.

But again, none of the verb’s roles may be left unexpressed:16 ( 121) a. * Pat heaped mash potatoes. b. *Pat heaped her plate. c. *The mash potatoes heaped onto her plate. The same is true o f cram: (122) a. b.

Pat cram m ed the pennies into the jar. Pat cram m ed the ja r with pennies.

(123) a. *Pat cram m ed the pennies. b. *Pat crammed the jar. c. *The pennies cram m ed into the jar. Thus, like the slather- class above, verbs of these classes must have three pro­ filed participant roles: heap (heaper, location, heaped-goods) cram (crammer, location, crammed-goods) The load-class o f verbs also occurs in both constructions: ( 124) a. b.

She loaded the wagon with the hay. She loaded the hay onto the wagon.

It is not as clear in this case lhat all roles need be overtly expressed. W hile the

178

Chapter Seven

agent and container roles are obligatory, as (125) shows, load can occur w ith­ out an overtly expressed theme role, as in (126): (125) a. *The hay loaded onto the truck. b.??Sam loaded the hay. (126) Sam loaded the truck. Can we still claim for this case that all roles are profiled? Notice that if the theme role is unexpressed, its filler m ust be presum ed to be known to both speaker and hearer. T hat is, unless the context tells us what was loaded onto the truck, exam ple (126) is infelicitous. Thus the theme role is allowed to be a definite null complement as discussed in scction 2.4.5; this type o f argument was argued to be lexically profiled despite the fact that it is not obligatorily expressed; it is indicated below by square brackets. T he load-class therefore has three profiled roles as well: load (lo ad e r, c o n ta in e r, [loaded-them e]) Verbs o f the load-c lass mark their theme roles as profiled, but optionally om issible if licensed by context to receive a definite interpretation. Verbs o f the spray-class, including spray, splash, splatter, sprinkle, work slightly differently. In the case o f splash, both liquid and target roles must be expressed: (127) a. b.

Chris splashed the w ater onto the floor, Chris splashed the floor with water.

(128) a. *C hris splashed the water, b. *C hris splashed the floor.

Spray itself allows the liquid role to be unexpressed if it is given a definite interpretation: (129) The skunk sprayed the car [ |. In this exam ple, the unexpressed liquid role is available to speaker and hearer through contextual given-ness; it is therefore still considered lo be profiled. Both these verbs can occur w ithout an overtly expressed agent, as in (130): ( 130) a. b.

W ater splashed onto the lawn, W ater sprayed onto the lawn.

T he follow ing lexical entries capture these observations: splash (splasher, ta rg e t, liquid) spray (sprayer, ta rg e t, [liquid |>

The English Caused-Motion Construction

I 79

The fact that the target can be construed as a type o f patient, in that the entity which is sprayed can be construed as totally affected, allow s spray's roles to fuse with the argum ent roles o f the causative construction. In particular, spray is licensed to occur in the caused-m otion construction since the sprayer can be construed as a cause, the liquid as a type o f theme, and the target as a type of goal-path. Similarly, the fact that the agent is not obligatory (i.e., non­ profiled) allows spray to occur in the intransitive m otion construction instan­ tiated by ( 130).17 7.6

C o n c l u s io n

It has been argued that the argum ent structure associated with the inter­ pretation of directly caused motion needs to be recognized as an English con­ struction, independently o f the lexical items w hich instantiate it. The evidence came from the fact that several aspects of the m eaning o f caused-m otion ex­ pressions (causation, motion) and o f their form (e.g., the direct object com ple­ ment) are not generally predictable from lexical requirem ents or from other constructions. The construction discussed in this chapter has as its basic sense a causer or agent directly causing a theme to move to a new location. The basic sense is extended in various ways, allow ing the construction to appear with a variety of systematically related interpretations. As noted in section 3.3.2, this polysem y is strikingly sim ilar to the polysem y argued to exist for the ditransitive con­ struction in chapter 2. In addition, specific sem antic constraints have been proposed in an attem pt to show principled patterns where there is apparent idiosyncrasy. These specific constraints can be interpreted as beginning to provide necessary conditions on the notion of “direct causation” (or o f a “ single event"). These constraints have been argued to involve a com bination o f lexical sem antics and general world knowledge. Finally, the load/spray alternation was discussed, and it was shown that both variants could be accounted for by understanding a single verb m eaning to be able to fuse with two distinct constructions, the caused-m otion construction and a causative construction plus with- adjunct.

8 The English Resultative Construction

8.1

In t r o d u c t io n

In this chapter the resultative construction, which was argued to be a m etaphorical extension o f the caused-m otion construction (cf. chapter 3), is discussed in more detail. A great deal o f attention has been focused recently on attem pting to delim it the class of expressions to which resultatives can be ap­ plied (Bresnan and Zaenen 1990; C arrier & Randall 1992; Hoekstra 1988; Rappaport Hovav & Levin, 1991; Jackendoff 1990a; Levin & Rappaport 1990b; Napoli 1992; Randall 1983; Sim pson 1983; Van Valin 1990a). T his chapter defends the position that the necessary constraint on the appear­ ance o f resultatives can be stated in sem antic terms; the resultative can only apply to argum ents that potentially (although not necessarily) undergo a change o f state as a result o f the action denoted by the verb. Such arguments are tra­ ditionally identified as patients. The traditional test for patienthood is that the expression can occur in the following frame (Lakoff 1976): ( 1)

a. b.

W hat X did to (patient) w a s ,. .. W hat happened to (patient) w a s ,. . .

T his idea would seem to be intuitive, given the fact that resultatives code a change of state caused by the verb. And in fact, this proposal has been approxi­ mated recently by Bresnan and Zaenen (1990), Jackendoff (1990a), and Van Valin (1990a). However, the existence o f so-called “ fake object” cases has been taken as an exception to the sem antic constraint. Cases involving "fake o bjects,” so named by Sim pson (1983), include exam ples such as the following (the attested exam ples here and below com e from V isser 1963): “ Paulo, w ho had roared him self hoarse, was very w illing to be si­ lent.” (OED: Mrs. Radcliff. Italian vii (1797)) b. *H e roared himself.

(2)

a.

(3)

a.

“The G erm ans c ri’d their throats dry with calling for a general Council.” (OED: Leighton (1674) in L auderdale Papers (1885)) b. *The G erm ans cried their throats.

The postverbal N P in these cases is said to bear no sem antic relation to the main verb, and therefore is viewed as being exceptional to the semantic con­ straint o f patienthood. The existence o f these cases has led several researchers 180

The English Resuliative Construction

18 1

to conclude lhai the phenom enon m ust be stated in syntactic term s (Sim pson 1983; C arrier & Randall 1992; Rappaport Hovav & Levin 1991). In what follows, I will continue to refer to these as “ fake object” cases d e­ spite the fact that I will argue that the “ fake object” should be treated as a semantic argument. In w hat follows, I restrict my attention to adjectival resul­ tatives, although I intend the term “ resultatives” to cover expressions which encode a resultant state with a PP as well. The facts which must be accounted for are the following: 1. Resultatives apply to direct objects o f som e transitive verbs: (4)

“This nice man probably just wanted M other to . . . kiss him uncon­ scious.” (D. Shields, Dead Tongues (1989))

(5)

“ I had brushed my hair very sm ooth.” (Ch. Bronte, Jane Eyre (1847))

(6)

“ You killed it stone-dead.” (W. Som erset M augham , “ A ltogether" (1910)) 2. But they do not apply to direct objects o f others:

(7) *He watched the TV broken. (8) *He believed the idea powerful. 3. Resultatives apply to subjects o f passives which correspond to ac­ ceptable actives: (9)

"I charged with them, and got knocked silly for my pains.” (R ider H ag­ gard, “ King Solom on’s M ines” (1889))

( 10) The tools w ere wiped clean. 4. They apply to the subjects o f particular intransitive verbs, often asso­ ciated with unaccusativity: (11) The river froze solid. (12) It broke apart. 5. But they do not apply to the subjects o f other intransitive verbs, o f­ ten associated with unergativity: (13)* He talked hoarse. (14) *At his wedding, he smiled sore. (15) *He coughed sick. 6. Finally, as mentioned above, resultatives occasionally occur with socalled fake objects— that is, postverbal NPs that do not bear the norm al argu-

182

Chapter ITight

meni relation to the m atrix verbs. Some additional exam ples are the following (see also section 8.6): (16)

“ W hose whole life is to eat, and drink . . . and laugh them selves fa t " (OED: Trapp, Com m , and Epist. and Rev. (1647))

(17)

“The dog would bite us all mad." (Dougl. Jerrold, Mrs. C audle’s Curt. L ect 4 (1846)) (This does not necessarily imply that the dog would bite us all)

(18)

“ She laughed herself crooked.” (Benson. “ Mr. Teddy" (1910))

8.2

T h e Status

o f the

Po stverbal

NP

Follow ing Simpson (1983), many researchers have assumed that the postverbal NP in the case o f fake object resultatives is not an argument of the verb, w hereas the postverbal NP o f transitive resultatives is (Rappaport Hovav & Levin 1991; Bresnan & Zaenen 1990; Jackendoff 1990a; Napoli 1992). Car­ rier and Randall (1992) explicitly argue this point. They observe that some processes that are taken to apply only to direct internal argum ents do not apply to fake object resultatives, although they do apply to regular resultative expres­ sions. Specifically, they argue lhat m iddle formation, adjectival passive for­ mation, and process nom inalization apply lo direct internal arguments. These processes are said to apply to “ transitive" resultatives: (19) Transitive Resultative: He ham m ered the metal (flat). a. M iddle Formation: This meial ham m ers flat easily. b. Adjectival Passive: the ham m ered-flat metal c. N om inalization: the ham m ering o f the metal flat However, they do not apply to fake objcct resultatives: (20) Fake O bject Resultative: He drove his tires *(bald). a. M iddle Form ation: *T hose tires drive bald easily. b. Adjectival Passive: *the driven-bald tires c. Nominalization: *lhe driving o f the tires bald Notice, though, that none o f these constructions occurs across the board with all transitive resultatives either. For exam ple, the following middles are based on transitive verbs, yet they pattern exactly like fake object cases in being ungram m atical: (21) M iddle Construction with transitive verbs a. *Pat kicks black and blue easily. b. *The w asher loads full easily. c. *H is face w ashes shiny clean easily.

The Fnglish Resultative Construction

183

And, as Jackendoff (1990a) has pointed out, m ost if not all adjectival passives and nominalizations based on transitive resultatives are also ungram m atical; this is exemplified in (22) and (23), respectively: (22) Adjectival Passive of Transitive Verbs a. * the w ashed-shiny-clean face b. *the shot-dead man c. *the kicked-black-and-blue dog (23) Nominalization o f Transitive Verbs a. *the shooting of the man dead b. *the washing of the face shiny clean c. *the driving o f him crazy Notice we cannot claim that these facts provide evidence that even transitive resultatives do not have an internal argum ent since as (19) show ed, som e tran ­ sitive resultatives do occur in these constructions. Exam ples with uncontroversial direct internal argum ents differ on w hether they can occur in these constructions. For example: (24) M iddle Construction a. This movie watches easily. b. *This movie sees easily. (25) Adjectival Passive Construction a. the murdered man b. *the killed man (Lakoff 1965:46) (26) NominaJization Construction a. the persuasion o f people to new faiths b. *the persuasion o f people to be quiet Therefore, although there may well be an implication that if X occurs in the middle construction and adjectival passive construction and nominalization construction, then X is an argum ent, the converse is clearly false. So we cannot use these constructions to argue that fake object cases are not argum ents. N ei­ ther Carrier and Randall nor the other researchers cited above provide other rea­ sons for, or benefits to, attributing a non-argum ent status to the postverbal NP. Each of the above constructions should be considered independently to see why some resultatives are com patible with them and others are not. To make this point, we will consider the case of middle form ation in som e detail below. 8.3

M

id d l e

F o r m a t io n

M iddles require that the unexpressed agent argum ent be indefinite, inter­ preted as “ people [or w hatever the agent is] in general.” M iddles also require

184

Chapter Eight

that the patient subject argum ent have a particular inherent quality which makes it prim arily responsible for the property expressed in the predicate phrase (van O osten 1977, 1984). Moreover, the unexpressed agent argum ent is typically interpreted as volitional, intending the result (if a result is entailed) as well as intending to perform the action denoted by the verb. To illustrate the fact that m iddles are norm ally interpreted as involving an (indefinite) volitional agent, notice the contrast between (27a, b) and between (28a, b): (27) a. T his car drives with the greatest o f ease, b. #T his car drives with the greatest difficulty. (28) a. This wine drinks like it was water, (van O osten 1977) b. #T his wine drinks like it was vinegar.' The same sem antic features which are characteristic of middles can be captured by an appropriate paraphrase. For exam ple, consider (29): (29) T he metal ham m ers fiat easily. T his sentence is interpreted to mean: (30) People can ham m er the metal flat easily because o f an inherent quaiity o f the metal. Several factors conspire to m ake fake object cases (as well as many transitive resultatives, and in fact, many sim ple transitives) incom patible with the middle construction. For one, fake object cases occur most readily with objects that are coreferential with the subject (Jackendoff 1990a); for example. (31) He cried him self asleep. (32) He talked him self blue in the face. T he restriction on middles that the unexpressed agent argum ent be indefinite suffices to rule the corresponding middles ungramm atical: (33) * He cries asleep easily. (34) *He talks blue in the face easily. A nother source o f incom patibility stems from the constraint that the patient argum ent m ust be interpreted to have a particular inherent quality that makes the predicate true. Fake object cases, however, are often used as hyperbole tn express the idea that the action perform ed was done to excess; in this use, it would be anom alous to attribute the predicate's holding to some particular property o f the fake object referent. For exam ple, consider (35): (35) The joggers ran the pavem ent thin. (C arrier & Randall 1992)

The Tnglish Resultative Construction

185

This statem ent would not be used to describe an actual change in the thickness of the pavement, let alone to convey the idea that the pavement bore som e kind o f particular property which caused it to becom e thin from people running on it. Notice that the sem antically analogous paraphrase (36c) is as unacceptable as the middle form (36b) itself: (36) a. #People can run that pavement thin easily because of an inherent qual­ ity of the pavement, b. #T hat pavem ent runs thin easily. M oreover, the fact that m iddles are typically used when the unexpressed indefi­ nite agent is understood to be volitional serves to render other possible fake object cases infelicitous. This is intuitively expected since fake object cases are often used to express a negative outcom e; therefore assigning volitionality to the unexpressed agent results in anomaly. Given the right context, we find that m iddles with fake objects can be greatly improved. For example, imagine lhat people in charge o f props on a movie set are asked to drive fifty tires bald for a stunt. Insofar as speakers find (37) ac­ ceptable. the corresponding m iddle form (38) is also acceptable: (37) He drove fifty tires bald. (38) Go buy some cheap tires for that scene, those inexpensive tires drive bald really quickly. Similarly, imagine lhat a farm er has had such trouble with stray dogs attacking his chickens that he breeds the chickens such that they wake up easily upon hearing any barking. In this context, insofar as speakers I have checked with accept (39), they also report (40) to be acceptable: (39) The dog barked the chickens aw ake.2 (40) His chickens bark awake easily. Thus, once closer attention is paid to the particular sem antics associated with the middle construction, we can account for why fake object resultatives are not normally acceptable as middles, and we find that it is possible to concoct a context in which the sem antics o f a particular expression in fact is com patible with the middle construction. 8.4

O

ther

A

ccounts

In the previous chapter (section 7.2), proposals by Pustejovsky ( 1991a), Rappaport Hovav and Levin (1991), and H oekstra (1992) lo com positionally derive the m eaning of resultative expressions from their com ponent parts were

186

Chapter Eight

critically discussed. In this section, several additional accounts which are in som e way related to the present proposal are reviewed.

8.4.1 Jackendoff 1990a Jackendoff, although rejecting C arrier and R andall’s specific arguments, also rejects the idea that the postverbal N P is an argum ent, instead suggesting that it is an adjunct. However, the postverbal NP fails traditional tests for adjuncthood. For instance, it can appear as the subject o f passives: (41) The baby was barked awake every m orning by the neighbor’s noisy dog. It m ust occur directly after the verb, with no intervening material: (42) *The dog barked ferociously the baby awake. O m ission o f the postverbal N P results in a radical change o f meaning. Also, only one postverbal N P can occur in a given clause: (43) *The dog barked us them awake. T herefore the claim that the postverbal NP of fake object cases is an adjunct and not an argum ent is unwarranted. In postulating an “ adjunct rule” which can add the postverbal NP to the basic sem antics of the verb, Jackendoff does, however, capture the basic insight that particular lexical items can be viewed as “ fitting into” a construction with its own inherent semantics. A detailed com parison o f Jackendoff’s general approach and the one suggested here is presented in section 10.1.1.

8.4.2

Bresnan and Zaenen 1990 Bresnan and Zaenen (1990) argue that the critical factor is lhat the resul­

tative be predicated o f an argum ent that is intrinsically marked with the [ —r] ( ‘unrestricted’) feature (cf. discussion in section 4.2.1). This feature is taken to be shared by subjects and direct objects, distinguishing them from preposi­ tional objects and secondary objects. The [ - r] feature is assigned as an intrin­ sic classification in either o f two cases: ( I ) the argum ent bears the thematic role o f patient (or “ them e"), or (2) the argum ent is not assigned a thematic role by the verb. T he first case, the argum ent bearing the them atic role o f patient, accounts for the m ajority o f resultatives: those that are predicated o f the direct objects o f transitives, the subjects o f passives, and the subjects o f unaccusatives. The second case, lhat the argum ent is assigned no sem antic role by Ihe verb, is designed to account for fake object cases. Fake objects are assumed to be un­

The English Resuitanve Construcnon

187

subcategorized for by the verb; therefore they are claim ed to bear no thematic role, and are assigned the critical ‘unrestricted’ feature. The problem with this account is that not only is the fake object not assigned a thematic role by the verb, it is not norm ally an argum ent of the verb, whether semantically empty or not; that is, it does not norm ally correspond to a com­ plement o f the verb. Bresnan and Zaenen fail to account for how it is that the internal object makes its way into the argum ent structure of the verb in order to receive its critical [ - r] marking. Both the approach and the problem with it can be stated more generally. Bresnan and Zaenen propose treating the verb with its fake object as a raising verb and, consequently, treating the postverbal NP as a com plem ent but not as an argument. The question is how to account for the existence of this postverbal complement. In order to deal with this issue, one could postulate a lexical rule which would add the internal argum ent to the argum ent structure o f the verb (as has been done for applicatives in A lsina & M chom bo 1990) and for the resultative adjective itself within Bresnan and Zaenen’s theory. And, if this were done, the additional argum ent could be assigned the thematic role o f patient, since it is in all cases an affected argum ent. Providing thematic roles to argum ents yielded by lexical rule is uncontroversial. This would allow Bresnan and Zaenen’s ac­ count to reduce to the straightforw ard sem antic constraint proposed here: re­ sultatives can only be predicated of patient arguments.

8.4.3

Van Valin 1990a The claim that resultatives can only be applied to patient argum ents

sounds on the face o f il much like the account recently proposed by Van Valin ( 1990a). Van Valin argues that the resultative must be predicated o f an “ undergoer." However, he notes lhat “ the label ‘undergoer’ should not be taken lit­ erally” (p. 226, fn. 6). In particular, undergoers do not correspond to patients in that it is not necessary that they potentially undergo a change of state: in­ stead, in English any argum ent which can be passivized is taken to be an un­ dergoer.3 Therefore the undergoer condition is underconstrained with respect to resultatives. and would falsely predict exam ples which have undergoer but nonpatient argum ents to be acceptable. There is a more serious problem with Van Valin’s account. He claim s that “ the constructions allow ing resultative phrases are either accom plishm ents or achievements, all o f which code a result state as part of their inherent meaning. Activity verbs, which are inherently atelic and therefore cannot in principle code a result state or have an undergoer argum ent, do not take resultative predi­

180

Chapter tight

cates" (p. 255). The problem stem s from the fact that when Van Valin tries to exem plify the claim that resultatives only occur with accom plishm ents or achievem ents, he cites the resulting construction, not the construction before the resultative is added. For exam ple, he notes that (44) is an accom plishm ent and that (55) is an achievement: (44) Terry w iped the table clean in/* for five minutes. (45) The river froze solid. He then argues that unergative verbs do not allow resultatives. citing the fol­ low ing example: (46) He talked * in/for ten minutes. However, this exam ple is not parallel to the earlier exam ples since those al­ ready contained resultative phrases. There is no disagreem ent about the fact that expressions with a resultative are accom plishm ents or achievements, since the resultative phrase itself serves to bound the event. However, it is not the case that only such independently classifiable accom plishm ents or achieve­ m ents occur with resultative predicates. For exam ple, push in the following is an activity verb: (47) Terry pushed the door * in an hour/for an hour. And yet push can occur with a resultative: (48) Terry pushed the door shut. Also, talk is an activity verb, and yet the following is sim ply ungrammatical: (49) *H e talked himself. In short. Van Valin’s account begs the question o f accounting for which predi­ cates can occur with resultatives, and o f how the fake object is related to the main verb. 8.5

T h e E x is t e n c e

of a

R e s u l t a t iv e C o n s t r u c t i o n

The generalization we wish to capture is that the occurrence o f resulta­ tives can be predicted in purely sem antic terms: Resultatives can only be applied to arguments which potentially undergo a change of state as a result o f the action denoted by the verb. As claim ed above, the argum ent m ust therefore be classifiable as a type of

patient. A gain, we can use the traditional tesL for patienthood in order to deter­ m ine w hether the argum ent is o f this sort.

The English Resultative Construction

(50) a. b.

189

W hat X did to (patient) w a s ,. . . W hat happened to (patient) w a s ,. . .

Notice there is no requirem ent that the predicate independently codes a change o f state, only that it potentially causes a change o f state. By now, the reader who has been following along might guess how we can account for the occurrence o f resultatives w ithin a constructional approach.4 A resultative construction is posited which exists independently o f particular verbs that instantiate it. In order to account for fake object cases, we need to recognize that the construction itself can add a patient argum ent, besides add­ ing the result argum ent to nonstative verbs which only have an “ instigator" as profiled argum ent. C onstructions as delined have sem antics and are capable of bearing arguments. Thus the postverbal NP o f the fake object cases is an ar­ gument o f the construction although not necessarily of the main verb. U nder this analysis, the verb retains its intrinsic sem antic representation, w hile being integrated with the m eaning directly associated with the construction. The resullative construction can be represented thus:

Resuliative-Construction

Sem

CAUSE-BECOME




I




means Syn

V

SUBJ OBJ OBLAP/PP

Figure 8.1

To see how the construction is able to add argum ents, consider the following cases. Verbs such as wipe and talk can integrate into the resultative construc­ tion because they have com patible roles: (51) a. b.

wipe (w iper wiped) He wiped the table clean, talk (talker) He talked him self blue in the face.

In general, for a verb to occur in a particular construction, the participant roles associated with the verb m ust fuse with the argum ent roles associated with the construction, according to the principles described in chapter 2. The participant

190

Chapter Eight

roles o f the verbs talk and wipe fuse with the argument roles o f the construction as follows: Composite Structure: Resultative + wipe

Sem CAUSE-BECOME | means WIPE

Syn




ADJ/PP

Composite Structure: Resultative + talk

Sem CAUSE-BECOM E < | means TALK

Syn

V

agt |

p at j

< talk e r

SUBJ OBJ

result-goal > i >

ADJ/PP

Figure 8.2

Thus the construction adds only the result-goal argum ent if the verb has a participant role which fuses with the patient argum ent o f the construction, as is the case with wipe.'1 Alternatively, the construction can contribute both pa­ tient and result-goal roles, as is done in the case of talk. Tw o other types o f cases are ruled out. The construction itself does not pro­ hibit a hypothetical verb with participant roles which are instances (types) of agent and result-goal from integrating into the construction, since the construc­ tion could presum ably add the patient argum ent. However, the existence of such a verb is disallow ed by the general constraint that instances o f the resultgoal role can only be predicated o f patient-like roles. A verb such as becom e with the participant roles ‘(p a tie n t resu lt-g o al)’ cannot integrate with the construction, because the construction specifies that the agent role m ust be fused with an independently existing participant role of the verb (this is indicated by the solid line from the construction’s agent role to the PRED role array). Intransitive resultatives (i.e., resultatives with unaccusative verbs) require a slightly different construction; however, the more general constraint on patient-

The English Resultative Construction

19 1

hood is shared by two-place resultatives and intransitive resultatives (cf. ex ­ am ples 11 -1 2 ): Resuliaiive-Consiruction

Sem CAUSE-BECOME




I

PRED

Syn

SUBJ OBJ

OBLPP/AP

ls : cause Intransitive Resultative Construction

Sem BECOME




SUBJ

OBLpp/Ap

Figure 8.3

In figure 8.3 a subpart inheritance link relates the tw o-place resultative con­ struction to the intransitive resultative construction. T hat two constructions are required is not necessarily a draw back o f the present proposal. It seem s that Italian allows only tw o-argum ent resultatives and does not allow resultatives with unaccusatives (cf. Napoli 1992). The constructional approach captures the insight w hich led other researchers to explore the possibility that the postverbal NP is not an argum ent o f the main verb, namely, that the postverbal NP does not intuitively correspond to any participant norm ally associated with the activity denoted by the main verb. The resultative construction is itself associated with a particular argum ent structure configuration, independently o f verbs w hich instantiate it. Particular verbs re­ tain their inherent semantics.

19?

Chapter Eight

T he analysis can m otivate the existence o f fake objects cases. Since the fake reflexive ca se s— the cases in which the resultative adjective is predicated of an argum ent which is coreferenlial with the subjcct— are the most common (ac­ cording to V isser’s survey), the most prototypical exam ples, and for some the only gram m atical cases, we can understand fake object eases as having arisen from an expressive desire to predicate a change o f state o f an agent or instigator argum ent. A construction which adds a patient argum ent to the inherent argu­ m ent structure o f the verb allow s the resultative to apply to a patient argument w hile allow ing the patient argum ent to be corefenlial with the agent argument. In addition, the syntactic expression o f the postverbal NP would follow from general principles. A ssum ing a ternary branching structure (see Green 1973; W illiam s 1983; and C arrier & Randall 1992 for argum ents against a small clausc analysis) the patient argum ent is linked with OBJ by the canonical link­ ing conventions o f English (as suggested recently by, e.g., Gropen et al. 1991; Pinker 1989; Dowty 1991). Further, an account which situates the possibility o f resultative expressions in the sem antics can naturally account for various sem antic constraints on the construction. T hese are discusscd in the following section. Finally, this approach also allows us to capture the fact that there is a great deal of idiosyncrasy involved (Green 1972; Dowty 1979). Resultatives are of­ ten part of collocations w ith particular verbs. For exam ple, eat is most collo­ quial with the resultative sick: (52) a. He ate him self sick. b. ?He ate him self ill/nauseous/full. Cry is most colloquial with the relative to sleep: (53) a. She cried herself to sleep. b. ?She cried herself asleep. c.??She cried herself calm /w et. T he follow ing minimal variants arc markedly odd: (54) ?He ate him self asleep. (55) ?She cried herself sick. W hat needs to be noted is that there arc gram m aticalized instances o f the con­ struction w hich are partially lexically filled. A dopting a usage-based model o f gram m ar as discussed in chapter 5 (which draw s on the work o f Bybee (1985) and L angacker (1987a)), novel extensions are acceptable to the degree that they conform to the semantic (and morphophonological) constraints on existing clusters o f cases.

The English Resultative Construction

e .6

193

C o n s t r a in t s o n th e R e su lta tiv e C o n s t r u c t io n

The construction suggested above only provides a necessary condition on the appearance of resultatives. Several other, co-occurring constraints are required in order to begin to triangulate sufficient conditions on resultatives. In this section, the following restrictions will be argued to hold of (adjectival) resultative expressions generally (m odulo cases in which the verb is lexically causative independently o f the construction): 1. The tw o-argum ent resultative construction m ust have an (anim ate) instigator argument. 2. The action denoted by the verb must be interpreted as directly caus­ ing the change of state: no intermediary tim e intervals are possible. 3. The resultative adjective m ust denote the endpoint o f a scale. 4. Resultative phrases cannot be headed by deverbal adjectives (Green 1972; C arrier & Randall 1992). 8 .6.1

(Animate) Instigator Constraint For m any speakers (including m yself), only anim ate instigator argu­

ments are acceptable as subjects in tw o-argum ent resultative constructions. The animate argum ent is not necessarily an agent, since no volitionality is required: (56) She coughed herself sick. (57) She slept herself sober. In some dialects, inanim ate instigator argum ents are also acceptable. For example: (58) The jackham m er pounded us deaf. (Randall 1983) (59) The alarm clock ticked the baby awake. (Randall 1983) However, no speakers I checked with find instrum ent subjects acceptable: (60) *The feather tickled her silly. (61) *The ham m er pounded the metal flat. This constraint does not hold o f lexical causatives, that is, verbs whose basic sense entails a change o f state independently o f the resultative: (62) Water filled the tub half full. (63) The sleeping pills made me sick. 8 .6.2

Aspectual Constraint There has been som e disagreem ent about which aspectual classes can

occur with resultative phrases. Van Valin (1990a) suggests that resultatives can

194

Chapter Eight

only occur with telic predicates. Dowty (1979) and Jackendoff (1990a), on the other hand, suggest that resultatives can only occur with activity, or “ un­ bounded," predicates. It is at least generally agreed that resultatives cannot occur with stative verbs (H ockstra 1988). In this section I will argue that there is an aspectual constraint, but that it does not coincide with a distinction between telic and atelic predicates, both of w hich can be seen to appear in the resultative construction: (64) a b.

H arry shot Sam dead. H arry shot Sam * for an hour, (telic, except on repetitive reading)

(65) a. b.

Sam talked him self hoarse. Sam talked for an hour, (atelic)

T he relevant constraint can be stated as follows: The change of state must occur simultaneously with the endpoint of the action denoted by the verb. Disallowed.

Allowed:

change o f state

change o f slate time

This constraint rules out cases in w hich there is any time delay between the action denoted by the verb and the subsequent change of state. N otice that in a neutral context, eat with an unexpressed argum ent normally im plies that the agent finished eating a meal: (66) He (already) ate. However, when eat occurs in the resultative construction, the eating is neces­ sarily interpreted as extending over the period of time leading up to the change into a state o f being sick. That is, (67) necessarily im plies that the agent’s con­ tinuous eating m ade him sick; it cannot imply that the meal he ate made him sick. O r consider (68): (67) He ate him self sick. (68) Sam cut him self free. This sentence cannot be used to mean that Sam cut himself, causing his captors to release him in order to clean him up. It must mean that he cut w hateverbonds w ere preventing him from being free, thereby imm ediately gaining his free­ dom. Sim ilarly with (69):

The English Resultative Construction

195

(69) Chris shot Pat dead. This cannot be used to mean that Chris shot Pat and Pat later died in the hos­ pital; instead it must mean that Pat died im m ediately from the shot. T his constraint can be interpreted as a consequence of a more general constraint that the causation must be direct: no intervening period is possible in a causal sequence (cf. discussion in 7.4.2. for other constraints on direct causation).

8.6.3

End-of-Scale Constraint The type of adjective that can occur as a resultative is fairly limited. W hile

adjectives “ asleep/aw ake,” “ open/shut,” “ flat/straight/sm ooth,” “ free," ‘‘full/ em pty," "dead/alive,” “ sick,” “ hoarse,” “ sober,” and “ crazy " occur fairly regularly, others occur rarely if at all: (70) *He drank him self funny/happy. (71) *He wiped it dam p/dirty. (Green 1972) (72) *The bear growled us afraid. (73) *He encouraged her confident. (74) *He ham mered the metal beautiful/safe/tubular. (Green 1972) Most o f the adjectives which can occur in the construction can be indepen­ dently classified as having a clearly delim ited lower bound and are therefore typically nongradabte (Sapir 1944). N ongradable adjectives are said to be un­ able to appear (ceteris paribus) with quantifying phrases: (75) ?a little sober (76) ?a little flat/smooth (77) ?a little alive/dead (78) ?a little asleep/aw ake (79) ?a little full/em pty (80) ?a little free Intuitively, one cannot be a little sober, because one is either entirely sober or not sober: there is, all things being equal, no grey area. Sick and hoarse, on the other hand, do not obviously code states with a clearly delim ited lower bound: (81) a little sick (82) a little hoarse However, when used in the fake object construction, they are interpreted as

196

Chapter Fight

delim iting ihe clear boundary beyond w hich the activity cannot continue. Con­ sider (83, 84): (83) He ate him self sick. (84) He talked him self hoarse. These expressions imply that the agent ate to the point where he could eat no m ore, or talked to the point w here he could talk no more. N otice how in this context the adjectives receive a nongradable interpretation: (85) ?H e talked him self a little hoarse. (86) ?She ate herself a little sick. The adjectives crazy and silly are sim ilar in this respect: (87) He drove her crazy/bananas/bonkers/m ad/insane. (88) He tickled her silly. They imply that the patient argum ent has “ gone over the edge,” beyond the point where normal functioning is possible (o f course they are typically used as hyperbole, not literally).6 R ender is interesting in that it lexicalizes this con­ straint, requiring a resultative adjective which codes a state o f loss o f function (that is, the property m ust be the negative end o f a scale): (89) a. It rendered them speechless/im potent/obsolete. b.??It rendered them alive/full/free. Therefore it is fair to say that the resultative o f the fake object construction codes a clearly delim ited endpoint.7 The endpoint may be on some absolute scale (in the case o f inherently nongradable adjectives) or on a scale of func­ tionality, in w hich case continued functioning is im possible beyond it:

old stale

new state lime

E xceptions to this generalization are o f tw o kinds. First, there are verbs w hich are lexically causative, independently o f the resultative construction. T hese verbs are much freer in the sem antic and syntactic type o f resultative phrase they may occur in than render and productive cases: (90) a. b.

He m ade the metal safe/pretty/tubular/dam p/dirty, He m ade her a queen.

The English Resultative Construction

(91) a. b.

197

He painted his house pink. He painted his house a bright shade o f red.

O ther exceptions to the above generalization have been attested, but aside from their apparent rarity, each can be seen to have a distinctly novel character. In general, exceptional cases tend to be from the sam e sem antic dom ain as more conventionalized cases, and can be seen as one-shot novel extensions from a gram m aticalized pattern: (92)

“ Bees will suck them selves tipsy upon varieties like the sops-of-w ine.” (OED: J. Burroughs, “ Locusts and Wild H oney" (1879))

(93) Till he had drunk him self sleepy. (R. L. Stevenson, Treasure Island (1893)) There is one attested case, though cited by Rappaport Hovav and Levin (1991), which truly seem s to fly in the face of this generalization: (94) “ Look, isn't it lovely? It’s the stale loaf I put out for the birds and th ey ’ve pecked it really pretty.” (cited by Rappaport Hovav and Levin 1991; from Z. W icomb, You C an't G et Lost in Cape Town (1987)) However, this exam ple is judged ungram m atical by A m erican English speakers I have asked. It is possible that South A frican English does not have the endof-scale restriction.

8.6.4

Restriction against Deverbal Adjectives A general constraint that is widely recognized is that resultatives cannot

be adjectives derived from either present or past participles (Green 1972; C ar­ rier and Randall 1992): (95) a. She painted the house red. b. *She painted the house reddened. c. *She painted the house reddening. (96) a. She shot him dead, b. *She shot him killed, b. *She shot him dying. (97) a. She kicked the door open. b. *She kicked the door opened. c. *She kicked the door opening. This restriction has been attributed to a sem antic clash o f aspect (C arrier & Randall 1992); however, the exact nature o f the cause clash— clash has proved elusive.

198

Chapter Eight

8.7

C o n c lu s io n

This chapter has argued that the sem antic restriction that resultatives can only apply to patient argum ents is viable, even in the case of fake object resul­ tative expressions, despite recent argum ents to the contrary. This analysis has the advantages o f (1) assim ilating fake object cases to other transitive resulta­ tive cases, (2) motivating the existence o f fake object cases, (3) allowing for the existing idiosyncrasy in a natural way, (4) predicting the syntax o f the con­ struction from canonical linking patterns and without ad hoc stipulations, and (5) accounting for semantic constraints in a natural way. The following specific sem antic constraints were proposed in order to re­ strict the applicability o f the lexical rule (or the instantiation o f the construc­ tion): (1) tw o-argum ent resultatives must have an instigator argument; (2) the causation involved must be direct, with no intervening tim e periods allowed; (3) the resultative adjective m ust have a clearly delim ited lower bound; and (4) the resultative adjective m ust be considered a type o f path phrase, which accounts for several co-occurrence restrictions.

9 The Way Construction

9. i

In t r o d u c t io n

The construction to be discussed in this chapter can be skeletally repre­ sented as follows (where V is a nonstative verb, and O B L codes a directional): [SUBJ, [V [POSS, way) OBL]] Several large corpora were searched for exam ples. The m ajority o f the ex ­ amples (1,050 out o f 1,177) are from the O xford University Press corpus (oup), which has been cited in earlier chapters already.1 A dditional exam ples have been culled from exhaustive searches o f the Wall Street Journal 1989 corpus (wsj), the Lund corpus consisting o f various spoken dialogs (lund), and the United States D epartm ent o f Agriculture corpus (usda). 9.2

T h e E x is t e n c e

o f the

C o n s t r u c t io n

Instances o f this construction imply that the subject referent moves along the path designated by the prepositional phrase. The construction’s sem antics cannot be fully predicted on the basis o f the constituent parts o f the construc­ tion. For exam ple, (1) entails that Frank moved through the created path out o f the prison. ( 1)

Frank dug his way out of the prison.

Similarly, (2) entails that Frank m anaged to travel to New York. (2)

Frank found his way to New York.

However, none of the lexical item s involved entails motion. To see this, com ­ pare (1) and (2) with (3) and (4) below: (3)

Frank dug his escape route out o f prison.

(4)

Frank found a way to New York.

The only interpretation for these exam ples is one in which the prepositional phrase modifies the direct object. N either (3) nor (4) entails motion: (4 ')

Frank dug his escape route out of prison, but he hasn’t gone yet.

(5 1)

Frank found a way to New York, but he hasn't gone yet.

199

200

Chaoter Nine

This is in contrast with exam ples (I) and (2), which do entail motion: (2 ') * Frank dug his way out o f prison, but he hasn’t gone yet. (3 ') * Frank found his way to New York, but he hasn't gone yet. T he only difference between (1) and (3) is that way is replaced by escape route. Exam ple (4) prevents us from postulating that way codes motion, because way is present in this exam ple and yet the sentence does not entail motion. Without belaboring the point, it should be pointed out that motion is not dictated by the com bination o f bound pronoun and way either, since the expression in (5) does not entail motion: (5)

He knows his way around town.

Here entailm ent o f motion is not present because the verb know is stative, and the construction requires a nonstative verb. Salkoff (1988) and Jackendoff (1990a) also point out that this construction provides evidence for the claim that verbs do not exclusively determ ine com­ plem ent configuration. O ne solution Jackendoff proposes is (hat examples such as those in (1 - 2 ) instantiate a particular clause-level construction: a pairing of form and m eaning that exists independently o f the particular verbs which in­ stantiate it. As he suggests, “ in a sense, the w av-construction can be thought o f as a kind o f ‘constructional idiom ,’ a specialized syntactic form with an idiom atic m eaning, marked by the noun w ay" (1990a: 221). Levin and Rapoport (1988) suggest instead that each verb in the construction has a special motion sense, perhaps generated by a lexical rule, which predicts its occurrence in this pattern. However, this pattern occurs with an enormous variety o f verbs. For exam ple, we would need to posit such a motion sense for each o f the verbs in the following: (6)

a. b. c.

“ . . . he’d bludgeoned his way through, right on the stroke o f halftim e.” (oup) “ [the players will] m aul their way up the middle o f the field.” (oup) “ . . . glaciers which had repeatedly nudged their way between En­ gland and W ales.” (oup)

T hat is, we would need a special sense o f bludgeon, ‘to move by bludgeoning,’ a special sense o f maul, ‘to move by m auling,’ and so forth. These senses are intuitively im plausible. The follow ing exam ples (presented in section 5.5.2 and repeated below) involving m etaphorical motion would be even more difficult to im agine as projections from a lexical subcategorization:

The Way Consrrucfion

(7)

a. b. c. d.

201

. . their custom ers snorted and injected their way to oblivion and sometim es died on the stairs.” (oup) "B ut he consum m ately ad-libbed his way through a largely secret press m eeting." (oup) “ 1 cannot inhabit his mind nor even imagine my way through the dark labyrinth o f its distortion.” (oup) “ Lord King craftily jo k e d a nd blustered his way out o f trouble at the m eeting.” (oup)

If new senses were involved, then it would follow that each of the verbs above would be am biguous between its basic sense and its sense in this syntactic pattern. Therefore we would expect that there would be som e language that would differentiate the two senses by having two independent verb stems. However, to my know ledge there is no language that has distinct verb stems for any o f the meanings that would be required for the exam ples in ( 6 - 7 ) . In addition to being im plausible, positing additional verb senses can be seen to be less parsim onious than associating the sem antic interpretation directly to the construction. The reason for this stem s from the fact that the proposed senses occur only in this construction; they are not available when these verbs are used with other valences: (8) *Chris bludgeoned/m auled/snorted and injected into the room. The same is not true of verbs w hich clearly do lexically code literal or m eta­ phorical motion, for exam ple, inch and worm: (9)

a. b.

Lucky may have inched ahead o f Black Stallion, He can’t worm out of that situation.

That is, both inch and worm can be used as (m etaphorical) motion verbs even when they are not used specifically in the way construction. Therefore, not only would we need to stipulate the existence o f additional senses for each o f the verbs in exam ples ( 6 - 7 ) , but we would have to further stipulate the fact that the new verb senses can only occur in this particular syntactic configuration. Clearly it is more parsim onious to attribute the motion interpretation directly to the construction itself. Given that the interpretation o f way expressions is not fully predictable from the sem antics o f the particular lexical items, a constructional analysis will be adopted here. An explicit statem ent o f the construction will be preceded by a more specific analysis o f the construction’s sem antics, since it will be argued that the syntax o f the construction is motivated by its semantics.

202

Chapter Nine

9.3

T he S e m a n tic s o f th e W a y C o n s t r u c t io n

9.3.1

Tw o Different Senses Both Levin and Rapoporl (1988) and Jackendoff (1990a) suggest two

distinct paraphrases o f this construction, one in which the verb designates the m eans o f motion, the other in w hich the verb designates some other coexten­ sive action or m anner.1 For exam ple, Jackendoff notes that (10) is interpretable in either o f the two ways given in (11): (10) Sam joked his way into the meeting. (11) a. b.

Sam got into the m eeting by joking, (means) Sam w ent into the meeting (while) joking, (manner)

These paraphrases together are taken to com prise a disjunctive interpretation. However, logical disjunction is com pletely sym m etric, and there are reasons to think that the m eans and m anner interpretations do not have equal status in the g ram m ar— that in fact the m eans interpretation is primary. In the O xford University Press, Lund, Wall Street Journal, and Department o f A griculture corpora, verbs which designated a coextensive action or manner, as opposed to the means o f m otion, were rare. In fact, the total number of occurrences o f verbs with m anner interpretation in the corpora was 40 out of 1,177, o r less than 4% o f the data (the percentage was no more than 4% for any o f the corpora). In addition, it is argued below that the syntactic form of the construction is m otivated by the sem antics associated with the m eans interpretation. The same cannot be said o f the m anner interpretation. T hat is, recognizing a cline of analyzability (or com positionality) o f idiom atic expressions (cf. Nunberg, Wasow & Sag 1982; G ibbs 1990), the means interpretation will be argued to be more analyzable than the m anner interpretation. Finally, not all speakers find the strict m anner interpretation acceptable. A case in point is (12), one o f Jackendoff’s examples: (12) H e belched his way out o f the restaurant. W hen asked for judgm ents o f this sentence, which was intended to have a man­ ner interpretation (the subject w ent out o f the restaurant w hile belching), sev­ eral speakers I checked with concocted situations in w hich the belching instead w as the m eans by w hich motion was achieved. For exam ple, one speaker sug­ gested that the sentence would be acceptable in the context that the other diners found the belching so objectionable that they cleared a path through which the offending party could exit. A nother speaker suggested that the sentence would be acceptable if the belching were understood to be a means o f propulsion. O thers, including myself, find the m anner interpretation only marginal.

The Way Construction

203

Since as w e've seen ii is natural for constructions to be associated with a central sense, and with extensions from that sense, these facts can be easily accounted for. We can analyze the m anner interpretation as an extension o f the more basic means interpretation. This analysis predicts, for exam ple, that there are no speakers who accept only the m anner interpretation and not the means interpretation. And to date, I have found none. Interestingly, there is diachronic evidence that the means interpretation of the construction predates the m anner interpretation by more than four centu­ ries. The first citation of this pattern in the O ED is from the year 1400: "I m ade my way . . . unto Rome. " 3 T he first citation with any other verb is from 1694: "[H e] h ew 'd out his way by the pow er o f the Sword. ” The first exam ple cited in the O ED that involves a pure m anner interpretation, “ The muffin-boy rings his way down the little street," is dated 1836, more than a century after the construction was first used productively with a m eans interpretation, and more than four centuries after the first citation with make. The diachronic data o f course do not directly support the claim that the m eans interpretation is synchronically more basic; however, they do provide evidence that the extension from means to m anner is a reasonable move for speakers to m ake, since at least one generation o f speakers was willing to extend the pattern in ju st this way. To summ arize, it has been argued in this section that the m eans interpretation is the more central, or basic, in te rre la tio n o f the construction. The m anner interpretation has been argued to be a less basic extension, on the grounds that (1) m anner exam ples were rare in each of the four corpora analyzed (account­ ing for less than 4% o f the data), (2) speakers’ judgm ents as to the acceptability o f the m anner cases range from unacceptable to marginal to acceptable, while the means cases are all fully acceptable, and (3) the means interpretation diachronically preceded the m anner interpretation by several centuries. A fourth reason for claim ing that the means interpretation is more basic follows from the observation, detailed below, that the syntactic form o f the construction can be motivated by the means interpretation but not by the m anner inteipretation. In the following, a particular sem antic constraint on the m eans inteipretation is proposed, namely, that the motion m ust be through a literal or m etaphorical self-created path. This constraint is argued to play a crucial role in m otivating the syntactic form o f the construction. 9.3.2

The M eans Interpretation: Creation of a Path

Jespersen (1949) had the basic insight that the direct object, POSS way, was a type of “ object o f result.” This can be interpreted to m ean that the path (the way) through which motion takes place is not preestablished, but rather is created by some action o f the subject referent. This observation can be used to

204

Chapter Nine

account for the fact that, with the exception of the pure m anner interpretation, the construction is used to convey that the subject moves despite some external difficulty, or in some indirect way: the path is not already established, but must in some sense be created by the mover. C onsider the following: (13) Sally m ade her way into the ballroom. This sentence is understood to imply that SaJly moved through a crowd or other obstacles. It cannot be used to mean that Sally sim ply walked into an empty ballroom . In the ease o f m etaphorical motion, the necessity o f creating a path im plies that there is some difficulty o r m etaphorical barrier involved. For ex ­ am ple, notice also the difference in acceptability between the following: (14) a.??Sally drank her way through the glass of lemonade, b. Sally drank her way through a case o f vodka. Exam ple ( 14b) is much more acceptable because it is much easier to construe drinking a case o f vodka as requiring that some barrier be overcome than drink­ ing a glass o f lemonade. In fact, the m ost com m on interpretation o f this construction involves motion through a crowd, mass, obstacle, or other difficulty

that is, there is some rea­

son why a path needs to be created. The verb either lexically subcategori7.es for the construction (e.g., m ake) or designates the means by which the motion is achieved. For example: (15)

"F o r the record, Mr. Klein, as lead clim ber for the Journal team, pushed his way past the others, tram pling the lunch o f two hikers in his black arm y boots, and won the race to the sum m it.” (wsj)

(16)

“ In some cases, passengers tried to fight their way through smokechoked hallways to get back to their cabins to get their safety ja c k e ts" (wsj)

(17)

"F or hours, troops have been shooting their way through angry, unarmed m obs.” (wsj)

Contain verbs, such as thread, wend, weave, encode a slightly different in­ terpretation. They involve deliberate, careful, m ethodical, or winding motion. In these cases as well as in the cases which involve some external difficulty, the subject is not m oving along a preestablished path. For example: ( 18)

“ This tim e, with no need to thread his way out, he simply left by the side door for a three-day outing." (wsj)

(19)

"A couple in fashionable spandex warm -up suits jogs by. headphones jauntily in place, w eaving their way along a street o f fractured and fallen houses." (wsj)

The UZsy Construction

20b

The fact that the construction enlail.s that a path is created to effect m o­ tion— that the motion takes place despite som e kind o f external difficulty or is winding and indirect— accounts for why high-frequency, m onom orphem ic (basic or superordinate level) motion verbs are typically unacceptable in this construction: (20 )* S hc w ent/w alked/ran her way to New York (Napoli cited by Jackendoff 1990a) (21) *She stepped/m oved her way to New York. These vanilla motion verbs do not norm ally imply that there is any difficulty or indirect motion involved, an implication which is required by the means interpretation o f the construction. (Note that the m anner interpretation is also unavailable, since these verbs do not code any salient manner.) If a context is provided in which a basic-level motion verb is understood to imply motion despite difficulty, these cases are decidedly better: (22) a. b.

The novice skier walked her way down the ski slope. The old man walked his way across the country to earn money for charity.

Another case in which a (m etaphorical) path may need to be created is given if there are social obstacles standing in the way. C ontrast the follow ing examples: (23) a. #W elcome our new daughter-in-law , who ju st m arried her way into our family. b. Welcome our new daughter-in-law , who ju st m arried into our family. Example (23a) is pragm atically odd because it implies that the daughter-in-law in question managed to get herself into the family by m arriage, and such an implication is incongraent with a sincere welcome. The follow ing exam ple is relevantly similar: (24) Joe bought his way into the exclusive country club. This example entails that Joe managed to get him self into the country club despite social obstacles. The necessity of the m etaphorical creation o f one's own path despite social obstacles can account for the implication that the sub­ ject referent used some unsanctioned m eans to attain his goal. T hat is, if there are social obstacles preventing one from attaining a goal, the only way to attain the goal is to violate the social constraints. Attested exam ples o f this class include bribe, bluff, crapshoot, wheedle, talk, trick, con, nose, sneak, weasel, cajole, inveigle. Several lexical items seem to lexicalize this sense, for ex ­ ample, worm, weasel, and wrangle.

206

Chapter Nine

T he claim then is that way is analyzable as a literal or metaphorical path that is created by the action denoted by the verb. This accounts for the semantic constraint that the motion is effected despite some external physical or social obstacles, by forging a path through or around those external obstacles. Support for the claim that way is analyzed as a meaningful elem ent comes from the fact that it can appear with modifiers. The following examples are attested: (25) a. b.

. . the goats w ending their fam iliar way across the graveyard . . .” (oup) ‘‘[He] decided from then onw ards that he could make his own way to s c h o o l. . . ” (oup)

In exam ple (25a), fa m ilia r is a m odifier o f w ay— that is, the path is familiar. Similarly, in (25b), way is internally modified by own. These facts argue that the phrase POSS way is not an arbitrary syntactic tag o f the construction, but rather plays a role in the sem antics o f the construction. Further support for the claim that the construction at least historically was associated with the creation o f a path com es from the fact noted above that the verb make, a verb which norm ally means “ create,” has had a privileged status with respect to this construction: this verb was used in the construction for alm ost three centuries before the construction was extended to be used with other verbs, according to citations in the OED. M ake continues to be closely associated with the construction insofar as it is used with greater frequency than any other single verb, accounting for 20% o f the tokens. This suggests that m ake may well have a privileged status synchronically as well. Finally, the recognition that the way is an effected entity motivates the syn­ tactic form o f the construction. As stated at the onset o f this chapter, Jackendoff notes that there are reasons to assign the construction the structure: [SUBJ, [V [POSS, way) OBL]] He argues that the noun phrase ‘POSS w ay’ is a direct object, rather than some kind o f syntactic adjunct or m easure phrase, because nothing can intervene between the V and this phrase: (26) *Bill belched noisily his way out of the restaurant. (1990a 212) The O BL phrase coding the path is argued to be a sister o f the verb, rather than a m odifier o f way, on the grounds that an adverb may intervene between the two com plem ents, indicating a constituency break:

The Way Construction

(27) a. b.

207

Bill belched his way noisily out o f the restaurant. ( 1990a: 2 12) “ He made his way cautiously along the path beside the lake." (oup)

Given the sem antics o f the means interpretation described above, the construc­ tion can be viewed as a kind of conventionalized am algam that com bines the syntax and semantics o f creation expressions such as (28), w hich have two argum ents— a creator and a “createe-w ay” — with the intransitive motion co n ­ struction exemplified by (29), which has two argum ents— a mover (them e) and a path. (28) He made a path. (29) He moved into the room. The way construction syntactically and sem antically am algam ates these two constructions into a structure with three com plem ents: the creator-them e, the createe-way, and the path. Thus the way construction can be viewed as inher­ iting aspects o f both the creation and the motion constructions, while never­ theless existing as an independent construction in its own right. The sem antics involves both the creation o f a path and m ovement along that path. As was true for the constructions discussed in previous chapters, the verb may, but need not necessarily, code the sem antics associated with the construc­ tion directly. Cases in which the verb does directly code the sem antics o f the construction include worm, inch, and work. In other cases, the verb may des­ ignate the means o f effecting the action designated by the construction; that is to say, the verb may code the means o f effecting motion through a self-created path. This is represented below by the means link between PRED, representing the verb sense, and the CREATE-M OVE predicate. Way Construction: Means Interpretation

Sem

CREATE-M OVE < c reato r-th em e createe-w ay, | means PRED

Syn

V

|
:59—112. Keenan, Edward L. 1972. On Semantically Based Grammar. Linguistic Inquiry 4(3): 413-462. Keenan, Edward L. 1976. Towards a Universal Definition of “ Subject." In C. N. Li, ed., Subject and Topic, 303-334. New York: Academic Press. Keenan, Edward I.. 1984. Semantic Correlates of the Ergative/Absolulive Distinction. Linguistics 2 2 :197-223. Kemmer, Suzanne. 1988. The Middle Voice. Ph.D. diss., Stanford University. Kiparsky, Paul. 1987. Morphology and Grammatical Relations. Unpublished manu­ script, Stanford University. Kirsner, Robert S. 1985. Iconicily and Grammatical Meaning. In J. Haiman, ed., Iconiciry in Syntax, 249-270. Amsterdam: John Benjamins. Koenig. Jean-Pierre. 1993. Linking Constructions vs. Linking Rules: Evidence from French. B L S 19, 217-231. Kuroda, Sige-Yukj. 1965. Generative Grammatical Studies in the Japanese Language. Ph.D. diss., MIT. Lakoff, George. 1965. On the Nature of Syntactic Irregularity. Ph.D. diss., Indiana Uni­ versity, Published as Irregularity in Syntax. New York: Holt, Rinehart and Winston, 1970. Lakoff, George. 1968. Instrumental Adverbs and the Concept of Deep Structure. Foun­ dations of Language 4:4-29. I-akoff, George. 1970a. Adverbs and Opacity: A Reply to Stalnaker. Unpublished manu­ script, University of California, Berkeley. Lakoff, George. 1970b. Global Rules. Language 46:627-639. Lakoff, George. 1971. On Generative Semantics. In D. D. Steinberg and L. A. Jakobovits, eds., Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, 232-296. London: Cambridge University Press.

252

Bibliography

Lakoff, George. 1972. Linguistics and Natural Logic. In D. Davidson and G. Harman, eds., Semantics of Natural Language, 545 - 665. Dordrecht: Reidel. Lakoff, George. 1976. Towards Generative Semantics. In J. D. McCawley. ed.. Syntax and Semantics 7: Notesfrom the Linguistic Underground. 43 62. New York" Academic Press. First circulated in 1963. Lakoff, George. 1977. Linguistic Gestalts. C LS 13, 225-235. Lakoff, George. 1984. There-Construcuons: A Case Study in Grammatical Construc­ tion Theory and Prototype Theory. Cognitive Science Technical Report 18. University of California, Berkeley. Revised version published as "Case Study" in Lakoff 1987. Lakoff, George. 1987. Women, Fire, and Dangerous Things: Wlutt Categories Reveal about the Mind. Chicago: University of Chicago Press. Lakoff, George. 1993. The Contemporary Theory of Metaphor. In A. Ortony, ed.. Meta­ phor and Thought, 2d ed. Cambridge: Cambridge University Press. Lakoff, George, and Mark Johnson. 1980. Metaphors We Live By. Chicago: Univcivity of Chicago Press. Lakoff, George, and John Robert Ross. 1976. Is Deep Structure Necessary? In J. D. McCawley, ed., Syntax and Semantics 7: Notes from the Linguistic Under­ ground, 159- 164. New York: Academic I’ress. Hirst circulated in 1967. Lakoff, Robin. 1968. Abstract Syntax and Latin Complementation. Cambridge. Mass.: M IT Press. Lambrecht. Knud. 1987. Sentence Focus, Information Structure, and the TheticCategorical Distinction. B IS 13, 366-382. Lambrecht, Knud. 1990. ‘What me worry?’ Mad Magazine Sentences Revisited. B IS 16, 215-228. Lambrecht, Knud. 1994. Information Structure and Sentence Form: A Theory o f Topic, Focus, and the Mental Representation of Discourse Referents. Cambridge Studies in Linguistics. Cambridge: Cambridge University Press. Landau, Barbara, and Lila R. Gleitman 1985. Language and Experience: Evidence from the Blind Child. Cambridge, Mass.: Harvard University Press. Langacker, Ronald W. 1969. Pronominalization and the Chain of Command. In D. Reibel and S. Schane, eds., Modern Studies in English: Readings in Transfor­ mational Grammar. Englewood Cliffs, N.J.: Prentice-Hall. Langacker, Ronald W. 1985. Observations and Speculations on Subjectivity. In J. Hai­ man, ed.. Iconicity in Syntax, 109- 150. Amsterdam: John Benjamins. Langacker, Ronald W. 1987a. Foundations of Cognitive Grammar. Vol. I : Theoretical Prerequisites. Stanford, Cal.: Stanford University Press. Langacker, Ronald W. 1987b. Grammatical Ramifications of the Setting/Participant Distinction. B L S 13, 383-394. Langacker, Ronald W. 1988. An Overview of Cognitive Grammar. In B. Rudzka-Ostyn, ed. Topics in Cognitive Linguistics, 127-161. Philadelphia: John Benjamins Langacker, Ronald W. 1991. Foundations o f Cognitive Grammar. Vol. 2: Descriptive Application. Stanford, Cal.: Stanford University Press. Laughren, Mary. 1988. Toward a Lexical Representation of Warlpiri Verbs. In W. W il­ kins, ed., Syntax and Semantics 21: Thematic Relations, 215-242. New YorkAcademic Press. Legendre. G6raldine, Yoshiro Miyata, and Paul Smolensky. 1990. Harmonic Gram­

Bibliography

253

mar— A Formal Mulii-level Conncctionist Theory of Linguistic Wellformedness: Theoretical Foundations. Technical report no. 90-5, Institute of Cognitive Science, University of Colorado at Boulder. Legendre, (ilraldine, Yoshiro Miyata, and Paul Smolensky. 1991. Unifying Syntactic and Semantic Approaches to Unaccusativity: A Connectionist Approach. B L S 17. 156-167. Le Roux. C. 1988. On the Interface of Morphology and Syntax: Evidence from Verb Particle Combinations in Afrikaans. Stellenbosch Papers in Linguistics 18. University of Stellenbosch, South Africa. Levin. Beth. 1985. Lexical Semantics in Review: An Introduction. In B. Levin, ed.. Lexical Semantics in Review. Lexicon Project Working Papers I . Cambridge, Mass.: M IT Center for Cognitive Science. Levin, Beth. 1993. English Verb Classes and Alternations. Chicago: University of Chi­ cago Press. l,evin, Beth, and T. Rapoport. 1988. Lexical Subordination. C LS 24, Part 1, 275-289. Levin, Beth, and Malka Rappaport. 1986. The Formation of Adjectival Passives. Lin­ guistic Inquiry 17:623-661. Levin, Beth, and Malka Rappaport Hovuv. 1990a. The Lexical Semantics of Verbs of Motion: The Perspective from Unaccusativity. In I. M. Roca, ed.. Thematic Structure: Its Role in Grammar, 247-269. Berlin; Mouton de Gruyier. Levin, Belh, and Malka Rappaport Hovav. 1990b. Wiping the Slate Clean: A Lexical Semantic Exploration. Cognition 41:123-155. Levin, Beth, and Malku Rappaport llovav. 1992. Classifying Single Argument Verbs. Unpublished manuscript. Northwestern University and Bar IIan University. Levin, Lori. 1987. Toward a Linking Theory of Relation Changing Rules in LFG . Re­ port CSLI-87-115. Stanford, Cal.: Center for the Study of Language and In­ formation, Stanford University. Levin, Lori, T. Mitamuru. and A. T. Mahmoud. 1988. Lexical Incorporation and Resultative Secondary Predicates. Presentation at the LS A Annual Meeting, New Orleans. Lieber, Rochelle. 1988. Phrasal Compounds in English and the Moqphology-Syntax Interface. C LS 24, 202-222. Lindner, Susan. 1981. A Lexico-Semantic Analysis o f Verb-Particle Constructions with Up and Out. Ph.D. diss.. University of California, San Diego. Locke, John. 1690. An Essay Concerning Human Understanding. Published 1964. Cleveland: Meridian Books. McCawley, James D. 1968a. The Role of Semantics in a Grammar. In E. Bach and R. T, Harms, eds.. UniversaLi in Linguistic Theory, 124-169. New York: Holt. Rinehart and Winston. McCawley, James D. I% 8 b . Lexical Insertion in a Transformational Grammar without Deep Structure. C LS 4, 71-80. McCuwley, James D. 1973. Syntactic and I ogical Arguments for Semantic Structures. In O. Farjimura, ed.. Three Dimensions in Linguistic Theoiy, 259-376. Tokyo: TEC.' Corporation. McCawley, James D., ed. 1976. Syntax and Semantics 7: Notes from the Linguistic Underground. New York: Academic Press.

254

Bibliography

McCawley, James D. 1978. Conversational Implicature and the Lexicon. In P. Cole, ed., Syntax and Semantics 9: Pragmatics, 245-259. New York: Academic Press. McCawley, James D. 1986. The Focus and Scope of Only. University of Chicago Work­ ing Papers in Linguistics. Linguistics Department, University of Chicago. McClelland, J. L , D. E. Rumelhart, and G. E. Hinton. 1986. The Appeal of Parallel Distributed Processing. In Rumelhart and McClelland, eds., 1986, vol. I: Foundations. Parallel Distributed Processing, 3-44. MacWhinney, Brian. 1978. The Acquisition of Morphophonology. Monographs of the Society for Research in Child Development, vol. 43. Chicago: Universily of Chicago Press. MacWhinney, Brian. 1989. Competition and Lexical Categorization. In R. Corrigan, F. Eckman, and M. Noonan, eds.. Current Issues in Linguistic Theory. Vol. 61: Linguistic Categorization. Amsterdam Studies in the Theory and History of Linguistic Science, series 4. Amsterdam: John Benjamins. MacWhinney, Brian. 1991. Connectionism as a Framework for Language Acquisition Theory. In J. Miller, Research on Child Language Disorders, 73-104. Aus­ tin, Texas: Pro-Ed. Makkai. Adam. 1972. Idiom Structure in English. The Hague: Moulon. Maldonado Soto. Ricardo. 1992. Middle Voice: The Case of Spanish Se. Ph.D. diss.. University of California, San Diego. Marantz, Alec P. 1984. On the Nature o f Grammatical Relations. Cambridge, Mass.: M IT Press. Marantz, Alec P. 1992. The Wczy-construction and the Semantics of Direct Arguments in English: A Reply to Jackendoff. In T. Stowell and E. Wehrli, eds., Syntax and Semantics 26: Syntax and the Lexicon, 179- 188. New York: Academic Press. Maratsos, M.. R. Gudeman, P. Gerard-Ngo, and G. DeHart. 1987. A Study in Novel Word Learning: The Productivity of the Causative. In B. MacWhinney. ed.. Mechanisms of Language Acquisition. Hillsdale, N.J.: Lawrence Erlbaum Associates. Matsumolo, Yo. 1991. Some Constraints on the Semantic Structures of Verbs: Ev i­ dence from Japanese Motion Predicates. Unpublished manuscript, Stanford University. Mchombo, Sam. 1978. A C ritical Appraisal o f the Place o f Derivational Morphology in Transformational Grammar. Ph.D. diss.. School of Oriental and African Studies, University of London. Mchombo, Sam. 1992. The Stative in Chichewa and the Relevance of Thematic Infor­ mation. Presentation at the Cognitive Science Colloquium, University of Cal­ ifornia, Berkeley. Meyer, D. E., and R. W. Schvaneveldt. 1971. Facilitation in Recognizing Pairs of Words: Evidence of a Dependence between Retrieval Operations. Journal of Experi­ mental Psychology 90:227 -234. Michaelis, Laura. 1993. Toward a Grammar of Aspect: The Case of the English Perfect Construction. Ph.D. diss., University of California, Berkeley. Michaelis, Laura. 1994. A Case of Constructional Polysemy in Latin. Studies in Lan­ guage 18:23-48.

Bibliography

255

Minsky, Marvin. 1975. A Framework for Representing Knowledge. In P. H. Winston, ed.. The Psychology of Computer Vision. New York: McGraw-Hill. Mithun, Marianne. 1991. Active/Ageniive Case Marking and Its Motivation. Language 67(3): 510-547. Mohanan, Tara. 1990. Arguments in Hindi. Ph.D. diss.. Stanford University. Montague, Richard. 1973. The Proper Treatment of Quantifiers in Ordinary English. In J. Hintikka. J. Moravecsik, and P. Suppes, eds.. Approaches to Natural Lan­ guage. Dordrecht: Reidel. Reprinted in R. Thomason, ed.. Formal Philoso­ phy: Selected Papers of Richard Montague. New Haven: Yale University Press, 1974. Morolong, Mailillo, and Larry Hyman. 1977. Animacy, Objects, and Clitics in Sesotho. Studies in African Linguistics 8:199-217. Mufwene, Salikoko. 1978. English Manner-of-Speaking Verbs Revisited. C LS 14, Pa­ pers from the Parasession on the Lexicon, 278-289. Na, Younghee. 1986. The Conventionalization of Semantic Distinctions. C LS 22, Part I, 166-178. Naigles, Lelilia. 1990. Children Use Syntax to Learn Verb Meanings. Journal of Child luxngiutge 357-374. Naigles, L.utitia, Henry Gleitman, and Lila Gleitman. 1993. Children Acquire Word Meaning Components from Syntactic Evidence. In E. Dromi, ed.. Language and Cognition: A Developmental Perspective, 104-140. Norwood, N.J.: Ablex Publishing. Napoli. Donna Jo. 1992. Secondary Kesultative Predicates in Italian. Journal o f Lin­ guistics 28:53-90. Norvig, Peter, and George Lakoff. 1987. Taking: A Study in Lexical Network Theory. B L S 13. 195-206. unberg, Geoffrey. 1979. The Non-Uniqueness of Semantic Solutions: Polysemy. Lin ­ guistics and Philosophy 3(2): 143- 184. Nunberg, Geofrey, and Annie Zaenen. 1992. Systematic Polysemy in Lexicology and Lexicography. In Hannu Tommola, Krista Varantola, Tarja-Salmi-Tolonen and Jurgen Schopp, eds.. Proceedings o f Euralex II. Tampere, Finland: Uni­ versity of Tampere. Nunberg, Geoffrey, Thomas Wasow, and Ivan Sag. 1982. Idioms: An Interim Report. In Proceedings of the Plenary Sessions, 13th International Congress o f Lin ­ guists, Tokyo 1982. Oehrle, Richard T. 1976. The Grammatical Status o f the English Dative Alternation. Ph.D. diss., MIT. Oosten, Jeanne van. 1977. Subjects and Agenthood in English. C IS 13, 459-471. Oo .ieii, Jeanne van. 1984. The Nature o f Subjects, Topics and Agents: A Cognitive Explanation. Ph.D. diss.. University of California, Berkeley. Distributed by Indiana University Linguistics Club, Bloomington. The I'ompacl Edition of the Oxford English Dictionary (OED). 1971. Oxford: Oxford University Press. Park, T.-Z. 1977. Emerging Language in Korean Children. Unpublished MS., Institute of Psychology, Bern. Parlee, Barbara Hall. 1965. Subject and Object in Modern English. Published in J. Han-

256

Bibliography kamer. ed.. Outstanding Dissertations in Linguistics Series. New York: Gar­

land, 1979. Partee, Barbara Hall. 1971. On the Requirement That Transformations Preserve Mean­ ing. In C. J. Fillmore and D. T. Langendoen, eds.. Studies in linguistic Se­ mantics. 1-21. New York: Holt, Rinehart and Winston. Perlmutter, David M. 1978. Impersonal Passives and the Unaccusativc Hypothesis. BLS 4, 157-189. Perlmutter. David M. 1989. Multiattachment and the Unaccusative Hypothesis: the Per­ fect Auxiliary in Italian. Probus 1(1):63 - 119. Perlmutter, David M., and Paul M. Postal. 1983a. Some Proposed Laws of Basic Clause Structure. In D. M. Perlmutter, eds., Studies in Relational Grammar, vol. I. 81-128. Chicago: University of Chicago Press. Perlmutter, David M., and Paul M. Postal. 1983b. Toward a Universal Characterization of Passivization. In D. M. Perlmutter, ed., Studies in Relational Grammar. vol. 1, 3-29. Chicago: University of Chicago Press. Pinker, Steven. 1981, Comments on the paper by Wexler. In C. L. rfaker and J. J. M c­ Carthy, eds., The Logical Problem of Language Acquisition, 53-78. Cam­ bridge, Mass.: M IT Press. Pinker, Steven. 1984. Language team ability and Language Development. Cambridge. Mass.: Harvard University Press. Pinker, Steven. 1987. The Bootstrapping Problem in Language Acquisition. In B. MacWhinney, ed., Mechanisms of Language Acquisition, 399 441. Hills­ dale, N.J.: Lawrence Erlbaum Associates. Pinker, Steven. 1989. I^earnability and Cognition: The Acquisition of Argument Struc­ ture. Cambridge, Mass.: M IT Press. Pinker, Steven, and Alan Prince. 1988. On Language and Connectionism: Analysis of a Parallel Distributed Processing Model of Language Acquisition. Cognition 28:73- 193. Pinker, Steven, and Alan Prince. 1991. Rules and Associations. BL S 17,230 251. Pinker, Steven, D. S. Lebeaux, and L. A. Frost. 1987. Productivity and Constraints in the Acquisition of the Passive. Cognition 26:195-267. Pollard, Carl, and Ivan A, Sag. 1987. Information-Based Syntax and Semantics I: Fun­ damentals. C S L I Lecture Notes Series no. 13. Stanford, Cal.: Center for the Study of Language and Information, Stanford University. Pollard, Carl, and Ivan A. Sag. 1994. Head-Driven Phrase Structure Grammar. Chi­ cago: University of Chicago Press and Stanford, Cal.: Center for the Study of Language and Information. Postal, Paul M. 1971. On the Surface Verb Remind. In C. J. Fillmore and D. T. Langen­ doen, eds., Studies in Linguistic Semantics, 18 1 270. New York: Holt, Rine­ hart and Winston. Also in Linguistic Inquiry 1:37- 120 (1970). Prince. Alan, and Paul Smolensky. 1991. Lecture Notes on Connectionism and Lingui stic Theory. Distributed at the Summer Institute of Linguistics. Santa Cru?. Cal. Pustejovsky, James. 1991a. The Syntax of Event Structure. Cognition 41:47-81. Pusiejovsky, James. 1991b. The Generative Lexicon. Computational Linguistics 17(4): 409-441. Quillian. M. Ross. 1968. Semantic Memory. In M. Minsky, eds., Semantic Information Processing, 227 -270. Cambridge. Mass.: M IT Press.

Bibliography

257

Quine, W. V. O. I960. Word and Object. Cambridge. Mass.: M IT Press. Randall, Janet H. 1983. A Lexical Approach to Causatives. Journal of Linguistic Re­ search 2(3) :77-105. Kappaport. Malka, and Beth Levin. 1985. A Study in Lexical Analysis: The Locative Alternation. Unpublished manuscript. Bar llan University and Northwestern University. Kappaport. Malka, and Beth Levin. 1988. What to Do with Theta Roles. In W. Wilkins, eds., Syntax and Senuzntics 21 Thematic Roles, 7-36. New York: Academic Press. Rappaport Hovav, Malka, and Beih Levin. 1991. Is There Evidence for Deep Unaccu­ sativity in English? An Analysis of Rcsultative Constructions. Unpublished manuscript. Bar llan University and Northwestern University. Ratcliff, R., and G. McKoon. 1978. Priming in Item Recognition: Evidence for the Prepositional Structure of Sentences. Journal of Verbal Learning and Verbal Behavior 17:403-417. Reddy Michael. 1979. The Conduit Metaphor. In A. Ortony, ed.. Metaphor and Thought, 284-324. Cambridge: Cambridge University Press. Ricc. Mabel 1... and John V. Bode. 1993. G A PS in the Verb Lexicons of Children with Specific Language Impairment. First Language 13, 113-131. Rice, Sally. 1987a. Transitivity and the Lexicon. Center fo r Research in Language Newsletter 2.2. University of California, San Diego. Rice, Sally. 1987b. Participants and Non-participants: Toward a Cognitive Model of Transitivity. Ph.D. diss., University of California, San Diego. Rice, Sally. 1988. Unlikely Lexical Entries. B L S 14, 202 - 212. Roberts. R. B., and I. P. Goldstein. 1977. The F R L Manual. Technical Report AIM-408. Cambridge, Mass.: M IT Artificial Intelligence Laboratory. Rosch, Eleanor. 1973. Natural Categories. Cognitive Psychology 4:328-50. Rosch, Eleanor, and Carolyn Mervis. 1975. Family Resemblances: Studies in the Inter­ nal Structure of Categories. Cognitive Psychology 7:573 —605. Rosch. Eleanor, Carolyn Mervis, Wayne Gray, David Johnson, and Penny BoyesBraem. 1976. Basic Objects in Natural Categories. Cognitive Psychology 8: 382-439. Rosen, Carol G. 1984. The Relations. In D. Grammar, vol. 2, Ross, John Robert. 1969. 77-102.

Interface between Semantic Roles and Initial Grammatical Perlmutter and C. G. Rosen, eds., Studies in Relational 38-77. Chicago: University of Chicago Press. Auxiliaries as Main Verbs. Journal of Linguistics 1(1):

Ross, John Robert. 1970. On Declarative Sentences. In R. A. Jacobs and P. S. Rosen­ baum. eds., Readings in English Transformational Grammar, 222-272. Wal­ tham, Mass.: Ginn. Rumelhart, David E., and James L. McClelland, eds., 1986. Parallel Distributed Pro­ cessing: Explorations in the Microstructure of Cognition. 2 vols. Cambridge, Mass.: M IT Press. Sag, Ivan A. and Carl Pollard. 1991. An Integrated Theory of Complement Control. Language 67( 1): 63—113. Salkoff, Morris. 1988. Analysis by Fusion. Lingvisticae Investigationes 12( 1): 49—84. Amsterdam: John Benjamins.

258

Bibliography

Sanches, M. 1978. On the Emergence of Multi-Element-Utterances in the Child’s Japa­ nese. Unpublished MS, University of Texas at Austin, Dept, of Anthropology. Sapir. Edward. 1944. On Grading: A Study in Semantics. Philosophy of Science 293-116. Saussure, Ferdinand de. 1916. Cours de linguistique generate. Paris: Payet, 1973. Translated by W. Baskin. New York: McGraw Hill, 1976. Schank, R. C., and R. P. Abelson. 1977, Scripts, Plans, Goals and Understanding: An Inquiry into Human Knowledge Structures. Hillsdale, N.J.: Lawrence Erlbaum. Schieffelin. B. B. 1985. The Acquisition of Kaluli. In D. I. Slobin. ed.. The Crosslinguistic Study of Language Acquisition. Vol. I: The Data, 525-593. Hillsdale, N.J.: Lawrence Erlbaum Associates. Schlesinger, I. M. 1971. Production of Utterances and Language Acquisition. In D. I. Slobin, ed.. The Ontogenesis o f Grammar. New York: Academic Press. Searle, John R. 1983. Intenlionalin;: An Essay in the Philosophy o f Mind. Cambridge: Cambridge University Press. Shibatani, Masayoshi. 1973. A Linguistic Study of Causative Constructions. Ph.D. diss.. University of California, Berkeley. Shibatani, Masayoshi, ed. 1976. Syntax and Semantics 6: The Grammar of Causative Constructions. New York: Academic Press. Shieber, S. 1986. An Introduction to Unification-Based Approaches to Grammar. C SLI Lecture Notes no 4. Stanford, Cal.: Center for the Study of Language and Information, Stanford University. Shieber, S., L. Karttunen, and F. Pereira, eds. 1984. Notes from the Unification Under­ ground: A Compilation of Papers on Unification-Based Grammar Formal­ isms. SR I Technical Report 327. Menlo Park, Cal.: S R I International. Simpson, J. 1983. Resultatives. In L. Levin, M. Rappaport. and A. Zaenen.eds., Papers in Lexical-Functional Grammar, 143- 157. Bloomington: Indiana University Linguistics Club. Slobin, Dan. 1970. Universals of Grammatical Development in Children. In W. J. M. Levelt and G. B. Flores d'Arcais, eds.. Advances in Psycholinguistic Re­ search. Amsterdam: North-Holland. Slobin, Dan. 1985. Crosslinguistic Evidence for the Language-Making Capacity. In D. Slobin, ed., A Crosslinguistic Study o f Language Acquisition. Vol. 2: Theoretical Issues. Hillsdale, N.J.: Lawrence Erlbaum. Smolensky, Paul. 1986. Information Processing in Dynamical Systems: Foundations of Harmony Theory. In D. E. Rumelhart and J. L. McClelland, eds., Parallel Distributed Processing. Vol. I: Foundations, 194 - 201. Sproat, Richard W. 1985. On Deriving the Lexicon, Ph.D. diss., MIT. Stowell. Timothy. 1981. Origins of Phrase Structure. Ph.D. diss., MIT. Sweetser. Eve. 1990. From Etymology to Pragmatics. Cambridge: Cambridge Univer­ sity Press. Talmy, Leonard. 1976. Semantic Causative Types. In Shibatani, M., ed., Syntax and Senmntics 6: The Grammar of Causative Constructions. New York: Aca­ demic Press. Talmy. Leonard. 1977. Rubber-Sheet Cognition in Language. C LS 13. Talmy, Leonard. 1978. The Relation of Grammar to Cognition. In D. Waltz, ed., Pro■

Bibliography

259

ceedings ofTINLAP-2 (Theoretical Issues in Natural language Processing). Champaign: Coordinated Science Laboratory, University of Illinois. Tjlmy, Leonard. 1983. How Language Structures Space. In H. Pick and L. Acredolo, eds., Spatial Orientation: Theory. Research, and Application. New York: Plenum Press. Talmy, Leonard. 1985a. Lexicalization Patterns: Semantic Structure in Lexical Forms. In T. Shopen, ed., Language Typology and Syntactic Description, vol. 3: Grammatical Categories and the Lexicon. 57- 149. Cambridge: Cambridge University Press. Talmy, Leonard. 1985b. Force Dynamics in Language and Thought. C LS 21.1. Parasession on Causatives and Agentiviry, 293 - 337. Tenny. Carol. 1987. Grammaticalizing Aspect and Affectedness. Ph.D. diss. MIT. Thomason, R. H. 1992. N ET L and Subsequent Path-Based Inheritance Theories. Com­ puters and Mathematics with Applications 23(2-5): 179-204. Touretzky, David. 1986. The Mathematics of Inheritance Systems. Los Altos, Cal.: Mor­ gan Kaufmann. Traugott. Elizabeth C. 1988. Pragmatic Strengthening and Grammaticalization. B L S 14, 406-416. Traugott, Elizabeth C. 1989. On the Rise of Epistemic Meanings in English: An Ex­ ample of Subjectification in Semantic Change. Language 65( I ): 3 1-55. Trechsel. Frank R. 1982. A Categorial Fragment of Quiche. Texas Linguistic Forum 20. Austin: Department of Linguistics, University of Texas. Tuggy, David 1988. Nahuatl Causalive/Applicatives in Cognitive Grammar. In B. Rudzka-Ostyn, ed.. Topics in Cognitive Linguistics. Philadelphia: John Benjamins. Van Valin, Robert D., Jr. 1990a. Semantic Parameters of Split Intrasitivity. Language 66(2): 221 -260. Van Valin, Robert D., Jr. 1990b. The Linking Theory in RRG. Presentation at the Center for the Study of Language and Information. Stanford University. Van Valin, Robert D., Jr. 1992. Incorporation in Universal Grammar: A Case Study in Theoretical Reductionism. Journal of Linguistics 28:199-220. Vel&squez-Castillo. Maura. 1993. The Grammar of Inalienability: Possession and Noun Incorporation in Paraguayan Guarani. Ph.D. diss.. University of California, San Diego. Visser. F. Th. 1963. An Historical Syntax of the English Language. Part I, Syntactical Units with One Verb. Leiden: E. J. Brill. Ward, Gregory. Richard Sproat, and Gail McKoon. 1991. A Pragmatic Analysis of SoCalled Anaphoric Islands. Language 67(3):439-474. Wasow, Thomas. 1977. Transformations and the Lexicon. In P. W. Culicover, T. Wasow, and A. Akmajian, eds.. Formal Syntax, 327-360. New York: Academic Press. Wasow, Thomas. 1981. Comments on the Paper by Baker. In C. L. laker and J. J. McCarthy, eds.. The Logical Problem of Language Acquisition, 324-329. Cambridge, Mass.: M IT Press. Watkins. R. V., M. L. Rice, and C. C. Moltz. 1993. Verb and Inflection Acquisition in Language-impaired and Normally Developing Preschoolers. First Language. Wheeler. Daniel. 1970. Processes in Word Recognition. Cognition Psychology 1: 59-85.

260

Bibliography

Wierzbicka, Anna. 1986. The Semantics of the ‘Internal Dative’: A Rejoinder. Quadem i di Semantica 7 : 155- 165. Wierzbicka, Anna. 1988. The Semantics o f Grammar. Amsterdam: John Benjamins. Wilensky, Robert. 1982. Points; A Theory of the Struc'.me of Stores in Memory. In W. Lehnert and M. Rengle, eds., Strategies fo r Natural-I^anguagc Processing. 345-374. Hillsdale, N.J.: Lawrence Erlbaum Associates. Wilensky, Robert. 1986. Some Problems and Proposals for Knowledge Representation. Cognitive Science Report 40, University of California, Berkeley. Wilensky, Robert. 1991. Extending the Lexicon by Exploiting Subregularities. Report UCB/CSD 91/618. Computer Science Division (EEC S ), University of Cali­ fornia, Berkeley. Williams, Edwin. 1983. Against Small Clauses. Linguistic Inquiry 14(2):287—308. Wittgenstein. Ludwig. 1953. Philosophical Investigations. New York: Macmillan. Zadrozny, Wlodek, and Alexis Manaster-Ramer. 1993. The Significance of Construc­ tions. Unpublished manuscript. IB M T. J. Watson Research Center and Wayne State University. Zaenen, Annie. 1991. Subcategorization and Pragmatics. Presentation at the Center for the Study of Language and Information, Stanford University. Zaenen, Annie. 1993. Unaccusativity in Dutch: Integrating Syntax and Lexical Seman­ tics. In J. Pustejovsky, ed.. Semantics and the i rxicon. 129- 161. Dordrecht: Kluwer. Zaenen, Annie, and Adele E. Goldberg. 1993. A Review of Grimshaw's Argument Structure, language 69(4): 807-817. Zubizarreta, Maria Luisa. 1987. Levels of Representation in the lexicon and Syntax. Dordrecht; Foris. Zwicky, Arnold. 1971. In a Manner of Speaking. Linguistic Inquiry 11f2 ): 223—233. Zwicky, Arnold. 1987. Constructions in Monostratal Syntax. C LS 14, Zwicky, Arnold. 1989. What's Become of Derivations? Defaults and Invocations. BLS 15, 303-320. Zwicky, Arnold. 1990. Syntactic Words and Morphological Words, Simple and Com­ posite. Yearbook of Morphology 3, 201-216.

Index

abduction, 71 Abelson, R. P., 26,31 accommodation. See coercion Ackerman, F„ 112, 113-15,215,236 Aissen.J., 105 Alsina. A.,61, 112-15, 187,230, 232 Anderson, J. R., 72 Anderson, S. R., 2, 103, 175 argument role, 43 -59, 65, 110, 231 Argument Selection Principle, 116 Aronoff, M „ 23. 74, 229 Aske.J., 153. 158 associative memory, 133-35, 139 Austin, J. L., 25, 32,44 Bach. E „ 222 Baker. C. L , 121 Baker. E.. 19 Baker, M. C., 24-25. 102, 103. 214. 235 ■ tartlett, F., 26, 31 Baler r 70 Bellelli, A., 231 benefuctive construction, 77, 141, 149-51 Bever, T. G., 165 Jiijection Principle. 11 uloom, L.. •*!, 42 Bobrow, D. G., 5,26,31,72 Bode, J. V., 42 Bolingcr, D. L . 2-3,6, 32,44,68,97. 219. 229 Borkin, A.. 2-3, 97 Bowerman. M „ 5. 9.40-42,66, 105, 122. 226 Boyes-Braem, P., 32.44 Braine, M. D. S . 120, 121 Bresnan, J „ 7, 11,20, 23, 28,74, 104, 112, 113, 122. 142, 180, 182, 186-87,217.230 Brody. R. E., 120 Brown, R , 19,42, 121 Brugman, C. M ., 4,6, 32.44. 159, 219 Burzio, L . 212, 231 Bybee, J , 29. 133, 134. 192, 226

Carlson, G. N„ 17 Carrie,, J., 180, 181. 182- 83, 184, 186, 192. 193. 197,237

Carroll, J.. 36 Carter. R.. 8, 110, 159, 229, 232, 238 Cattell, R.. 142 Causal Relation Hypothesis, 62-65 causative construction, 61 caused-motion construction, 3. 52- 54.60, 100, 107, 152-79, 234; polysemy of, 76; relation of to resultative construction, 81-97 Channon, R., 106 Chomsky, N., 1,2, 8, 11, 23, 101, 104, 106, 230 Clark. E. V.. 3.5, 27,40-41.48. 66, 67, 68. 70,226, 231,233 Clark. H. H.,27 cleft construction, 229 co-predication. 155 coercion. 159. 238 cognitive decision, 166-67, 174 compositionalily, 13-16.202 Comrie, B.,49 conative construction, 4,63-64 conceptual archetype, 40.226 conduit metaphor, 148 conflation pattern, 60 conncclinnism, 25, 71,74, 135 construction, definition of, 4 constructional polysemy. See polysemy, constructional conventionalized scenario, 168-69, 174 correspondence principle, 50. 53, 177, 239 Croft, W , 61-62,66, 159, 240 cutting, 57-58 dative construction (prepositional paraphrase of ditransitive construction), 89-97, 104-6, 123, 130 Davis, A., 72 DeHart, G.. 120 DeLancey, S., 6 depictive predicate. 156 Dinsmore, J., 175 direct causation. 152, 165-75, 179

261

262

Index

directional phrases, 81-89, 152, 158-59, 169-71, 199. 208 DiSciullo. A.-M., 4, 21,215 dilransitive construction, 2. 3, 8, 49 -51, 53-55,60.75-77,82, 104-6, 141 51. 230, 235: morphophonological constrain) on, 128-29; polysemy of. 32-39; prag­ matic constraints on, 92-97; productivity of, 120-36; relation of to prepositional paraphrase, 89-97; semantic change in, 132, verb classcs, 125-36 Dixon. R. M. W „ 117 dominating construction, 74,226 Dowiy, D.,6, 14,21,57, 101. 116-17, 132. J92. 194,222, 223,231.232.236 Dryer, M.. 104-5, 106, 235 elaboration link, 233 Emanatian, M „ 23, 32 Emonds, J., 219, 230 encoding idiom. 155 England, N . 117 Erteschik-Shir, N.,92. 123. 125 Espenson, J.. 239 extension links, 233 Fahlman, S.. 5, 72. 74 Fait?., L „ 235 Farrell. P., 105 Fauconnier. G., 103 Filip, H.,4 Fillmore. C. J., 2. 4,6,7. 21.25, 26. 28.32, 39,43-45,48, 49, 57-59,66. 74, 102, 103, 108, 110, 132, 153. 159, 170, 175. 219,222, 229,241 Fisch, S. M .. 120 Fisher, C.. 19,44,57.230 Flickinger, D., 72, 73 focus, 92 97. 123-25, 234 Fodor, J. A.. 25, 165, 235 Fodor. J. D.. 25 Foley, W. A., 6, 7,20,28,49, 110, 112. 116 force dynamics. 76. 161-63 frame semantics, 7. 25-31, 40,43-44,47, 221.132 Frege, G., 13-14 Freidin, R., 20, 230 French, 134 Frost, L. A.. 120 funeiiona\ word, 2 \5 fusion, 50. 58, 65

Garrett. M. F., 25 Gawron.J.M., 26, 154-56. 158, 165.237 Gazdar. G . 6, 7, 14, 222 Geeraerls, D.. 237 Gelman.S. A . 70, 233 generative semantics. 6, 60. 101-4, 112 Gerard-Ngo, P.. 120 Gergely, G., 165 Gibbs. R „ 202 Gibson, J. J., 24 Giv6n, T., 3,6 Gleitman, H., 19 Gleitman, L „ 18-20,44.45. 57, 230, 233 Goldberg, A. E „ 4, 81, 128, 231, 233,239 Goldberg, R „ 8, 9.49. 105-6, 110. 112, 120, 122, 124. 126, 128-29, 133, 137. 139. 142, 175, 192. 237 Goldsmith. J., 35. 71 Goldstein. I. P., 26 Gordon, D., 21 Gray, W „ 32, 44 Green, G „ 2, 126, 129, 131, 142. 146, 150, 153. 192, 193, 195, 197 Greenfield, P. M.. 48 GnJgoire, A., 41 Grimshaw, J., 57, 131, 217, 231 Gropen. J , 8.9. 49, 105-6. 110. 112. 120. 122, 124. 126, 128-29, 133, 137, 139, 142, 175, 192, 236.237 Gruber. J. S.. 43. 89, 233,234 Gudeman, R., 120 Guerssel. M , 63 Guillaume, P., 134 Haiman, J „ 6. 32.44, 67 -69. 166 Hale. K.. 63. 230 Hall, G., 19,44, 57, 230 Halliday, M. A. K..43.234 Hanks, P.. 240 Hanlon, C.. 121 Higginbotham, J , 231 Hinton, G. E „ 25 Hobbes. T , 235 Hoekstra, T„ 153, 154, 156-58, 180, 185, 194, 237, 238 Hollander. M.. 8, 9, 49, 105-6, 110, 112. 120, 122, 124, 126, 128-29. 133, 137, 139, 142, 175, 192.237 Hood, I... 41 Hopper, P. 1. \ \&,Tits Hovav, M. Rappaport See Rappaport Hovav, M

Index Hudson. R , 72, 74. 235 humanly relevant scene, 39-43,225 Humboldt, W von. 68 Hyman, L,, 113 indirect negative evidence, 121-27, 139, 226 inheritance, 5,67, 72,73, 81,98-100, 108-10, 117,226,229; complete mode, 73; link, 5.71-81,99-100, 108. 138.226; multiple, 73,97-98,100; normal mode, 73-73,98-99 inslancc link. 79-81, 138,233 intransitive motion construction. 3. 207,218 Israel, 240 JackendolT. R.. 2,6. 14, 21, 28. .17.49, 50,57, 60.62,74,76, 89, 103, 104, 110, 112. 122, 143, 148, 165. 169, 175, 180, 182-84, 186. 194.200, 202, 205,206. 212.214-17. 219-22, 234.240.241 Jakobson, R , 148 Jespersen, O., 203 Johnson. D.. 32,44 Johnson, M „ 33, 143,234 Jurafsky. t).. 5, 72-74 Kanerva, J., 28. 112. 217 Kapur, S., 124 Karttunen, L., 14 Kat/, E „ 19 Katz, J. J., 101 Kay. M.. 74 Kay. P., 4.6, 21,34, 35.49.74. 108. 110. 165. 171,219,222, m 238 Keenan. E. L.,6,49. 101. 116,219 Kemmer, S., 219, 232 Keyser, S., 231 Kiparsky, P.. 28, 57 Kirsner, R. S., 3 Klein, E,. 6. 7, 14,222 Koenig, J.-P.. 4, 111,219,232 Kuroda, S .Y ., 229 Lahey, M., 41 LakofT, G.. 4, 5.6, 7. 20,21. 26-28,32. 33, 44, 60. 67, 70, 72-74, 81. 87, 97, 100-103. 118. 122, 132, 143, 165, 167. 180.219,222,230.234 Lakoff. R., 102 Lambrecht, K „ 4, 6.43, 219, 234 Landau, B.. 18-20.45,233 Lane, N., 233

263

Langacker, R W „ 3, 6, 7, 21. 26. 28. 40. 44, 49,66,70, 101, 103, 133. 148. 175. 192. 219, 226, 230, 233,234 Laughren, M „ 63 Lebeaux, D. S., 120 Legendre, G., 71, 241 Leibnitz. G. W „ 224 Lemmens. M., 231 Le Roux, C., 23 Levin, B., 1.6,8, 10. 12-13, 16, 28,49,55, 60,62,63.82,85-86, 107, 111, 112-13, 122, 126-27, 134. 137, 154, 156-57, 175. 176, 180-82, 185, 197, 200.202, 212, 217, 231.234,238 Levin, L.. 112. 234 lexical causative, 165, 174 lexical rule approach, 7-9, 21 - 23,39, 224-25.230,232 Lieber, R.. 23 Lindner. S.,32,44 lisleme, 4, 215 locative alternation, 106-7, 175-79 Locke. J., 230 McCawley. J D.,6.60,87, 101. 102. 165 McClelland, J. L.,25,71 McKoon, G., 23, 72 McNamara, J., 19 MacWhinney, B., 67, 70. 71. 134,219 Mahmoud, A. T., 234 Mokkai, A., 155 Maldonado Soto. R., 23,219, 232 Monaster-Ramer, A „ 74, 219, 229 Marant/., A. P., 16. 20, 120, 138,216 Muratsos, M., 120 Matsumoto, Y., 61.65, 66,233 Mchombo, S. 20, 58,61, 104, 112-15, 187, 236 merging. 58 Mervis. C,, 32,44, 116 metaphorical extension, 33. 75. 81,88-89, 231 Meyer, D. E., 72 Michaelis, L.,4,219. 232 middle construction, 58, 183-85 Miller. P., 41 Minsky, M . 26. 31 Mitamura.T., 234 Mithun, M., 132 Miyata, Y.. 71

Mohanan, T., 215 Moltz. C . C . 4 2

264

Index

Montague Grammar, 13, 14,219.222-23 Morolong, M . 113 morphologically-marked alternation, 22-23 Moshi, L „ 112, 113 motion construction. 207 motivation, 67,68-73,99-100. 108 motivation link. See inheritance Mufwene, S.. 134 Mulder, R.. 237 Na, Y., 241 Naigles. L.. 19 Napoli. D. J., 180. 182. 191. 205,237,240 negative evidence, 121-25 Norvig, P., 5 null complement. 58-59. 178 Nunberg, G., 202,221, 241 O'Connor. C..4, 6.21.219 Oehrlc.R.T.,2. 36. 106, 130, 142, 150 Ogden, C. K.. 68 Oosten, J. van, 184 Park, T.-Z.. 4 1 Parkes.C. H „ 25 Partec, B. H „ 2, 101. 103, 142, 175, 229 participant role, 43-59,65. 110.231 patient role, 48, 180, 188.231,236 Pereira. F., 14 Perlmutter, D. M.. 102, 105, 106. 212. 232 Peters, S., 222 Pinker, S.. 6, 8, 9, 16, 19. 28-29,47. 49, 70. 103, 105-7, 110, 112, 113, 118. 120, 122-31, 133-35, 137, 139. 142, 175. 176, 192,217, 223, 226, 230,232, 236, 237, 238,240 Pollard. C.. 6,7,20. 72-74, 159, 238 polysemy, constructional, 32-39. 161 -64. 210-12,218. 225 polysemy link, 74-77, 118, 233 Postal. P. M , 6. 101, 102,105. 106 preemption, 29-30 Prince. A., 129 Principle of Avoiding Synonymy, 3. 67, 68, 91,229 Principle of Correspondence, 50,53, 177, 239 Principle of Maximized Economy. 67-69 Principle of Maximized Motivation, 67, 69 Principle of Maximizing Expressive Power. 67, 68-70

Principle of Semantic Coherence, 50, 53 - 54, 177, 239 productivity. 77. 100. 120-40,226 profiling, 26.44-49, 56-59, 225, 239; con­ structional, 48-53,56-59; mismatches. 52-53; test for, 45 Projection Principle, 11 proto-roles. 116, 232 Pullum. C.. 6, 7, 14,222 Pustejovsky, J.. 154-56,159, 185.219 Quillian, M. R.. 5 Quine. W. V. O., 18 Rakowitz. Susan. 19, 44, 57. 230 Randall, J. H.. 153, 180-84. 186. 192, 193. 197,237 Rapoport, T., 6.8. 10, 55,60.62, 113.200. 202,217 Rappaport Hovav, M., 8. 16, 49.55,62, 82. 84-86, 107, 110-12, 126, 137, 154. 156-57,175, 176, 180-82, 185, 197, 212, 231.234,238 Ratcliff. R „ 72 recipient role. I I I . 146-48 Reddy. M.. 128. 131. 148 reflexive, 232 resultative construction, 3, 80 - 89,97-100. 136-37, 154, 180-98, 217, 234; end-ofscalf constraint on, 195-97; metaphorical interpretation, 81 -87; productivity of. 136-37; relation of to caused-motion con­ struction, 87-89 Rice, M. L „ 42 Ricc, S., 21, 22, 118, 148,219, 232 Richards, I. A., 68 Rizzi. L „ 231 Roberts. R. B . 26 Rosch. E „ 32.44, 116 Rosen, C .G .. 102, 132 Ross.J. R.,6, 102 rule-to-rule hypothesis, 222 Rumelhart, D. E.. 25, 71 Sag. I A., 6, 7, 14, 20, 72, 74, 159, 202. 222. 238 SalkofT, M.. 200 Sanchcs, M , 41 Sapir. E . 195 Saussure, F. de. 4.69, 229 scene, 25, 229

Index Schank, R. C., 26, 31 Schieffelin, B. H.,42 Schlesinger, I. M.. 42 Schvaneveldt, R. W., 72 Schwartz, A.. 237 Senrle, J. R.. 32. 127. 161 semantic coherence principle. See Principle of Semantic Coherence shading. 57 Shared Participant Condition, 65 Shibatani. M.. 165, 168, 174,238 Shieber, S., 14 Simpson, J., 82, 86, 180. 181, 182 Slobin, D , 5.40, 42.66, 118. 162, 226 Smith, J., 48 Smolensky, P.. 71, 241 Sproat, R. V.., 23 S towel I, T., 23 subpan link, 78-79,81, 234 Sweetser, E „ 32,33,44, 162 syntactic bootstrapping, 19 Talmy. L . 29. 37,60,61.66, 76, 103, 153, 155, 159-62, 173,232,240 Tanenhaus. M. K.. 17 taxonomic relation, 70, 74 Tenny, C.. 84-86 theme role, linking of. 112 Thomason. R. H.. 72 Thompson, S. A., 118, 236 token frequency, definition of, 134 Touretzky, D.. 72 transformation, structure-preserving, 230 transitive construction, 117-19 TraugoU, E. C\, 33, 211 Trechsel, F. R „ 117 truth-functional synonymy, 103 Tuggy. D..219 type frequency, 77, 100. 134-39, 226; defini­ tion of, 134 unexpressed role, 58 unification, 14, 229 unique path constraint, 81-86

265

Universal Alignment Hypothesis, 102, 235 Universal Theta Assignment Hypothesis, 102 usage-based account, 133-39, 192, 226 Van der Leek. F., 34 Van Valin, R. D „ Jr.. 6, 7, 20, 28, 49, 104, 110. 112, 116, 180. 187-88. 193.212, 234, 235 Velazquez-Castillo, M., 216 Vendryes,J.,68 verb + particle construction, 97 Visser, F. Th.. 137, 180. 192 Walker, E. C. T.. 25 Wall, R „ 222 Ward, G., 23 Wasow, T., 21.71, 72, 120. 202, 230 Watkins. R. V„ 42 way construction, 16,62-63, 199 -218,236; productivity of, 137 Webber, B., 72 Webelhuth. G.. 215 Weisberger. M. J,, 120 Wheeler, D., 24 White Eagle. J., 63 Wierzbicka. A., 2-3,6, 34. 130. 219, 223-24. 229 Wilcox, S. A.. 70 Wilensky, R „ 5, 26,71-73,75, 98. 146, 164 Williams. E.,4,21, 192,215 Wilson, R „ 8, 9, 105-6, 110, 112. 120, 122. 124, 126, 128-29, 133, 137, 139, 142, 237 Winograd, T „ 5, 26, 31 Wittgenstein, L., 32, 44 word order, 109-10,229 Yazdani, A., 231 Zadrozny, W „ 74. 219, 229 Zaenen, A., 15- 16, 112, 175, 180, 182, 186-87,212,217, 231,239, 240,241 Zubizarreta, M. L., 231 Zwicky. A.. 120, 126, 215, 219. 236