Particle Verbs and Local Domains (Linguistik Aktuell Linguistics Today, LA 41)

Particle verbs and local domains Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) pro...

Author: Jochen Zeller

46 downloads 442 Views 4MB Size Report

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

DOWNLOAD PDF

Particle verbs and local domains

Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.

Series Editor Werner Abraham University of California at Berkeley University of Vienna Rijksuniversiteit Groningen

Advisory Editorial Board Guglielmo Cinque (University of Venice) Günther Grewendorf (J.W. Goethe-University, Frankfurt) Liliane Haegeman (University of Lille, France) Hubert Haider (University of Salzburg) Christer Platzack (University of Lund) Ian Roberts (University of Stuttgart) Ken Safir (Rutgers University, New Brunswick NJ) Lisa deMena Travis (McGill University) Sten Vikner (University of Stuttgart) C. Jan-Wouter Zwart (University of Groningen)

Volume 41 Particle verbs and local domains by Jochen Zeller


Jochen Zeller Johann Wolfgang Goethe Universität Frankfurt

John Benjamins Publishing Company Amsterdam/Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Zeller, Jochen. Particle verbs and local domains / Jochen Zeller. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 41) Includes bibliographical references and index. 1. Grammar, Comparative and general--Verb. 2. Grammar, Comparative and general--Particles. 3. Grammar, Comparative and general--Syntax. 4. Grammar, Comparative and general--Morphology. 5. Semantics. 6. German language--Verb. I. Title. II. Linguistik aktuell ; Bd. 41. P281.Z4 2001 415--dc21 isbn 90 272 2762 4 (Eur.) / 1 58811 036 2 (US) (Hb; alk. paper)

00-140117

© 2001 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Acknowledgements Introduction The topic of this study1 The problem1 The solution3 Particles5 Preliminaries7 Outline of the book9 Chapter 1 Syntax, morphology and lexical licensing 1.1 The lexicon and the architecture of grammar15 1.2 Syntactic movement and “late” lexical insertion23 1.2.1 Movement and interface level(s)24 1.2.2 The status of LF27 1.2.3 Distributed Morphology28 1.3 The domain of morphology33 1.3.1 On the notion “word”33 1.3.2 The rules of word formation37 1.3.3 Parallel Morphology44 1.4 Conclusion50 Chapter 2 The syntax of particle verbs 2.1 Verb movement to Comp54 2.1.1 Syntactic separability55 2.1.2 The morphological approach61 2.1.3 The incorporation approach68

ix 1

13

51

vi

Table of contents

2.2 Verb movement to Infl69 2.2.1 “Morphological” separability70 2.2.2 Backformations76 2.2.3 Particle verbs and finite Infl80 2.3 Gapping83 2.4 Topicalization88 2.4.1 The syntactic side of particle topicalization89 2.4.2 Conditions on particle topicalization93 2.5 Modification100 2.6 Structural restrictions101 2.7 Typological remarks104 2.8 Conclusion107 Chapter 3 The non-functional nature of particle phrases 3.1 Extended projections109 3.1.1 The extended projection of verbs110 3.1.2 The extended projection of nouns112 3.1.3 The extended projection of adjectives114 3.1.4 The extended projection of prepositions116 3.2 Particle phrases as non-functional XPs127 3.2.1 Referential noun phrases and functional structure127 3.2.2 Prepositional phrases and referentiality135 3.2.3 Adjectival particles and resultatives143 3.3 Conclusion147 Chapter 4 The lexical representation of particle verbs 4.1 The semantics of particle verbs152 4.1.1 Particles as arguments153 4.1.2 Semantic adjunction and eventive particles155 4.1.3 Aspectual particles and meaning variation158 4.1.4 Non-transparent particle verbs as phrasal idioms161 4.2 Local domains and class-based meaning166 4.2.1 Special meaning and syntactic contexts166 4.2.2 Class-based semantics and possible hosts173 4.2.3 Particle verbs derived by semiproductive rules179 4.2.4 More irregularities186

109

151

Table of contents

4.3 The lexical licensing of prepositions189 4.4 Conclusion192 Chapter 5 Local domains and morphology 5.1 Particles as semantic affixes196 5.1.1 “Semantic” allomorphy196 5.1.2 “Semantic” suppletion200 5.2 Morphological properties of particle verbs209 5.2.1 Particle verbs with “non-existing” base verbs211 5.2.2 Auxiliary selection215 5.2.3 Case assignment218 5.3 Local domains and the licensing of P0225 5.3.1 The morphological structure of postpositions225 5.3.2 Functional heads and local domains230 5.3.3 Postpositions as semi-lexical elements232 5.4 Conclusion236

195

Chapter 6 Particle verbs and word formation 239 6.1 Particle verbs in words240 6.1.1 The morphological bracketing of words derived from particle verbs240 6.1.2 The “No Phrase Constraint”, Compounding and Affixation243 6.1.3 The particle verb as a V0 and word formation by Move248 6.2 Structural adjacency and morphological structures255 6.2.1 Reanalysis255 6.2.2 Morphological Merger and Parallel Morphology258 6.3 Word formation with resultatives and h-verbs263 6.4 Conclusion269 Chapter 7 Typological remarks and reanalysis 7.1 On Verb Raising in Dutch271 7.2 On the verb-particle alternation in English and Mainland Scandinavian283 7.3 Conclusion296

271

vii

viii Table of contents

Conclusion

297

References

305

Name Index

315

Subject Index

319

Acknowledgements

A number of people have contributed to the writing and completion of this book. It’s a real pleasure for me to thank them here. A special thanks goes to my thesis supervisor Günther Grewendorf for his support, his encouragement, and his confidence in me. I also want to mention the major role he played in my becoming a linguist. He stimulated my interest in linguistics and in language as a fruitful topic of investigation. I am grateful to him for letting me profit from his broad linguistic knowledge and for allowing me to find my own way and pursue my own ideas at the same time. Much that I have achieved as a linguist, I owe to him. I also want to give a special thanks to my other two supervisors, Ian Roberts and Ede Zimmermann, who spent much time at numerous appointments discussing my work. I profited greatly from these discussions and from their helpful suggestions and valuable comments. Without my (linguist and non-linguist) colleagues Katharina Hartmann, Heike Schmitt-Spall, Eric Fuß, and Joachim Sabel, my time at Frankfurt university, both intellectually and socially, would not have been half as good as it has been. They were always willing to share their time to discuss various linguistic and non-linguistic issues. Whenever I needed to talk about a particular problem or a new idea, they offered me comments and criticism that shaped the various stages of my work. Working with and talking to them never failed to be both inspiring and challenging. From 1995 to 1996, I had the opportunity to spend a year as a visiting scholar at the University of Massachusetts at Amherst. The department of linguistics at South College was extremely stimulating. A special thanks goes to Hagit Borer, who made my stay in Amherst possible, and whose help, linguistic advise, and unconditional support made me believe in my ability to start, continue, and finish writing a dissertation. I had many fruitful discussions and constructive arguments with her; her work on the syntax-morphology interface had a profound influence on my own work. Many thanks to the faculty at South College for making linguistics fascinating, especially Angelika Kratzer, Barbara

x

Acknowledgements

Partee, Kyle Johnson, and Tom Roeper, who read and commented on very early drafts of parts of this book. I have had occasions to come to every one of them with questions, and they have been exceptionally approachable and helpful. Thanks also to my fellow students, classmates and friends, as well as to many other people that made my stay in the US worthwhile, particularly John Alderete, Mark and Eugenia Cassielles, Mike Dickey, Bart Hollebrandse, Angeliek van Hout, Winnie Lechner, Maribel Romero, Ed and Erica Rubin, Bernhard Schwarz, Satoshi Tomioka, Mercedes Valle, and Elisabeth Villalta. I also want to thank Alec Marantz and Ray Jackendoff, whose theories have been extremely relevant for the topic of this study. Jackendoff’s 1997-book on the architecture of the language faculty has been a continuous source of inspiration throughout my work. Marantz’s insight that syntactic domains are highly relevant for the determination of special meanings and special sounds lies at the heart of the proposal that I make in this book. When I contacted them to ask for appointments, they didn’t hesitate to meet with me, although they didn’t really know who I was. I profited tremendously from talking to them. I have met a number of linguists who share my interest in particle verbs and who have been (not surprisingly) very interesting and friendly people: Anke Lüdeling, with whom I had the pleasure to give my first joint talk; Karen Ferret, who explained the difference between Serge Gainsbourg and Allen Ginsberg to me; Martin Haiden, who once presented the truth about particle verbs in a talk entitled “The truth about particle verbs”, and Andrew McIntyre (“particle verbs are good for you”), with whom I share my enthusiasm for particle verbs and for Australia’s greatest rock band (Cold Chisel), and who sent me approximately 13 versions of selected parts of his 300-pages dissertation (single spaced, 9-point type, no margin). I thank all of them for our fruitful and stimulating conversations. At several stages I have benefited from comments, discussions and occasional appointments with Werner Abraham, Daniel Büring, Damir Cavar, Haydar Celik, Christine Erb, Hans-Martin Gärtner, Susanne Glück, Michael Grabski, Roland Pfau, Andrew Simpson, Susanne Laudenklos, Markus Steinbach, Ralf Vogel, Chris Wilder, Susi Wurmbrand, and Malte Zimmermann. I also gratefully acknowledge the opportunity to present parts of the material in this study to the audiences of various conferences in Europe and the United States and, at many times, to the colloquium of the linguistic department at Frankfurt University. I thank the audiences for their comments and criticism and my hosts in different cities for their hospitality. I would also like to express my gratitude for the financial support I have received from the Studienstiftung

Acknowledgements

des deutschen Volkes and the Deutsche Forschungsgemeinschaft. Work on this book was supported by DFG grants # GR 559/5–1 and GR 559/5–2. I have named many people, and yet I have left out even more, especially my parents and my brother, and all my non-linguist friends. I want to thank them here, and I apologize to those who I should have mentioned, but have not. Finally, I want to give a very special thanks to Dori for the love, the fun and the warmth she brought into my life, and for the support and understanding she has always provided. Without her, I never would have been able to write this book. I dedicate this work to her. Jochen Zeller Frankfurt, Germany August 2000

xi

Introduction

The topic of this study In this book, I show that the syntactic, semantic, and morphological properties of particle verbs follow from the specific, strictly local relation that holds between the particle and its base verb. Importantly, I argue that this local relation is not to be understood in morphological terms; the particle and the verb do not form a complex word. Rather, I suggest that a particle and a verb are structurally adjacent, with “structural adjacency” being defined as the relation between a head and the head of its complement. This means that a particle is a head whose maximal projection is the sister of the verb in syntactic structure. Since I assume that particles are members of the lexical categories P, N, and A, the verb-particle construction is characterized by two lexical heads that are structurally adjacent. In this respect, particle verbs differ from “normal” verb-complement constructions, where the verb is structurally adjacent to a functional head, since the verb’s complement is a functional projection. Therefore, a verb is normally separated from every lexical head inside its complement by at least one intervening layer of functional structure. However, the relation between a particle and a verb is not disturbed by a functional boundary; both heads are part of the same locality domain which is defined by the structural adjacency-relation. The major goal of this book is to show that the “word-like” properties of particle verbs follow from this situation; the local relation between terminal nodes that causes an element to behave like a word may also be established by a non-morphological structure.

The problem It is well-known that particle verbs in German and Dutch have properties of both complex words and of syntactic phrases. Consider the bracketed constituents in the following sentences:

2

Introduction

(1) prefix verb: weil er sich dem Gegner [unterwirft] because he refl the enemy [pref-throws ‘because he surrenders to the enemy’ (2) particle verb: weil er ihm seine Verfehlungen [vorwirft] because he him his lapses [part-throws ‘because he reproaches him with his lapses’ (3) verb-complement weil er ihm den Brief [in den Briefkasten wirft] because he him the letter [into the letterbox throws ‘because he throws the letter into his letterbox’

In (1), we have the prefix verb unterwerfen; (2) shows the particle verb vorwerfen, and (3) is a phrasal construction where the verb werfen combines with the full prepositional phrase in den Briefkasten. At first glance, the particle verb in (2) seems to be a word like the prefix verb in (1); both constructions have “word-like” properties. For example, consider the morphological rule that derives nouns from verbs. It seems that both the combination of a prefix and a verb and of a particle and a verb can be used as the appropriate input to this rule. The noun Unterwerfung is derived from the prefix verb unterwerfen in (1), while the noun Vorwurf is derived from the particle verb vorwerfen in (2). In contrast, it is impossible to derive a noun from the phrasal construction in (3); there is no word like *in den Briefkasten-Werfung or *in den Briefkasten-Wurf. It seems that particle verbs, like prefix verbs, are complex morphological elements, whereas the combination of the verb and the prepositional phrase in (3) is larger than a word; it is a phrasal constituent. Prefix verbs and particle verbs also behave like words in another respect. It can be observed that the lexical meanings of unterwerfen and vorwerfen are to a certain extent unpredictable; they are not based on the literal meaning of the verb werfen, but rather are idiosyncratically associated with the complex verb as a whole. In contrast, the semantic combination of the verb werfen and its phrasal complement in (3) is semantically transparent; the meaning of in den Briefkasten werfen is derived by combining the regular meaning of the verb werfen with the meaning of the complement-PP. According to standard assumptions, semantic idiosyncrasy is a property of words. Again, this observation seems to support the conclusion that the particle verb, like the prefix verb, is a word, whereas the complex construction in (3) is not.

Introduction

The surprising fact about particle verbs is that they sometimes do not behave like words: (4) Er [unterwirft]i sich dem Gegner ti he [pref-throws refl the enemy ‘He surrenders to the enemy’ (5) Er wirfti ihm seine Verfehlungen [vor ti] he throws him his lapses [part ‘He reproaches him with his lapses’ (6) Er wirfti ihm den Brief [in den Briefkasten ti] he throws him the letter [into the letterbox ‘He throws the letter into his letterbox’

In (4)–(6), the verb has undergone movement to the left in order to derive the verb second (V2) order of German main clauses. The prefix verb in (4) moves as a whole, which is what we expect from a word. Curiously, the particle verb in (5) does not behave like the prefix verb; instead, only the verbal part of the particle verb moves, and the particle is left behind. This resembles the situation that we find in (6). Here, the main verb moves and also leaves the prepositional phrase in den Briefkasten behind. This suggests that the particle vor in (2) has the same syntactic status as the PP in den Briefkasten, i.e. that particles are phrasal complements of their base verbs. This assumption accounts for the separation of the particle and verb in (5). But how would the assumption explain the word-like properties of particle verbs?

The solution The answer that I offer in this study is that the verb-particle construction shares important properties of both phrasal constructions and morphological objects. I argue that the particle is in fact represented as a phrasal complement of the verb. Consequently, the particle verb in (2) patterns with the construction in (3) with respect to a number of syntactic properties (like, for example, the fact that the particle is stranded when the verb moves). However, I will show that there is a crucial difference between particle phrases and phrases like in den Briefkasten. I assume that every regular phrasal complement of a verb consists of the phrasal projection of a lexical head and its functional extended projection that intervenes between the lexical head and the verb. This additional functional structure is absent in verb-particle constructions. Therefore, the relation between a particle

3

4

Introduction

and a verb is different from the relation between a verb and a lexical head inside a normal phrasal complement; crucially, it is more local, because no functional structure intervenes between the verb and the particle. This is where particle verbs begin to show parallels with morphological objects — the relation between two terminal nodes that form a complex morphological element is also strictly local, and no functional structure intervenes. (7) illustrates these differences and parallels with respect to the constructions in (1)–(6): (7) a.

Particle verb

VP

0

P vor V0 + complement

P0 in

P0 unter

V0 werf-

VP V0

FP PP

Prefix verb

V0 V0 werf-

PP

c.

b.

F0

werf-

NP den Briefkasten

The dashed lines in (7a) and (7b) illustrate that the relation between the lexical nodes V0 and P0 is strictly local in both prefix verbs and particle verbs. In contrast, V0 and P0 in (7c) are separated by the functional head F0. The locality domain in (7a) is different from the locality domain that is defined by the complex word in (7b), and many studies on particle verbs are primarily concerned with this structural difference. In contrast, the analysis that I propose in this book emphasizes the similarities between (7a) and (7b) with respect to locality. I argue that particle verbs pattern with morphological objects like prefix verbs because both constructions consist of two terminal nodes in a specific local relation. This is why particle verbs exhibit properties that are otherwise only attested with words. Locality domains are also relevant when it comes to the way lexical information is associated with syntactic nodes. My proposal requires a particular

Introduction

view about the way lexical information is associated with (morpho-)syntactic structures. The view that I defend in this book is based on the idea that the information contributed by a lexical item is a means to license the interaction between the phonological, the syntactic, and the semantic module of grammar. I argue that the possibility of associating special lexical meanings with terminal nodes in a syntactic tree is restricted and determined by the local syntactic environment of the node. In contrast to the traditional view, a complex word like a prefix verb is not the only structure that may define the local environment that is required for the assignment of special meanings. Phrasal structures can determine special meanings as well. I argue that particle verbs may exhibit idiosyncratic semantic properties, because the verb and the particle are structurally adjacent and hence part of the same locality domain.

Particles In the following, I want to show which elements I consider particles in this study (and which I do not). To do this, I simply list (sets of) examples; a definition of the notion “particle” that is compatible with this illustrative description is given in Chapter 3. It is well-known that most particles are prepositional in nature. In German, this class includes at least the following elements (cf. Lüdeling 1998a): (8) ab (‘off, from’), an (‘on, at’), auf (‘on, up’), aus (‘from, out of ’), bei (‘at, near’), durch (‘through’), ein (‘in, into’), nach (‘after, to’), über (‘over’), um (‘around’), unter (‘under’), vor (‘before, in front of ’), weg (‘away’), wider (‘against’), zu (‘to, at, for’) (9) Prepositional particle verbs: a. einbrechen (lit. in-break, ‘break in’) b. nachdenken (lit. after-think, ‘think about’) c. abwaschen (lit. off-wash, ‘wash up’) d. vorführen (lit. before-guide, ‘demonstrate’) e. austrinken (lit. out-drink, ‘drink up’)

My main focus will be on prepositional particle verbs, in particular when I discuss the lexical representation of particles in Chapter 4. In contrast to the prepositional elements in (8), I will argue that the following elements do not qualify as particles: (10) hinab, herab, heran, hinaus, heraus, herbei, hindurch …

5

6

Introduction

The elements in (10) are morphologically complex; they consist of a prepositional part and a (deictic) prefix her or hin. In traditional German grammars, these elements are called pronominal adverbs. Some linguists treat the elements in (10) as particles; however, I adopt van Riemsdijk’s (1990) view, according to which the elements in (10) are postpositions that realize the functional head of the extended projection of prepositional phrases. Although I show that postpositions share many of the properties of particles, postpositions and particles are essentially different elements with different structural and lexical properties. I assume with Booij (1990) and Stiebels & Wunderlich (1994) that there are also nominal particles: (11) Nominal particle verbs: a. Rad fahren (lit. bicycle-drive, ‘ride a bike’) b. Klavier spielen (lit. piano-play, ‘play the piano’) c. heimfahren (lit. home-drive, ‘drive home’) d. standhalten (lit. stand-hold, ‘resist’) e. teilnehmen (lit. part-take, ‘take part’)

The nominals in (11) (Rad, ‘bike’, Klavier, ‘piano’, etc.) are determiner-less singular count nouns. Since singular count nouns in German usually cannot occur without a determiner, the examples in (11) cannot be analyzed as normal verb-object constructions. Rather, I argue in Chapter 3 that the nominals in (11) are bare NP-complements of their verbs. Since determiners are associated with functional structure, this explains why the determiner is absent. The nonfunctionality of the nominal’s maximal projection makes it a particle by definition. Finally, I assume that there are also adjectival particles in German. As noted by Stiebels & Wunderlich (1994), there are only a few particle verbs derived from adjectives. Some examples are listed in (12): (12) Adjectival particle verbs: a. kurz treten (lit. short-kick, ‘go easy’) b. krankfeiern (lit. sick-celebrate, ‘play hooky’) c. schieflaufen (lit. askew-run, ‘go wrong’) d. ernst nehmen (lit. serious(ly) take, ‘take seriously’) e. schwarzfahren (lit. black-drive, ‘travel without a ticket’)

It is possible that some of the elements in (12) are in fact adverbs rather than adjectives (they modify the verb’s semantics rather than expressing a property that is predicated of an individual). Therefore, instead of labeling the elements

Introduction

in (12) adjectival particles, it is probably more appropriate to speak of particles of category A. In Chapter 3, I show that the particles in (12) are represented as bare AP-complements of V0. I distinguish them from functional adjectival complements of the verb which are generally classified as adjectival resultative constructions. A brief comment is also necessary on the question of whether there are also particles of category V. Stiebels & Wunderlich (1994) assume that the constructions in (13) are particle verbs: (13) Verbal particle verbs ?? a. spazieren gehen (lit. stroll-go, ‘stroll’) b. sitzen bleiben (lit. sit-remain, ‘stay down’)

The question is whether there is any evidence that the serial verb constructions in (13) are particle verbs. According to the claim that I make in this study, the examples in (13) would qualify as particle verbs if the base verbs take bare VP-complements. However, in order to answer the question of whether the infinitives in (13) are bare VPs, one would have to provide a general theory of the syntax of infinitival constructions. Since such a theory is clearly beyond the scope of my study, I have decided to remain agnostic about verbal particles and to restrict my attention to prepositional, adjectival, and nominal particles. Although I focus primarily on particle verbs in German, Dutch particle verbs, which share many of the properties of their German counterparts, are also discussed at various points in this book. Chapter 7 deals exclusively with verb-particle constructions in Germanic languages like Dutch, English, and Norwegian.

Preliminaries I now provide a short introduction of the main theoretical assumptions that I adopt as my point of departure. Some of these assumptions are modified or extended throughout the study; wherever this happens, I provide discussion in the text. With respect to syntax, I presuppose familiarity with the Principles-andParameters approach (cf. Chomsky 1981, 1986a, b, 1995) and with the core assumptions of X-bar- and Government-Binding theory (Chomsky 1981). More recent versions of the Principles-and-Parameters approach, as they follow from the Minimalist Program of Chomsky (1995), as well as questions and

7

8

Introduction

issues that arise with respect to these theoretical innovations, will be discussed in Chapters 1 and 3. I start out by assuming a simple clause structure as in (14); a more articulated phrase structure is then introduced and motivated in Chapter 3: CP

(14)

C'

Spec C0

IP Spec

I' Infl0

VP XP

V0

I assume that in both German and Dutch, IP and VP are right-headed, which gives rise to the SOV-word order in verb-final subclauses. Furthermore, I assume that the verb obligatorily moves to Infl0 in embedded clauses; the verb second (V2)-order of main clauses is derived by further movement of Infl0 (which includes the finite verb) to Comp0 (cf. den Besten 1983). Following Borer (1991), I take all X0-movement to be adjunction to the target head position. The SpecCP-position that precedes the verb in V2 can be filled by topicalized constituents. I adopt the semantic framework introduced in Jackendoff (1983, 1990). In Jackendoff’s framework, the meaning of a lexical entry is given in its Lexical Conceptual Structure (LCS). Jackendoff assumes a basic repertoire of major conceptual categories, such as Event, Thing, Place, Path etc. These categories can be elaborated as function-argument structures that contain primitive semantic predicates such as CAUSE, GO etc. The arguments of these predicates are represented as variables in the LCS of a lexical entry. For example, (15) is the LCS of drink: (15) [EventCAUSE ([Thing ]αA, [EventGO ([ThingLIQUID], [PathTO ([PlaceIN ([ThingMOUTH OF([α])])])])])] (cf. Jackendoff 1990: 53)

Introduction

(15) says that drink expresses a conceptual entity of the category “Event”, which is specified as the “movement” of something into the mouth of someone (caused by the Agent). Jackendoff represents the unity of two conceptual arguments through conceptual binding indices. The Greek letter α in (15) therefore specifies that in a drinking-event, the liquid ends up in the mouth of the Agent (and not in somebody else’s mouth, cf. the difference between eat and feed). Furthermore, (15) represents the fact that the object of drink must be a liquid as a selectional restriction on the first argument of GO. In Jackendoff’s theory, theta-roles are defined as structural positions in a conceptual structure. The Agent is the first argument of CAUSE; the Theme is the first argument of GO or BE; the Source is the first argument of FROM etc. The linking of these thematic roles to syntactic positions is stipulated as being part of the lexical entry of the verb. Jackendoff (1990) uses the subscript A to indicate argumenthood of a conceptual constituent; brackets around this subscript indicate optionality. The LCS of drink in (15) therefore specifies that this verb is transitive. It links the Agent to the subject position and the Theme (optionally) to the direct object position. (For details of the linking mechanism, see Jackendoff 1990, Chapter 11.) Finally, with respect to morphology, I presuppose basic notions like allomorphy and suppletion, derivation and inflection, compounding and affixation etc. The relation between morphology and the lexicon on the one hand, and between morphology and syntax on the other is discussed in great detail in Chapter 1. Here I introduce Borer’s (1988, 1991, 1993) theory of Parallel Morphology and the Distributed Morphology-approach of Halle & Marantz (1993). Both frameworks are highly relevant for my analysis of particle verbs.

Outline of the book In Chapter 1, I formulate my core assumptions about the way the syntactic, the morphological, and the lexical component of grammar interact. In particular, I argue against so-called “lexicalist” theories, according to which morphological objects are elements that are “taken from” the lexicon and “inserted” into syntax. I think that the apparent conflict between the morphological and the syntactic properties of particle verbs that I illustrated in (1)–(6) above is an unfortunate consequence of this lexicalist view. The alternative view of grammar that I defend in the first chapter of this book combines various proposals made in work by Borer (1988, 1991, 1993), Halle & Marantz (1993), Brody

9

10

Introduction

(1995), Jackendoff (1997), and Marantz (1997, in prep.). I argue that the lexicon does not precede syntax, but has access to syntactic representations formed and constrained by syntactic rules. It is part of the interface between syntax, semantics, and phonology; lexical items link representations of these three components to each other. The way this link is established may depend on the structural relations determined by syntactic structures. Importantly, the structures that can be associated with special meanings are not restricted to morphological domains (i.e. complex X0-elements), but may also refer to domains larger than the word. My analysis of particle verbs is based on this view of grammar. In Chapter 2, I discuss the syntactic representation of particle verbs, and I show that particles project phrases in syntax. I hence take issue with so-called morphological approaches that assume that particle verbs are morphologically complex verbal heads, but I also argue against those syntactic theories that assume particle incorporation. Instead, I show that particles are heads of phrasal complements of the verb and do not leave this position to combine with the verb. At the syntactic level that is accessed by the rules of the lexical interface component, the particle is still the head of a phrase. The syntactic and morphological separability of particle verbs, data from particle topicalization, modification, gapping, as well as typological observations provide evidence for this view. In Chapter 3, Grimshaw’s (1991) theory of “extended projections” is introduced, according to which the phrasal projection of the major lexical categories N, V, A, and P is always dominated by a layer of functional structure (the extended projection of the respective lexical element). In elaborating proposals made by Koopman (1993) and Haiden (1997), I provide evidence that particle phrases are exceptional in that they lack functional structure. Particle phrases are bare NP-, AP-, or PP-complements of V. This claim leads to the following definition: (16) Particles are heads of non-functional phrasal complements of the verb and do not leave their base position in overt syntax

According to (16), the particle and the verb are two lexical elements that are structurally adjacent; a lexical category becomes a particle by virtue of its local verbal environment. In Chapter 4, I argue that the specific local relation between the verb and the particle that is expressed in (16) is listed as a structural requirement in the lexical entry of the particle. In discussing the semantics of prepositional particles in German, I argue that particle meanings can only be associated with terminal nodes in syntactic structure if these nodes are structurally adjacent to

Introduction

the verb. Furthermore, I show that a particular particle element (= a syntactic P0-node with a particular phonological form which is structurally adjacent to V0) can be assigned different meanings, depending on the class of the verb with which the particle is combined. Chapter 5 addresses the morphological properties of particle verbs by comparing the two locality domains represented in (7a) and (7b) above, i.e. the domain defined by structural adjacency and the domain defined by complex words. I argue that the former domain, which is characteristic of particle verbs, has much in common with the local domain that is constituted by morphological objects. For example, I argue that particles are “semantic affixes”, because the licensing of their lexical semantics depends on the presence of a verbal host in a particular locality domain in the same way as the licensing of the phonological form of certain affixes depends on the presence of an appropriate host (in the locality domain defined by the word). The discussion of other phenomena, such as the case-assignment properties of particle verbs, further supports the view that the specific relation between a head and the head of its complement is similar to the morphological relation between terminal nodes that form a complex word. In Chapter 6, I address the observation that particle verbs may provide the input to further operations of derivational morphology. I provide evidence that particle verbs that appear inside words (i.e. as parts of complex A0s or N0s) must be words (= V0s) themselves. This fact apparently contradicts the results of the preceding chapters. However, I suggest that the “phrasal” representation of particle verbs in a syntactic context is compatible with a word-like representation of particle verbs in a morphological context. I argue that morphological operations of word formation can only apply to genuine morphological objects. This requires a reanalysis of the particle verb as a V0 whenever it is the input to derivational morphology. Importantly, I assume that reanalysis is only possible because of the similarities between the local domain defined by morphological objects and the specific local relation between particles and verbs. Finally, Chapter 7 provides a discussion of the consequences of my proposal for the analysis of particle verbs in other Germanic languages. In particular, I look at the behavior of particle verbs in Dutch Verb Raising constructions, and I address the verb-particle alternation in English and in the Mainland Scandinavian languages. This comparative discussion will show that my analysis offers a new perspective on some of the problems that are raised by a cross-linguistic investigation of particle verbs.

11

Chapter 1

Syntax, morphology and lexical licensing

In trying to answer the question “What are particle verbs?”, linguists have developed many (sometimes fundamentally different) analyses which are often strongly influenced by the theoretical perspective from which this phenomenon has been looked at. In particular, researchers have articulated diverging opinions about the rules that are operative in syntax and morphology and about the principles that govern their interaction. Despite these differences, however, what most (if not all) accounts have in common is that they are based on the linear view of so-called lexicalist models of grammar. According to these lexicalist models, the “output” of morphology and/or the lexicon is the “input” to syntax. Lexical items are drawn from the lexical/morphological component and inserted into syntactic structures. Regardless of whether morphology is seen as a theory of or about the lexicon or whether it is assumed that the rules of a separate morphological component operate on the basic entries taken from a “core” lexicon, the notions “morphology” and “lexicon” are always used to refer to the module of grammar that provides the things that the syntax operates on. The syntax uses words, complex or simple, as its basic entities, and since the morphological component is considered to be part of the lexicon, word formation always precedes syntactic derivations. (1) gives a rough illustration of this conception (cf. le Roux 1988; Borer 1991): (1) Lexical Insertion in lexicalist theories:

Lexicon (Lexical elements + word formationcomponent)

Syntax Lexical Insertion

On the basis of models like (1), the question “What are particle verbs?” has often been rephrased as “Where are particle verbs?” — on which side of the arrow in (1) are they formed? On the one hand, syntactic separability suggests

14


that particle verbs are formed in syntax, because the particle and the verb behave like independent syntactic elements that can be split by syntactic movement rules: (2) Peter hobi das Buch auf ti Peter picked the book up ‘Peter picked the book up’

On the other hand, particle verbs have properties of morphological objects, since they provide the input to further operations of derivational morphology, as shown in (3). Furthermore, (4) illustrates that the semantics of many particle verbs is non-transparent, which shows that at least some particle verbs must be stored as complex elements in the lexicon: (3) a. anhebenV (‘to lift’) b. annehmenV (‘to accept’)

Æ Æ

Anheb-ungN (‘raising’) annehm-barA (‘acceptable’)

(4) a. anfangen, lit. on-catch, ‘to begin’ b. aufhören, lit. up-hear, ‘to stop’

For linguists that adopt the linear perspective of lexicalist theories, the properties of particle verbs illustrated in (2)–(4) pose a problem. Syntactic separability suggests that they are syntactic constructions formed on the right side of the arrow in (1), whereas their word-like properties suggest that they are formed on the left side of the arrow. Consequently, the main issue concerning particle verbs has always been whether they should be analyzed as morphological objects (verbal compounds) or syntactic constructions (verb plus phrasal complement of which the particle is the head). In this study, I adopt a different perspective. I propose that the properties of the verb-particle construction must be taken as strong evidence against a lexicalist model of grammar. I argue that it is only the “linear” view of lexical insertion that makes the situation with particle verbs look paradoxical. In the first chapter of this book, I therefore introduce an alternative to the lexicalist model in (1) which will be based on work by Jackendoff (1997) and Marantz (1997) (see also Zeller 1997b). The conception of grammar that I defend draws a distinction between the notions “lexicon” and “morphology”; it dispenses with the view that syntax comes “after” the lexicon and morphology, and it provides the basis for the analysis of particle verbs that I offer in this study. In Section 1.1 I focus on the place of the lexicon in grammar. Following Jackendoff (1997), I assume that the lexicon is located at the interfaces between syntactic structure, phonological structure, and conceptual structure. A lexical


item is a correspondence rule that establishes a link between the structural representations generated in these three different components. It provides the link between idiosyncratic phonological and semantic information and pieces of (morpho-)syntactic structure. Jackendoff’s model of lexical licensing is a formal implementation of the way this link is established. In Section 1.2 I look at similar conceptions of grammar as they have been proposed in the Principles-and-Parameters framework. Most importantly, I discuss the idea of “late” lexical insertion and the framework of Distributed Morphology (Halle & Marantz 1993; Marantz 1997, in prep.) I show that “late” lexical insertion in the Principles-and-Parameters framework can be reinterpreted as lexical licensing. Section 1.3 focusses on morphology and its place in the model proposed in Section 1.1 and 1.2. I regard (the syntactic side of) morphology as the module of grammar that is responsible for relations between terminal nodes. Since usually, the local domain in which two terminal nodes occur is the syntactic word, i.e. the X0-element, the rules of morphology govern operations that take place below the level of X0. In addition, I argue in Section 1.3 that the combinatorial rules that form words and those that derive phrases are the same, and I briefly discuss Borer’s (1988, 1991, 1993) framework of Parallel Morphology, which is based on the idea that words can be derived by syntactic movement.

1.1 The lexicon and the architecture of grammar A lexical item combines syntactic, phonological, and semantic information. For example, by knowing a lexical item like cat, one knows its particular phonological representation, its meaning, and that it corresponds to a syntactic node of category N. According to lexicalist theories, lexical items are the basic entities of syntax; they are taken from the lexicon and inserted into X0-slots in the syntactic tree. The lexicon is seen as “feeding” the syntax, and as a result, the phonological and semantic features of lexical items are present when syntax operates on them. They are only interpreted “later”, at the interfaces to phonological and conceptual structure. The alternative proposal that I defend in this chapter is based on a conception of the architecture of the mind that is illustrated in (5):

15

16


(5) Overall organization of the mental information structure (Jackendoff 1990: 16) phonological formation rules

syntactic formation rules

conceptual formation rules

auditory input

vision phonological structures

syntactic structures

conceptual structures

action etc.

motor input

rules of inference

(5) is based on the view that is proposed in various work by Jackendoff (1983, 1987, 1990, 1993, 1997). The main hypothesis behind Jackendoff’s analysis in (5) is called Representational Modularity. According to this hypothesis, the mind/brain consists of different modules, each a formal system that encodes information according to its own specific primitives and principles. These “languages of the mind” are individuated by the representations they generate. The language faculty, which is embedded in the information structure of the mind, is also organized in different modules. As shown in (5), grammar consists of three components, phonological structure, syntactic structure, and conceptual structure, all three of them being generative systems with different atoms and different rules and operations. The nature of the representations generated by the rules of one component is entirely different from the nature of representations created in another module. The three modules in (5) constitute what Jackendoff (1997) calls the tripartite parallel architecture of the language faculty. As Jackendoff (1997) emphasizes, the tripartite model implies that grammar excludes “mixed” representations. There cannot be phonological information in the syntax, or conceptual information at phonological structure, simply because this information can only be processed in the respective domains. The model in (5) is hence incompatible with theories that assume lexical insertion of elements with phonological and semantic features present in the syntax. Lexical insertion, as advocated in lexicalist theories, would automatically create mixed representations, which are not tolerated by the grammar. If syntax, semantics, and phonology are separate modules that generate representations of an essentially different nature, the question arises how these


representations can interact. In Jackendoff’s (1997) theory, the link between the different modules of the language faculty is established by a separate interface module. Jackendoff gives the following definition of the general structure of an interface between two systems A and B (1997: 24): (6) Components of an interface between system A and system B (an A-B interface) a. A set of representations in system A to which the interface has access (the system B interface level(s) of system A or BILA) b. A set of representations in system B to which the interface has access (the system A interface level(s) of B or AILB) c. A set of A-to-B correspondence rules, of the general form illustrated in [(7)], which create the correlation between BILA and AILB. (7) General form of correspondence rules Configuration X in BILA {must/may/preferably does} correspond to configuration Y in AILB.

Each interface between two modules involves (at least) two interface levels, one for each representation. For example, some level of syntactic structure is linked to conceptual structure; it is the conceptual interface level of syntactic structure or CILSS. Furthermore, some level of conceptual structure is linked to syntax, the syntactic interface level of conceptual structure SILCS. Both interface levels are part of the interface between syntax and conceptual structure. According to Jackendoff (1997), the level of syntax that is linked to conceptual structure is simultaneously part of the syntax-phonology interface; he argues that CILSS and PILSS are the same syntactic level. This is a departure from the standard assumptions of the Principles-and-Parameters framework (Chomsky 1981, 1986a, b, 1995), where it is assumed that the level of syntax which is interpreted by phonology is different from the level that provides the interface to conceptual structure. However, I show in Section 1.2 that both views are not as incompatible as it may seem at first glance. Let me postpone the discussion of this point to the next section, and proceed. (6c) states that an interface that relates the interface levels of two modules to each other is furthermore characterized by a set of correspondence rules (indicated by the double arrows in (5)) that mediate between these two distinct forms of representations. For example, correspondence rules are necessary to relate the representation at the phonological interface level of syntax in (8a) to the intonational phrases in (8b) (cf. Jackendoff 1997: 26):

17

18


(8) a. SS: [this [is [the cat [that [ate [the rat [that [ate [the cheese]]]]]]]]] b. PS: [this is the cat] [that ate the rat] [that ate the cheese]

The interface between syntax and phonology must provide readjustment rules that guarantee that the syntactic structure in (8a) corresponds to the prosodic structure in (8b). In the same way, correspondence rules between syntactic structure and conceptual structure establish the relation between these two components. For example, there may be general correspondence rules like (9) that relate particular syntactic categories to particular conceptual categories: (9) The conceptual category Event preferably corresponds to the syntactic category V

Importantly, (9) includes the expression “preferably corresponds”. This is so because correspondence rules like (9) are subject to exceptions. The concept of an Event is not always expressed by a verb; for example, the nouns concert or earthquake express Events as well. As Jackendoff (1997) points out, many aspects of the lexical meaning of a word are invisible to syntax, and some distinctions reflected in syntax (such as grammatical gender) correspond only sporadically to respective distinctions in conceptual structure. The interface module of grammar must provide the rules that establish the appropriate correspondence.1 So-called bracketing paradoxes provide another set of examples that illustrate the general necessity of correspondence rules. Consider the word unhappier. Phonological conditions that constrain the affixation of the comparative suffix -er in English suggest that its structural representation looks like (10); its meaning, however, corresponds to a structure like (11) (cf. Pesetsky 1985; Marantz 1988): (10) [Aun [A happy er] ] (11) [A [A un happy] er ]

The suffix -er may only attach to monosyllabic adjectives and a few disyllabic adjectives with a very light second syllable, but never to trisyllabic adjectives

1.According to Jackendoff (1990, Section 1.4), grammar also includes correspondence rules that directly mediate between phonological structure and conceptual structure. In Jackendoff (1997), however, this assumption is abandoned; the correspondence between phonological structure and conceptual structure is directly mediated through syntactic structure.


(*redundanter). This suggests that the structural representation of unhappier is as in (10). However, its meaning is not what a structure like (10) would predict; unhappier does not mean “not more happy”, but “more not happy”. This is represented through the structure in (11). It is clear that correspondence rules are needed in order to link the meaning of unhappier to a phonological structure that meets the condition on -er-suffixation. The model of grammar illustrated in (5) opens two possibilities of dealing with bracketing paradoxes like the one illustrated by (10) and (11). The first possibility is to assume that the morphosyntactic bracketing mirrors the phonology. Then the structural representation of unhappier is as in (10). A correspondence rule is required that operates at the interface between syntactic structure and conceptual structure and links the representation in (11) to its correct interpretation. An analysis that exploits this strategy is presented by Pesetsky (1985). Pesetsky suggests that the morphosyntactic representation of the word unhappier looks like (12a). At LF, however, the comparative morpheme raises and adjoins to the highest A0-node, (12b): A0

(12) a.

A0

b.

A0

A0 A0 happy

A0

er

A0

A0 0

A happy

A0 eri ti

Whereas the structure in (12a) is linked to phonology and obeys the phonological constraints on -er-affixation, the LF-structure in (12b) yields the correct interpretation, since the comparative morpheme has scope over the complex adjective unhappy. The alternative route towards an analysis of unhappier is taken by Marantz (1988). Marantz suggests that the syntactic representation of unhappier parallels the semantics and therefore looks like (11). A correspondence rule between syntactic structure and phonological structure (Marantz’s (1988: 258) Mapping Principle) determines that the structure in (11) is mapped onto a phonological structure where the phonological requirement of -er can be met. Regardless of the (eventually empirical) question of which alternative should be chosen to analyze the bracketing paradox illustrated in (10) and (11), it is clear that correspondence rules are a necessary part of grammar. They are not so

19

20


much a specific characteristic of Jackendoff’s model but an instrument that is required by any linguistic theory that seeks to account for mismatches between syntactic, phonological, and conceptual structures. In acknowledging their necessity, one is led directly to a new and more accurate view of the lexicon. Jackendoff (1997) points out that lexical items are mixed representations by their very nature, since they combine syntactic, semantic, and phonological information. Insertion of these elements with all this information present at once would automatically violate Representational Modularity. This means that there cannot be anything like the traditional (lexicalist) concept of lexical insertion. Therefore, Jackendoff (1997) substitutes his conception of lexical licensing for lexical insertion. According to this model, lexical items are not “taken” from the lexicon and inserted; instead, a lexical entry is nothing other than a (small) correspondence rule, and the lexicon as a whole is part of the interface module. A lexical item includes information from all three generative systems, and it shows at the same time how this information is combined. It provides a small chunk of phonology (lexical phonological structure; LPS), a small chunk of syntax (lexical syntactic structure; LSS), and a small chunk of semantics (lexical conceptual structure; LCS), and it licenses these parts as the results of three independent derivations performed in phonological structure, syntactic structure, and conceptual structure. The link between the three structures is established by a lexical index that states which conceptual constituent corresponds to which syntactic category etc. Consider a concrete example. The lexical entry for the word cat is represented by three parts of information that are linked through a subscripted index (cf. Jackendoff 1997: 89): (13) a.

LPS Wordb

σ

b.

LSS Nb

c.

LCS [Thing TYPE:CAT]b

count sing

k æ t

The specific information contributed by a lexical item are the lexical linking indices. In (13), the lexical index b relates the phonological information in (13a) to its syntactic representation N and to its conceptual interpretation as “cat”. Notice that the lexical item does not produce these structures; they are


derived by the formation rules of the three generative modules (but see note 2 below). The actual content of a lexical entry is the information about where to put the lexical indices, i.e. which part of syntactic structure is linked to what constituent in conceptual structure and phonological structure. Jackendoff (1997) borrows the terminology of Levelt (1989) to refer to the connections between syntactic, conceptual, and phonological structure expressed by the linking indices. The linkup between syntactic structure and a conceptual constituent is a lemma; the linkup between the LSS and the LPS of a lexical element is a lexical form. As Jackendoff (1997: 96) points out, lexical licensing can be regarded as an operation of unification. A lexical item like (13) is unified simultaneously with structures independently generated in syntax, phonology, and conceptual structure. For example, the three structural descriptions in (13) may be part of larger structures created by the formation rules of the three components. Lexical licensing means that the lexical entry cat in (13) adds its lexical indices to these structures, thereby providing the information that a particular part of phonological structure corresponds to a particular part of syntactic structure and to a particular part of conceptual structure. The knowledge of the word cat enables the language user to relate the syntactic representation in (13b) to a special sound and a special meaning.2 The LSS in (13b) is a single terminal element N0. The lexical representation in (13) is therefore in accordance with the traditional understanding of lexicalist theories, which assume that special meanings are exclusively associated with X0elements. However, as is well-known, syntactic structures that correspond to special meanings can also be larger than X0. The examples in (14)–(17) are taken from Beth Jackendoff’s Wheel of Fortune Corpus (cf. Jackendoff 1997): (14) a. kick the bucket b. hit the road c. give it your best shot (15) a. son of a gun b. bridge over troubled water c. point of view

2.A potential disadvantage of the idea that lexical licensing is an operation of unification may be that unification requires lexical entries to reproduce the structures independently generated by grammar. Otherwise, a lexical entry would include no information with which these structures could be unified (cf. Marantz in prep. for a similar criticism).

21

22


(16) a. in touch with one’s feelings b. down in the dumps c. off the top of my head (17) a. hold everything b. that’s a laugh c. that takes the cake

The phrasal idioms and fixed expressions in (14)–(17) associate idiosyncratic meaning with a non-minimal syntactic structure (VPs in (14), NPs in (15), PPs in (16), and full sentences in (17)). As pointed out and emphasized by Di Sciullo & Williams (1987), Jackendoff (1997), Marantz (1997, in prep.), and many others, the special meaning of idioms must be learned. The information associated with the examples in (14)–(17) is lexically listed; idioms are lexical entries like the words cat and walk. It is clear that the examples in (14)–(17) provide a very strong argument against lexicalist theories of grammar which assume that the operation of lexical insertion substitutes lexical items for X0-categories in the syntax. There is no direct correspondence between lexical items and syntactic atoms; idiomaticity is not a sufficient argument for wordhood. In contrast, phrasal idioms do not raise problems for the theory of lexical licensing. An expression like kick the bucket is listed in the lexicon as a phrasal lexical item. It is a correspondence rule like every other lexical entry, with the only difference that it associates three elements of phonological structure with the three elements in syntax that form a VP, but with only one element in conceptual structure (cf. Jackendoff 1997: 169):3

3.I have slightly modified Jackendoff’s entry, because Jackendoff associates the lexical index x that links syntactic structure to conceptual structure with V0 rather than with VP, for reasons that are not quite clear to me.


(18) a.

LPS aWd

k i k

b. bCl

ð 6

LSS

VPx

cWd

b % k 6t

aV

0

NP bDet

0

NP 0 cN

c.

[Event DIE([ ])A]x

The lexical entry in (18) licenses the correspondence between a special meaning and a complex syntactic phrase. A VP like (18b) is generated by the syntax, and the lexical entry provides the information that a syntactic VP, whose parts are associated with the phonological features of the verb kick, the determiner the, and the noun bucket, is associated with the concept expressed in (18c). Notice that V0, N0, and Det0 in the LSS of (18) do not correspond to any conceptual constituents in the LCS, whereas they are each linked to their regular phonological forms. This means that a phrasal idiom like kick the bucket involves a single lemma expressed by three lexical forms. One might object that Jackendoff’s analysis of idioms is too unconstrained, because in his system of lexical licensing, every syntactic structure could in principle mean anything. However, as has been observed by Marantz (1984), in contrast to the vast number of VP-idioms like the ones in (14), there are no idioms with a fixed external argument and varying internal arguments. Marantz (1997, in prep.) emphasizes that the assignment of idiosyncratic meanings to syntactic structures is constrained by syntax itself. I will come back to Marantz’s (1997, in prep.) analysis of idioms in Chapter 4.

1.2 Syntactic movement and “late” lexical insertion The view of grammar that I outlined in Section 1.1 may look strange to the eye of the linguist who is used to working in the Principles-and-Parameters framework (Chomsky 1981, 1986a, b, 1995). First, both Government Binding theory (Chomsky 1981, 1986a) and the Minimalist Program (Chomsky 1995, 1998) are

23

24


based on the linear view of lexical insertion, according to which the lexicon precedes the syntax. Second, Jackendoff’s view that the level of syntax which is linked to phonology is the same level as the one that is linked to conceptual structure goes against the view that syntax is connected with conceptual structure via the abstract level LF — a representation that is different from the syntactic level that is linked to PF. In this section, I try to relate Jackendoff’s proposal to some similar ideas that have emerged within the Principles-and-Parameters framework. In Section 1.2.1, I show that Jackendoff’s tripartite architecture model is well compatible with a syntactic approach like Brody’s (1995) where movement operations are represented through chains at one single syntactic interface level. The related question about the status of LF is discussed in Section 1.2.2. In Section 1.2.3 I discuss the framework of Distributed Morphology, introduced by Halle & Marantz (1993) and elaborated in Marantz (1997, in prep.). I will show that the core ideas of Distributed Morphology are well compatible with the tripartite parallel architecture of grammar. 1.2.1 Movement and interface level(s) According to the Principles-and-Parameters approach, the syntax performs derivations, with Move α changing the position of an element in the syntactic tree. The operation Spell-Out applies to a structure at some point and maps the structure to the phonology, while the computation continues “covertly” and derives LF. LF is taken to be the interface level of syntax that is related to conceptual structure.4 Consequently, this view seems contrary to the claim made by Jackendoff (1997), who argues that there is only one interface level of syntax that is linked to both phonology and conceptual structure. However, Jackendoff’s claim is not entirely new, not even from a Minimalist perspective. It is reminiscent of the representational theory proposed in

4.The inconsistent use of terminology in Chomsky (1995) renders his exact view about the status of LF unclear. Chomsky (1995: 2, 22) describes PF and LF as interface levels, i.e. as levels of linguistic representation that are located at the interfaces to the performance systems. Occasionally (1995: 130), however, Chomsky calls PF and LF “the interfaces”, and he raises the question whether the “LF interface” involves a conceptual system in the spirit of Jackendoff’s conceptual structure (1995: 21). The discussion in Section 1.1 should have made clear that the interface level of syntax, the interface between syntactic and conceptual structure, and conceptual structure itself are three different conceptions that play different roles in grammar. They must not be confused.


Brody’s (1995) “radical minimalist” approach. Brody argues against a theory that includes both the concept of movement and the concept of a chain. He provides evidence that the latter concept is needed on general grounds, and claims that assuming movement transformations in addition would be redundant. Brody therefore does away with movement altogether and represents the relations between syntactic positions that were traditionally expressed by movement exclusively through chains. His proposal entails that there are no syntactic derivations that transform one level of syntax into another. According to Brody (1995), the syntactic module does not perform derivations, but rather projects one single level of representation (Lexico-Logical Form LLF) that is assembled out of syntactically well-formed chains. LLF is the only syntactic interface level. It serves as the target for both phonological and semantic interpretation. Furthermore, Brody adopts the copy theory of chains that is proposed in Chomsky (1995). In the copy theory, a “moved” element is no longer represented through a trace, but by an identical copy in the base position of its chain. A “moved” element may thus occur in multiple positions in syntactic structure. For example, movement of a subject wh-phrase to Comp is represented through the LLF in (19): CP

(19)

Spec NPwh

C′

C0 links of wh-chain

IP Spec NPwh

I′

Brody’s (1995) proposal is well compatible with the model of grammar illustrated above. If there is only one single syntactic level, the representations of the conceptual system can be directly mapped onto their phonological forms via the syntax. At the same time, the fact that LLF is assembled out of chains allows for the recovery of grammatical and lexical relations that are established between the feet of these chains. Notice that chains may also receive special interpretations as such. For example, the NP-chain in (19) is

25

26


interpreted semantically as an operator-variable relation, with the operator meaning associated with the head of the chain and the variable with its foot position. Phonology associates the LPSs of the NP with either the head or the foot of the chain, depending on whether the language has “overt” or “covert” wh-movement (see the next section). There has been some debate about the question whether a representational system like Brody’s and a derivational approach are in fact intertranslateable. As a potential argument against a representational theory, Chomsky (1995: 223) points out that “in syntax, crucial relations are typically local, but a sequence of operations may yield a representation in which locality is obscured”. Chomsky illustrates this statement through the following head movement-example. If N0 incorporates into V0, it leaves a trace tN. If the complex V0 moves further to Infl0, it leaves a trace tV. Viewed derivationally, both steps are local, but this locality is obscured by the resulting output representation, where the N0-part of V0 and its trace are separated by the intervening verbal trace: (20) [Infl0 + [V N0 V0 ]] … tV … tN ….

However, Brody (1995) addresses Chomsky’s argument, pointing out that the copy theory of movement provides a straightforward means to capture the local relation between N and its base position representationally: (21) [Infl0 + [V N0 V0 ]] … [V N0 V0] … N0 ….

As Brody (1995: 40) points out, “[t]here is no a priori reason why the fundamental property of syntactic computation must be the existence of local relations destroyed before the output stage … or the presence of these local relations in the output representations”. Given the independent motivation of the copy theory that is provided in Brody (1995), I agree with his conclusion and assume that head movement does not provide a valid argument against a representational theory, since it is always possible to formulate syntactic constraints on movement in terms of outputs as well.5 In the light of Brody’s theory, then, the difference between Chomsky’s derivational model, which postulates more than one interface level of syntax, and Jackendoff’s view, which argues that there is only one interface representation, becomes less significant. In fact, I think that if one abandons the view that operations of grammar are linearly ordered with respect to each other, the

5.Chomsky (1995: 224) is aware that his argument is invalid under the copy theory; however, he rejects the copy theory in the case of head movement as a “coding trick”.


difference may even reduce to some minor technical issues. We tend to think of grammar as the successive application of rules, where one step follows another. This linear view may be helpful in describing and understanding the properties of the grammatical system; however, it is the wrong way to look at how grammar works. This point is also addressed by Chomsky (1995: 380; note 3) who points out that “the ordering of operations is abstract, expressing postulated properties of the language faculty of the brain, with no temporal interpretation implied”. Therefore, I think it is unproblematic to continue to talk about “derivations” and “movement” if one keeps in mind that these abstract notions do not imply that the output “follows” the input in any meaningful way. 1.2.2 The status of LF The notion of “covert” movement that was briefly mentioned above raises the question about the status of LF and the operations that are associated with this level. Traditionally, LF is assumed to be different from the syntactic representation that is interpreted by the phonological component. However, it is clear that a representational system like Brody’s allows us to incorporate LF into the theory. Like overt movement, LF movement can be expressed by chains. Overt and covert movement only differ with respect to the position of the chain that is associated with the phonological features of the moved element. For example, suppose that the syntactic representation of wh-movement is as in (19) in all languages. In languages with overt wh-movement, like German and English, the moved NP in (19) is pronounced in SpecCP. In contrast, languages like Japanese with wh-phrases in situ associate the phonological features of a wh-phrase with the foot of the chain. The interface between syntax and phonology has to incorporate rules that state which element of a chain is linked to the LPS of a lexical item. For example, one could postulate a general PS-SS correspondence rule that establishes that the phonological features of a “moved” element are preferably associated with the foot of its chain. Such a rule would correspond to Chomsky’s (1995) Procrastinate-principle that states that covert movement operations are preferred by grammar. Further correspondence rules then have to be motivated that specify under which conditions phonological features are associated with the head of a chain. In the Minimalist Program, overt movement is triggered by “strong” grammatical features of functional categories; it seems to be relatively unproblematic to formulate Chomsky’s feature checking mechanism as a correspondence rule that captures this idea. Alternatively, a correspondence rule could be formulated that forces LPS-

27

28


entries to be preferably associated with the head position of chains. Brody’s (1995, Chapter 4) principle of Transparency is a proposal in this spirit. There is a alternative possibility of incorporating LF into the theory. Instead of representing LF-movement as a chain at a single syntactic interface level, one may assume that LF-operations are essentially correspondence rules that establish a link between syntactic structure and conceptual interpretation. According to this possibility, LF and the operations that characterize the “covert” component of grammar are not really part of syntax, but of the interface module between syntactic structure and conceptual structure. One example was already given in Section 1.1. According to Pesetsky’s (1985) analysis of bracketing paradoxes, LF-movement relates the meaning of a word like unhappier to a different structural representation that is motivated on phonological grounds. This LF-movement rule essentially is a correspondence rule. The important insight about correspondence rules that are expressed through LF operations is that they are formulated on the basis of syntactic constraints. LF-“movement” is a syntactic way of representing the SS-CS correspondence rules. LF-correspondence rules directly refer to syntactic domains, since LF is a syntactic level, and hence LF-representations are constrained by the same locality conditions that govern “overt” syntactic operations. This could be taken to be a welcome result in the light of the aforementioned objection that the system of correspondence rules, although empirically motivated, is quite unconstrained. Correspondence rules between syntactic structure and conceptual structure that are expressed as LF (movement) operations allow us to reduce the set of possible correspondences that can be established between the three modules on the basis of syntactic principles. The question whether or not LF exists is not essential for the analysis that I motivate and propose in this and in the following chapters. The crucial movement operations that I discuss in this study are always “overt”; i.e. the respective chains are always spelled-out in their head positions. However, I do not strictly exclude the possibility that some syntactic relations are expressed by LF-movement, i.e. through chains whose foot positions are linked to phonology. 1.2.3 Distributed Morphology The tripartite architecture of grammar takes the lexicon to be a set of correspondence rules that access the interface levels of the three components of grammar. The interface level of syntax consists of chains that represent the results of syntactic movement. Therefore, the LSS of a lexical entry is always


unified with a syntactic structure which represents the “output” of a derivation through chains. Consequently, lexical licensing implies that the lexicon is not ordered “prior” to syntax, as lexicalist theories claim. Instead, if we continue to use the “movement”-metaphor and think of chains at the interface level of syntactic structure as being formed by syntactic transformations, the lexicon rather “follows” these syntactic movement rules. The lexical link between the three modules of grammar is established at some “later” point, i.e. “after” syntactic operations have been performed. The idea that lexical information has access to syntactic structures that have been formed by syntactic rules has emerged in the Principles-and-Parameters framework as well, where it goes under the label “late lexical insertion” (cf. e.g. Marantz 1993; Bobaljik 1995). The notion of late lexical insertion emphasizes the difference between the assumption that phonological and semantic features are associated with syntactic structures derived by syntactic rules on the one hand and the traditional conception of lexical insertion on the other, which claims that lexical entries are inserted into X0-positions with all their features present. “Late” insertion refers to the idea that the lexicon operates at the interface(s) between the syntactic, the phonological, and the semantic part of grammar; it is hence possible to reinterpret late insertion as lexical licensing. A model of grammar that incorporates late lexical insertion is the theory of Distributed Morphology, introduced by Halle & Marantz (1993) and elaborated in Marantz (1997, in prep.). Halle and Marantz (1993) argue that there are no phonological features in syntax. Instead, the phonological realization of syntactic nodes is executed through an operation called Vocabulary Insertion. In the (1993)-version of Distributed Morphology, Vocabulary Insertion takes place at a separate interface with the phonological component, a syntactic level that Halle and Marantz call Morphological Structure: (22)

DS (D-Structure) SS (S-Structure) Logical Form (LF)

MS (Morphological Structure) PF (Phonological Form)

29

30


As can be seen in (22), Halle & Marantz’s (1993) architecture of Distributed Morphology still includes levels like D-structure and S-structure that are abandoned in Marantz’s (1997) later version (see below). In Distributed Morphology, the phonological representation of a verb and the phonological representation of its inflectional affixes are associated with separate terminal nodes (X0s) that are combined by syntactic (and post-syntactic6) rules. For example, the root of the German verb lachen, ‘laugh’, is represented in syntactic structure as a terminal node of category V0, but the corresponding phonological information is only “late” inserted at Morphological Structure. Suppose the functional head Infl0 bears the grammatical feature [+past] in syntactic structure. In the syntax, the main verb moves to Infl0, and at Morphological Structure, Vocabulary Insertion connects the phonological feature bundles of the lexical entries (the “Vocabulary items”) of V and Infl with bundles of grammatical features that are associated with these nodes in the syntax. All that is required for insertion is that the features of the lexical item are non-distinct from the features of the syntactic node:7 (23) a.

D-Structure IP

VP

V0

b.

InX0 [+past]

Morphological Structure IP

InX0i/j VP

V0i

Infl0j [+past]

V0i Vocabulary Insertion of e.g. lach- and -t

6.In Halle & Marantz (1993), complex heads may also be created by post-syntactic operations that take place at Morphological Structure. These operations (merger, fusion, or fission) allow morphological relations to be established between terminal nodes that are not combined in overt syntax. I will briefly discuss the merger-operation in Chapter 6; the reader is refered to Halle & Marantz (1993) for a detailed discussion of post-syntactic operations in Distributed Morphology. 7.I have omitted the realization of agreement morphemes here, which are independently added at Morphological Structure in the Distributed Morphology-framework.


Since the verb lachen, whose Vocabulary item is inserted in (23b), is inflected regularly, the phonological features of the tense-morpheme are associated with the terminal node Infl0j at Morphological Structure. Typical morphological operations, like the choice of the right past tense-allomorph, also apply at Morphological Structure. In addition, “late” insertion in Distributed Morphology has the crucial advantage that it also allows us to incorporate cases of suppletion into a theory where inflection is done via head movement in the syntax. Instead of inserting the phonological matrix of individual lexical entries into the terminal nodes V0i and Infl0j, the complex head that dominates V0 and Infl0 at Morphological Structure may also be fully supplied with the phonological features of a completely idiosyncratic entry. For example, if the German copula verb sein, ‘be’, is chosen, the complex Infl0i/j in (23b) may be directly associated with the phonological features of the verb’s past tense stem war-.8 In Distributed Morphology, “word formation” comprises both late insertion of PF-features and the syntactic and post-syntactic operations that create complex heads. The notion “morpheme” is used to refer to a terminal element both before and after it is associated with its phonological features. Morphology is hence “distributed” among syntax and phonology. I will come back to the assumption that syntactic operations like head movement count as operations of word formation in Section 1.3.2. So far, Distributed Morphology is concerned only with the relation between syntax and phonology. Marantz’s (1997, in prep.) elaboration of the Distributed Morphology-framework, however, is also concerned with the semantic interpretation of syntactic structures. Marantz raises similar objections against the traditional concept of lexical insertion as Jackendoff (1997). He also reaches the conclusion that the different pieces of information provided by a lexical entry do not coincide in a single terminal node via an operation of lexical insertion. According to Marantz, what traditionally is called “the lexicon” is in fact three sorts of lists:

8.Halle & Marantz (1993) assume that Vocabulary Insertion can only apply to atomic nodes. Therefore, suppletion of the complex Infl0-head in (23) is preceded by a fusionoperation which joins the lower Infl-node and the verbal node into one. However, as is also pointed out by Bobaljik (1995: 34, note 10), one could assume alternatively that Vocabulary Insertion actually replaces the whole complex Infl0-node, such that suppletion can apply without V0 and Infl0 being fused. I have adopted Bobaljik’s proposal in the text, mainly for ease of exposition.

31

32


(24) The structure of grammar according to Distributed Morphology (Marantz 1997: 204) List 1 narrow lexicon

List 2 Vocabulary

Computational system (Syntax) = Merge and Move

Phonology

Phonetic interface

LF

Semantic interface

List 1 Encyclopedia

List 1 in (24) is what Marantz calls the “narrow” or “pure” lexicon; it provides the elements on which the syntax operates. Importantly, and in contrast to lexicalist views, these elements only consist of the atomic roots of the language and the atomic bundles of grammatical features that serve as terminal nodes in the syntax. They do not include any phonological or semantic features. In Jackendoff’s terms, the “narrow” lexicon is the list of all LSS representations. Notice that the syntactic module in (24) no longer connects levels of representations like D-structure and S-structure, but is considered a bare computational system with the two major operations Merge and Move (see Section 1.3.2 below). Consequently, the level of Morphological Structure has disappeared. Instead, the operations that were assumed to take place at this level are now taken to be part of Phonology in general. Phonology is the component that relates the output of the computational system to what Marantz calls the “phonetic interface”. In present terms, phonology is the interface module between syntactic structure and phonological structure, and “phonetic interface” in (24) refers to what Jackendoff calls SILPS, i.e. the syntactic interface level of phonological structure. According to Marantz (1997, in prep.), elements from List 2, the Vocabulary items, are added in phonology. In terms of the theory presented in Section 1.1, List 2 comprises all LPS-entries. Vocabulary Insertion is defined as the association of phonological feature bundles with terminal nodes in the syntax. This operation licenses the correspondence between a piece of syntactic structure and a piece of phonological structure through lexical information. The “insertion” of a Vocabulary item in Phonology hence corresponds to the idea that lexical forms (the linkups between LSS and LPS) are rules that operate at the interface between syntactic structure and phonological structure.


Finally, List 3 is the “Encyclopedia”, the list of idiosyncratic, non-compositional meanings — LCS-entries in Jackendoff’s theory. According to Marantz (1997, in prep.), special meanings from the Encyclopedia are associated with syntactic structures at the interface level. Marantz (1997, in prep.), like Jackendoff (1997), emphasizes that the syntactic structures with which special meanings are associated may be larger than X0. He even extends Jackendoff’s proposal in one crucial respect. Marantz points out that special meanings are not arbitrarily associated with arbitrary syntactic structures. Rather, he assumes that special meanings are always associated with terminal nodes, but on the basis of the syntactic environment of the respective head. In this way, the syntax-semantics interface is similar to the interface between syntax and phonological structure, where the special phonological realization of a particular terminal node may also depend on elements in its local domain (allomorphy). Marantz’s claim will be discussed in more detail in Chapters 4 and 5; it is crucial for the lexical representation of particles that I propose. The theory of late lexical insertion and the theory of lexical licensing refer to the same idea, looked at from slightly different perspectives. The fact that this idea has emerged within different theoretical frameworks attests to the need for an alternative model of the lexicon that dispenses with the traditional form of lexical insertion.

1.3 The domain of morphology In this section I want to discuss the status of morphology in the light of the theory that I introduced in Sections 1.1 and 1.2. The notion “word” is discussed in Section 1.3.1. In Section 1.3.2 I argue that complex X0-elements are derived by the same combinatorial rules that also form phrases. Finally, I discuss Borer’s (1988, 1991, 1993) theory of Parallel Morphology in Section 1.3.3, which is based on the idea that words can be derived by head movement. 1.3.1

On the notion “word”

Di Sciullo & Williams (1987) distinguish three different ideas of what constitutes a word. The first idea assigns word status to morphological objects, i.e. the set of elements that are built out of morphological atoms (morphemes) by morphological rules. According to the second view, words are syntactic atoms, i.e. a word is what is associated with an X0-slot in a syntactic structure. Finally,

33

34


there is the third notion of words as the listed units (or listemes in Di Sciullo & Williams’ terminology) of a language. However, we have already seen in Section 1.1 that the latter view is certainly incorrect; phrasal idioms are listemes that are neither morphological objects nor X0s. Therefore, I will not use the word “word” in this way in this study. Another use of “word” that must be mentioned is the notion of “phonological word” which is determined by phonological conditions, such as the presence of at least one foot (cf. McCarthy & Prince 1986). In many languages, the status of a syntactic atom is irrelevant for the definition of the domain of the phonological word (cf. Nespor & Vogel 1986); there are X0-elements that are both smaller and larger than phonological words (cf. Jackendoff 1997; Marantz 1997). Since the distinction between X0s and phonological words is irrelevant for the concerns of this study, I will ignore it here as well. This leaves us with the distinction between “morphological objects” and “syntactic atoms”. Di Sciullo & Williams (1987) draw this distinction primarily on the basis of French compounds like the ones in (25) that behave like syntactic atoms (i.e. they are X0s), but whose internal structure seems to be non-morphological, (26):9 (25) a.

arc-en-ciel (lit. arc-in-sky = rainbow) b. hors-la-loi (lit. outside-the-law = outlaw)

N0

(26)

PP

P0 hors

NP la loi

(27) N Æ XP

Di Sciullo & Williams assume that elements like the ones in (25) are formed by the rule in (27), which they regard as a “non-morphological word-creating rule of the periphery of the grammar” (1987: 82). Although these compounds are

9.Keep in mind that syntactic heads do not dominate phonological and semantic features. However, throughout this study, I will occasionally use the traditional representation (with “words” dominated by syntactic categories) for purposes of exposition.


syntactic atoms, they do not qualify as morphological objects.10 Di Sciullo & Williams’ (1987) distinction between syntactic atoms and morphological objects is not crucial for this study. But it emphasizes an important point about (the syntactic side of) morphology. A morphological object is not only defined by the structural domain for which morphology is responsible (= X0). It must also consist of elements that morphology operates with (i.e. terminal nodes, morphemes). In most cases, if one of the two conditions is fulfilled, the second condition is automatically fulfilled as well. The combination of two terminal nodes creates a complex X0-element; this element may then project a phrase and function as a syntactic atom (it may of course also provide the input for further word formation): XP

(28)

YP

X Z

0

0

Domain of morphology

X

0

However, what the examples in (25) show is that both conditions are not always met simultaneously. Compounds which include full phrases are X0s, but not proper morphological objects, because they do not exclusively consist of morphological atoms. This is because morphology is defined in two ways: by considering the set of elements with which it operates, and by looking at the structural domain where its rules apply. Now suppose that there are constructions that constitute something like the mirror image of the compounds in (25). Imagine an element that consists of two terminal nodes, but these two nodes are not located within the domain of morphology (they do not combine to form a complex X0). In this case, we expect the element to have certain word-like properties that follow from the fact that it consists of two terminal elements. Nevertheless, this construction does not qualify as a genuine morphological object, because it is not a complex X0. This is precisely what I will argue is the case with particle verbs. The central claim of my study is that particle verbs consist of the two terminal nodes that

10.Di Sciullo & Williams’ (1987) analysis is not uncontroversial. See Anderson (1992) for counterarguments.

35

36


correspond to the particle and the verb, and that the properties of the verb-particle construction are essentially determined by this situation. However, the particle and the verbal node are not part of the same X0, but realized within a different, non-morphological local domain that I define as “structural adjacency”: (29) Structural adjacency A head X and the head Y of its complement YP are structurally adjacent.

VP

(30)

PrtP

V0

Prt0

Structural adjacency

In the Principles-and-Parameters framework, the relation between V0 and Prt0 in (30) is traditionally characterized through the notion of head government (cf. Chomsky 1981; Rizzi 1990): an X0 properly head-governs the head of its complement; consequently, a verb properly head-governs a particle.11 The reason why I decided to describe this relation through the notion of structural adjacency rather than through proper head-government is that structural adjacency defines the relation between a verb and a particle in a symmetric way: V0 is structurally adjacent to Prt0, and Prt0 is structurally adjacent to V0. In contrast, proper head-government is asymmetric: V0 head-governs Prt0, but Prt0 does not head-govern V0. In Chapters 4 and 5, I discuss a number of lexical dependencies between verbs and particles that are described and explained more elegantly if it is assumed that the relation between both elements defines a locality domain that determines mutual dependencies. Expressing this relation through head-government might create the (false) impression that the particle (the governee) depends in a stronger way on the verb (the governor) than the verb depends on the particle. Therefore, I chose a term that is more neutral in this respect. I assume that structural adjacency defines a local domain established by two elements; the fact that the syntax represents this domain through a structure in which one element head-governs the other is a consequence of the particular way in which this domain is defined. On the one hand, the local domain expressed in (30) is not the domain of

11.The term “structural adjacency” is used in Halle & Marantz (1993); cf. also the notion of “structural government” in Marantz (1984).


morphology. Therefore, the two framed nodes Prt0 and V0 do not form a proper morphological object. Instead, particle verbs behave in many respects like independent syntactic constructions. On the other hand, particle verbs still share with genuine morphological objects the property that they consist of two terminal nodes in a particular locality domain. As I will show in Chapters 5 and 6, this explains why particle verbs exhibit many properties that are otherwise only attested in the morphological domain. Since X0-elements that do not qualify as morphological objects play virtually no role in this study, I will ignore the distinction between syntactic atoms and morphological objects. I will henceforth use the notions “word”, “syntactic atom”, or “morphological object” to refer to both simple and complex elements that are syntactically represented as X0s. However, it must be kept in mind that word-like properties of complex elements do not always and necessarily imply that these elements are genuine morphological objects. 1.3.2 The rules of word formation According to Chomsky (1995), syntactic derivations are based on two operations; viz. Merge and Move. Merge is defined as an operation that takes a pair of syntactic objects (α, β) and replaces it by a new syntactic object K. Merge is asymmetric; K may be a projection of either α or β. For example, the combination of V0 and NP creates the syntactic object V¢, a projection of V0:12 (31) syntactic object V0

V′ Merge

syntactic object NP

syntactic object: V0

NP

The operation Move is a combination of Copy and Merge. An element already existing in the structure is copied and merged with a targeted category K. This operation creates a chain that consists of the moved element α merged with K and its copy (or trace) in the base position of α. For example, Move may copy the NP in (32), target Infl¢, and merge the NP with Infl¢ into the specifier of IP:

12.Chomsky’s (1994, 1995) bare phrase structure theory dispenses with X-bar theory and projection levels like X0, X¢, and X-max as primitives. I address this point shortly below.

37

38


(32) target

Infl′ Infl

0

IP VP Move

V0

NP

Infl′

Spec NPi Infl

0

VP V0

NPi

Recall that Move is a metaphor. Syntactic structure is characterized through chains that link two or more different positions, and the interface component has access to all positions in these chains, depending on the correspondence rules. Merge and Move are operations that form structural objects. Morphologically complex X0s are structural objects as well. Therefore, I do not see any reasons why these operations should not also be responsible for the creation of complex X0s. For example, a word like the noun bushpig has been formed by applying the operation Merge to two N0-elements: N0

(33) syntactic object N0 Merge syntactic object N

0

syntactic object:

N0 bush

N0 pig

The complex N0 has been formed by the same operation that forms phrases. What makes the N0 bushpig a morphological object is not the nature of the operation that combines the two terminal nodes, but only the X0-status of these two elements. Although the notions “Merge” and “Move” are adopted from Chomsky (1995), I have represented all syntactic structures in an X-bar theoretical format. This goes against the spirit of Chomsky (1994, 1995), where standard X-bar theory, and elements like XP or X¢, are eliminated in favor of a “bare phrase structure theory”. However, I will continue to use the X-bar theoretical notions X0, XP etc. in this study. One may simply look at them as informal notations, as convenient metaphors for expressing the difference between “words” and “phrases”. As illustrated in (33), I will also represent terminal nodes as X0s in order to distinguish between morphological structures like (33) and phrasal structures like (31); this distinction cannot easily be drawn in bare phrase structure terms. Chomsky (1995) argues that the notions “minimal” and “maximal projection” are relational properties of categories, not properties inherent to them. He assumes that, “[a] category that does not project any


further is a maximal projection XP, and one that is not a projection at all is a minimal projection Xmin” (1995: 242). However, notice that this view implies that the nominal bush in (33), which does not project, is a maximal projection. But then it is not quite clear why the combination of this maximal projection and the noun pig in (33) derives an X0, whereas the combination of an NP and a verb in (31) does not. I think that the difference between (31) and (33) is that in (31), an X0-element combines with a phrase, while in (33), two X0s are combined. In order to capture the difference between complementation and compounding, I will therefore represent terminal nodes inside complex words as non-maximal elements, i.e. X0s, whereas I will assume that outside the morphological domain, X0-elements can only be merged with phrasal complements, i.e. XPs. However, I will briefly return to bare phrase structure theory in the conclusion to this book, where I discuss some of the consequences of my analysis for a theory of syntactic projection. Notice furthermore that from a morphosyntactic point of view, I do not draw a distinction between compounding and affixation (cf. Höhle 1982). I assume that affixes, like free morphemes, are members of a particular syntactic category (cf. Lieber 1980, 1992), and the syntactic X0-node that the affix corresponds to merges with the terminal node of its host in the process of syntactic affixation. The only difference between affixes and free morphemes is that the former require a particular context which is specified in their lexical entry. For example, the English plural suffix /z/ has a lexical representation like (34) (cf. Jackendoff 1997: 118):13 LPS

(34) a.

LSS

b.

N0

Wd

Wd

Cla

z c.

N0

N0a

 sing   count 

plur

LCS [Entity PLUR ([Entity])]a

13.Cases of phonologically conditioned or class-based allomorphy can also be captured by the representational format in (34); see Jackendoff (1997) and Chapter 5.

39

40


The plural suffix is represented as a terminal node of category N which is linked to its corresponding semantic and phonological representations through the lexical index a. In this respect, the suffix is similar to a noun like cat, whose lexical entry was given in (13) in Section 1.1.1. However, in contrast to the lexical information associated with a free morpheme, (34) also specifies the context which has to be created by the rules of the syntactic and phonological component such that (34) can be unified with it. Phonologically, the plural suffix is a clitical element that must attach to the right of a phonological word. The LSS in (34b) captures the morphosyntactic contextual requirement of the plural-N0 (its “morphological subcategorization frame”; cf. Lieber 1980, 1982); it must attach to a singular count noun. Only if the syntactic structure and the phonological structure match this contextual information can (34) be unified with them.14 Let me come back to the idea that the combinatorial rules of syntax and morphology are the same. This claim is also made by Marantz (1997), who calls it a “natural assumption that whether you get a ‘zero-level category’ (word-like unit) or a phrasal category by merging two constituents is a function of the (categories of the) constituents involved, not of the merger operation itself” (1997: 205). In other words, Merge can apply to morphemes as well as to phrases; the character of the newly formed element only depends on the nature of its parts. The same reasoning applies to the operation Move (= Copy + Merge). If an X0-element already present in the structure is copied and then merged with another category Y0, this latter step is not different from the step illustrated in (33). For example, if the verb is copied and merged with Infl0, we create a complex Infl0-head:

14.(34) differs slightly from Jackendoff’s (1997) original example. In Jackendoff’s representation, all terminal nodes in the LSS of the plural are linked to their corresponding elements in the LPS, and the LCS of the plural is associated with the highest N-node in the LSS. However, as far as I can see, the way I represent the linking indices in (34) also establishes the right correspondences. Suppose (34) and the lexical entry of a singular count noun are unified with a syntactic structure, such that the LSS of the noun fulfills the contextual condition in (34b). Then the LPS of the host automatically fulfills the phonological requirement stated in (34a), and its LCS automatically fills the conceptual Entity-slot in (34c). This follows from general correspondence rules at the SS-PS and SS-CS-interface and need not be stated in the lexical entry of the affix.


Infl’

(35)

target

Infl0

Infl′ VP

V

0

Infl0

Move NP

V0i

VP Infl0

V0i

NP

The status of a complex element is determined on the basis of the X0-status of its parts. If we combine two terminal nodes, we derive a complex X0-category. Every complex X0 that consist of X0-elements is a morphological object; the combination of two heads by movement counts as word formation as well. The structural side of morphology is simply concerned with the formation of complex X0-structures. It is immaterial for morphology whether the relevant rule that forms these structures is Merge or Move. This position is in accordance with the traditional assumption of the (preMinimalist) Principles-and-Parameters approach, according to which inflectional morphology is done by moving verbs to functional head positions and by attaching them to the inflectional morphemes that are associated with these positions. Baker’s (1988) work on incorporation applies the same idea to word formation in the domain of derivational morphology, arguing that verbal compounds can be derived by movement of a verbal, a prepositional, or a nominal head and subsequent merging of this head and the verb. The resulting element is a word, because it consists of two X0-elements. Of course, the same idea is crucial to the Distributed Morphology-program, where it is assumed that syntactic head movement creates the complex X0-elements to which morphological operations apply. The idea that morphological objects are formed in syntax must not be confused with proposals such as those of e.g. Lieber (1992) or Ackema (1995), who claim that morphological principles can be reduced to syntactic principles. Even though the structure of morphological objects is derived by syntactic rules, there is still a morphological component, in the sense of “component” as a coherent system of principles and rules that operate on particular objects in particular domains. Even though the mechanisms that form a morphological object are the same as the ones that construct phrases, the rules and regularities that apply to this newly formed object may essentially be different. Syntax and morphology are different domains of grammar; morphological and syntactic regularities hold in different structural environments (i.e. words vs. phrases). The rules that determine the internal properties of a complex X0-element are

41

42


different from the rules that constrain operations on phrases. This is essentially the view that is defended by Baker (1988, Section 2.2.5). One example of morphological principles that govern the formation of morphological objects regardless of whether they are derived by Move or Merge is the so-called Right-hand Head Rule (Williams 1978, 1981). Williams argues that morphological objects, like phrases, have heads; according to Williams (1981), the head of a word is the rightmost element inside a complex X0category. The Right-hand Head Rule predicts that syntactic head movement that creates morphological structures always adjoins the moved element to the left of its host. Another example is provided by percolation conventions that are predicted to hold inside complex words. Williams (1981) argues that only heads can pass their features on to the dominating node. However, Selkirk (1982) introduces the idea that in the morphological domain, features of non-heads may also percolate (see also Di Sciullo & Williams 1987; Lieber 1992). If a morphological head X is not specified with respect to the feature expressed by the non-head Y, Y’s feature can percolate as well. Baker (1988) assumes that a complex X0 derived via head movement of an element Y0 bears both X’s and Y’s index. (This assumption is necessary in order to guarantee that a head that has undergone successive cyclic movement can still be related to its trace or copy.) This kind of “index percolation” from the non-head may be seen as the result of percolation conventions that are operative inside complex X0s derived by Move. A final example which illustrates that syntactically derived X0s form morphological domains is provided by the observation that the internal structure of X0s is invisible to rules of syntax, regardless of the operation that has formed this element. For example, if the compound bushpig in (33) is accompanied by a syntactic modifier like red, this adjective must modify the whole complex word; it cannot refer to only a part of the word (a red bushpig is not a pig out of the red bushes, neither is it a red pig out of the bushes; it is a bushpig which is red). As further evidence, consider the following contrast, noted for German in Stiebels & Wunderlich (1994, 918) (see also the discussion of similar examples from English, Spanish, and French in Horn (1989)): (36) a.

Er war mit dem Computer nicht zufrieden, sondern wollte einen he was with the computer not satiesfied but wanted a neuen new ‘He was not satisfied with the computer but wanted a new one’


b. *Er war mit dem Computer unzufrieden, sondern wollte einen he was with the computer unsatiesfied but wanted a neuen new ‘He was dissatisfied with the computer but wanted a new one’

The connector sondern, ‘but’, in German requires preceding negation with wide syntactic scope. (36a) is only grammatical if nicht, ‘not’, has wide scope. The wide scope reading can be established by the syntactic negation; for example, one could assume that wide scope is derived via LF-movement of nicht. However, sondern is not licensed by the morphological negation, expressed by the prefix un- in (36b), which has always narrow, word internal scope. The ungrammaticality of (36b) shows that the syntactic operation that applies to nicht and derives the wide scope reading in (36a) is not available to word internal elements like un-. The constraint that syntax cannot look inside words is known as the Principle of Lexical Integrity, which has developed out of Chomsky’s (1970) lexicalist hypothesis and which is formulated as in (37a) by Lapointe (1980) and as in (37b) by Di Sciullo & Williams (1987) (see also Bresnan & Mchombo 1995): (37) Principle of Lexical Integrity: a. Generalized Lexical Hypothesis (Lapointe 1980): No syntactic rule can refer to elements of morphological structure. b. Thesis of the atomicity of words (Di Sciullo & Williams 1987): Words are “atomic” at the level of phrasal syntax and phrasal semantics. The words have “features”, or properties, but these features have no structure, and the relation of these features to the internal composition of the word cannot be relevant in syntax.

Di Sciullo & Williams (1987) assume that Lexical Integrity is not an independent principle of grammar but rather a natural consequence of the fact that syntax and morphology are two subtheories of grammar. According to Di Sciullo & Williams, syntax simply lacks the vocabulary for analyzing morphological objects. Therefore, the rules of the syntactic component cannot “look inside” complex words. This implies that it is impossible to apply the operation Move to an element which is part of a complex morphological structure. Again, this holds for both words derived by Move and words derived by Merge. If a terminal element X0 has combined with another element Y0, Lexical Integrity prevents syntax from

43

44


accessing one part of the derived head; only the complex head itself should be able to be moved further. In fact, it is generally assumed that so-called excorporation, i.e. movement of a part of a complex X0, is not licensed by grammar.15 Baker (1988) explicitly rules out excorporation by the following filter: (38) A trace can never be nonexhaustively dominated by a zero-level category (Baker 1988: 73)

In the copy theory, which does not have the notion of “trace”, (38) would have to be modified; for example, it could be stated as a principle that prevents the operation Move from applying to parts of an X0 or as a syntactic constraint on chain formation (such that a chain is illicit if its foot is part of a complex X0category). The condition behind (38) seems to be morphological in spirit; it is a rule that prevents Move from triggering an X0-element inside a word. The close relationship between (38) and the principle of Lexical Integrity in (37) follows from the assumption that syntactic head movement in fact creates morphological structures; excorporation therefore counts as a violation of Lexical Integrity (see also Baker 1988: 73). Both the atomicity of words and the ban on excorporation are actually the result of one and the same principle. Notice that Lexical Integrity does not affect the idea that Move can derive words: Although Move can be applied to an X0-element, and create a new word by combining it with another Y0-element, it seems to be a general constraint that Move cannot be applied to parts of X0-elements. A distinction therefore must be drawn between morphological rules that operate inside X0-elements, and general combinatorial rules like Move and Merge that form X0-elements. We can continue to call Move and Merge syntactic rules; however, in that case, all structural operations that govern word formation are syntactic. 1.3.3 Parallel Morphology The idea that X0-structures derived by head movement are subject to morphological rules and regularities in the same way as X0-structures derived by Merge is the basis of Borer’s (1988, 1991, 1993) theory of Parallel Morphology. Borer coins the term “parallel” morphology in order to emphasize that morphology

15.However, see e.g. Guasti (1991), or Neeleman (1994) who suggest that excorporation is not generally impossible, but licensed under certain conditions. See also Roberts (1991), who argues that excorporation is impossible in morphological cases of X0-movement, but possible with cases like cliticization.


is not located prior to syntax, as lexicalist models claim, but runs parallel to syntax and has access to complex heads regardless of how they are derived. In this section, I will briefly illustrate how Parallel Morphology works. Borer (1991) discusses causative and inchoative verbs derived from adjectives. For example, the English suffix -en may attach to adjectives and derive inchoative verbs as in (39b): (39) a. The canal is [A0 wide] b. The canal [V0 widened]

The affix -en is of the syntactic category V; its lexical entry specifies that it must combine with a host of category A (cf. the lexical entry for the plural suffix, given in (34) above). As I argued in the preceding section, complex words like the complex V0 in (39b) can generally be derived by either Merge or Move. This means that the adjective wide and the verbal suffix -en may combine as in (40a) or as in (40b): (40)

a.

A0 does not project: A0 and V0 merge V′ (...)

V0 0

V0

A

b. A0 projects AP: A0 moves and merges with V0 V′ V0 0

A

AP V0 0

A

In (40a), the adjective and the verbal affix combine via Merge. In (40b), the adjective first projects an AP, then it must incorporate into the verb, because the lexical entry for -en requires the affix to be part of the same syntactic word as the adjective. This word has been derived via Move in (40b). In principle, the verb widen may be derived as in (40a) or (40b); generally,

45

46


both structures are possible. However, Borer (1991) observes that in certain circumstances, there is clear evidence that a word must have been derived by Move. Consider (41): (41) The canal widened more than a river (after the flood) (Borer 1991: 133)

Borer notes that the modifier more than a river in (41) modifies the adjective, not the derived verb. The meaning of (41) is “the canal widened more than a river is wide”, not “the canal widened more than the river widened”.16 This is unexpected; the principle of Lexical Integrity in (37) tells us that parts of words cannot be modified. Furthermore, modifiers like more than a river must be adjoined to the phrasal projections of the predicates they modify. We must conclude that (41) includes an AP, i.e. corresponds to a structure like (40b). In addition, Borer (1991) provides evidence from Hebrew that the occurrence of modifiers like more than a river is contingent on the presence of an AP. Consider the examples in (42): (42) a.

Ha-sadin Seli haki laban Se’epSar the-sheet mine most white that.possible ‘My sheet is as white as possible’ b. Ha-sadin Seli hilbin haki Se’epSar the-sheet mine whitened most that.possible ‘My sheet whitened as (much) as possible’ (Borer 1991:132)

The inchoative verb hilbin, ‘whiten’, in (42b) is derived from the adjective laban, ‘white’, in (42a) by what Borer calls the “hiCCiC-pattern”. The capitalized Cs stand for the consonants of the root that are combined with the prefixvocalic pattern hi-i-. In (42b), they are replaced by the consonants l, b, and n of the corresponding adjective laban. Borer notes that the adjectival superlative modifier haki Adj. Se’epSar, ‘as Adj. as possible’, is never independently attested in the verbal system; it is only licensed with adjectival phrases. Nevertheless, it can occur with the inchoative in (42b). Therefore, the underlying structure must include an AP. If no modifier is present, an inchoative verb like widen still can be derived by either Merge or Move. However, Borer (1991) shows that in some specific instances, word formation via Move is excluded. Consider (43):

16.However, Borer (1991: note 9) notes that for “a significant minority of native speakers”, (41) is ungrammatical under the intended reading.


(43) a. The flood widened the canal b. *The flood widened the canal like/more than a river (Borer 1991: 133)

In (43), we have the causative form of the verb widen. As (43b) shows, the causative verb cannot occur with a phrase that modifies the adjective. This suggests that there is no AP in (43b), i.e. the causative verb widen in (43a) must be derived by Merge, because on this assumption, modification would correctly be ruled out by the principle of Lexical Integrity. A second example that illustrates that words derived by Move may be different from words derived by Merge is Borer’s (1993) analysis of derived nominals in English and Hebrew (see Schoorlemmer (1995) for a Parallel Morphology-analysis of derived nominals in Polish). Compare (44a) and (b): (44) a.

The (frequent) collection of mushrooms for six months finally gave rise to a heavenly meal b. The collection was complete

(44) illustrates the difference between process and result derived nominals (cf. Grimshaw 1990). In both sentences, a noun collection is derived from the verbal stem collect-. In (44a), the noun refers to the event of collecting. In this reading, the internal argument of the verb is realized, and adverbials like frequent can modify the singular form of the noun. In contrast, the noun in (44b) refers to the result of the collection; no arguments of the verb are possible, and adverbials like frequent can only occur with the plural form of the noun. According to Borer, the difference between (44a) and (44b) follows from a structural difference. The noun collection in (44a) is derived by Move. Both the verbal stem collect- and the nominal suffix -tion project phrases in the syntax. The verbal head projects a VP. Since the nominal head is an affix, the verb must move and adjoin to N0:17

17.In Chapter 3 I will argue, following Koopman (1993), Borer (1998), and many others, that the case assignment properties of a lexical head are licensed by functional structure. Since the respective functional structure of the verb is absent in (45), the verb cannot assign case to its argument. Therefore, this argument must be realized inside a prepositional phrase in order to receive case from P0.

47

48


NP

(45) N0i/j

V0i collect-

N0j -ion

VP

V0i

PP

(of mushrooms)

The presence of the VP in (45) accounts for the verb-like properties of process nominals i.e. that they have an event semantics, that they inherit the verb’s arguments, and that adjectives like frequent in (44a) can quantify over events expressed by the verb. In contrast, if V0 and N0 are merged before the verb projects a phrase, the syntactic tree does not contain any verbal projection of collect, and the resulting N0 is a result nominal. Certain properties of process nominals follow directly from the structure in (45). For example, the arguments of the base verb are generated below VP and hence remain part of the structure when a process nominal is derived. However, Borer’s analysis leaves open the question of why the complex N0 in (45) is associated with an eventive reading in the context of a full VP, whereas the nominal refers to the result of the event expressed by the base verb if no VP is present.18 It seems that the rules at the interface between syntax and conceptual structure do not only consider the morphological structure of the nominal by itself. The whole structure and the syntactic environment of the terminal nodes also seem to play a role. The fact that in the interpretation of process nominals, the verb is at the same time part of a noun and present inside a full VP, seems to have a direct impact on the interpretation of the nominal. I cannot present an analysis here that captures this situation without stipulations. I merely intend to show that the tripartite architecture of grammar provides the means to relate the semantic difference between process and result nominals to the syntactic difference between word formation by Move and word formation by Merge. It was argued that correspondence rules at the interface between syntactic structure and conceptual structure establish the link between syntactic representations and interpretation. Now suppose that the following correspondence rule holds at the syntax-semantics interface:

18.I am indebted to Kyle Johnson (p.c.) for making me aware of this question.


(46) Interpretation of derived nominals a. If V0i and N0j form a complex N0i/j, and N0i/j is structurally adjacent to a copy of V0, then N0i/j refers to the Event-concept expressed by V0 in conceptual structure b. If V0i and N0j form a complex N0i/j, and N0i/j is not structurally adjacent to a copy of V0, then N0i/j refers to the concept that qualifies as a possible result of the Event expressed by V0 in conceptual structure

(46) incorporates the notion “structural adjacency” that I introduced in Section 1.3.1. The rule in (46a) links the eventive reading of process nominals to an N0-head that is structurally adjacent to the verb. The definition of structural adjacency then requires the verb to project a phrase which is the complement of N0 in syntactic structure. Importantly, the fact that V0 must also form a chain whose head merges with N0 is not required by the interpretation, but by the morphological subcategorization frame of the nominal suffix. The entry for -tion requires this affix to form a complex word with its verbal host; this requirement can now only be fulfilled if word formation proceeds via Move. Fortunately, the copy-theory of movement provides the appropriate tool to represent the fact that the syntactic structure of process nominals satisfies both (46a) and the morphological subcategorization frame of the affix. (46b) stipulates that a complex N0 which is not structurally adjacent to the verb’s copy is interpreted as a result nominal.19 In sum, the examples in (41) and (44) show that structures with words formed by Move do not always yield the same semantics as structures with words derived by Merge. This situation can be captured through correspondence rules at the interface between conceptual structure and syntax. Importantly, correspondence rules like (46) do not exclusively refer to the actual words that occur in these structures, but also take into account local relations that refer to structures that go beyond the X0-level. This observation is highly relevant for the analysis of particle verbs that I offer in the following chapters.

19.In Chapter 6, I will come back to the differences between derived process and result nominals when I investigate the properties of nominals and adjectives that are derived from particle verbs.

49

50


1.4 Conclusion In this section I have outlined my basic assumptions about the interaction of syntax, morphology and the lexicon. In Section 1.1, I introduced the model of grammar and the view of the lexicon on which my analysis of particle verbs will be based. According to this model, the representations of syntax, phonology, and semantics are linked through correspondence rules at the interfaces; the lexicon is part of these interfaces, and lexical items are correspondence rules. In Section 1.2 I showed how this model can be incorporated into the Principlesand-Parameters framework. It is possible to assume that there is one level of syntax that is assembled out of chains; these chains represent the results of what traditionally is called “syntactic movement”. This means that lexical entries establish a link between phonological/conceptual structures and syntactic structures that have been derived by syntactic rules, and I have shown that this is the core idea behind generative models that assume “late lexical insertion”. In Section 1.3 I have argued that morphology is concerned with complex words that are built exclusively out of other X0-elements and that the combinatorial rules that create these morphological structures are the same ones that also form phrases, i.e. Merge and Move. I also emphasized that the domain of morphology is not the only locality domain, and I introduced the notion of structural adjacency, which refers to the relation between a head and the head of its complement. In the following two chapters, I will show that particle verbs are characterized by the structural adjacency-relation.

Chapter 2

The syntax of particle verbs

In this chapter I argue that the basic syntactic representation of particle verbs is phrasal. The particle projects a phrase which is generated as the sister of the verb in syntactic structure. I discuss the structural properties of the verbparticle construction in detail, and I show that they follow directly from the assumption that the particle is the head of the complement of the base verb. When a particle verb appears in a verb-final clause in German and Dutch, the particle is adjacent to the verb (I will gloss particles as “Part” from now on): (1) weil Peter die Tür abschließt because P. the door part-locks ‘because Peter locks the door’ (2) dat Jan Marie opbelt that J. M. part-calls ‘that Jan calls up Marie’

(German)

(Dutch)

There are two major approaches to particle verbs in the literature. The first position, that I refer to as the morphological approach, holds that particle verbs are verbal compounds and are represented as verbal heads in syntactic structure (cf. e.g. Koster 1975; le Roux 1988;1 Johnson 1991; Neeleman & Weerman 1993; Stiebels & Wunderlich 1994; Neeleman 1994; Stiebels 1996; Olsen 1997b; McIntyre 2001). This means that the sentence in (1) would have a structure as in (3): (3) morphological approach: [C¢ weil [IP Peter [VP die Tür [V0 abschließt] ]]]

1.Strictly speaking, le Roux’s (1988) proposal is not morphological, because she explicitly analyses particle verbs as syntactically derived compounds, i.e. complex verbal heads that are formed in syntax rather than in morphology. However, according to what I argued in Chapter 1, such a distinction cannot be drawn, since morphology is the domain below X0, and any complex head is a morphological object by definition. In this sense, le Roux’s theory can be subsumed under the morphological approach.

52


According to the morphological approach, the adjacency of the particle and the verb in verb-final sentences simply follows from the fact that the particle and the verb are generated as parts of the same word. In contrast, a second position has been defended that considers the particle to be the head of a phrasal complement to the left of the verb. The proponents of this syntactic approach are split into two camps. On the one hand, it has been argued by van Riemsdijk (1978), Hoekstra, Lansu & Westerduin (1987), Grewendorf (1990), Mulder (1992), Koopman (1995), and others that the particle undergoes syntactic head movement and incorporates into the verb. According to the incorporation approach, the adjacency of the particle and the verb in (1) is the result of a structure like (4): (4) incorporation approach: [C¢ weil [IP Peter [VP die Tür [V¢ [PrtP Prt0i] [V0 abi-schließt] ] ] ]]

The morphological approach and the incorporation approach share the assumption that particle verbs are V0s. They only differ in that the former assumes that they are derived by Merge, whereas the latter argues that they are formed by Move. However, in contrast to what is assumed by the advocates of the incorporation approach, I have argued (Zeller 1997a, b) that the particle does not move in overt syntax, but stays in situ. A similar idea has been suggested by Drach (1963)2 and has also been advocated by Haiden (1997), Wurmbrand (1998), and Lüdeling (1998a). According to the in situ-approach, the adjacency of the particle and verb in embedded clauses is the result of the structure of the VP in Dutch and German. Since both languages are SOV, the PrtP-complement of which the particle is the head precedes the verb. The sentence in (1) is then represented as (5):3

2.“Es gibt keine trennbaren Verben. Es gibt […] einerseits echte Zusammensetzungen, andererseits verbale Gefüge mit Klammerfähigkeit und irreführender Rechtschreibung. […] Komposita sind immer untrennbar; sonst sind es keine.” (“There is no such thing as a separable verb. There are, on the one hand, genuine compounds, and, on the other, verbal structures whose parts occupy both sentence brackets [occupy C0 and the right periphery of VP respectively, J.Z.], and, confusingly, form an orthographic word. There are no separable compounds.”) This passage from Drach (1963: 59; first printed in 1940) illustrates that the view defended in this chapter is actually quite old. 3.In Zeller (1997a, b) I assume that the particle incorporates covertly on the way to LF. With respect to the claim that particles do not move overtly, this assumption and the in situapproach are identical.


(5) in situ-approach: [C¢ weil [IP Peter [VP die Tür [V¢ [PrtP ab] [V0 schließt] ] ] ]]

My main goal in this chapter is to substantiate the validity of the in situapproach. I argue that the particle and the base verb are independent heads in syntactic structure and do not form a word:4 VP

(6)

Spec NP (argument)

V′

PrtP

V0 (verb)

Prt0 (particle)

I show in the following sections that the properties of the verb-particle construction follow from a representation as in (6). I present the arguments that can be given in favor of the syntactic in situ-approach, and I evaluate their conclusiveness on the basis of the morphological alternative, according to which particle verbs are morphologically complex V0s. The discussion will show that the morphological approach is not equipped with an appropriate means to account for the syntactic properties that are attested with particle verbs. Most of the arguments that can be given against a morphological approach directly carry over to the incorporation alternative, which shares with the morphological approach the assumption that particle verbs are complex V0s at the interface level of syntactic structure (cf. (3) and (4)). The claim that particle verbs are realized as phrasal constructions holds with one qualification. The reader should be forewarned that the word-formation

4.There are different opinions about the position of the internal argument of the particle verb. According to those proponents of the syntactic approach who follow Stowell (1981) and Kayne (1985) and adopt a Small Clause analysis (cf. Hoekstra, Lansu & Westerduin 1987; Grewendorf 1990; den Dikken 1995), the internal argument is located inside the particle’s maximal projection. Others (cf. e.g. Wurmbrand 1998) assume that the internal argument is located inside the VP, but outside the PrtP. In this chapter I adopt the latter position, which I motivate in more detail in Chapter 3.

53

54


properties of particle verbs that I discuss in Chapter 6 can only be accounted for by assuming that in these specific contexts, particle verbs may also occur as V0s. However, I show that this idea cannot be generalized; it does not apply to the representation of particle verbs that are not part of derived words. I therefore assume that the syntactic status of particle verbs is “basic” in the sense that the alternative representation of particle verbs as words is only realized in specific (morpho-) syntactic environments. This chapter provides evidence that in all other contexts, the properties of particle verbs in German follow from the syntactic in situ-approach. The following discussion will reveal the shortcomings of the morphological approach and the incorporation approach; furthermore, it provides me with a good opportunity to illustrate and discuss some of the most influential theories on particle verbs in German and Dutch that have been developed in recent years. In Section 2.1 I discuss the fact that particle verbs are split by verb movement to Comp0. Section 2.2 shows that particle verbs are also split when the verb does not move further than Infl0; inflectional elements like the infinitival marker zu that are associated with Infl0 may thus intervene between the particle and the verb. In Section 2.3, I look at the behavior of particle verbs with respect to Gapping, and Sections 2.4 and 2.5 show that particles can be topicalized and modified by adverbs. In Section 2.6, some further structural properties of the verb-particle construction are discussed that provide further evidence in favor of the in situ-approach, and Section 2.7 shows how this approach also accounts for the properties of the verb-particle constructions in other Germanic languages like English, Norwegian, or Danish.

2.1 Verb movement to Comp In traditional grammars, particles are sometimes called “separable prefixes”, because they can be separated from the verb in particular syntactic contexts. In this section I show that the separability of particle verbs can only satisfactorily be explained by the syntactic approach. Section 2.1.1 compares separable particles to inseparable prefixes and resultative predicates; it will become clear that the structure of particle verbs bears a much closer resemblance to resultative constructions than to prefix verbs. Section 2.1.2 shows that the morphological approach runs into serious problems in trying to account for separability, and Section 2.1.3 shows the same thing with respect to the incorporation approach.


2.1.1 Syntactic separability The strongest argument in favor of a syntactic approach to particle verbs is the observation that the verb and the particle can be split by syntactic rules. The particle appears adjacent to the verb in verb-final clauses, but it is separated from the verb in verb second (V2) contexts. (Although I adopt the copy theory of movement in this study, I sometimes informally represent the base position of a moved element through a trace if the example is not structured by syntactic brackets): (7) a.

weil Peter in den Bus einsteigt because P. in the bus part-climbs ‘because Peter gets on the bus’ b. Peter steigti in den Bus ein ti P. climbs in the bus part ‘Peter gets on the bus’ (cf. *Peter einsteigt in den Bus)

(8) a.

weil Peter die Nachricht aufschreibt because P. the note part-writes ‘because Peter makes a note’ b. Peter schreibti die Nachricht auf ti P. writes the note part ‘Peter makes a note’ (cf. *Peter aufschreibt die Nachricht)

(9) a.

weil Peter sein Bier austrinkt because P. his beer part-drinks ‘because Peter drinks up his beer’ b. Peter trinkti sein Bier aus ti Peter drinks his beer part ‘Peter drinks up his beer’ (cf. *Peter austrinkt sein Bier)

Regardless of their semantic properties, all particle verbs are split under V2.5 A second syntactic operation that separates the verb and the particle occurs in Dutch. In Dutch, a certain class of verbs (modals, perception and causative verbs etc.) triggers so-called Verb Raising of an embedded infinitive into the matrix clause (cf. Evers 1975; van Riemsdijk 1978). If the embedded infinitive is a particle verb, as in (10), two options exist:

5.In Section 2.2.2, I discuss so-called backformations in German, which I will analyze as particle verbs that fail to undergo V2 altogether. Dutch verbs with similar properties are discussed in Koopman (1995).

55

56


(10) a.

dat ik Jan op ti wil belleni that I J. part want call b. dat ik Jan ti wil opbelleni that I J. want part-call ‘that I want to call Jan up’

In (10a), the matrix verb willen has triggered movement of the base verb, stranding the particle. Again, the verb moves without carrying the particle along. (10b) shows that the whole particle verb may be moved as well. However, as pointed out by van Riemsdijk (1978), postpositions in Dutch, which are undoubtedly heads of phrasal complements of the verb, can also be moved together with the verb in Verb Raising constructions. (10b) hence does not provide evidence in favor of a morphological approach. The analysis of the alternation in (10) requires some assumptions that I have not motivated yet; I therefore postpone the discussion of the behavior of Dutch particle verbs in Verb Raising constructions to Chapter 7. In this section, I restrict myself to a discussion of syntactic separability under V2. The separability of particles in V2 is striking because nothing comparable is attested with genuine prefix verbs. In German, there is a second important class of verbal modifiers, the so-called inseparable prefixes. The set of inseparable prefixes in German is a closed class. It consists of two subsets. The first one includes the deaccented variants of former prepositions (be-, ent-, er-, ver; and ge-, miß-, voll-, zer-). As shown in (11b) and (12b), these prefixes must move together with their base verb: (11) a.

weil Peter den Berg besteigt because P. the mountain pref-climbs ‘because Peter climbs the mountain’ b. Peter besteigti den Berg ti P. pref-climbs the mountain ‘Peter climbs the mountain’ (cf. *Peter steigt den Berg be-)

(12) a.

weil Peter die Munition verschießt because P. the ammunition pref-shoots ‘that Peter gets rid of the ammunition by shooting’ b. Peter verschießti die Munition ti P. pref-shoots the ammunition ‘Peter gets rid of the ammunition by shooting’ (cf. *Peter schießt die Munition ver-)


The second subclass includes variants of existing prepositions (so-called P-prefixes: durch-, hinter-, über-, um-, unter-, wider-). P-prefixes must also move with the verb in V2 (cf. Zeller 1997a; Wurmbrand 1998): (13) a.

weil Peter den Brief unterschreibt because P. the letter P-pref-writes ‘because Peter signs the letter’ b. Peter unterschreibti den Brief ti P. P-pref-writes the letter ‘Peter signs the letter’ (cf. *Peter schreibt den Brief unter-)

(14) a.

weil Peter die Stadt überfliegt because P. the city P-pref-flies ‘because Peter flies over the city’ b. Peter überfliegt die Stadt P. P-pref-flies the city ‘Peter flies over the city’ (cf. *Peter fliegt die Stadt über)

Some prepositional elements (like um, ‘around’, durch, ‘through’, or über, ‘over’ etc.) are included both in the list of particles and in the list of P-prefixes. This gives rise to the existence of contrasts like the following:6 (15) a.

weil Peter den Mann UMfährt (particle verb) because P. the man part-drives ‘because Peter runs the man over (knocking him down)’ b. weil Peter den Mann umFÄHRT (P-prefix verb) because P. the man P-pref-drives ‘because Peter drives around the man’

(16) a.

Peter fährt den Mann um P. drives the man part ‘Peter runs the man over (knocking him down)’ b. Peter umfährt den Mann P. pref-drives the man ‘Peter drives around the man’

(particle verb)

(P-prefix verb)

6.It is sometimes argued (particularly by standard handbooks for German, but see also e.g. Abraham 1995) that pairs like (16) exhibit the following tendency: particle verbs tend to have “literal” meanings, whereas inseparable P-prefix verbs tend to be more “figurative”. This analysis is contradicted by a number of examples, including (16). See Dewell (1996) for more discussion of this point.

57

58


In standard written language, the verb umfahren is ambiguous; it may be a particle verb or a prefix verb (cf. Dewell (1996) for more examples). However, particles bear stress, whereas prefixes are unstressed. Therefore, verbs like umfahren are always disambiguated in spoken language, as shown in (15). Since the verbal modifier um is a particle in (15a), it is stranded if the verb undergoes V2, (16a), whereas um is a prefix in (15b) and therefore moves with the verb to Comp0, (16b). It is clear that the difference between separable particles and inseparable prefixes must be the result of a structural difference between particle verbs and prefix verbs. The most natural assumption is that only prefix verbs are words. As stated by the principle of Lexical Integrity (see Section 1.3.2), the internal structure of words cannot be accessed by syntactic rules; hence, it is predicted that prefix verbs are inseparable. In contrast, if particle verbs are phrasal constructions in syntactic structure, it is predicted that the particle and the verb are separated if the verb moves. The verb’s sister is a phrase whose head is the particle, and when the verb moves, the particle is stranded inside its PrtP. The contrast between particle verbs and prefix verbs is illustrated in (17):7 (17) a.

particle verbs VP

Spec

V′ verb movement PrtP

V0 (verb)

Prt0 (particle)

7.(17b) represents prefix verbs as being derived by Merge. However, I do not exclude the possibility that prefix verbs may also be derived by Move, as argued by e.g. Büring (1991), Zeller (1997a, 2001), and Wurmbrand (1998).


b. prefix verbs VP

Spec

V′

verb movement

V0

Pref 0 (preWx)

V0 (verb)

I assume that (17a) is the structure of the verb-particle construction in German and Dutch. The syntactic representation of particle verbs is hence similar to that of regular verb-complement constructions. For example, prepositional particles, like auf in (18a), are syntactically similar to full prepositional phrases like auf den Wagen in (18b). The only difference is that the internal argument of the preposition is omitted in (18a): (18) a.

Peter lädti das Heu [auf] ti P. loads the hay [part ‘Peter loads the hay (onto something)’ b. Peter lädti das Heu [auf den Wagen] ti P. loads the hay [onto the wagon ‘Peter loads the hay onto the wagon’

(PrtP-complement)

(PP-complement)

Furthermore, the verb-particle construction is structurally comparable to resultative constructions (I will occasionally refer to these constructions simply as “resultatives”):8 (19) a.

Peter lief seine Sohlen [AP kaputt] P. ran his soles broken ‘Peter ran off his soles’

8.Resultative constructions are discussed by Simpson (1983), Hoekstra (1988), Jackendoff (1990), Carrier & Randall (1992), Levin & Rappaport (1995), and many others. For German, see in particular Kaufmann (1995) and Lüdeling (1998b).

59

60


b. Peter strich die Tür [AP rot] P. painted the door red ‘Peter painted the door red’ (20) a.

Peter trank Hans [PP unter den Tisch] P. drank H. under the table ‘Peter drank Hans under the table’ b. Peter wäscht sich die Seife [PP aus den Augen] P. washes refl the soap out.of the eyes ‘Peter washes the soap out of his eyes’

The base verb in a resultative construction combines with a “resultative predicate”; an AP-complement in (19), a full PP-complement in (20). In combination with the direct object, a resultative predicate always expresses the “result state”9 of the event expressed by the verb. For example, the meaning of (19b) can roughly be paraphrased as “As a result of Peter’s painting (the door), the door ended up red”. Although there is disagreement in the literature about the question of whether the resultative predicate and the direct object form a syntactic unit (as is argued by proponents of a Small Clause-analysis, like e.g. Hoekstra 1988 or von Stechow 1995), it is generally acknowledged that resultative predicates are phrases. However, notice that in (19), the phrasal resultative predicate only consists of the head of the AP; this makes the examples in (19) look very much like the verb-particle constructions in (21): (21) a.

Peter lief seine Sohlen [PrtP ab] P. ran his soles part ‘Peter ran off his soles’ b. Peter strich die Tür [PrtP an] P. painted the door part ‘Peter painted the door’

As full phrases, the adjectival predicates in (19) are left behind when the verb moves to Comp0. The separability of a verb and a resultative predicate follows in a straightforward manner. If the syntactic approach is adopted, and particles are represented as phrases as well, the separability of particle verbs, which is

9.I use Kratzer’s (1995) notion of “result state” to describe the terminal state of an event. Similar terms that are mentioned in the literature are “resultant state” or “target state” (see Parsons 1990 for discussion).


shown in (21) and above, also comes as a natural consequence of their structural representation. Despite the parallel between (19) and (21), many linguists assume that particle verbs are words. The main reason is that the semantics of particle verbs is not as regular as the semantics of resultative constructions. As I show in Chapter 4, the meaning of particle verbs cannot be analyzed in a uniform way. However, the idea that semantic irregularities are only associated with X0elements is based on the lexicalist assumption that the lexicon precedes syntax and feeds it with words. In Chapter 1, I have argued that this assumption is incorrect. Since semantic irregularities are also attested with structures larger than X0, there is no reason to take the semantic difference between resultative constructions and particle verbs as evidence for the claim that only the former are phrases, whereas the latter are words. 2.1.2 The morphological approach According to the morphological approach, particle verbs are words. This implies that when the verb moves in V2, a complex V0 is split: VP

(22)

NP

V Prt 0

part of a verb moves

0

V0

A first problem for the morphological approach is raised by the principle of Lexical Integrity, which has been introduced in Chapter 1, Section 1.3.2. According to this principle, the internal structure of words is not accessible for syntactic movement rules; the syntactic component cannot “look inside” complex words. If Lexical Integrity holds, particle verbs cannot be complex morphological objects. However, the principle of Lexical Integrity has been challenged by a number of authors in recent years (cf. Lieber 1992; Neeleman 1994; Ackema 1995; McIntyre 2001). These authors argue that the empirical effects of Lexical Integrity can be reduced to independent principles of grammar, and that parts of words can be moved out of words as long as these principles are not violated. On the basis of this assumption, a morphological analysis like (22) is offered by Neeleman (1994). Neeleman captures the difference between inseparable prefix

61

62


verbs and particle verbs by postulating a fundamental difference between two kinds of morphological structures, stem compounds and word compounds. Whereas the former are derived by combining affixal elements (that Neeleman, following Selkirk (1982), labels X−1-elements), the latter are built out of independent words, i.e. X0-elements. Neeleman assumes that verbal prefixes are X−1-elements (because they are phonological affixes), whereas particles are X0s. In this approach, particle verbs therefore consist of two X0-elements, whereas inseparable prefix verbs are formed out of two X−1-elements: (23) a.

prefix verb (cf. (13a))

V Pref -1 unter

0

V -1 schreib-

b.

particle verb (cf. (9a)) V0

Prt0 aus

V0 trink-

According to Neeleman, the fact that particle verbs can be split whereas prefix verbs cannot follows from the fact that the particle is an X0-element. Neeleman argues that independent principles of grammar require X−1-elements, but not X0-elements, to be part of a complex word in syntax. Consequently, words that consist of X−1-elements cannot be split, but nothing prevents movement of a particle out of a morphological object that consists of only X0-elements. Neeleman’s analysis illustrates the general dilemma that all versions of the morphological approach run into. In order to maintain the claim that the particle verb is a V0, its syntactic separability has to be restated in morphological terms. The strategy that has been chosen by most authors, including Neeleman, is to assume that the structure of a particle verb is formally different from the structure of other complex words. However, recall that according to the view of morphology that I motivated in Chapter 1, Section 1.3.2, the morphosyntactic structure of words is always the same, regardless of whether they consist of affixal elements or free morphemes. Terminal nodes are always represented as X0s, and the boundedness of affixes is simply listed as a particular contextual requirement in the affix’s lexical entry. From this perspective, the postulation of two morphologically different structures must be considered a rather unwelcome result. However, even if one adopts Neeleman’s distinction between two morphologically different modes of word formation, Neeleman still has to explain why the particle is stranded, i.e. why the particle verb does not move as a whole — this option is not excluded a priori, given that the particle verb is analyzed as a


complex V0. Neeleman’s explanation is based on the rather dubious Complexity Constraint originally proposed in Neeleman & Weerman (1993): (24) Complexity Constraint (Neeleman & Weerman 1993: 460) X0 should either be a lexical head or have a lexical head (25) Complexity Constraint (Neeleman 1994: 304) The head of an X0 may not be complex

Let me explain the central idea behind (24) and (25). Consider the morphological structure in (26): X10

(26)

Z

0

X2 0 Y0

X3 0

In a structure like (26), X30 is the morphological head of X20; X20 is the head of X10. According to Neeleman and Weerman, a “lexical head” is a head that immediately dominates phonological material. In terms of the framework elaborated in Chapter 1, a lexical head is a terminal node that is associated with a particular Vocabulary item. If the terminal nodes in (26) are regularly associated with phonological features, then X20 is not a lexical head. This shows why (24) is essentially equivalent to (25). If the head of a word is itself complex, and its terminal nodes are regularly associated with LPS-entries, no phonological information is associated with this head itself. Now consider how (24) and (25) are supposed to rule out movement of the whole particle verb to Comp0 in V2. According to Neeleman & Weerman (1993) and Neeleman (1994), if the particle moves to Comp0, the resulting structure looks as in (27): *Comp 0

(27)

V0 Prt 0 aus

V0 trink-

63

64


(27) is how Neeleman & Weerman (1993) and Neeleman (1994) represent substitution. Comp0 directly dominates the V0 that has been moved; as a result, a moved verb is the head of Comp0. If the verb is complex, it violates (24) and (25), because in this case, the head of Comp0 itself (the complex V0 in (27)) is not associated with phonological features. This is how Neeleman accounts for the fact that a particle verb may never move to Comp0 as a whole. However, Neeleman’s account is highly problematic. First, it predicts that prefix verbs cannot move to Comp0 either, because a structure like (28) also violates the Complexity Constraint, as formulated in (24) and (25): C0

(28)

V0 Pref -1 durch

V-1 lauf-

In order to not wrongly exclude V2 with prefix verbs, Neeleman (1994) is forced to give a definition of complexity which basically states that V0 in (27) is complex, but V0 in (28) is not. But this is barely more than a descriptive restatement of the facts which lacks real explanatory content. Second, the Complexity Constraint is empirically incorrect, as pointed out by Stiebels & Wunderlich (1994). They note that adjectives with complex heads do exist in German (cf. superfederleicht, “super-featherlight”). Furthermore, as Neeleman & Weerman (1993) and Neeleman (1994) admit, the Complexity Constraint does not hold for nouns, since right-branching compounds exist both in German and in Dutch (cf. Holzfensterbank, ‘wood-window ledge’). It therefore seems unlikely that syntactic and morphological derivations are in fact constrained by something like the Complexity Constraint. But if the Complexity Constraint does not exist, it remains unclear why particle verbs cannot move to Comp0. If they were V0s, as claimed by Neeleman’s (1994) morphological approach, we would also expect them to move as V0s. Similar objections must be raised against the morphological account proposed by Stiebels & Wunderlich (1994). According to their theory, particles, but not prefixes, are X0-elements inside V0 that carry a feature [+max]. This feature makes the particle visible to syntax, and if a part of a word is visible, the word can be split by syntactic rules. Furthermore, Stiebels and Wunderlich stipulate a target principle according to which only one [+max]-category is


licensed in Comp0. This prevents particle verbs from being licensed in this position; since the base verb is marked [+max] as well, a particle verb consists of two [+max]-elements. In contrast, prefix verbs, which include only the [+max]-feature of the verb, can freely move to Comp0. However, the feature [+max] and the target principle, like Neeleman’s Complexity Constraint, are mere stipulations; they are necessary because otherwise, the morphological approach makes the wrong predictions about particle verbs in V2. Furthermore, the feature [+max] causes some theory-internal problems for Stiebels and Wunderlich that considerably reduce the explanatory force of their account (for details, see the critical discussion in McIntyre in prep.). A strong morphological analysis of particle verbs that does away with unmotivated stipulations is proposed in McIntyre (2001). McIntyre does not assume that the morphological structure of particle verbs is essentially different from that of prefix verbs. Instead, he tries to explain the difference between separable and inseparable prefixes on the basis of an audible difference between particle verbs and prefix verbs. Whereas prefixes are unaccented, particles bear stress. According to McIntyre’s Stress Hypothesis, it is the accentuation of the particle that triggers separability. McIntyre (2001) supports the Stress Hypothesis with the following observation. Consider the verbs in (29): (29) a.

weil Peter seine Freundin mißversteht because P. his girlfriend pref-pref-stands ‘because Peter misunderstands his girlfriend’ b. weil Peter die Auseinandersetzung überbewertet because P. the argument pref-pref-estimates ‘because Peter overestimates the argument’

The verbs mißverstehen and überbewerten are formed by affixation of miß- and über-. These elements are genuine prefixes and usually unstressed. However, if the first syllable of the base verb is unstressed as well (as is the case with the base verbs verstehen and bewerten, which are themselves prefix verbs), the prefixes miß- and über- are stressed in order to avoid a succession of unstressed syllables. Interestingly, in these contexts, the prefixes miß- and über- in (29) are separable in V2: (30) a. ??Peter versteht seine Freundin mißP. pref-stands his girlfriend pref ‘Peter misunderstands his girl friend’

65

66


b. ?Peter bewertet die Auseinandersetzung überP. pref-estimates the argument pref ‘Peter overestimates the argument’

McIntyre (2001) concludes from the examples in (30) that (i) genuine morphological objects can be separated, and (ii) complex verbs with stress on the first syllable force the base verb to be separated from the prefix or the particle. However, there also are problems with McIntyre’s approach. First, if separation is really stress-related, it is unclear why stranding of the prefixes in (30) is less natural than stranding of particles. In fact, most speakers tend to avoid V2 with verbs like mißverstehen and überbewerten, while others also accept V2 with the whole verb moved to Comp: (31) a. ?Peter mißversteht seine Freundin b. ??Peter überbewertet die Auseinandersetzung

In contrast, there is nothing strange about stranding particles in V2; speakers do not try to avoid V2, and crucially, they never move the whole particle verb. Even if separability of the prefixes in (30) followed from the fact that they exceptionally receive stress, this does not yet explain why particles are separated more easily. Second, it is unclear why phonological conditions should prevent a base verb and a prefix or particle from appearing in Comp0, whereas no such conditions preclude the adjacency of the two terminal nodes in verb-final position. How does stress assignment, which is established at phonological structure, affect the syntactic position of a complex element? Third, it is curious that this phonological condition only constrains the distribution of a certain class of verbs. As McIntyre (2001) himself notes, the verbs in (32), which are converted from compound nouns which are initially accented, are perfectly acceptable in V2: (32) a.

FRÜHstücken ‘to have breakfast’ b. SCHRIFTstellern ‘work as a writer’

– –

Wir frühstücken zu Hause ‘We have breakfast at home’ Peter schriftstellert wieder ‘Peter works as a writer again’

The verbs in (32) with accentuation of the first constituent are located in Comp0. If it was really the case that phonological conditions keep the whole particle verb from moving to Comp0, we would expect the same conditions to also rule out examples like (32).


Finally, systematic counterexamples to the Stress Hypothesis are observed when contrastive stress (indicated by capital letters in (33)) is assigned to a prefix: (33) a.

Peter UNTERschätzt die Frauen, aber Hans ÜBERschätzt sie P. pref-estimates the women but H. pref-estimates them ‘Peter underestimates women, but Hans overestimates them’ b. *Peter schätzt die Frauen UNTER, aber Hans schätzt sie ÜBER

Like any other prefix verb, the prefix verbs unterschätzen and überschätzen usually bear main stress on the base verbs, but if one prefix is contrasted with another, contrastive stress is assigned to the prefix. The Stress Hypothesis predicts that the prefix verbs should now either be split under V2, (33b), or that V2 is excluded altogether. But (33a) clearly shows that V2 is possible with these prefix verbs inspite of the accentuation of the prefix, contrary to what the Stress Hypothesis predicts. Since McIntyre remains vague about the precise phonological mechanisms that rule out complex verbs with stress on the first constituent in Comp0, and in the light of the problems raised by (31)–(33), I also reject his Stress Hypothesis. The fact that particles receive stress does not cause separation; rather, stress on the particle and separability both follow from the structural representation of particle verbs. In German, the main accent of a clause is placed on the element which is left-adjacent to the base position of the verb (cf. Abraham 1995). As shown in (34a–c), this element is the most deeply embedded head. Since particle verbs are phrasal constructions, the particle is left-adjacent to the verb and therefore receives stress (cf. Zeller 1997a): (34) a.

nach HAUSE gehen to home go b. TRAURIG sein sad be c. ein BUCH lesen a book read d. AUF-legen part-put, ‘put on’

(stress inside PP) (stress inside AP) (stress inside NP) (stress inside PrtP)

In contrast to what is argued by McIntyre (2001), I do not think that in (30), morphological objects are separated. Instead, I assume that the stress pattern of these particular verbs (which is triggered by the tendency to avoid two unstressed syllables) may tempt speakers to reanalyze these prefixes as particles, i.e.

67

68


to represent them as phrasal complements of the verb. In this case, the complex verbs in (30) behave like particle verbs, i.e. they are split when the base verb moves. However, since this kind of reanalysis creates structures that are not in accordance with the syntactic context that is specified by the lexical entry of the respective prefix, they must be regarded as a marked option. Therefore, sentences like (30) are judged as slightly odd, and speakers tend to avoid V2 altogether. Syntactic separability of particle verbs under V2 remains a mystery if it is assumed that particle verbs are words. In contrast, separability follows directly from the syntactic representation of the verb-particle construction that I have proposed in (17a) above. The particle is the head of an independent phrase; it is therefore expected to be stranded when the verb moves. Particle verbs are not complex morphological objects that are split, but complex syntactic constructions for which separability is the rule. 2.1.3 The incorporation approach The syntactic separability of particle verbs also raises problems for a syntactic approach that assumes overt particle incorporation. The incorporation approach patterns with the morphological approach in that it falsely predicts that the whole particle verb moves to Comp0 in V2. In order to account for the syntactic separability of particle verbs, the incorporation approach would have to assume that the verbal part of the complex V0 derived by incorporation can excorporate and strand the particle inside V0. However, recall that a V0 derived by Move is a word like a V0 derived by Merge, and excorporation is a violation of Lexical Integrity. But even if excorporation was permitted, the incorporation approach would fail to explain why the particle does not move with the verb in V2 after incorporation has combined the two terminal nodes. As was the case with the morphological approach, the incorporation approach predicts that the complex particle verb created by incorporation can move to Comp0 as a whole. Although the incorporation approach is based on the syntactic approach and represents the particle inside a phrase, nothing is gained by this assumption if the particle is assumed to incorporate, because V2 still breaks up a complex verb. A possible way out of this dilemma would be to assume that particle incorporation is not obligatory, but optional, as suggested by van Riemsdijk (1978). One would have to assume that only in V2, the particle stays in situ, whereas it incorporates otherwise. This view, however, is problematic. According to Chomsky (1991, 1995), the operations of the computational system are


governed by principles of economy that require these operations to be driven by conditions on representations that would otherwise be violated. Movement cannot really be optional, but is a Last Resort-operation that is always triggered by morphological properties of the elements that undergo or attract movement. For example, in the theory of lexical licensing that I adopt in this study, head movement might be triggered by the LSS-entry of an affixal head that requires syntax to produce the right context (see the discussion of Parallel Morphology in Chapter 1, Section 1.3.3). But if an element is only licensed in a particular syntactic environment, then this environment must always be created by syntax in order to make unification of this element’s LSS and syntactic structure possible. If particle incorporation occurs as a result of some affixal property of V0 or Prt0, it is obligatory and must apply regardless of the syntactic context in which the particle verb will appear. It is therefore impossible to assume that in V2, the particle does not combine with the verb if it does otherwise. In sum, syntactic separability of particle verbs is best explained by the syntactic in situ-approach. On the assumption that the particle is represented as the head of a phrasal complement of the verb, the fact that the verb moves to Comp0 without the particle follows straightforwardly. In contrast, the morphological approach and the incorporation approach both fail to explain the separability of particle verbs without additional assumptions and stipulations. I hence take syntactic separability as a strong argument in favor of my claim that particles are the heads of their own phrases in syntactic structure.

2.2 Verb movement to Infl The fact that particle verbs are split when the verb moves to Comp0 provides strong evidence for the in situ-approach. However, it remains to be shown that the same analysis can be maintained with respect to particle verbs in nonV2-contexts. I show in Section 2.2.1 that so-called “morphological” separability of particle verbs (cf. Stiebels 1996; Lüdeling 1998a) is in fact another instance of their syntactic separability. In Section 2.2.2 I discuss backformations, which provide a potential problem for this claim, and the properties of particle verbs in finite verb-final sentences are discussed in Section 2.2.3.

69

70


2.2.1 “Morphological” separability The infinitival marker zu in German is a prefix; zu precedes the infinitival form of the verb: (35) a.

Peter versucht heute zu kommen P. tries today to come ‘Peter tries to come today’ b. Peter versucht eine Frau zu finden P. tries a woman to find ‘Peter tries to find a woman’

I assume, following Giusti (1989), Sternefeld (1990), Grewendorf (1990), Grewendorf & Sabel (1994), Sabel (1996a, b, 2000), and others, that zu does not combine with the infinitival form of the verb in the VP, but is associated with a [−finite]-Infl0. It therefore patterns with English to, which is associated with Infl0 (cf. Pollock 1989), as is suggested by the examples in (36)–(38) (cf. Sabel 1996a, 2000, and references cited therein): (36) …and [leave]i we tried to ti (37) Mary likes to tour art galleries, but Bill hates *(to) (Sabel 1996a: 27) (38) a. John wants not to go b. John wants to not go (Pollock 1989: 375)

(36) shows that VP-topicalization leaves the infinitival marker to behind. VP-ellipsis in (37) does not delete the LPS of to, and (38b) shows that negation may intervene between to and the base verb. These data follow from the claim that to is a realization of Infl0 and that the verb does not leave the VP in English. The fact that examples like (36)–(38) are not possible in German is due to the affixal character of zu. As a prefix, zu cannot survive without its host, which means that the verb must move and attach to Infl0 in order to fulfill the affixal requirement of the infinitival element. However, aside from this difference, zu and to behave alike. For example, as shown in Sabel (1996a, 2000), both elements can be left out in the same contexts: (39) a. I help my plants to grow by using Ortho b. I help my plants grow by using Ortho (Gee 1977) (40) a.

Ich helfe meinen Pflanzen zu wachsen, indem ich Ortho verwende I help my plants to grow in.that I O. use ‘I help my plants to grow by using Ortho’


b. Ich helfe meinen Pflanzen wachsen, indem ich Ortho verwende I help my plants grow in.that I O. use ‘I help my plants grow by using Ortho’

If German zu was directly merged with the verb, although to is located in Infl0, the similarity between the non-appearance of to in (39b) and the non-appearance of zu in (40b) would remain unexplained. However, suppose that it is a lexical property of helfen/help that this verb may optionally select for either an IP or a VP. Then (39) and (40) follow directly from the assumption that both to and zu are located in Infl0. Finally, as Grewendorf (1990) notes, the contrast in (41) provides further evidence for this assumption: (41) a.

Peter scheint ins Kino zu gehen P. seems to.the cinema to go ‘Peter seems to go to the movies’ b. Ich höre Peter ins Kino gehen I hear P. to.the cinema go ‘I hear Peter going to the movies’

The raising verb in (41a) selects an infinitive with zu, but the ACI-verb in (41b) does not. If zu is associated with Infl0, the difference between (41a) and (41b) reduces to different c-selectional properties of the matrix verbs in (41) (scheinen selects an IP; hören a VP). In contrast, if zu and the verb were merged inside the VP, the difference between (41a) and (41b) becomes more problematic; it cannot be accounted for without additional stipulations. I hence conclude that the zu-infinitival forms of verbs are derived by movement of the verb to Infl0. When the Vocabulary items of the verb and [−finite]Infl0 are inserted, the infinitival prefix is attached to the left of the phonological realization of the verb.10 It is therefore predicted that zu precedes all verbs.

10.There are two ways of making this process of zu-prefixation more precise. One possibility is to argue that V0 adjoins to the right of [−finite]-Infl0. This would directly account for the order infinitival prefix — verb stem. However, this idea contradicts the claim that heads always left-adjoin to their targets (see Chapter 1, Section 1.3.2). An alternative assumption would be to argue that V0 adjoins to the left of [−finite]-Infl0, but that the phonological realization of [−finite]-Infl0 is a circumfix [zu + V-en], with the [−finite]-feature associated with the suffix -en. However, although this analysis allows us to maintain the assumption that the structure of infinitives derived by verb movement obeys the Right-hand Head Rule, it requires some additional assumptions about bare VP-infinitives (where the verbal stem is suffixed with -en as well). Notice that the choice between the two analyses does not affect the

71

72


This prediction is borne out with respect to prefix verbs, as shown in (42): (42) a.

Peter versucht einen Tisch zu bestellen P. tries a table to pref-put ‘Peter tries to order a table’ b. Peter versucht die Stadt zu überfliegen P. tries the city to pref-fly ‘Peter tries to overfly the city’

Prefix verbs are V0s; consequently, they attach to Infl0 as a whole, and when the LPS of zu is “late” inserted, the infinitival element zu precedes the complex verb. However, particle verbs are different. The infinitival prefix zu does not precede the particle verb, but intervenes between the particle and the verb (cf. Booij 1990 and Neeleman & Weerman 1993 for Dutch): (43) a.

Peter versucht die Tür ab-zu-schließen (*zu abschließen) P. tries the door part-to-lock ‘Peter tries to lock the door’ b. Peter versucht sein Bier aus-zu-trinken (*zu austrinken) P. tries his beer part-to-drink ‘Peter tries to drink up his beer’

This so-called “morphological” separability of particle verbs would be hard to explain if particle verbs were words. If the particle and the verb formed a complex V0 in syntax, we would expect that zu precedes the particle verb, as is the case with the prefix verbs in (42). The surprising contrast between prefix verbs and particle verbs with respect to zu follows directly from a syntactic in situ-approach. V0 moves to Infl0, and zu is prefixed to the base verb, while the particle remains in situ and therefore precedes zu in the phonological string (cf. Zeller 1997a): (44) [CP [IP PRO [I¢ [VP die Tür [PrtP ab] Vi] [I0 zu [schließ-]i -en]]]]

According to (44), what is called “morphological” separability is in fact the result of the phrasal syntactic representation of particle verbs. As I have argued in Section 2.1, particle verbs receive the same syntactic analysis as other verb + complement-constructions — notice that resultative adjectives and PP-complements also precede the infinitival prefix:

argument that I make in this section, since according to both views, the infinitival prefix attaches to the left of the (LPS of) the verb that has undergone movement to Infl0.


(45) a.

Peter versucht sich nicht [heiser] zu reden (P. tries refl not [hoarse to talk Peter tries not to talk himself hoarse (resultative predicate) b. Peter versucht das Heu [auf den Wagen] zu laden P. tries the hay [onto the wagon to load ‘Peter tries to load the hay onto the wagon’ (PP-complement)

The similarity between (43) and (45) confirms the in situ-approach. The particle is a phrasal complement of the verb and remains in situ while only the verb moves. In contrast, since prefix verbs are V0s, they move as one complex, and (42) follows directly. Notice that the participial prefix ge- also separates the verbal stem and the particle: (46) a.

Das Heu wurde ab-ge-laden the hay was part-ge-loaded ‘The hay was unloaded’ b. Das Bier wurde aus-ge-trunken the beer was part-ge-drunk ‘The beer was drunk’

Baker & Johnson & Roberts (1989) argue that the morpheme that derives passive participles is associated with Infl0. If we adopt this idea, ge-V-en in (46) would be an Infl0-element, and (46) could be analyzed in the same way as (43). However, this proposal implies that the passive morpheme is a functional element. This assumption is problematic in the light of adjectival participles like those in (47): (47) a.

Der Brief ist [geschrieben]A0 the letter is [written ‘The letter is written’ b. Das Brot ist [geschnitten]A0 the bread is [cut ‘The bread is cut’

The participles in the adjectival passives in (47) are analyzed as adjectives derived from verbal participles (cf. Levin & Rappaport 1986; Kratzer 1994). On the one hand, it is unlikely that derived adjectives include functional material; therefore, ge- in (47) is probably attached to V0 via Merge.11 On the other hand,

11.The assumption that ge- may attach directly to the verb stem is independently needed to explain examples where ge- occurs as a derivational affix (cf. Stiebels & Wunderlich 1994):

73

74


it is not unlikely that the ge-morphemes in (46) and (47) are different elements (or different realizations of the same element, cf. Schoorlemmer 1995). In that case, one could maintain the idea that ge- in (46) is associated with Infl0, and (46) and (43) could be explained along the same lines. How do the morphological approach and the incorporation approach explain morphological separability? Regardless of whether the complex V0 that corresponds to a particle verb is derived by Merge or Move, it is expected that the whole complex verb moves to Infl0, and zu should attach to the left of the particle, yielding the wrong order. A derivation for (43) based on the idea that the particle overtly incorporates into the verb is suggested by Grewendorf (1990), who adopts Kayne’s (1985) Small Clause-analysis. Grewendorf assumes that the object of the particle verb is located inside the particle phrase (the Small Clause); it is θ-marked by the particle. The Small Clause is θ-marked by the verb. Grewendorf suggests that before the particle incorporates, the verb moves and right-adjoins to zu in Infl0 in a first step; (48a). In a second step, the particle follows and adjoins to the left of Infl0, thereby skipping the copy of the moved verb inside the VP; (48b): (48) a.

b.

[CP [IP PRO [I’ [VP [SC die Tür ab] V 0i] [I0 zu [schließ-]i-en] ] ]] [CP [IP PRO [I’ [VP [SC die Tür Prt 0i] V0] [I0 abi [zu schließen]] ]]]

The first problem with this account is that it violates the Head Movement Constraint (Travis 1984), according to which head movement cannot skip intermediate head positions. Furthermore, as Grewendorf himself admits, the movement exhibited in (48b) is a violation of the principle of strict cyclicity. But even if a derivation like (48) was permitted, it remains unclear why the particle has to “wait” until the verb has combined with Infl0 before it combines with the verb. Grewendorf ’s explanation is based on his claim that a θ-role assigning head is not allowed to incorporate into a verb if this verb assigns a θrole to the head’s maximal projection. This rules out incorporation of the

(i)

a. b.

Rumgerede Angequatsche

from rumreden, ‘chat away’ from anquatschen, ‘twaddle’

In the derived nominals in (i), the particle is separated from the verb by ge-. This shows that at least in its derivational use, ge- is part of the verb stem.


particle into the verb, because the verb θ-marks the Small Clause. Instead, the particle is forced to incorporate into Infl0 after the verb has moved to this position. However, Grewendorf ’s claim is not substantiated empirically. As Baker (1988) shows, in languages such as Oneida, Mohawk, or Greenlandic Eskimo, we find examples of noun incorporation where the nominal head of a noun phrase that is θ-marked by the verb is allowed to incorporate into the verb, although this nominal head assigns a θ-role to a possessor-NP inside its maximal projection: (49) Wa-hi-[nuhs]i-ahni:nu: [NP John N0i] Wa-hi-[house-buy J. ‘I bought John’s house’ (lit. ‘I house-bought John’) (Oneida; Baker 1988: 96)

I conclude that Grewendorf’s (1990) explanation for (43) cannot be maintained. Notice that the problems with Grewendorf ’s approach only emerge as a result of step (48b). In contrast, the in situ-approach argues that the derivation of the infinitival form of a particle verb is completed after step (48a). As a comparison of (48b) and (44) reveals, the linear order particle-zu-verb is predicted by both the incorporation approach and the in situ-approach. It is only that the latter avoids the problems that the former automatically encounters. According to the morphological approach, the particle verb is a V0 derived by Merge. On the basis of the claim that zu is associated with Infl0, the morphological approach would predict that it ends up as a syntactic prefix to the left of the whole particle verb. In order to avoid this result, the morphological approach would have to postulate a specific phonological rule that reverses the order of the particle and the infinitival marker at phonological structure. At first glance, such a rule looks quite ad hoc.12

12.Ackerman & Webelhuth (1998: 331) provide additional evidence against any kind of “infixation” process that applies to a complex V0 consisting of the particle and the verb. They note that the participal prefix ge- in German (cf. (46)) only attaches to verbs with main stress on the first syllable. Prefix verbs like beziehen, ‘cover with’, with non-initial main stress omit the prefix, (i). (ii) shows that the ge-prefix is also absent with a particle verb like einbeziehen, ‘consider’, which is derived from the combination of the prefix verb and a particle ein: (i) infinitive: beziehen (ii) infinitive: einbeziehen (iii) *eingebezogen

– –

past participle: bezogen past participle: einbezogen

(cf. ziehen–gezogen)

If the particle verb in (ii) was a V0 that moves to Infl0 as a whole, we would expect that (iii) is the correct past participle form of the verb, because the particle verb bears stress on its initial syllable. However, the fact that the particle verb in (ii) lacks the prefix ge- follows

75

76


However, although the postulation of a rule that changes the order of zu and the particle looks like a mere stipulation, there seems to be evidence that zu can in fact split other complex words as well. This evidence is provided by the properties of so-called backformations. Since it might be used as a potential argument against the in situ-approach, I discuss backformations in the next section. 2.2.2 Backformations In backformations in German, a deverbal compound is reanalyzed as a verb (cf. Stiebels & Wunderlich 1994: 944): (50) verb Æ noun/ adjectival participle

Æ

deverbal compound

Æ

compound Æ verb reanalysis

As an example, consider the derivation of the verb voranmelden, ‘announce beforehand’. The verb [anmelden]V, ‘announce’, is turned into the noun [[Anmeld]V-ung]N, ‘announcement’, which forms the compound [Vor[anmeldung]]N. The non-head member of this compound and the verb are reanalyzed, yielding [[Voranmeld]V-ung]N, and finally, the backformed verb [[voran]melden]V is derived. Backformation converts a complex nominal into a verb; it therefore has been concluded that the derived verb must be a complex V0. But now consider (51): (51) a.

Sie versuchten, ihre Bestellung voran-zu-melden they tried their reservation voran-to-announce ‘They tried to announce their reservation beforehand’ b. Sie versuchten das Stück urauf-zu-führen they tried the piece urauf-to-perform ‘They tried to give the first performance of the piece’

(51) shows that backformed verbs are split by the affix zu in the same way as particle verbs are. If backformed verbs are V0s, then the intervention of the infinitival prefix cannot be taken as evidence that particle verbs are syntactic constructions.

directly from the in situ-approach. The base verb beziehen moves to Infl0 and is inflected in the same way as the prefix verb in (i), whereas the particle remains in its left-adjacent base position inside the particle phrase.


However, the analysis of backformed verbs that I offer approaches the problem from a different perspective. I take the examples in (51) as evidence for the contrary assumption, i.e. that backformed verbs are not V0s, but phrasal constructions like particle verbs. I have shown in Section 2.2.1 that morphological separability of complex verbs follows from the assumption that zu is located in Infl0 and combines with the base verb, while the verbal modifier is stranded inside its phrasal projection. Since backformed verbs are also morphologically separable, I draw the conclusion that in examples like (51), the non-head elements voran and urauf project phrases as well. I suggest that (51b) has a structure like (52): (52) [CP [IP PRO [I¢ [VP das Stück [PrtP urauf] Vi] [I0 zu [führen]i]]]]

I follow Stiebels & Wunderlich (1994) in assuming that the elements voran and urauf in (51) are reanalyzed as particles. But what does it mean to reanalyze an element as a particle? The answer depends of course on the theory that one has about particles. I suggest that particles project syntactic phrases; consequently, I assume that reanalyzing the non-head of a backformed verb as a particle means that this element now projects a phrase. If a speaker changes the mental representation of a compound nominal to a backformed verb, she assigns the respective element a different syntactic structure. This explains the examples in (51); only the verb moves to Infl0, and zu attaches to its left. The reanalyzed non-heads are stranded, exactly like particles (in fact, they are particles in (51)), and a similar pattern with respect to morphological separability follows. The proposal illustrated in (52) also explains why a backformed verb cannot move to Comp0 as a whole: (53) a.

weil sie ihre Bestellung voranmeldeten because they their reservation voran-announced ‘because they announced their reservation beforehand’ b. *Sie voranmeldeten ihre Bestellung

(54) a.

weil sie das Stück uraufführten because they the piece urauf-performed ‘because they gave the first performance of the piece’ b. *Sie uraufführten das Stück

If backformed verbs were really V0s, the ungrammaticality of (53b) and (54b) would remain unexplained. In contrast, if the non-verbal parts of the verbs in (53) and (54) are reanalyzed as particles, such that they project phrases, the impossibility of V2 follows directly.

77

78


However, this proposal faces a problem. On the one hand, in order to account for (51), the internal structure of backformations must be similar to that of particle verbs. On the other hand, however, backformations are different from particle verbs, because V2 with backformed verbs is entirely impossible in most cases (cf. Stiebels & Wunderlich 1994: 946): (55) a. *Sie meldeten ihre Bestellung voran b. *Sie führten das Stück urauf

If backformed verbs are syntactically similar to particle verbs, we expect V2 to be possible. So why are the sentences in (55) ungrammatical? The hypothesis that is frequently advocated in the literature is that backformed verbs are defective because they must still maintain a close relation to the corresponding noun. For example, McIntyre (2001) notes that the acceptability of a backformed verb increases if its occurrence is “formally similar” to the source nominal. Of course, the syntactic representation of the backformed verb that I propose in (52) is formally completely different from the representation of the source noun (which is a complex N0). However, the tripartite model of grammar allows it to draw a distinction between conditions that govern the syntactic representation of an element and conditions that hold with respect to its phonological form. I assume that “formal similarity” is only relevant with respect to the phonological representation of backformed verbs. It is a particularity of backformations that at phonological structure, the backformed verb must be represented in such a way that it still resembles the phonological form of the nominal compound from which it is derived.13 Importantly, if an element is reanalyzed as a particle that projects a phrase, the phonological string is basically left unaffected in infinitives; in (52), urauf still precedes the verb. Even though reanalysis leads to a linear order where the backformed verb is split by zu, the infinitival form of a backformed verb remains formally similar

13.The following contrast provides further evidence for this assumption: (i) (ii)

N0: Uraufführung Infinitive: uraufzuführen N0: Wettrennen Infinitive: ?wettzurennen

–

Past Tense: uraufführten

–

Past Tense: *wettrannten

As (i) shows, both the zu-infinitive and the past tense form of uraufführen are grammatical. However, although the zu-infinitive of the backformed verb wettrennen, ‘run a race’, is also grammatical, the past tense form is ungrammatical. This is due to the Ablaut-process that changes the root of the verb rennen and hence leads to a phonological form which is different from the form of the underlying noun.


to a source noun like Uraufführung because the linear order is preserved (the verbal form and the particle are only separated by a clitic). Of course, the situation changes in V2. Verb movement to Comp0 would completely destroy the similarity between the nominal compound and the backformed verb: (56) [CP sie [führten]i/j [IP [I¢ [VP das Stück [PrtP urauf] Vi] [Vi+Infl0j]i/j]]]

In (56), the verbal part of the backformed verb is not even adjacent to the particle element. Since backformed verbs do not tolerate a phonological structure like this, (56) is ungrammatical. It would be interesting to test the relevance of linear order for V2 with backformed verbs by topicalizing the particle phrase in (56). The resulting order would be (57): (57) [CP [PrtP urauf]i [führten] [IP sie [I¢ [VP das Stück [PrtP]i V] [V+Infl0]]]]

In (57), the particle-like element urauf ends up in front of the verb. Therefore, the phonological string matches with the linear order of the noun Uraufführung, and (57) should be grammatical. Unfortunately, topicalization of the particle phrase in (57) is excluded for independent reasons (which will be discussed in Section 2.4), and (57) cannot be used to test the explanation I offer above. In a nutshell, backformations are licensed under two conditions. Their nonheads must be reanalyzed as particles that project phrases, and the syntactic representation must create a phonological string that mirrors the phonological structure of the original nominal compound. Since both requirements cannot be met at the same time in V2, this option is excluded for backformed verbs.14 According to the view defended here, backformed verbs and particle verbs have the same syntactic representation, which explains their parallel behavior with respect to morphological separability. Particles and the non-heads of backformations are heads of verbal complements and do not move. Syntactically, neither particle verbs nor backformed verbs are words; therefore, verb movement that separates the particle element and the verbal stem does not

14.Stiebels & Wunderlich (1994: 946) note that in some rare cases, backformed verbs allow V2. In examples like Sie turnte ihre Übung probe, ‘she did her exercise for trial’, the condition that the backformed verb and the nominal compound Probeturnen must be formally similar has been overwritten, and probe has become a full particle. In examples like Er staubsaugt, ‘he sucks dust (he vacuums)’, the non-head of the compound Staubsauger, ‘vacuum cleaner’, has not been reanalyzed as a particle, but as a prefix.

79

80


count as a violation of Lexical Integrity. However, the special condition that the phonological structure of backformed verbs must resemble the phonological representation of the source noun as closely as possible prevents backformed verbs from being entirely split by V2. Since no comparable condition constrains the phonological representation of particle verbs, the base verb of a verbparticle construction can move to Comp0, and the difference between particle verbs and backformed verbs with respect to V2 follows.15 2.2.3 Particle verbs and finite Infl Thus far, I have used the separability of particle verbs as evidence in favor of the syntactic in situ-approach. But what about examples like (58), where the finite particle verb remains in sentence-final position? (58) a.

weil Peter in den Bus einsteigt because P. in the bus part-climbs ‘because Peter gets on the bus’ b. weil Peter die Nachricht aufschreibt because P. the note part-writes ‘because Peter takes a note’

In contrast to the examples discussed in Sections 2.1 and 2.2.1, the particle and the verb in the examples in (58) are adjacent. Are these data compatible with the in situ-approach as well? I have argued that a sentence like (58b) is represented through a structure as in (59): (59) [C¢ weil [IP Peter [I¢ [VP die Nachricht [V¢ [PrtP auf] V0i]] [I0 schreibi-t]]]]

A potential problem is raised by the strong adjacency requirement that is attested with particle verbs in sentence-final position. According to the in situ-approach, there are two maximal projections (namely, PrtP and VP) that intervene between the verb and particle. The structure in (59) hence seems to predict that extraposed phrases that right-adjoin to e.g. VP can appear between a particle and the verb. However, (60c) shows that this is impossible:

15.As Ian Roberts (p.c.) has pointed out to me, the special phonological condition that holds for backformed verbs could be looked at as a phonological version of Lexical Integrity (if Lexical Integrity is more generally understood as a condition that disallows the splitting of both syntactic and phonological words).


(60) a.

dass Peter das Heu mit der Heugabel ablädt that P. the hay with the fork part-loads b. dass Peter das Heu ti ablädt [mit der Heugabel]i that P. the hay part-loads [with the fork c. *dass Peter das Heu ti ab [mit der Heugabel]i lädt that P. the hay part [with the fork loads ‘that Peter loads down the hay with a fork’

The PP mit der Heugabel, ‘with the fork’, must either precede or follow the particle verb, but it cannot intervene between the two nodes. This has been taken as evidence that particle verbs are words, since the internal structure of a V0 does not provide a landing site for an extraposed phrase. Of course, (60c) can only be used as an argument against the in situ-approach if extraposition to VP in verb-final sentences is otherwise possible. However, as argued in Büring & Hartmann (1997), extraposition and adjunction to VP is impossible as long as the VP stays in situ and does not move out of the governing domain of Infl0, as in the case of VP-topicalization. Extraposed phrases must therefore always be adjoined to the right of IP. Notice that extraposed constituents cannot intervene between other nonminimal secondary predicates and the verb, either: (61) a. *dass Peter das Bild ti fertig [in seinem Zimmer]i malte that P. the picture ready [in his room painted ‘that Peter finished the painting in his room’ b. *dass Peter ti nach Hause [mit seinem Hund]i ging that P. to home [with his dog went ‘that Peter went home with his dog’ (German) (62) a. *dat hij probeert de emmer ti leeg [met een lepel]i te scheppen that he tries the bucket empty [with a spoon to scoop ‘that he tries to scoop the bucket empty with a spoon’ (Dutch, van Riemsdijk 1998a: 640) b. *dat Jan ti de sloot in [gisteren]i sprong that J. the ditch into [yesterday jumped ‘that Jan jumped into the ditch yesterday’ (Dutch, Zwart 1994: 398)

The data in (61) and (62) of course cannot be explained by assuming that the resultative APs in (61a) and (62a) and the directional PPs in (61b) and (62b) form a complex V0 with the verb. There must be another reason why (61) and (62) are ungrammatical, but if this reason is found, it can also be used to explain the

81

82


ungrammaticality of (60c). In Zeller (1997a), I assume, following Truckenbrodt (1995), that extraposition is phonologically constrained, such that phonological phrases cannot be separated by extraposed material. Since in both (60) and (61)–(62), the verb and its respective complement phrase form a phonological phrase, extraposition must move the PPs in (60)–(62) to the right boundary of this phrase. Hence, the PPs must right adjoin to IP, as in the grammatical example (60b). This means that (60c) is not excluded because of the syntactic structure of the particle verb; this example does not contradict the in situ-approach. The following data provide further evidence that what moves to Infl0 in (58) is only the verbal part of the particle verb. Booij (1989) notes that certain non-separable complex verbs in Dutch are inflected regularly, whereas a particle verb always adopts the verb’s irregular inflection (cf. also Paulissen & Zonnefeld 1988; Bennis 1992; Neeleman 1994):16 (63)

zuigen zoog suck sucked

(64) a.

stofzuigen dust-suck b. stofzuigen dust-suck

(65) a.

gezogen sucked

*stofzoog dust-sucked stofzuigde dust-sucked

uitzuigen uitzoog out-suck out-sucked b. uitzuigen *uitzuigde out-suck out-sucked (Neeleman 1994: 294)

(base verb) *stofgezogen (non-separable prefix verb) dust-sucked gestofzuigd dust-sucked uitgezogen out-sucked *uitgezuigd out-sucked

(particle verb)

The complex verb in (64) is a prefix verb; it moves to Infl0 as a whole. Therefore, the Vocabulary entries that are assigned to the terminal nodes below Infl0 may differ from the entries that are assigned to the base verb in isolation. For example, the LPS of Infl0 may be regular in the context of a prefix verb, although the base verb triggers the insertion of an irregular stem (cf. (63) vs. (64b)). In contrast, in (65a), the verb adopts the irregular inflection of the base verb.

16.In Dutch, there is a number of prefix verbs that are obligatorily or optionally inflected as weak verbs and differ from the inflectional class of the simplex verb, but there are also prefix verbs that are inflected like their base verbs. In German, the inflection class of the base verb is always preserved with prefix verbs (cf. Toman 1983; Lieber 1992; Stiebels & Wunderlich 1994). The following argument can therefore only be constructed for a certain type of prefix verbs in Dutch.


This follows from my claim that only the base verb moves to Infl0; the input to Vocabulary Insertion in (65) is the same as in (63). The in situ-approach correctly predicts that the verbal part of (65b) is inflected in the same way as the verb in (63). To conclude, the discussion in this section has shown that the behavior of particle verbs with respect to morphological separability and inflection follows directly from the idea that the particle is a syntactically independent head that remains in its base position when the verb moves. In explaining the difference between particle verbs and prefix verbs, the in situ-approach has proved to be superior to both the morphological and the incorporation approach. The in situ-approach does not draw a distinction between different kinds of morphological structures. Instead, the contrasting properties of prefix verbs and particle verbs are simply the result of the difference between words and phrasal constructions.

2.3 Gapping The next argument in favor of a syntactic in situ-approach comes from Gapping constructions (for a recent analysis of Gapping, cf. Hartmann 2000). The rule of Gapping deletes the (non-leftmost) occurrences of finite verbs in conjoined sentences. For Gapping to be licensed, the non-deleted constituents of the VP in which deletion occurs must be non-identical to the corresponding constituents in the leftmost conjunct. Apart from the subject, the non-deleted constituent is a PP in (66b), (67a), an NP in (67b), and an AP in (66a): (66) a. Max spoke fluently, and Albert (spoke) haltingly b. Max eats with chopsticks, and Albert (eats) with a fork (Hankamer 1973: 18) (67) a.

weil Peter auf den Berg steigt und Maria ins Tal because P. up the mountain climbs and M. into.the valley (steigt) (– ‘because Peter climbs up the mountain and Mary down into the valley’ b. weil Peter ein Buch liest und Klaus die Zeitung (liest) because P. a book reads and K. the newspaper (– ‘because Peter is reading a book and Klaus is reading the newspaper’

83

84


In addition to the finite verb, Gapping can also delete more constituents together with the verb, like, for example, the direct object: (68) a.

weil Hans das Buch auf den Tisch legt und Maria (das Buch) in because H. the book on the table puts and M. (– in den Schrank (legt) the cupboard (– ‘because Hans puts the book on the table and Mary in the cupboard’ b. weil Hans das Fleisch in der Pfanne brät und Maria because H. the meat in the pan fries and M. (das Fleisch) im Topf (brät) (– in.the pot (– ‘because Hans fries the meat in the pan and Mary fries it in the pot’

Crucially, Gapping must delete the whole finite verb. A part of a complex finite verb cannot be deleted: (69) a. *weil Jan übertreibt und Hans unter(treibt) because J. pref(over)-forces and H. pref(under) (– ‘because Jan exaggerates and Hans understates’ b. *weil Max die Franzosen überschätzt und Jan die because M. the French pref(over)estimates and J. the Brasilianer unter(schätzt) Brazilians pref(under) (– ‘because Max overestimates the French and Jan underestimates the Brazilians’ c. *weil Martin den Wald durchfährt und Hans die Stadt because M. the forest pref(through)drives and H. the city um(fährt) pref(around) (– ‘because Martin drives through the forest and Hans around the city’

In (69a), the verbal base of the P-prefix verb untertreiben, ‘understate’, has been deleted, as have the verbal parts of unterschätzen, ‘underestimate’, in (69b) and of umfahren, ‘drive around’ in (69c). The deleted verbal part is identical to the corresponding part of the verb in the leftmost conjunct; all non-deleted constituents, including the P-prefix, are non-identical to their counterparts.


Nevertheless, the data in (69) are ungrammatical.17 Now consider (70): (70) a.

weil Peter einsteigt und Hans aus (steigt) because P. part(in)-climbs and H. part(out) (– ‘because Peter gets in and Hans gets off’ b. weil Karl die Tür aufmacht und Fritz das Fenster because K. the door part(open)-makes and F. the window zu (macht) part(close) (– ‘because Karl opens the door and Fritz shuts the window’ c. weil Karl seine Freundin anlacht und Maria den Lehrer because K. his girl-friend part-laughs and M. the teacher aus (lacht) part (– ‘because Karl smiles at his girl-friend and Mary laughs at the teacher’ d. weil Fritz das Bild aufhängt und Maria das Poster ab because F. the picture part-hangs and M. the poster part (hängt) (– ‘because Fritz hangs the picture on the wall and Mary takes the poster off’ e. weil Franz Klavier spielt und Maria Geige (spielt) because F. part(piano) plays and M. part(violin) (– ‘because Franz plays the piano and Mary plays the violin’

(70a–d) show prepositional particle verbs; (70e) is a nominal particle verb. As the examples show, Gapping of the finite verb in the verb-particle construction

17.The unacceptability of the examples in (69) illustrates an interesting difference between Gapping and Right Node Raising. In Right Node Raising, parts of complex verbs can be deleted: (i)

a.

b.

Peter hat Hans erst über-(schätzt) und dann unterschätzt P. has H. first prefand then pref-estimated ‘Peter first overestimated Hans and then underestimated him’ Peter hat den Wagen erst ent-(laden) und dann wieder beladen P. has the wagon first pref and then again pref-loaded ‘Peter first unloaded the wagon and then reloaded it’

If some “deletion-under identity”-contexts allow deletion of parts of a word, it must be seen as a specific aspect of the rule of Gapping that deletion of parts of words is not allowed.

85

86


without deletion of the particle is grammatical.18 The examples in (70) contrast with the examples in (69) and therefore provide strong evidence in favor of my claim that the particle projects a phrase and is located in its base position in syntactic structure. Deletion of only the finite verb is therefore possible for the same reason that the data in (67) are grammatical. In contrast, if the particle and the verb were parts of the same V0, as the morphological approach and the incorporation approach claim, we would expect (70) to be ungrammatical, in the same way as the examples in (69) are. The contrast between the examples in (70) and the ungrammatical sentences in (69) is a strong argument for the syntactic status of particles. To my knowledge, the Gapping data from German have not been addressed by proponents of the morphological approach. Nevertheless, we have to ask if it is possible to account for the contrast between (69) and (70) on the basis of the claim that the particle verb is a complex V0. Recall from the discussion in Section 2.1.2 that Neeleman (1994) accounts for the difference between particle verbs and inseparable prefix verbs by assigning these elements different morphological structures:

18.Surprisingly, Höhle (1982) claims that Gapping of the finite part of the German particle verb was impossible for many speakers. This opinion is neither confirmed by my own judgement nor by the judgements of my informants. The data in (70) are definitely acceptable. It is possible, however, that Höhle does not take into account the fact that gapping of the verb in verb-particle constructions is subject to semantic conditions similar to those that restrict topicalization of the particle (cf. the next section). Gapping becomes progressively more acceptable the more transparent the meanings of the respective particle verbs are (cf. Le Roux 1988 who observes the same thing about gapping of the verb in verb-particle constructions in Afrikaans). However, Fraser (1976) and Johnson (1991) note that Gapping of the finite verb in English verb-particle constructions is generally impossible. When I consulted speakers about the following transparent particleverbs, theyfound themat mostmarginally acceptable: (i)

a. ??John climbed in and Mary out. b. ?Mary looked up and Peter down. c. ?*John turned the TV set on and Mary the radio off.

I do not have an explanation to offer for the contrast between German (and Afrikaans) and English.


(71) a.

prefix verb (cf. (69a)) (stem compound)

b.

V0 Pref -1 über

particle verb (cf. (70a)) (word compound)

V0

V -1 treib-

Prt0 ein

V0 steig-

Is it possible to attribute the difference between (69) and (70) to different morphological structures of the complex verbs? Advocates of a morphological approach would have to argue that Gapping can delete a part of a complex head if the remaining part is still an X0-element, and not an X−1-element. Since according to (71), particles are words, but the P-prefixes in (69) are stems, only particle verbs would allow Gapping. However, this explanation is not adequate, as it makes false predictions. Recall that Gapping allows for the deletion of more constituents than the finite verb (cf. (68)). If parts of a word could be deleted as long as the remnant is still an X0, we would expect that a part of a nominal compound that is the object of the verb can also be deleted together with the finite verb. However, (72c) shows that this is impossible: (72) a.

weil Hans die Plastiktüte auf den Tisch legte und Peter die because H. the plastic bag on the table put and P. the Papiertüte in den Schrank (legte) paper bag in the cupboard (– ‘because Hans put the plastic bag on the table and Peter put the paper bag in the cupboard’ b. weil Hans die Plastiktüte auf den Tisch legte und Peter because H. the plastic bag on the table put and P. (die Plastiktüte) in den Schrank (legte) (– in the cupboard (– ‘because Hans put the plastic bag on the table and Peter put it in the cupboard’ c. *weil Hans die Plastiktüte auf den Tisch legte und Peter die because H. the plastic bag on the table put and P. the Papier(tüte) in den Schrank (legte) paperin the cupboard (–

In (72a), only the finite verb has been dropped. In (72b), the finite verb and the direct object (which is identical to the direct object of the left conjunct) have

87

88


been deleted. As (72c) shows, however, deletion of the finite verb and the head of the nominal compound is impossible, although the head of this compound is identical to the head of the compound in the first conjunct. Importantly, the remnant part of the compound in (72c) is an N0.19 Therefore, the ungrammaticality of (72c) proves that parts of words can never be deleted in Gapping constructions, even if the remaining part is still an X0-element. The contrast between the ungrammatical sentences in (69) and the grammatical examples in (70) can only be explained by the syntactic in situ-approach.20

2.4 Topicalization In contrast to what is sometimes claimed in the literature, many particles in German and Dutch can be topicalized. In Section 2.4.1, I discuss the relevant data that illustrate this possibility. Since only syntactic phrases can be topicalized, these data provide a strong argument in favor of a syntactic approach. In Section 2.4.2 I take a closer look at various semantic conditions that seem to affect the possibility of moving the particle phrase to SpecCP. Although that section essentially includes descriptive remarks, the discussion will reveal some interesting constraints on topicalization of particles which (to my knowledge) have not been noted before.

19.Notice that the gender of the nominal compound Papiertüte, ‘paper bag’, is determined by the head Tüte, ‘bag’. According to Neeleman’s (1994: 296f.) definition, this property makes Papiertüte a word compound. 20.However, Kyle Johnson (p.c.) pointed out to me that there is another way in which the morphological approach could be saved. The Gapping data can be explained on the assumption that particle verbs are V0s if the following three assumptions are adopted as well: (i) (ii) (iii)

The verbal part of the particle verb can be moved out, violating Lexical Integrity. In verb-final sentences, the verbal part of the particle verb moves to a functional head above VP (e.g. Infl0) and strands the particle. A V0 including a trace/copy of the moved verb does not have to be deleted with the moved verb.

If the finite verb in the examples in (70) has moved to Infl ((i) and (ii)), it strands the particle inside the head of VP. If we now also adopt the proposal (iii), then the data are explained as well. However, in the light of the problems that come with (i) and (ii) that I discussed in Sections 2.1 and 2.2, I reject the alternative explanation proposed by Johnson.


2.4.1 The syntactic side of particle topicalization The examples in (73)–(76) illustrate that particles in German can be topicalized (see Hoeksema 1991 and Bennis 1991 for Dutch): (73) a.

(An der Haltestelle stiegen hübsche Frauen ein). (at the bus-stop climbed pretty women part(in) [Aus] stiegen nur Männer part(out) climbed only men ‘At the bus-stop, pretty women got in. Only men got off’ b. [Ganz ab] hat er sich bloß die Haare geschnitten [completely part(off) has he refl just the hair cut ‘He completely cut off only his hair’

(74) a.

[Klavier] konnte Peter noch nie spielen [part(piano) could P. still never play ‘Peter never knew how to play the piano’ b. [Auto] fahre ich nicht gern, aber Fahrrad [part(car) drive I not with-pleasure but part(bicycle) ‘I don’t like driving a car, but I like riding a bicycle’

(75) a.

[Zu] hat er die Tür gemacht [part(close) has he the door made ‘He locked the door’ b. [Weg] ist er gegangen [part(away) is he gone ‘He left’(Grewendorf 1990)

(76) a.

[Auf] geht die Sonne im Osten (aber [unter] geht sie [part(up) goes the sun in.the east (but [part(down) goes it im Westen) in.the west ‘The sun rises in the east but sets in the west’ b. [Sehr nahe] ging Max der Tod seines Kanarienvogels [very part(close) went M. the death of.his canary ‘The death of his canary upset Max very much’ (Lüdeling 1998b)

The topicalized particles in (73)–(76) have moved to SpecCP. This syntactic process is only available for full phrases; accordingly, these examples suggest

89

90


that particles must be dominated by a maximal projection.21 The morphological approach fails to account for the data in (73)–(76), because even if particles could be moved out of complex V0s, they would still remain heads, and topicalization would be predicted to be impossible. The same holds for the incorporation approach — if the particle must combine with the base verb in overt syntax, it cannot be topicalized with its phrase. (73)–(76) therefore provide further evidence for the syntactic in situ-approach; only if the particle remains in its base position inside the PrtP can it be moved to SpecCP. Interestingly, however, topicalization is usually presented as an argument in favor of the view that the particle and the verb form a complex V0. Consider (77): (77) a. *Aus hat das Mädchen gut gesehen part has the girl good seen ‘The girl looked good’ b. *Auf hat Peter mit dem Trinken gehört part has P. with the drinking heard ‘Peter stopped drinking’ c. *Ein ist der Dieb gebrochen part is the thief broken ‘The thief broke in’

(77) shows that not all particles can be topicalized. Therefore, authors like Neeleman & Weerman (1993), Stiebels & Wunderlich (1994), and Neeleman (1994), among others, have used data like those in (77) as evidence against a syntactic treatment of particle verbs. However, this argument is inconclusive, because it is based on the assumption that syntactic movement can only be excluded syntactically. Since only phrases can be topicalized, it has been argued that elements that do not allow topicalization cannot be phrases. This is of course an invalid conclusion —

21.According to Grewendorf (1990: 96), certain particles can also be topicalized together with the direct object: (i)

[Die Tür zu] hat er gemacht [the door part has he made ‘He has closed the door’

However, this (slightly marginal) possibility is restricted to a very few constructions. It is possible that zu in (i) is reanalyzed as a resultative adjective; in contrast to particle phrases, I assume that the direct object of resultatives is located inside the phrasal projection of the resultative predicate (see Chapter 3, Section 3.2.3) and hence can be moved to SpecCP together with the predicate.


although XP-status of the moved element is a necessary condition for topicalization, it is by no means a sufficient condition. The fact that an element cannot be topicalized does not imply that it cannot be an XP. The following contrast illustrates this point:22 (78) a. *[Andato a casa]i, Gianni è ti [gone home G. has (lit. is) ‘Gianni has gone home’ b. [Andato a casa], Gianni non è ti [gone home G. not has (lit. is) ‘Gianni has not gone home’ (Cinque 1991: 84f.)

(78) shows that VP-preposing in Italian is contingent on the presence of negation. It would now be very odd to argue that, because (78a) is ungrammatical, andato a casa in (78a) cannot be a VP. It is clear that the ungrammaticality of (78a) has nothing to do with the syntactic status of the moved constituent, but rather must be related to the fact that (78a) is not negated. The following example suggests that the ungrammaticality of (78a) follows from semantic rather than syntactic properties of the sentence. As Cinque (1991) points out, mere presence of a negation that c-commands the trace of the fronted VP does not suffice to license VP-preposing: (79) *[Andato a casa]i, Gianni non è ti in macchina, ma a piedi [gone home G. not has by car but by foot ‘Gianni has not gone home by car, but by foot’ (Cinque 1991: 85)

In (79), negation c-commands the trace of the VP, as in (78b). However, the VP in (79) is not understood in the scope of the negation and therefore does not receive contrastive focus. In this respect, (79) contrasts with (78b), where the moved VP is interpreted as being in the scope of the negation. This difference suggests that VP-preposing is only licensed under a particular interpretation of a negated VP. If this is in fact the case, then the data in (78) and (79) show that syntactic movement operations are also conditioned by semantic factors. A similar point is made by Grimshaw and Mester (1988) in their discussion of Japanese light verb constructions. They show that the Japanese light verb suru subcategorizes and case-marks a direct object NP without assigning it a θrole. According to Grimshaw and Mester, this object NP is θ-transparent; the

22.Thanks to Ian Roberts for making me aware of the Italian data.

91

92


nominal realizes its arguments outside the NP and makes them arguments of the light verb: (80) John-ga [CP ookami-ga kuru-to] hookoku-o shita John-nom wolf-nom come report-acc suru ‘John reported (lit. made a report) that the wolf was coming’ (Grimshaw & Mester 1988: 208)

In (80), the θ-transparent nominal hookoku is assigned accusative case by the light verb and hence is a full NP. As Grimshaw and Mester note, normal NPs can freely be topicalized in Japanese. However, in (80), the CP-argument of the accusative NP is turned into an argument of the light verb. According to Grimshaw and Mester, this semantic operation makes it impossible to topicalize the NP hookoku: (81) *Hookoku-wa John-ga [ookami-ga kuru-to] shita report-top John-nom [wolf-nom come suru (Grimshaw & Mester 1988: 208)

The accusative NP in (81) cannot be topicalized because it is not a semantically independent element. Although the NP fulfills the syntactic condition for topicalization, the semantic properties of the Japanese light verb construction prevent this phrase from being moved. I conclude from the data in (78)–(81) that there is no reason to believe that the possibility of topicalizing certain particles can only be excluded on syntactic grounds. If the ungrammaticality of examples like (77) was really caused by structural constraints, then the grammaticality of the data in (73)–(76) would be hard to explain. Therefore, I instead follow Bayer (1993), Hinterhölzl (1996), Zeller (1997b), Wurmbrand (1998), and Lüdeling (1998a, b), and assume that the incapability of certain particles to undergo topicalization follows from semantic rather than syntactic properties of the relevant constructions. In the next section, I take a closer look at the semantic conditions that license the topicalization of particles. Stiebels (1996), a proponent of the morphological approach, acknowledges that some particles can be topicalized and therefore must have the status of a “syntactic predicate” (I suspect that her notion “syntactic” predicate is tantamount to “phrasal” predicate). However, she maintains her assumption that particles are usually parts of words, but suggests that only those particles that allow topicalization are phrases. This is of course a highly unwelcome result, because it denies the possibility of giving a uniform syntactic account for all


particles (the same critique must be formulated with respect to similar proposals made by le Roux 1988: 248 and Neeleman & Weerman 1993: 471, note 21). Stiebels’ proposal blurs the distinction between semantic and morphosyntactic properties of particle verbs. In order to maintain a syntactic explanation for the impossibility of topicalizing certain particles, she is forced to account for the semantic differences that cause some particles to behave differently than others through distinct structural representations. However, as the Japanese light verb examples have shown, topicalization of an element can be excluded on semantic grounds alone; this does not mean that this element is structurally different from other XPs that allow topicalization. Moreover, I show below that it is not always the semantic properties of the verb or the particle, but properties of the discourse that make it possible for a particle to be topicalized. It would be very strange to assume that the syntactic status of a particle depends on information provided by the context. 2.4.2 Conditions on particle topicalization Bayer (1993) and Wurmbrand (1998) assume that topicalization is restricted to phrases that are contrastively focused.23 This means that topicalization of a particle requires the existence of an alternative that the particle can be contrasted with. For example, the particle auf in (76a) has the meaning ‘up, upwards’ and can therefore be contrasted with the alternative unter, ‘down, downwards’. Consequently, auf can be topicalized. However, the meaning of the particle verb aufhören, ‘stop’, in (77b) is completely idiomatic; therefore, the meaning of auf cannot even be defined in isolation, let alone be contrasted with an alternative. Therefore, topicalization of auf in (77a) is impossible.

23.The examples (73)–(76) differ with respect to the semantic status of the topicalized particle. In (75), the particle is the Focus of the sentence. The sentences in (73), (74), and (76), however, are Topic-Focus-constructions; indicated by a rising pitch accent on the Topic and a falling accent on the Focus, as shown in (i) (cf. Büring 1996). The topicalized particles in these examples are the Topics of these sentences: (i)

/AUS stiegen nur MÄNner\

(cf. (73a))

The semantic difference between Focus and Topic is extensively discussed in Büring (1996). What is crucial here is that a contrastive interpretation of the particle is a necessary condition for topicalization, regardless of whether the particle is the Topic or the Focus. Since the further differences between Topic and Focus are of no relevance here, I will ignore this difference.

93

94


The ability of a particle to be topicalized further depends on the semantic properties of the whole particle verb. Stiebels (1996) observes the following contrast: (82) a.

Auf hat sie die Tür gemacht part has she the door made ‘She opened the door’ b. *Auf hat er das Schloß gebrochen part has he the lock broken ‘He broke open the lock’ (Stiebels 1996: 161)

Apart from the particle verb aufmachen, ‘open’, in (82a), there is the verb zumachen, ‘close’, that can be contrasted with aufmachen. As a consequence of this contrast, the particle zu counts as a possible alternative to auf, and auf can be topicalized. However, since there is no particle verb *zubrechen which could be contrasted with aufbrechen, ‘break open’, there is no particle that auf in (82b) could be contrasted with, and topicalization is not possible. This observation also accounts for the contrast between (73a), repeated here as (83a), and (83b): (83) a.

(An der Haltestelle stiegen hübsche Frauen ein). (at the bus-stop climbed pretty women part(in) Aus stiegen nur Männer part(out) climbed only men ‘At the bus-stop, pretty women got in. Only men got off’ b. (Die Blumen sind eingegangen.) *Aus ging nur das (the flowers are part-gone part went only the Dienstmädchen. maid ‘The flowers have died. Only the maid went out’

The particles ein and aus can only be contrasted in the context of (83a), because the particle verb einsteigen, ‘get in’, contrasts with aussteigen, ‘get off’. In contrast, the idiosyncratic verb eingehen, ‘melt, die’, in (83b) does not contrast with ausgehen, ‘go out’. Therefore, there is no alternative that aus in (83b) could be contrasted with, and the particle cannot be topicalized.24

24.Interestingly, (83b) is not so bad if it is meant as a play on words; it could be uttered by a comedian or by the owner of the flowers who sarcastically mentions that the maid has gone out instead of looking after his plants. The possibility to use an odd case of topicalization as


It is interesting to note that a particle can count as an alternative to a topicalized particle even if its own semantic contribution to the meaning of a particle verb cannot be clearly defined. This can be illustrated by the following example from Lüdeling (1998a): (84) (Lachst du mich aus?) Nein, an lache ich dich (laugh you me part no part laugh I you ‘Are you laughing at me? No, I’m smiling at you’ (Lüdeling 1998a: 57)

The particle an in (84) has the meaning “the event (expressed by the base verb) is directed towards an individual” (cf. Stiebels 1996 and Chapter 4). It is one of the most productive particles in German. The particle verb anlachen, ‘smile at’, can therefore be analyzed as being derived compositionally. However, although the meaning of lachen, ‘laugh’, in auslachen, ‘laugh at’, is regular, the meaning of aus in (84) is not related to any meaning that this particle contributes in transparent verb-particle combinations. So we actually would not expect that aus can be contrasted with an. However, we can certainly contrast the whole particle verb auslachen with the complex verb anlachen. This is sufficient to make aus an appropriate alternative for an in the context of the base verb lachen; consequently, an can be topicalized. However, as Günther Grewendorf (p.c.) has pointed out, the idea that particle topicalization is possible if the particle verb can be contrasted with another particle verb derived from the same base verb does not yet explain the ungrammaticality of (85): (85) a. *An hat er den Pullover gezogen part has he the sweater pulled ‘He put on the sweater’ b. *Aus hat er den Pullover gezogen part has he the sweater pulled ‘He took off his sweater’

The two verbs anziehen, ‘put on’, and ausziehen, ‘take off’, can certainly be contrasted with each other. Nevertheless, topicalization of the particle is

a play on words further shows that these examples are in fact semantically, and not syntactically, defective.

95

96


impossible.25 Notice that the meaning of the base verb ziehen in (85) is different from its regular, more literal, meaning ‘pull’. If the verb maintains this regular meaning in combination with a particle, this particle can be topicalized: (86) Zu hat er die Tür gezogen part(close) has he the door pulled ‘He pulled the door close’

The contrast between (85) and (86) suggests that it is the special meaning of the verb ziehen in (86) that makes particle movement impossible. In Chapter 5 I will come back to this point, arguing that the meaning of ziehen in (85) is only licensed if the verb and the particle stand in a particular local relation. This relation is destroyed if the particle undergoes topicalization. Since the regular meaning of ziehen in (86) does not depend on the presence of a particle, topicalization is possible. Topicalization of particle phrases does not always require that there is an actual alternative to the particle. For example, the sentences in (73b) and (76b), repeated in (87), include particles modified by adverbs (see also the next section): (87) a.

Ganz ab hat er sich bloß die Haare geschnitten completely part(off) has he refl just the hair cut ‘He completely cut off only his hair’ b. Sehr nahe ging Max der Tod seines Kanarienvogels very part(close) went M. the death of.his canary ‘The death of his canary upset Max very much’

The alternatives to the topicalized phrases in (87) do not have to include different particles. The topicalized constituents in (87) may also be contrasted with phrases which include the same particles, but different adverbial modifiers (cf. halb ab, ‘half off’; weniger nahe, ‘less close’). In this way, adverbials increase the number of possible alternatives for a topicalized constituent; therefore, adverbial modification usually increases the acceptability of constructions with topicalized particles (cf. Lüdeling 1998a, b; Wurmbrand 1998). It should be noted that different speakers sometimes have different judgements about the topicalization data, and that these judgements vary with the

25.According to Grewendorf, (85) is completely ungrammatical. Other speaker’s judgements, including my own, are more liberal. However, there is a contrast between the acceptability of (85) and the acceptability of (86).


context. For example, topicalization of a particle is generally graded more obligingly if the right contrast is already provided by the context. For example, although topicalization in (84) is fairly acceptable in the context of a question, the same sentence is much worse in isolation: (88) ??An lache ich dich part laugh I you ‘I’m smiling at you’

So far, I have only focused on semantic and discourse properties that have an impact on the possibility of topicalizing a particle. However, there are also structural factors that influence the acceptability of sentences with topicalized particles. For example, topicalization of ab in (89b) is better than in (89a): (89) a. ?*Ab ist Nixon 1974 getreten part is N. 1974 stepped b. ?Ab trat Nixon 1974 part stepped N. 1974 ‘Nixon resigned in 1974’

In (89b), the particle verb is used in the simple past — as a result, the finite verb in Comp0 is the verbal part of abtreten. In contrast, the particle verb in (89a) occurs in the present perfect tense, and what has been moved to Comp0 is the auxiliary. There is a clear contrast between (89a) and (89b); most speakers accept (89b) much more willingly than (89a). I offer the following explanation to account for this phenomenon. Notice that what is focused in (89) is actually not the particle, but the whole particle verb ((89b) might be contrasted with something like und gestorben ist er in 1994, ‘and he died in 1994’). Therefore, what actually should have been topicalized in (89) is not the particle, but the whole particle verb. Now, if we have a periphrastic present perfect tense, this possibility does indeed exist. The VP that includes the participle can be moved to SpecCP, and the auxiliary moves to Comp0: (90) [VP Abgetreten]i ist Nixon 1974 ti part-stepped is Nixon 1974

The particle verb in (89a) is also used in the present perfect. I now suggest that (89a) is odd, because (90) is a better alternative. Since the choice of the present perfect provides the possibility of topicalizing the whole constituent that is actually focused (i.e. the verb), (89a) is inappropriate, with only one part of the

97

98


verb in SpecCP. Since the particle alone cannot be contrasted with a possible alternative, (89a) is judged as ungrammatical. The situation changes in (89b), where the particle verb occurs in the simple past, an analytic tense form in German. The problem is that the finite verb must leave the VP and move to Infl0 to pick up the inflectional features. Therefore, moving the whole particle verb in front via VP-topicalization is not possible. Furthermore, topicalization requires the Comp0-position to be filled with the finite verb. If we have an analytic tense form, topicalization and V-to-Comp usually cannot occur simultaneously. However, the separability of particle verbs is helpful here. What happens in (89b) is that the speaker uses the finite part of the particle verb to fulfill the conditions on verb movement (V-to-Infl-toComp), while at the same time, at least a part of the verb (the particle) can be topicalized to establish focus on the verb. Since (89b) does not offer any other opportunity to establish focus via syntactic movement, the sentence is acceptable (see Wurmbrand 2000b for a recent elaboration of this idea). Notice that strictly speaking, (89b) should be as bad as (89a), since in both examples, only a part of the complex element that is focused has been topicalized. However, relatively speaking, (89b) is better than (89a), because it lacks a better alternative. It might be a promising objective for future research to investigate the contrast in (89) in the light of Optimality Theory (cf. Prince & Smolensky 1993; Archangeli & Langendoen 1997), where grammaticality is evaluated on the basis of a ranking of different (violable) constraints.26 An analysis that is both flexible and restrictive enough to deal with the complicated and idiosyncratic factors that influence the phenomenon of particle topicalization is far beyond the scope of this study. However, I think that the framework of grammar that I illustrated in Chapter 1 provides the right means to develop an adequate theory, since the interaction between syntactic operations and semantic conditions can be captured in the tripartite architecture through correspondence rules. Furthermore, I am convinced that the copy

26.In addition, I suspect that the adjacency of the particle and finite verb in (89b) makes it easier for the hearer to assign focus to the whole particle verb, although only the particle is topicalized. Certain idiomatic expressions in German are frozen topicalized particle constructions. Interestingly, they also have the verbal part in Comp0: (i) (ii)

Ab geht die Post! lit. part goes the mail, ‘Let’s go!’ Fest steht, dass … lit. part(tight) stands, that …, ‘It is clear that …’


theory of movement, according to which a “moved” element is present in two different positions at the same time, allows one to offer suitable explanations for examples like (89). The data discussed in this section have shown that, contrary to common belief, particles can be topicalized. The inability of some particles to undergo topicalization does not follow from the wordhood of particle verbs, but is the result of their semantic properties. Since only phrases can be topicalized, elements that do allow topicalization must be phrases. We can conclude that the topicalization pattern of particle verbs strongly supports the claim that particle verbs are complex phrasal constructions, with the particle being represented as the head of a complement of the verb.27

27.It must be noted that the argument from topicalization is vacuous if one believes that in the respective examples, we are actually not dealing with PrtP-movement, but with remnant VP-topicalization (see Hoeksema 1991 for a discussion of and for arguments against such an approach). One could assume that all constituents including the verb, but with the exception of the particle, leave the VP, and that the remnant VP which includes only the particle is topicalized. However, this possibility faces some problems. First, the copies of the constituents moved out of VP would not be bound by their antecedents. However, it has been argued by Sabel (1996) that copies of verbs must not be unbound (see also Haverkort 1990). Second, examples like (i) provide evidence that what has been moved is really a particle phrase: (i)

[PrtP ab]i hat er sich die Haare [PrtP ganz PrtPi] geschnitten part(off) has he refl the hair completely cut ‘He has cut off his hair completely’

The adverb ganz is adjoined to the particle phrase, but left behind when the lower segment of the particle phrase is moved to SpecCP. (However, one could alternatively assume that the adverb is scrambled and adjoined to VP; the topicalized constituent would then again be the lower VP-segment). Third, notice that examples like (i) would require the controversial assumption that participles have to leave the VP as well. However, although the remnant movement-alternative is conceptually not very attractive, it can technically be implemented. In that case, the (im-)possibility of particle topicalization cannot be used as an argument in favor of the morphological approach either, because from a syntactic point of view, particle topicalization would then be predicted to be possible even if particle verbs are words. Eventually, if remnant VP-topicalization really accounts for the grammatical examples, the ungrammatical examples must be excluded on semantic grounds in any theory.

99

100 Particle verbs and local domains

2.5 Modification Another phenomenon which provides evidence for the syntactic status of the particle is illustrated by (91)–(93) (see Koopman 1993; Neeleman 1994; den Dikken 1995 for Dutch): (91) a.

Peter hat die Tür [ganz weit auf]-gemacht P. has the door (completely wide part(open)-made b. [Ganz weit auf] hat Peter die Tür gemacht (completely wide part has P. the door made ‘Peter opened the door completely’

(92) a.

Max ging der Tod seines Kanarienvogels [sehr nahe] Max went the death of.his canary [very part(close) b. [Sehr nahe] ging Max der Tod seines Kanarienvogels [very part went M. the death of.his canary ‘The death of his canary upset Max very much’

(93) a.

Sie nimmt mich nicht [sehr ernst] she takes me not [very part b. [Sehr ernst] nimmt sie mich nicht [very part takes she me not ‘She does not take me very seriously’

The (a)-examples show that some particles can be modified by adverbs; topicalization in the (b)-examples illustrates that the APs ganz weit, ‘completely wide’, and sehr, ‘very’, modify the particle phrase, not the VP (cf. the example (87) in the preceding section).28 Regardless of whether the adverbial modifiers in (91)–(93) are analyzed as adjuncts or as specifiers, the element that they modify must be larger than a word. This implies that the particles in these examples cannot be part of complex V0s. Speakers may even exploit the phrasal status of the particle and adjoin a phrase that semantically modifies the whole VP to the particle phrase: (94) a.

Das Kleid da hinten sieht besser aus this dress there behind sees better part

28.Den Dikken (1995: 108) argues that bare modifiers like Dutch pal or vlak, ‘right’, are X0adjuncts. However, ganz weit, ‘completely wide’ is not a bare modifier but a phrase and therefore must be adjoined to a maximal projection. See also Neeleman (1994: 67, note 10) for arguments against den Dikken’s assumption.


b. [Besser aus] sieht das Kleid da hinten [better part sees this dress there behind ‘This dress over there in the back looks better’

In (94a), the adverb besser, ‘better’, may be adjoined to either PrtP or to VP; the linear order does not specify which. Yet, (94b) shows that it must be adjoined to the PrtP. If the particle was part of a complex word, sentences like (94b) would be absolutely impossible. However, since the particle is the head of a phrase, it can both be modified and topicalized together with the adverb. The behavior of particle verbs with respect to modification therefore provides further evidence against a morphological analysis. (Notice that modification does not count as an argument against the incorporation approach, since the adverbs in (91)–(94) would be expected to be licensed even if the particle could leave the particle phrase).

2.6 Structural restrictions Stiebels & Wunderlich (1994) and Stiebels (1996) observe that the iteration of particles in German is generally excluded; (95c); Neeleman & Weerman (1993) note that particle verbs are not licensed in resultative constructions; (96c):29 (95) a.

weil Peter seine Schulden abarbeitet because P. his debts part-works ‘because Peter works off his debts’ b. weil Peter losarbeitet because P. part-works ‘because Peter starts working’ c. *weil Peter (seine Schulden) losabarbeitet because P. (his debts part-part-works (intended meaning: ‘Peter starts working off his debts’)

29.Examples like (i) seem to provide a counterargument to the latter claim: (i)

Wir malten die Scheune rot an we painted the barn red part ‘We painted the barn up red’

However, I am not sure whether the adjective red is a genuine resultative predicate or rather a modifier (see also the debate between Neeleman (1994: 331–32, note 7) and den Dikken (1995: 76, note 46)). If it is the former, I admit that I do not have any interesting explanation to offer for why (i) is grammatical, but (96c) is not.

101


(96) a.

weil Peter einkauft because P. part-buys ‘because Peter goes shopping’ b. weil Peter den Laden leer kauft (resultative construction) because P. the shop empty buys ‘because Peter shops until the shop is empty’ c. *weil Peter den Laden leer einkauft because P. the shop empty part-buys (intended meaning: ‘Peter goes shopping until the shop is empty’)

(95) and (96) follow from the syntactic approach defended here if one adopts the assumption that a syntactic head can only take one single complement (cf. Kayne’s (1984) Binary Branching Requirement, Larson’s (1988a) Single Complement Hypothesis, Mulder’s (1992) Single Object Corollary). Since particles and resultative predicates can only be selected by verbal heads, the ungrammaticality of the (c)-examples follows. However, these examples cannot be excluded by a morphological approach; if particle verbs are V0s, then, in principle, they should be able to select another particle or combine with a resultative phrase (the impossibility of iteration is also correctly excluded by the incorporation approach). Therefore, the proponents of the morphological approach need to make additional claims to exclude the data in (95c) and (96c). For example, Stiebels & Wunderlich (1994) and Stiebels (1996) exclude particle iteration morphologically on the basis of the following Adjacency Condition (Stiebels & Wunderlich 1994: 933): (97) Adjacency No affixation can apply across more than one bracket

A detailed exposition of Stiebels and Wunderlich’s analysis would be too much of a digression here, and I will only briefly summarize the main idea behind their explanation. Stiebels and Wunderlich argue that the particle verb in (95c) is assigned the morphological structure in (98a), whereas simple particle verbs like abarbeiten in (95a) have the morphological structure (98b): (98) a. [los [[ab [arbeit-]]-et]] b. [ab [[arbeit-]-et]]

The ungrammatical word in (98a) is now excluded as a violation of (97), since two morphological brackets (printed in boldface in (98)) intervene between the inflectional affix -et and the verb.

The syntax of particle verbs 103

However, although the Adjacency Condition can account for (95c), it is not clear how it would exclude the example in (96c). Stiebels and Wunderlich assume that resultative predicates are phrases; the impossibility of realizing resultative phrases as complements of a particle verb cannot be derived from (97). This observation follows, however, from Neeleman and Weerman’s Complexity Constraint (Neeleman & Weerman 1993: 460), which was already discussed in Section 2.1.2: (99) Complexity Constraint X0 should either be a lexical head or have a lexical head

First note that Neeleman & Weerman (1993) analyze resultative constructions as V0s that include the maximal projection of the resultative predicate. They assign the following structures to the unacceptable verbs in (95c) and (96c): V1 0

(100) a. *

V2 0

Prt0 los

V3 0 arbeit-

Prt0 ab

V1 0

b. *

V2 0

AP leer Prt0 ein

V3 0 kauf-

The Prt0- and V30-heads in (100) are lexical heads, since they immediately dominate lexical material. V20 is not a lexical head, but it has one (i.e. V30). Therefore, V20 satisfies the Complexity Constraint in both (100a) and (100b). However, V10 does not; it neither is a lexical head; nor does it have one: the head of the V10 is V20, but V20 is not lexical. Therefore, both structures in (100) violate (99) and hence are illicit. Although the principle in (99) correctly rules out the data in (95c) and (96c), the stipulative and empirically problematic character of (99) that I discussed in Section 2.1.2 above sheds doubt on the validity of Neeleman & Weerman’s (1993) explanation. In the light of the objections that I raised


against (99), the explanation offered by the syntactic approach seems at least conceptually more attractive. Another structural argument in favor of the syntactic approach comes from Dutch and is provided by van Riemsdijk (1978). As noted briefly in Section 2.1.1, in Dutch Verb Raising constructions, the infinitival verb is raised out of a clausal complement to the right of the matrix verb, cf. (101): (101) omdat Jan ti begon de auto te wasseni because J. began the car to wash ‘because Jan began to wash the car’ (van Riemsdijk 1978: 104)

In (101), the trigger of Verb Raising is the matrix verb beginnen, ‘begin’. However, van Riemsdijk (1978) notes that a verb that triggers Verb Raising can never take any complement other than the clause out of which the infinitive is raised. A matrix verb does not allow Verb Raising if it takes another syntactic complement. Crucially, as first observed by Evers (1975), particle verbs can never trigger Verb Raising: (102) a.

omdat Jan aanving de auto te wassen because J. part-catch the car to wash ‘because Jan began to wash the car’ b. *omdat Jan ti de auto aanving te wasseni (van Riemsdijk 1978: 104)

Although the selectional properties and the meaning of the matrix verbs in (101) and (102) are the same, the possibility of raising the infinitive is excluded if the matrix verb is a particle verb. Instead, the whole complement clause must be extraposed (cf. (102a)). On the assumption that the particle in (102) is a syntactic complement of the base verb, the ungrammaticality of (102b) follows straightforwardly from van Riemsdijk’s generalization that Verb Raising can only be triggered by verbs with one single complement. This fact provides further strong evidence for the claim that the particle is realized inside a phrasal complement of the verb.

2.7 Typological remarks A last observation that supports the syntactic approach is provided by verbparticle constructions in other Germanic languages, such as English, Danish, Norwegian etc. It is commonly assumed that particle verbs in these languages


are syntactic constructions.30 For English, this has been first suggested by Emonds (1972). Emonds points out that particle verbs license the occurrence of the adverbial right: (103) a. John came right in b. He put the toys right back c. They looked it right up and left (Emonds 1972: 552)

As den Dikken (1995) shows, right can occur with PPs, (104a), but not with verbs, (104b): (104) a. John threw the ball right through the window b. *John right threw the ball through the window c. John threw the ball right back/up/down (Den Dikken 1995: 38f.)

If the particle verb in (104c) was a V0, and (104c) was derived by moving the verbal part of this complex verbal head, as a morphological analysis would predict, we would expect ungrammaticality, since right would still modify a verb. However, (104c) is grammatical. This follows only from a syntactic analysis that associates particles in English with a maximal projection of P. The assumption that the English verb-particle construction is syntactic has been adopted and elaborated by Di Sciullo & Williams (1987), den Dikken (1995), Svenonius (1996), Jackendoff (1997), and many others. Syntactic analyses along similar lines have also been suggested for Norwegian (Taraldsen 1983; Åfarli 1985; Svenonius 1996) and for other Scandinavian languages like Swedish, Danish, or Icelandic (cf. Svenonius 1996). I discuss these languages in more detail in Chapter 7. Given the (almost) uncontroversial syntactic status of the verb-particle construction in these Germanic languages, it is surely conceptually attractive to assume that particle verbs in German and Dutch are syntactic constructions as well. In particular, one important typological difference between these languages would follow immediately. Whereas the particle precedes the verb in its base position in German and Dutch, it follows the verb in English and in the Scandinavian languages: (105) a.

weil Peter ausgeht because P. part-goes ‘because Peter goes out’

30.But see Johnson (1991) as a notable exception.

(German)


b. dat Jan zijn moeder opbelt that J. his mother part-phones ‘that Jan calls up his mother’ (106) a. Peter goes out b. Vi kastet hunden ut we threw the.dog part ‘We threw the dog out’ c. Boris skruede musikken ned Boris screwed the.music part ‘Boris turned the music down’

(Dutch)

(English) (Norwegian; Svenonius 1996: 10)

(Danish; Svenonius 1996: 12)

If the syntactic in situ-approach is adopted for German and Dutch, the difference between the examples in (105) on the one hand and the data in (106) on the other can easily be attributed to a well-known parametric difference between the respective languages. Whereas in German and Dutch, the verb selects its complements to the left, in English and the Scandinavian languages, the verb has its sister on the right. Since the particle is the head of the verb’s complement and remains in situ, the language-specific properties of the verbparticle construction follow straightforwardly: (107) a.

German/Dutch:

b.

English/Norwegian/Danish…

VP PrtP

Prt 0

VP

V0

V0

PrtP

Prt 0

The SOV-structure of Dutch and German is responsible for the fact that the particle precedes the verb at phonological structure. This fact has misled many researchers in arguing that particle verbs are words, since the linear order Prt-V is in accordance with Williams’ (1981) Right-hand head Rule. However, the phrasal status of the verb-particle construction in SVO-languages provides further evidence that two morphemes that are adjacent at phonological structure do not necessarily correspond to a complex X0-element in syntactic structure.


2.8 Conclusion In this chapter I have tried to provide arguments in favor of the following assumption: (108) Particles are heads of phrasal complements of the verb and do not leave their base position

As I have shown, the idea expressed in (108) can explain why particle verbs are split when the verb moves to Comp0 or to Infl0 (Sections 2.1 and 2.2) and why it is possible to modify the particle in isolation (Section 2.5). Furthermore, particles can be topicalized (Section 2.4), and only the verbal part of a particle verb is deleted in Gapping-constructions (Section 2.3). These properties of particle verbs are only explicable on the basis of (108). The behavior of particle verbs with respect to iteration (Section 2.6) and the parallels between the verb-particle construction in German and Dutch on the one hand and in English, Norwegian etc. on the other (Section 2.7) have provided further support for this view. At various points in this chapter, the view of the structure of grammar that I defended in Chapter 1 became relevant for the discussion. The model of the tripartite parallel architecture emphasizes that the form of linguistic expressions is constrained by rules of all three components, i.e. syntactic, phonological, and conceptual structure. Consequently, the ungrammaticality of a construction is not necessarily the result of a violation of a syntactic principle. It is also possible that semantic or phonological factors exclude a particular derivation. With respect to particle verbs, it was this insight that allowed me to solve some of the problems that apparently arose from a phrasal analysis of particles. For example, I showed that the impossibility of topicalizing certain particles is not the result of the syntactic status of the particle verb, but rather follows from its semantic properties. Similarly, the fact that extraposed phrases cannot separate the particle and the verb follows from phonological conditions on extraposition and is compatible with the view that particles are heads of phrasal complements of the verb that do not undergo incorporation. Many of the alleged V0-properties of particle verbs result from the incorrect view that only the syntactic module of grammar affects the application of syntactic rules. In the next chapter, I propose a definition of the notion “particle” which combines the assumption in (108) with the insights that can be gained by investigating the properties of particle verbs on the basis of a more articulated syntactic structure.

Chapter 3

The non-functional nature of particle phrases

I will now extend and elaborate the syntactic approach to particle verbs that I defended in Chapter 2. I will show that the phrase of which the particle is the head is syntactically incomplete in the sense that it lacks functional structure. In Section 3.1, I discuss the functional projections that have been motivated for the four major lexical categories V, N, A, and P. In Section 3.2, I claim that particle phrases are non-functional, and I provide evidence for this claim by showing that full functional phrases are referential, whereas particle phrases are non-referential. I conclude from this observation that a lexical element can be defined as a particle if it is structurally adjacent to the verb. This particular local relation between the two terminal nodes V0 and Prt0 is the characteristic property of the particle verb.

3.1 Extended projections In Chapters 1 and 2 I have simply represented functional structure above VP through the inflectional categories Infl and Comp (cf. Chomsky 1986b). However, in recent years, a more elaborated functional structure for verbs has been proposed, and the hypothesis that lexical heads are dominated by functional structure has been extended to other lexical categories. It is this proposal that I will discuss in this section. The relation between lexical and functional structure is discussed in detail by Grimshaw (1991). Grimshaw assumes that a functional head F that selects the maximal projection of a lexical head L has the same categorial status as L, but differs from it with respect to its functional feature value. For example, both V and Infl are [+V −N], but V’s functional specification is F0 (it is non-functional), whereas Infl is specified as F1. Grimshaw assumes that the functional structure that dominates a lexical head is part of this head’s extended projection. In Sections 3.1.1–3.1.3 I briefly discuss the extended projections of verbs,

110


nouns, and adjectives. The status of functional prepositional phrases is discussed in more detail in Section 3.1.4. 3.1.1

The extended projection of verbs

Following Pollock (1989) and Chomsky (1991), many linguists have adopted the so-called split-Infl hypothesis, according to which the traditional Inflprojection between Comp and VP is divided into a Tns-projection and several agreement (Agr-) or aspectual (Asp-) projections. However, in Chapter 4 of the Minimalist Program, Chomsky (1995) dispenses with Agr-projections. According to Chomsky (1995), there are only two functional projections between VP and CP; i.e. TP and vP:1 TP

(1)

Spec

T′

Tns 0

vP Spec

v′ v0

VP Spec

V′ V0

XP

The Tns-projection in (1) replaces what has traditionally been called Infl. In syntactic structure, Tns0 bears grammatical features like [±finite] and [±past], which correspond to the respective temporal interpretation in conceptual structure. The grammatical features determine Vocabulary Insertion of the right tense morphemes (see Chapter 1, Section 1.2.3). Furthermore, according to Chomsky (1995), Tns0 carries a strong “EPP-feature”, which triggers

1.Although I adopt Chomsky’s proposal here with respect to the verbal domain, I do not strictly exclude the possibility of having syntactic agreement projections in syntactic structure. For example, contrary to Chomsky (1995), I will assume in Section 3.1.3 that adjectives are dominated by a functional projection that determines agreement between adjectives and their arguments.


movement of the subject to SpecTP where it receives nominative case.2 In addition, Chomsky (1995) postulates a light verb phrase vP that intervenes between VP and Tns. vP is a functional category whose head is phonologically empty in languages like German or English. The light verb projection goes back to similar proposals by Larson (1988), Hale & Keyser (1993), and Kratzer (1994); it is based on the idea that the so-called external argument is not a true argument of the verb, but is introduced by a separate head. For example, in Kratzer’s (1994) theory, the light verb is semantically interpreted as a predicate whose only (individual-type) argument is the Agent. (Notice that Kratzer (1994) labels the functional light verb “Voice”; but as far as I can see, Voice0 and v0 are equivalent.) If the semantics of the VP is combined with Voice0/v0, the result is a complex predicate with an external argument. Therefore, the first important function of v0 is that it is responsible for the introduction of (agentive) subjects, which are merged as arguments of v0 in SpecvP. Second, v0 determines accusative case on the object. According to Chomsky (1995), the direct object (which is merged as an argument of V0 in SpecVP) moves into a second specifier position of vP above the subject where it receives accusative case. Notice that the possibility of assigning structural case to the direct object depends on both the presence of v0 and lexical properties of the verb. On the one hand, accusative case is assigned structurally by v0 in (the higher) SpecvP. On the other hand, the verb must be lexically specified as being able to assign accusative case; Chomsky (1995: 368) assumes that the verb bears a feature [(assign) accusative case]. Accusative case assignment is contingent on verb movement; if V0 moves to v0, it can structurally assign its accusative case to the direct object in SpecvP. The idea that both the Agent-argument and accusative case are related to the existence of v is a structural implementation of Burzio’s generalization (Burzio 1986), which states that the ability to assign accusative case depends on the assignment of an external θ-role. In present terms, both properties are associated with v0. If v0 is absent, the verb cannot assign its accusative case, and no external θ-role is assigned. Therefore, unaccusative verbs are represented as bare VPs; the difference between the transitive verb in (2a) and the unaccusative verb in (2b) can hence be captured structurally:

2.In the Minimalist Program, case is not “received” or “assigned”, but “checked”. The main idea of the checking mechanism is that noun phrases bear case features when they are merged into the tree. These case features have to be checked against a matching feature on the verb. However, I will continue to talk about case assignment in this study.

111

112


(2) a. Peter broke the glass b. The glass broke

In both (2a) and (2b), the NP the glass is merged in SpecVP as an argument of break. In (2a), it moves to the higher SpecvP and receives accusative case from the verb that has moved to v0. At the same time, the subject is introduced by v0 in SpecvP; it moves to SpecTP to receive nominative case. In (2b), no Agentrole is assigned, because v0 is absent. The verb cannot assign its accusative case, due to the lack of the v-projection, and consequently, its argument must move to SpecTP to receive nominative case from Tns0. Kratzer (1994) assumes that Voice0/v0 bears a grammatical feature [±active]. If the light verb is [−active], the result is a passive sentence. A [−active] light verb assigns the Agent-θ-role, but prevents the verb from assigning accusative case. The direct object receives nominative case in SpecTP; external arguments are semantically, but not syntactically, present in passive clauses, due to the [−active]-specification of Voice0/v0. In sum, the extended projection of the verb includes two functional projections below Comp with increasing F-values: (3) V

= [+V −N F0], lexical category; introduces the internal argument(s); lexically specified for case assigning feature Voice/v = [+V −N F1], functional category; introduces the subject; [+active]-Voice/v projection determines structural accusative case assignment Tns = [+V −N F2], functional category; [+finite]-Tns determines structural nominative case assignment

The functional projections in (3) are verbal functional projections; they share Vs categorial features, but have higher F-values. 3.1.2 The extended projection of nouns Abney (1987) argues that the internal structure of noun phrases mimics certain important aspects of the structure of clauses. He suggests that noun phrases should be represented as projections of a functional element D0 (for determiner) that takes the NP as its complement:


DP

(4)

Spec

D′

D0 the

NP

N0 book

The structure in (4) represents the determiner as the head of the noun phrase. On the assumption that it is usually functional heads of phrases that enter into agreement relations, this captures the fact that grammatical relations between noun phrases and verbs are marked on the determiner in many languages: (5) Number: a. This book reads easily b. These books read easily (6) Gender: a. Le deputé a été élu the-masc member has been elected ‘The member has been elected’ b. La porte parole a été arrêtée the-fem spokeswoman has been arrested ‘The spokeswoman has been arrested’ (French, Ouhalla 1991: 157) (7) Case: a. der Liebhaber the-nom lover b. den Liebhaber the-acc lover c. dem Liebhaber the-dat lover ‘the lover’ (German)

The idea that noun phrases are DPs has opened the possibility of providing new analyses for a number of syntactic phenomena. For example, the well-known mixed properties of gerunds, which have the external distribution of noun phrases, but internal properties of VPs, are explained by Abney’s (1987)

113

114


proposal to analyze gerunds as DPs, with D0 selecting a VP-complement instead of an NP. Another phenomenon which has been analyzed on the basis of a complex noun phrase structure like (4) is the so-called Construct State nominal in Hebrew and Arabic, whose properties have been argued to follow from DP-internal movement operations (cf. e.g. Ritter 1988, 1991; Hazout 1990; Borer 1995, 1999 and references cited therein). Taraldsen (1990) argues that the optional N-initial order in Norwegian noun phrases is derived via movement of N0 to D0. Stowell (1989) argues that the D-projection prevents a PRO-subject in SpecNP from being governed by the verb in examples like (8): (8) John disapproves of [DP the [NP PROi hatred of oneselfi]

Longobardi’s (1994) analysis of noun phrases is based on the assumption that N0 raises to D0; for example, he provides evidence that in the case of proper names, only the head of the N-to-D chain is interpreted, whereas in the case of generics, only the foot is interpreted. In sum, the fruitful insights provided by these analyses yield strong evidence that the DP-hypothesis is correct (see also the contributions in Alexiadou & Wilder (1998)). It has been suggested by various authors that the extended projection of a noun consists of additional functional projections between D and N. For example, Bhatt (1990) and Löbel (1990) argue that quantifiers in German are associated with a functional Q-projection between D and N, whereas Ritter (1991) and Koopman (1999a) postulate the existence of a functional projection Num that determines number agreement on the noun (in Ritter’s theory, N0 is therefore forced to move to Num0). Corver (1990) suggests that adjectival modifiers of nouns are represented as specifiers and heads of an additional DegP (degree phrase) between N and D. However, since the precise number and status of the functional projections below DP are irrelevant for the concerns of this chapter, I will henceforth represent the extended projection of N simply as a DP whose head selects the NP, as illustrated in (4). As in the case of the extended projection of verbs, D and N have the same categorial features, but different functional values. 3.1.3 The extended projection of adjectives The extended projection of adjectives has not received as much attention from generative linguists as the extended projections of nouns and verbs. However, in recent years, some proposals about functional structure associated with APs have been made.


In Romance languages like French, adjectives agree with their arguments in post-nominal constructions like (9): (9) a.

Jean est petit J.-sg-masc is small-sg-masc ‘Jean is small’ b. Marie est petite M.-sg-fem is small-sg-fem ‘Marie is small’ c. [Anne et Marie sont petites [A. and M.]-pl-fem are small-pl-fem ‘Anne and Marie are small’

In order to account for the presence of an agreement morpheme on the adjective, I assume that the extended projection of adjectives includes at least one functional projection which determines an agreement relationship between the adjective and the DP that it agrees with. I assume, following a proposal in Chomsky (1995) (which he later abandons), that the subject DP in a sentence like (9b) is generated inside AP as an argument of the adjective and moves to the specifier of a functional Agr-projection. The adjective also moves and combines with its agreement element in Agr0, and agreement is established via the resulting Spec-head configuration (cf. Chomsky 1995: 283). A sentence like (9b) is therefore structurally represented as in (10): AgrP

(10)

Spec DPi Marie A0j petit-

Agr′

Agr 0 Agr 0 -e DPi

AP A0j

(10) illustrates the relevant movement operations that derive the right configuration for agreement between the adjective and the DP Marie. (Notice that the subject DP must raise further to SpecTP in order to receive nominative case.) The structure in (10) is in line with recent work on the functional structure of adjectival phrases in Dutch presented by Corver (1997). Corver also argues

115

116


that the lexical AP is dominated by a functional Agr-projection which is rightheaded in Dutch. Corver suggests that two more functional projections (DegP and QP) intervene between Agr and AP. However, as in the case of the extended projection of nouns, it is sufficient for my concerns to assume that lexical adjectives are dominated by functional structure and that the extended projection of an adjective is an AgrP. This idea is also adopted by van Riemsdijk (1998a), who assumes that the grammatical features of an adjective are associated with a head-final Agr-projection above AP. AgrP has the same categorial feature as the lexical category A, but a different functional specification. 3.1.4 The extended projection of prepositions The possibility that prepositions, just like the lexical categories V, A, and N, are dominated by functional structure, is discussed by van Riemsdijk (1990, 1998b) and Koopman (1993, 1999b). Both authors share the assumption that regular pre- and postpositional phrases should not be represented as bare PPs, but by a more articulated extended projection above the PP. Van Riemsdijk (1990) proposes the following structure for prepositional phrases: FPPrep

(11)

Spec

F′Prep PP

P0

F

0 Prep

DP

Before I discuss the properties of the functional structure in (11), let me briefly comment on the thematic properties of a preposition and the syntactic representation of the PP. Traditionally, prepositions are treated as transitive elements; conceptually, they express a relation between two arguments. The “internal” argument of the preposition has been called the reference object (cf. Jackendoff 1983), the landmark (cf. Dewell 1996; McIntyre 2001), or the Ground (cf. Talmy 1985; Svenonius 1996). The “external” argument is the Theme which is in motion or located with respect to the reference object. For example, in The coat is in the closet, the preposition in expresses that the Theme (the coat) stands in a particular relation to the landmark (the closet); this relation is then specified by the meaning of the preposition. However, in


Jackendoff’s (1983, 1990) semantic framework, prepositions do not express relations, but local concepts like Places (locative PPs) and Paths (directional PPs). In line with this view, most linguists, including van Riemsdijk (1990) and Koopman (1993, 1999b), assume that only the reference object of a preposition is located inside the PP, whereas the Theme of P0 is realized outside the predicate’s maximal projection and merged as an argument of the verb into the specifier of V. The structure in (11) is based on this latter assumption. In order not to complicate the discussion unnecessarily, I will adopt this traditional assumption in my study. I will therefore treat the Theme of a preposition as the direct object of the verb; it is merged in SpecVP, (12), and either moves to SpecvP (as in (13a)) or to SpecTP (as in (13b)) in order to receive case: (12) [VP Peter [V¢ V0 [FP to school]]] (13) a. Mary takes Peter to school b. Peter walks to school

I think that an alternative analysis could as well represent both the Theme and the reference object of the preposition inside the PP.3 However, since the Theme has to move for case reasons anyway, the discussion can be simplified by assuming that the Theme of prepositions is located outside the prepositional phrase.4 Let me come back to the structure in (11). In (11), the lexical PP is selected by a functional head that I have simply labeled FPrep. According to van Riemsdijk (1990), this functional head to the right of the lexical PP can be filled with postpositions. In particular, van Riemsdijk (1990) focusses on complex postpositional elements in German like the ones in (14):

3.In fact, Hoekstra & Mulder’s (1990) analysis of directional PPs is based on this assumption. They take the PP to be the predicate of a Small Clause (SC); it is predicated of the Theme which is located in the SC’s specifier position. 4.Notice that the “base position” of the Theme is irrelevant with respect to the semantics. It follows from the theory of grammar outlined in Chapter 1 that the thematic interpretation of a sentence is determined solely by the structural configuration at the conceptual interface level of syntactic structure (cf. Chomsky 1995; Neeleman 1997). At this level, which represents the results of syntactic movement, the Theme of a preposition is always located (and will therefore be thematically interpreted) in a specifier position of the verb’s extended projection.

117

118


(14) a.

Sie kletterte [FP [PP auf den Berg] hinauf] she climbed on the mountain H-up ‘She climbed up the mountain’ b. Sie kletterte [FP [PP über den Zaun] hinüber] she climbed over the fence H-over ‘She climbed over the fence’

The postpositions in (14) consist of a deictic h-element her (‘towards the speaker’) or hin (‘away from the speaker’)5 and a prepositional element. Henceforth, I will refer to these complex elements as h-postpositions; their precise morphological structure will be discussed in Chapter 5, Section 5.3.1.6 Van Riemsdijk (1990) calls the constructions in (14a) and (14b) “circumpositional phrases”. Notice that h-postpositions can never occur as prepositions, which suggests that they are in fact merged into the tree as the heads of FPPrep: (15) a. *Sie ist hinauf den Berg geklettert she is H-up the mountain climbed ‘She climbed up the mountain’ b. *Sie ist hinüber den Zaun geklettert she is H-over the fence climbed ‘She climbed over the fence’

Both van Riemsdijk (1990) and Olsen (1999) assume that PPs in circumpositional phrases like (14) are complements of the h-postposition. (16) provides evidence that the h-postposition is in fact the head of the circumpositional phrase: (16) Auf den Berg ist sie hinauf geklettert on the mountain is she H-up climbed ‘She climbed up the mountain’ (17) [Auf den Berg]i ist sie [FP PPi hinauf] geklettert

5.However, as is correctly observed by McIntyre (2001), the distinction between her and hin is to a certain extent artificial, and in many cases, the h-elements have ceased to express a deictic function. 6.Abraham (1995) argues that German and Dutch do not have genuine postpositions; he labels the complex elements in (14) “adverbial adpositions” instead. Following van Riemsdijk (1990, 1998b), I have chosen the term “h-postposition” in this study in order to emphasize that I consider these elements to be the realization of a functional head of P’s extended projection, but nothing crucial hinges on this terminological decision.


The leftmost preposition can be topicalized together with the DP, which shows that the DP forms a constituent with the preposition (i.e. a PP) and not with the h-postposition. In the light of these data, I will henceforth follow Olsen (1999) and van Riemsdijk (1990) and assume that h-postpositions in circumpositional phrases take PP-complements. Furthermore, I will adopt van Riemsdijk’s (1990) proposal and assume that h-postpositions are realizations of a functional head FPrep. In this study, I will restrict my attention to circumpositional phrases of the kind exhibited in (14), where the head of the PP is homophonous with the prepositional element that is part of the h-postposition. Notice that there are also circumpositional phrases where the prepositional element that the h-postposition consists of does not pattern with the phonological realization of P0: Sie sprang zum Fenster hinaus she jumped to.the window H-out ‘She jumped out of the window’ b. Sie ging unter der Brücke hindurch she went under the bridge H-through ‘She walked through under the bridge’

(18) a.

Constructions like (18) will be ignored in this study. Following Olsen (1996) and McIntyre (2001), I will call circumpositional structures like (14) “pleonastic”. As shown in (19), h-postpositions do not only take PP-complements, but can also combine with simple DP-complements: (19) a.

Sie ist den Berg hinauf geklettert she is the mountain H-up climbed ‘She climbed up the mountain’ b. Sie ist den Zaun hinüber geklettert she is the fence H-over climbed ‘She climbed over the fence’

There are several possible ways of analyzing these examples. One might assume that the DP in (19) is syntactically and semantically an argument of the h-postposition: (20) [FP [DP den Berg] hinauf]

This proposal requires the additional assumption that the DP receives accusative case from the element in F0Prep. This assumption is supported by the following contrast:

119


(21) a.

Sie ist [entlang des Flusses] gefahren she is [along the river-gen driven ‘She drove along the river’ b. Sie ist [den Fluß entlang] gefahren she is [the river-acc along driven ‘She drove along the river’

The preposition entlang is one of the (very few) prepositions in German that also occurs as a postposition. The examples in (21) show that the case-assigning properties of the postposition entlang (accusative in (21b)) are different from those of the preposition (genitive in (21a)). One might take this as evidence that accusative case assignment in (21b) and (19) is a property of F0Prep and independent of the element that occupies this position. An alternative proposal is made in McIntyre (2001). McIntyre suggests that postpositional phrases like the ones in (19) are in fact circumpositional phrases, with the DP-argument being generated as the argument of a phonologically empty preposition: (22) [FP [F¢ [PP Ø [DP den Berg]] hinauf]]

(22) is compatible with (21) as well if the empty preposition is taken to assign accusative case. I think that both structural alternatives (20) and (22) are possible; I do not want to commit myself to one of these views (but see Zeller (forthcoming a) for an analysis of postpositional structures in German and Dutch which is based on (22)). It is interesting that in Dutch, the order DP-postposition is more common than in German, while pleonastic structures like those in (14) do not exist (cf. (23) vs. (24) and Abraham 1995: 357). Furthermore, most postpositions in Dutch are simple elements and homophonous with prepositions, (25): (23) a.

het huis uit the house out b. *uit het huis uit out the house out ‘out of the house’ (Dutch)

(24) a. *das Haus hinaus the house H-out b. aus dem Haus hinaus out the house H-out ‘out of the house’ (German)


(25) a.

Zij is meteen [in het water] gesprongen she is immediately [into the water jumped ‘She jumped immediately into the water’ b. Zij sprong meteen [het water in] she jumped immediately [the water in ‘She jumped immediately into the water’ (Dutch; Koopman 1993: 21)

Without going into details, I will assume that the Dutch postpositions in (23) and (25) also realize the head of a functional phrase FPPrep. They therefore pattern in this respect with their morphologically complex German counterparts in (14). In Chapter 5, Section 5.3.1, I will argue that, despite a morphosyntactic difference, postpositions in both German and Dutch are prepositional elements of category F0Prep.7 Let me now take a closer look at the representation of simple prepositional phrases. Here I take issue with the proposal made by Koopman (1993, 1999b), who argues that there is more than one functional projection on top of the PP. According to Koopman, the projection that selects the PP determines the locative reading of a PP; it is called “Place”, in imitation of Jackendoff’s (1983, 1990) name for the concept expressed by a locative PP: (26) Das Buch ist auf dem Tisch the book is on the table ‘The book is on the table’ (27) [PlaceP [Place’ Place0 [PP auf [DP dem Tisch]]]]

According to Koopman’s proposal, PlaceP is left-headed. PlaceP is dominated by a functional layer “Path” whose maximal projection determines the directional reading of a PP (directional PPs express Path-concepts in Jackendoff’s terminology):8

7.(25) raises the question whether in Dutch, the order DP-postposition is derived from the order preposition-DP via movement (cf. van Riemsdijk 1978, 1990; Koopman 1993). This question is addressed in Zeller (forthcoming a) in the light of the observation that pleonastic circumpositional phrases do not exist in Dutch. 8.In effect, Koopman assumes that there are two more functional projections intervening between Path and Place; i.e. a DegP and a C-PlaceP. I ignore these additional structures here.

121


(28) Peter legt das Buch auf den Tisch P. puts the book on the table ‘Peter puts the book on the table’ (29) [PathP [Path¢ Path0 [PlaceP [Place¢ Place0 [PP auf [DP den Tisch]]]]]]

(29) implies that every directional prepositional phrase includes the projection of a locative prepositional phrase. The problem with Koopman’s account is that it makes a wrong prediction for German. Postpositional phrases in Dutch always receive a directional interpretation; the same holds for postpositional phrases headed by h-postpositions in German. However, in contrast to Dutch, there is a class of German postpositions that is also compatible with a Place-reading: (30) a.

Das Kleid hängt an dem Haken dran the dress hangs on the hook dr-on ‘The dress is on the hook’ b. Die Decke liegt über dem Stuhl drüber the blanket lies over the chair dr-over ‘The blanket lies over the chair’

Like the h-postpositions in the examples in (14), the postpositions dran and drüber in (30) must be heads of a functional prepositional phrase, since they take full PP-complements. Importantly, the prepositional phrases in (30) receive a locative meaning. Therefore, they must be represented as the heads of a PlaceP in Koopman’s theory. (Note that Koopman assumes that PP-complements in circumpositional phrases move to SpecPlaceP in order to establish the right order PP-Place0). There is a serious problem with Koopman’s proposal, however. Since, according to her theory, directional phrases always include a Place-projection, we would expect that a directional h-postposition can take the prepositional phrases in (30) as its complement. However, as (31) shows, this is impossible: (31) a. *Sie kletterte über [dem Zaun drüber hinüber she climbed over [the fence]-dat dr-over H-over ‘They climbed over the fence’ b. *Sie kletterte über [den Zaun drüber hinüber she climbed over [the fence]-acc dr-over H-over ‘They climbed over the fence’ (32) *[PathP [PlaceP [PP über dem/den Zaun] drüber] hinüber]


Locative postpositions like drüber and dran in (30) and h-postpositions stand in complementary distribution. This does not follow from Koopman’s theory, according to which both postpositions are associated with different functional heads. I therefore make an alternative proposal which is based on the structure in (11) suggested by van Riemsdijk (1990). I assume that all (spatial) prepositional phrases consist of one layer of functional structure whose head is specified for the grammatical feature [±directional]. This feature specification corresponds to the interpretation of the prepositional phrase as directional or locative. Hence, FPrep bears a certain resemblance to the functional category Tns, which may carry the grammatical feature specification [±past]. In the same way that this morphosyntactic feature corresponds to a particular interpretation of the clause as present or past tense, the feature [±directional] corresponds to the respective semantic distinction between directional and locative prepositional phrases. Importantly, the head of the functional prepositional phrase can only be filled with one postpositional element, depending on its feature specification as [+directional] or [−directional]. The ungrammaticality of (31) hence follows from the fact that both elements realize the same functional head. In simple prepositional phrases like (33a) and (33b), the PP is selected by an empty functional head F0Prep: (33) a.

Peter hängt das Kleid an [den Haken P. hangs the dress on [the hook]-acc ‘Peter hangs the dress on the hook’ b. Das Kleid hängt an [dem Haken the dress hangs on [the hook]-dat ‘The dress hangs on the hook’

(34) [FP [F¢ [PP an [DP den Haken]] F0 [+directional]]] (35) [FP [F¢ [PP an [DP dem Haken]] F0 [−directional]]]

Van Riemsdijk (1990) compares the relation between a PP and an h-postposition in circumpositional phrases to the relation between a determiner and the head of an NP inside DP. In terms of this comparison, simple prepositional phrases correspond to bare plurals or mass nouns, which are represented as DPs with phonologically null determiners (see Section 3.2.1). As I have shown in Sections 3.1.1–3.1.3, a number of syntactic properties of a lexical item are licensed through the functional structure of its extended projection (cf. e.g. van Riemsdijk 1990; Grimshaw 1991; Koopman 1993; Borer 1993, 1998). For example, recall that a verb may be lexically marked as an

123


accusative case assigner, but it can only assign case via its functional v-/Voiceprojection. Furthermore, structural case assignment depends on the right feature in v0/Voice0. According to Kratzer (1995), if Voice is specified as [−active], accusative case cannot be assigned. Therefore, it makes sense to assume that the grammatical properties of a lexical preposition in P0 are licensed through the presence of functional structure as well. This view is important with respect to the case assignment properties of P0. I assume that a lexical preposition can only assign case to its DP-complement if functional structure is present (cf. Haiden (1997) for a similar proposal). This assumption is supported by the following observation. As is well known, the case assigned by many of the so-called “topological” prepositions in German depends on whether they express a Place or a Path. For example, an, ‘at’, in (33) assigns dative case to its reference object if the prepositional phrase denotes a location, but it assigns accusative case if it expresses a Path. The same is illustrated for auf, ‘on’, in (36) (other prepositions that show the same alternation as auf and an are neben, ‘next to’, unter, ‘under’, in, ‘in’, hinter, ‘behind’, über, ‘over’, and many more): (36) a.

Peter legt das Buch auf [das Regal P. puts the book on [the shelf]-acc ‘Peter puts the book on the shelf ’ b. Das Buch ist auf [dem Regal the book is on [the shelf]-dat ‘The book is on the shelf ’

I suggest that the case that is assigned to the complement of P0 depends on the feature value of the functional head. If F0Prep is [−directional], P0 assigns dative case. If F0Prep is [+directional], P0 assigns accusative case: FPPrep

(37) a.

F0Prep [+directional]

PP P0

DPacc

= directional FP; Path-reading


FPPrep

b.

F0Prep [–directional]

PP P0

DPdat

= locative FP; Place-reading

It may seem surprising at first glance that a functional head can determine the case-assigning properties of the head of its complement without the lexical head being forced to incorporate. Recall from Section 3.1.1 that the case-assigning properties of a verb were also licensed through v0, but only on the assumption that V0 moves and combines with v0. However, in Chapter 1, Section 1.2, I have discussed the idea that elements may move “covertly”, i.e. that a chain at the phonological interface level of syntactic structure is not pronounced in its head position, but in situ. Although a preposition does not visibly incorporate into F0Prep, one might assume that it does so “invisibly”; i.e. that it combines with F0Prep in syntactic structure, but its Vocabulary item is associated with the base position of its chains. The licensing of case assignment can now be established through the chain created by incorporation.9 However, I think that besides “abstract” incorporation (cf. Baker 1988), there is another way of capturing the situation that we find in (37). We can assume that the local relation between the two heads P0 and F0Prep licenses the case-assignment properties of P0, even though both heads are not part of the same word or linked through a chain. Notice that the relation between P0 and the head of its extended projection is strictly local, because no other head intervenes between the two. In fact, head movement, which can only apply locally, is licensed in precisely this domain; this is why the abstract-movement metaphor is useful to account for cases like (37). In Chapter 1, I have given a definition of the local domain that is established between a head and the head of its sister. According to this definition, P0 and the functional head F0Prep are structurally adjacent. Therefore, it is sufficient to assume that the case assignment property of a lexical head may be triggered by properties of a functional

9.As noted by Spencer (1991: 29), in languages like Welsh, certain prepositions are inflected. It could be assumed that in these languages, F0Prep is the host of prepositional inflectional morphology, and P0 moves to F0Prep overtly in order to combine with the respective morpheme.

125


head within the same locality domain, as long as this domain may also be defined by structural adjacency. I assume that a case assigner and its licenser do not have to combine in a word in order to make case-assignment possible. Similar “local” phenomena and the notion of “local domain” will be discussed in Chapter 5, Section 5.2. Finally, a comment on intransitive prepositions like the ones in (38) is necessary: (38) a. Hij is binnen/buiten/boven/beneden (Dutch; Koopman 1993: 9) b. Er ist drinnen / draußen / oben / unten (German) he is inside / outside / upstairs / downstairs ‘He is inside/outside/upstairs/downstairs’

I assume that intransitive prepositional phrases like those in (38) are full FPPrep-projections. This is confirmed by the observation that intransitive prepositions are compatible with postpositions like drin and hinein; the latter realize F0Prep, which is phonologically empty in (38): (39) a.

Sie sind oben drin they are up dr-in ‘They are up in there’ b. Ich lege sie oben hinein I put them up H-in ‘I put them up in there’

The prepositions in (39) are heads of intransitive PPs; these PPs are selected by a functional head that can be filled with a directional or a locative postposition. In sum, I assume that the structural representation of prepositional phrases includes a layer of functional structure FPrep whose head determines the caseassigning properties of the preposition. F0Prep can be filled by a postposition or, as in the case of simple intransitive or transitive prepositional phrases, can remain phonologically unrealized. As was the case with the light verb v0, nominal determiners, or adjectival agreement morphemes, F0Prep is a functional element whose categorial features are the same as those of its lexical head. F0Prep only differs from lexical prepositions in that it has a higher F-value.

The non-functional nature of particle phrases 127

3.2 Particle phrases as non-functional XPs In this section, I provide arguments for a definition of the notion “particle” that I formulate as follows: (40) Particle definition Particles are heads of non-functional phrasal complements of the verb and do not leave their base position

The definition in (40) is based on the syntactic approach that I defended in Chapter 2; it emphasizes the specific character of the verb-particle construction. As I have shown, heads of phrasal complements of V are usually functional heads (F0Prep, D0, Comp0 etc.). According to (40), whenever the head of a lexical complement of the verb lacks an extended projection and does not incorporate into V0, it is a particle. The notion of particle is therefore defined by the syntactic environment of a lexical head. Using the definition that I gave in Chapter 1, Section 1.3.1, (40) states that whenever a non-affixal lexical head is structurally adjacent to the verb, it is a particle. It is generally assumed that there is a correspondence between the presence of functional structure and referentiality (cf. Miller 1993). On the basis of this assumption, the following discussion will provide evidence for the definition in (40). In Section 3.2.1 I look at the relation between referentiality and functional structure with respect to noun phrases. I show that those elements that are traditionally analyzed as nominal particles are always determiner-less nouns with a non-referential interpretation. I conclude that nominal particles are heads of bare NP-complements of V0. In Section 3.2.2 I extend this claim to prepositional particle verbs. The semantic contrast between these constructions on the one hand, and verbs with functional FPPrep-complements on the other suggests that prepositional particles are heads of bare PP-complements of V0. Finally, Section 3.2.3 discusses the consequences of this view for the analysis of adjectival particles. Again, I reach the conclusion that these elements are heads of bare AP-complements of V0. In this respect, they contrast with functional adjectival complements of verbs; I argue that the latter are resultative predicates. 3.2.1 Referential noun phrases and functional structure Jackendoff (1983) introduces the notion of “projected world”. The projected world consists of the mental representations of entities that the mind creates in experiencing the real world. Jackendoff argues that the information conveyed


by language is not about the real world, but about this projected world. According to Jackendoff, the projected world defines the reference of linguistic items. A linguistic expression refers if it corresponds to a constituent of conceptual structure that is projected; in that case, the language user is aware of this concept. In Jackendoff’s view, projectability and hence reference is a property of tokens; tokens are concepts that project into awareness. In contrast, types are concepts that do not project at all. A type-concept rather includes rules and principles that may be used for categorization. Linguistic expressions that express types are non-referring; they express a certain category or a kind. In (41), for example, the DP a dog is used as a type; the thing pointed out by that is established as an instance of this type: (41) That is a dog

The DP a dog in (41) is a non-referring expression. In contrast, the same DP is referential in (42), where it expresses a token: (42) A dog bit me

With respect to the nominal domain, many researchers have noted the relation between referentiality and functional structure. For example, Stowell (1989, 1991) argues that bare NPs are always predicates and non-referential, whereas DPs are usually referential expressions.10 Similarly, Borer (1995) notes that D0 is the locus of reference in noun phrases and that bare nouns are always nonreferential. Longobardi (1994: 648) argues that “the N position is interpreted as referring to universal concepts, that is, to kinds; the D position, instead, determines the particular designation of the whole DP”. Longobardi (1994) argues that D0 determines the reference of the DP either by hosting a pronoun that refers to a particular entity, or by hosting a determiner that stands in an operator-variable relation to the noun in N0. In the latter case, the range of the variable is determined by the type expressed by the nominal. According to Longobardi (1994), proper names are generated in N0, but move (overtly or covertly) to D0.

10.As the example in (41) shows, not every DP is referential. Stowell (1989, 1991) is also concerned with cases where a DP is used non-referentially: (i)

Bob called Stan a fool

In (i), a DP is used as a Small Clause predicate and is therefore non-referential as well. However, even though DPs may be non-referential, bare NPs can never refer, due to the lack of functional structure.


Notice that singular count nouns usually have to occur with a phonologically overt determiner in German: (43) a.

Peter repariert das Auto P. repairs the car ‘Peter repairs the car’ (cf. *Peter repariert Auto) b. Peter putzt das Klavier P. cleans the piano ‘Peter cleans the piano’ (cf. *Peter putzt Klavier)

Longobardi (1994) suggests that the plural or mass interpretation of N0 is unmarked. Therefore, a singular interpretation requires the presence of an overt determiner. However, the following examples show that in German, certain verbs can take singular count nouns without a determiner: (44) a.

Peter fährt Auto P. drives car ‘Peter is a car-driver’ b. Peter spielt Klavier P. plays piano ‘Peter plays the piano’

The constructions in (44) are traditionally analyzed as nominal particle verbs; the nouns in (44) are nominal particles (cf. Stiebels & Wunderlich 1994). In Chapter 2 I have argued that particles project phrases in syntactic structure. If these phrases were DPs, we would expect that the singular count nouns in (44) take a determiner, like the nouns in (43). I will now assume that the absence of a determiner follows from the absence of the D-projection; in other words, the nominal particle phrases in (44) are bare NPs. Nominal particle verbs are hence represented as in (45): VP

(45)

NP

V0 fahr-

N0 Auto

As expected, the interpretation of the noun phrases in (44) is different from their interpretation in (43). Whereas (43a) means that Peter repairs a particular


instance of the type “car”, and (43b) means that he cleans a specific member of the set of pianos, the nominal expressions in (44) do not express particular tokens of the types “car” and “piano”; they do not refer. On the assumption that referentiality of a phrase is contingent on the presence of functional structure, the lack of functional structure automatically leads to a non-referential interpretation. Therefore, the non-referentiality of nominal particles follows from the fact that the noun phrases in (44) are bare NPs. I will henceforth assume that whenever a verb occurs with a bare NP complement, we are dealing with a nominal verb-particle construction. This claim is confirmed by the observation that adjectival modifiers of nominal particles are excluded: (46) a. *Peter fährt schnelles Auto P. drives fast car ‘Peter is a driver of fast cars’ b. *Peter spielt schwarzes Klavier P. plays black piano ‘Peter is a player of black pianos’

Since prenominal adjectives in German agree with their nouns, it is reasonable to assume that they are associated with functional structure above the NP. Because this structure is absent in nominal particle constructions, (46a) and (46b) are ungrammatical.11 The data in (47) seem problematic with respect to the claim that all bare NPs are nominal particles: (47) a.

Peter aß Fisch P. ate fish-sg ‘Peter ate fish’ b. Nixon ist Präsident N. is president-sg ‘Nixon is president’

In both (47a) and (47b), a singular count noun appears without a determiner. We therefore may analyze Fisch essen, ‘eat fish’, and Präsident sein, ‘be president’, as V+NP-constructions, i.e. as nominal particle verbs. However, although

11.The adjectives in (46) contrast with adverbial modifiers of prepositions, which do not agree with prepositions and are adjoined to PP (see Chapter 2, Section 2.5). Therefore, prepositional particles can be modified, but nominal particles cannot.


in principle nothing is wrong with this view, there is reason to believe that the nominals in (47) are in fact DPs. Notice that the interpretation of the noun phrase in (47a) differs from the interpretation of the nominals in (44). Fisch in (47a) is not non-referential; rather, the use of this singular count noun allows a mass interpretation. As pointed out by Longobardi (1994), Fisch in (47a) rather behaves like the mass noun “fish meat”. Longobardi therefore suggests that the nominal in (47a) be analyzed as a DP with an empty determiner (see below). Like a phonologically overt D0, the empty determiner binds a variable in N0, whose range is constituted by the extension of N0. However, the range is not defined by members of N0’s extension, but by parts of its members. According to Longobardi, this particular “parts of”-reading is a specific property of empty determiners. Stowell (1991) argues that predicates in constructions like (47b) refer to titles rather than kinds and hence have been reanalyzed as adjectival predicates. In the present theory, this means that (47b) is an NP which is selected by a functional head of category A; i.e. it is an AgrP. This is confirmed by the observation that the count noun in (47b) can only occur without a determiner in the context of a copula like sein, ‘be’. With main verbs, the determiner is obligatory: (48) Nixon berät *(den) Präsident N. advises *(the president ‘Nixon advises the president’

I conclude that the particle definition given in (40) above does not apply to the examples in (47). Only if a nominal is the head of a non-functional complement of the verb is it a particle. The assumption that nominal particles are bare NPs explains the following observation made by Booij (1990: 49f.). Booij argues that the nominal particle verb adem halen, ‘breathe’, can be negated in two ways: (49) a.

Hij kon niet adem halen he could not breath take b. Hij kon geen adem halen he could no breath take ‘He could not breathe’

Note that (49) contrasts with (50): (50) a. *Hij kon niet adem krijgen he could not breath get

131

132


b. Hij kon geen adem krijgen he could no breath get ‘He could not catch his breath’

According to Booij (1990), the contrast between (49a) and (50a) shows that particle verbs must be words, and not phrasal constructions, since they behave differently to other verb-complement constructions with respect to negation. However, the definition of particles that I provided in (40) accounts for this contrast in a different way. The occurrence of the negative element geen is immediately triggered if negation precedes a full DP; geen is a combination of negation and the indefinite article een. Therefore, we can conclude that neither (49b) nor (50b) is a particle verb; both constructions must be analyzed as regular DP-verb-combinations. The absence of geen in (50a) does not imply that the nominal adem does not project a phrase, it only implies that functional structure is absent. Therefore, (49a) is a particle verb; the negative element must be adverbial and is associated with the verb’s functional structure. The ungrammaticality of (50a) merely illustrates that a particle verb adem krijgen does not exist; the singular count noun phrase in (50) is only licensed when functional structure is present. It is clear that the question of whether a D-projection is present is not easy to answer for bare plurals and mass nouns, since they may generally appear without a determiner: (51) a.

Peter trinkt Bier P. drinks beer ‘Peter drinks beer’ b. Peter schreibt Briefe P. writes letters ‘Peter writes letters’

We certainly do not want to analyze every verb that takes a bare plural or mass noun as its argument as a particle verb. Fortunately, this assumption does not seem to be necessary in the light of Longobardi’s (1994) analysis of bare plurals and mass nouns. Longobardi argues that noun phrases can only function as arguments if they are DPs and hence assumes that bare plurals and mass nouns in argument position are full DPs, with D0 filled with a phonologically “empty”


determiner.12 However, if we adopt this idea, how can we decide whether nominal particle verbs with bare plurals or mass nouns exist at all? Consider (52). The verb spielen, ‘play’, normally does not take bare plural DPs as internal arguments:13 (52) a. *Peter spielt Puppen P. plays dolls b. *Peter spielt (Bauklötze P. plays (children’s).building.bricks

The ungrammaticality of (52) follows from the assumption that DPs are subject to the Case Filter (or whatever corresponding principle requires that DPs bear case). The verb spielen in (52) does not assign accusative case and therefore does not combine with regular noun phrases. However, (56) is an exception: (53) Peter spielt Karten P. plays cards ‘Peter plays cards’

If the noun phrase in (53) could receive case from the verb, other bare plural DPs should be licensed in this position as well. However, (52) shows that this is impossible. Therefore, it does not seem unreasonable to assume that the noun phrase in (53) is a bare NP that does not need case; the construction in (53) is

12.According to Longobardi (1994), the empty determiner can license the existential interpretation of certain bare plurals in contexts like (i): (i)

Hat Peter Kinder? — Ja, eins has P. childrenpl ‘Does Peter have children?’ — ‘Yes, one’

In order for the answer to the question in (i) to be appropriate, Peter does not have to have more than one child. This shows that the use of the bare plural Kinder in (i) has a mere existential interpretation. According to Longobardi, this interpretation is determined by the phonologically empty operator in D0. 13.If spielen, ‘play’, is used with instruments or songs, the transitive use and accusative case assignment are possible, however: (i)

(ii)

Peter spielt ein Klavier P. plays a piano ‘Peter plays a piano’ Peter spielt ein Lied P. plays a song ‘Peter plays a song’

(i) of course contrasts with (44b) in the text in that the DP has a fully referential meaning.

133


a particle verb. This assumption receives strong support from the contrast between (54) and (55): (54) a.

Peter kaufte neue Karten P. bought new cards ‘Peter bought new cards’ b. Es wurden Karten gekauft it were-pl cards-pl bought ‘Cards were bought’

(55) a.

Peter spielte (*neue) Karten P. played (*new cards ‘Peter played with new cards’ b. *Es wurden Karten gespielt it were-pl cards-pl played c. Es wurde Karten gespielt it was-sg cards-pl played ‘Card games were played’

The verb kaufen, ‘buy’, in (54) takes a bare plural DP Karten as its internal argument. Consequently, the noun phrase can be modified with an adjective, and it receives structural accusative case. Therefore, (54a) can form a passive construction, as shown in (54b). Here, the DP Karten receives nominative case and agrees with the finite verb in Tns0. However, the noun phrase Karten in (55) behaves differently. First, as shown by (55a), adjectives are impossible, which indicates the absence of functional structure. And second, a regular passive formation like (55b) is ruled out. This again follows from the assumption that only DPs receive structural case and can undergo the passive transformation. However, Karten in (55) is a bare NP; therefore, it can not receive nominative case in (55b) and does not agree with the verb. (55c) is grammatical, since it is an impersonal passive. The absence of agreement between the verb and the noun phrase shows that the latter is not a proper argument of the verb and has not been promoted to subject (cf. Kroch & Santorini 1991). Finally, the assumption that Karten spielen is a particle verb is confirmed by the fact that the verb spielen, ‘play’, participates in the formation of nominal particle verbs quite productively: (56) a.

Peter spielt Fußball P. plays soccer ‘Peter is a soccer player’


b. Peter spielt Tennis P. plays tennis ‘Peter is a tennis player’

Although in most cases, the difference between a nominal particle and a bare plural-DP with an empty D0 is hard to establish, examples like (55) at least show that bare plurals also occur as bare NPs (and therefore as nominal particles). Consequently, I analyze both the noun in the example in (55) and the nouns in (44) as nominal particles; they only project a bare NP. 3.2.2 Prepositional phrases and referentiality I have shown that nominal particles are heads of non-functional complements of the verb. On the basis of the definition of “particle” that I have given in (40) above, I will now generalize this conclusion and make the following claim: (57) Prepositional particles do not project functional structure

An important prediction follows from (57). If referentiality can only be established through the presence of functional structure, we expect prepositional particle phrases to be non-referential. In order to show that this prediction is in fact borne out, I will first investigate the referential properties of functional prepositional phrases that are headed by an h-postposition. In Section 3.1.4 I have argued that F0Prep bears a relation to the head of its PP-complement which is comparable to the relation between D0 and N0. The claim that F0Prep is comparable to D0 has another interesting consequence. In line with Postal’s (1969) idea that pronouns are in fact definite articles, it is widely assumed that pronouns are best analyzed as realizations of the functional head D0 (cf. Abney 1987; Stowell 1991; Longobardi 1994): (58) a.

Peter hat [DP sie] geküßt P. has her kissed ‘Peter has kissed her’ b. Peter hat [DP ihn] nicht gesehen P. has him not seen ‘Peter has not seen him’

This hypothesis is confirmed by the occurrence of d-pronouns in colloquial German. The phonological form of the pronouns in (59) is that of determiners:

135


(59) a.

Peter hat [DP die] geküßt P. has the kissed ‘Peter has kissed her’ b. Peter hat [DP den] nicht gesehen P. has the not seen ‘Peter has not seen him’

Similarly, Romance clitics are pronominal elements with the phonological form of articles. Therefore, they are analyzed as D0s (cf. e.g. Chomsky 1995: 249): (60) a.

Jean ne voit pas [DP les [NP femmes]] J. not see neg the women ‘Jean doesn’t see the women’ b. Jean ne [DP les] voit pas J. not them see neg ‘Jean doesn’t see them’

Pronouns are referential DPs, but they do not completely specify what they refer to; this information must be inferred from the current discourse domain. For example, the gender of die in (59a) specifies that the person kissed by Peter is female, but the precise reference of the pronoun depends on the context. It is an interesting property of h-postpositions that they can be used intransitively; i.e. they can occur without a DP- or PP-complement: (61) a.

Peter kommt heraus P. comes H-out ‘Peter comes out of something’ b. Peter klettert hinüber P. climbs H-over ‘Peter climbs over something’

(62) a.

Peter holte den Hund herein P. brought the dog H-in ‘Peter brought the dog into something’ b. Peter zog den Wagen heran P. dragged the cart H-at ‘Peter pulled the cart close to some place’

(63) a. [VP Peter [V¢ [FP heraus] komm-]] b. [vP Peter [VP den Hund [V¢ [FP herein] hol-]] v0]

The examples in (61) are intransitive; the Theme of the unaccusative verb of motion is generated in SpecVP, (63a), and becomes the subject. The FPPrep is


solely represented through the intransitive h-postposition. In (62), the subject is the Agent introduced by v0; the Themes of the h-postpositions in (62) are also generated in SpecVP, (63b). However, (62) patterns with (61) in that the whole prepositional phrase is solely represented through the h-postposition.14 Importantly, the intransitive h-postpositions in (61) and (62) are fully referential; like nominal pronouns, they express tokens. For example, heraus in (61a) refers to the Path that Peter traverses towards the speaker.15 On the basis of this observation, McIntyre (2001) suggests that intransitive h-postpositions are prepositional proforms, i.e. proforms that stand for full prepositional phrases. (Notice that traditionally, intransitive h-postpositions are called “pronominal adverbs”). They have the same function as the pronouns in the sentences in (58) and (59); however, they refer to Paths and not to Things. The fact that intransitive postpositions are proforms, and therefore referential expressions, has an important implication for the interpretation of the implicit reference objects of these intransitive elements. In order for an FPPrep to refer to a Path-token, the reference object must also be a token; it must be “projected into awareness”. In referential pre-, post-, or circumpositional phrases, the reference object is explicitly identified through a DP-argument.

14.The data in (62) contrast with the data in (i), where the accusative object is located inside the FPPrep: (i)

(ii)

Ein Tourist kletterte den Berg hinauf (Ein Tourist = Theme, a tourist climbed the mountain H-up den Berg = internal argument) ‘A tourist climbed up the mountain’ [VP Ein Tourist [V¢ [FP den Berg hinauf] kletter-]]

In contrast to (62), the h-postposition in (i) is not pronominal. As shown in Section 3.1.4, the DP den Berg in (i) is located inside the FP (either inside a PP with an empty head or as a DP-argument of hinauf). The structural difference between (62) and (i) is also confirmed by the following contrast: (iii)

a.

b.

*Den Hund herein hat sein Besitzer geholt the dog H-in has his owner brought ‘His owner took the dog inside’ Den Berg hinauf ist ein Tourist geklettert the mountain H-up is a tourist climbed ‘A tourist climed up the mountain’

The difference between (iii-a) and (iii-b) follows from the claim that only in (iii-b), the DP and the h-postposition form a constituent (i.e. an FPPrep) and therefore can be topicalized together. 15.Jackendoff (1983, 1990) shows that the type-token distinction (referentiality/nonreferentiality) can be applied to Paths and Places as well as to Things.

137

138


However, if the reference object remains implicit, as in (61) and (62), it must be identified on the basis of contextual information. Therefore, as Dewell (1996) and McIntyre (2001) point out, the reference of intransitive h-postpositions can only be determined if the implicit reference object of the prepositional proform is pragmatically recoverable from the context. Examples like (61) and (62) can only be uttered felicitously if the context or the discourse provide the hearer with a plausible reference object that she can use to identify the Path that the FPPrep refers to. For instance, (61a) does not only mean that Peter traverses a Path that leads from out of somewhere towards the speaker. It must also be clear to the hearer what landmark-token the speaker refers to in using the proform. The Source of Peter’s motion must be contextually given, and this contextual givenness of the implicit reference object of a prepositional proform is a direct consequence of the referential nature of the whole phrase. This observation provides a means to test the aforementioned prediction with respect to the interpretation of prepositional particle verbs. If prepositional particles lack functional structure, the Paths expressed by these elements should be non-referential. Given the correspondence between the referentiality of prepositional phrases and the interpretation of implicit reference objects, we now predict that implicit reference objects of particle verbs are non-referential. We therefore expect a clear contrast between particle verbs with implicit internal arguments and corresponding constructions with verbs that combine with intransitive h-postpositions (henceforth h-verbs). Whereas the implicit reference objects of the latter are always contextually given, no such interpretation is expected with particle verbs. This expectation is borne out. McIntyre (2001), who offers a detailed comparison of particle verbs and h-verbs, observes an important difference between the interpretation of a (syntactically not realized) reference object of a particle verb and a (syntactically not realized) reference object of an h-verb. He formulates this difference in his Landmark Referentiality Generalisation, which is given in (64) (I have substituted the notions “h-verbs” and “particle verbs” for McIntyre’s terms “double particle verbs” and “single particle verbs”): (64) Landmark Referentiality Generalisation (McIntyre 2001) The underlying reference objects of h-verbs are referential, specific and are tokens. Those of particle verbs are non-referential, non-specific, generic and are types.

The generalization expressed by (64) is illustrated by the examples in (65) and (66) (cf. McIntyre 2001):


(65) a.

Peter will einen Kreis herausschneiden P. wants a circle H-out-cut ‘Peter wants to cut a circle (out of some unspecified entity)’ b. Hier strömt Gas heraus here streams gas H-out ‘Gas escapes (out of a some unspecified entity)’

(66) a.

Peter will einen Kreis ausschneiden P. wants a circle part-cut ‘Peter wants to cut out a circle’ b. Hier strömt Gas aus here streams gas part ‘Gas is escaping here’

In the examples in (65) and (66), the reference object is syntactically unexpressed; as in the examples in (61) and (62), it must be recovered from the context. If a speaker utters (65a), the hearer will select a contextually salient element as the reference object. For example, he may identify the newspaper or a piece of fabric as the entity out of which Peter wants to cut a circle, or a particular hole in a pipe as the Source of the escaping gas. Because of this contextual information, the hearer is mentally aware of the entity that corresponds to the reference objects of the elements in (65), and this in turn gives rise to a referential interpretation of the FPPrep headed by heraus. In contrast, the reference objects of the particles in (66) do not mentally project; the hearer is not aware of the entities that the reference objects of the particles in (66) correspond to. McIntyre (2001) argues that if the speaker’s intention does not place any importance on the nature or identity of the reference object, she chooses a particle verb rather than an h-verb. In (66), what is communicatively important are the activities denoted by the particle verbs, not the reference objects. For example, in a situation expressed by (66b), it is not relevant from where exactly gas is escaping, but that it is escaping. The conceptual argument of the particle that corresponds to the reference object is not present in the projected world of the hearer. Since the landmark is nonreferential, it is clear that the whole Path expressed in (66) is non-referential as well. We hence find the expected contrast between functional h-postpositions and prepositional particles, which follows from my claim that the latter do not project functional structure. The Landmark Referentiality Generalization also explains another curious property of particle verbs. McIntyre (2001) shows that with many particle verbs, the character and nature of the reference object of the particle is implicitly


known by, and part of, the conceptual knowledge of the hearer, although it is not projected. As an example, consider the difference between (67a) and (67b): (67) a.

Ich habe heute drei Briefe hineingeworfen I have today three letters H-in-thrown ‘Today, I have thrown three letters (into something)’ b. Ich habe heute drei Briefe eingeworfen I have today three letters part-thrown ‘I have posted three letters today’

The particle verb einwerfen, ‘post’, has a specific meaning. It can only be interpreted as the activity of throwing something into a letterbox. The specific interpretation of einwerfen and similar verbs has often been argued to be a lexical accident; however, McIntyre argues convincingly that it follows directly and compositionally from the type-nature of the particle’s reference object. If a hearer is confronted with a sentence like (67a), he is forced to search the context or the discourse domain for the right token that the implicit reference object of hinein in (67a) refers to. This is why the sentence is odd in isolation; an appropriate context is required in order to establish this reference. In contrast, a sentence like (67b) is perfect even when uttered in isolation, because the hearer may select the appropriate type from the group of types that are part of his concept knowledge and on the basis of the Event expressed by the particle verb. The type he will chose is the Thing that prototypically occurs as the reference object in the activities expressed by the particle verb, i.e. a letterbox in (67b).16 (67a) can be paraphrased as “I have thrown three letters into this specific token”, whereas the meaning of (67b) could be formulated as “I have thrown three letters into instances of the kind of thing that one usually throws letters into”. Furthermore, as McIntyre (2001) observes, (67) also illustrates this difference in a second interesting way. Sentence (67a) can only mean that the speaker dumped the three letters into one single thing, which may be a letterbox, a rubbish bin, a lake etc. The context helps to identify the specific token expressed by the reference object, but as a consequence, the implicit argument of hinein refers to a single entity. In contrast, the only constraint on the reference object in (67b) is categorial. It must be something of the type

16.The “prototypical” nature of the reference object of particle verbs is discussed in great detail by McIntyre (2001). He also notes other aspects that determine the choice of the type expressed by the reference object.


“letterbox”, but it does not have to be a single entity. Therefore, it is possible that the speaker has put every letter into a different instance of this type; i.e. in the situation described by (67b), three letterboxes may have been involved. This difference also follows from the fact that the reference objects of particle verbs express types and are non-referential, whereas the reference objects of h-verbs express tokens and refer. These data provide evidence for the claim in (57). Since functional structure is the locus of referentiality, it follows from (57) that implicit reference objects of particles never refer. According to (40), the particle phrase is a bare PP that lacks functional structure. This means that the Path expressed by the particle is not a token, and this in turn implies that the particle’s implicit reference object cannot be a token either. Therefore, it cannot be contextually manifest. We can therefore conclude that prepositional particle verbs are represented as in (68):17 VP

(68)

PP

V0 werf-

P0 ein

The claim that particle phrases are bare PP-complements of V has another important implication. As I argued in Section 3.1.4, the case assigning properties of a preposition are only realized if functional structure is present. This means that particles cannot assign case to their complement DPs; they are “unaccusative” (cf. Koopman 1993). This explains why the reference object of a particle verb cannot be realized as the complement of P0; in this position, it simply could not receive case. Therefore, even if the direct object of a prepositional

17.The claim that particle phrases lack functional structure is also made in Koopman (1993) (see also Haiden 1997). However, Koopman’s proposal differs from mine in that she draws a distinction between “idiomatic” and “directional” particles. She argues that only the former are syntactically represented as bare PPs, whereas the latter are still associated with functional structure. This assumption is problematic in the light of the examples that I discussed in (66) and (67b) in order to illustrate McIntyre’s (2001) Landmark Referentiality Generalization. The particles ein and aus in these examples obviously have a directional meaning. Nevertheless, these particles are clearly non-referential, which suggests that functional structure is absent in these examples as well, contrary to what is claimed by Koopman.

141


particle verb is semantically interpreted as the reference object of the particle, as in the examples in (69), it can only be realized outside the particle phrase, as shown in (70). (69) a.

Peter eilt dem Bahnhof zu (der Bahnhof = reference object of zu) P. runs the station part ‘Peter runs towards the station’ b. Peter schüttet den Eimer aus (der Eimer = reference object of aus) P. pours the bucket part ‘Peter pours out the bucket’

(70) [vP den Eimer [VP [V¢ [PP aus] schütt-]]]

In the following, I represent particle phrases as in (68), i.e. as “bare” projections of the particle that do not include any arguments. The internal arguments of particle verbs, regardless of whether they are semantically arguments of the particle or of the verb, will syntactically always end up outside the particle phrase, as shown in (70).18,19 The following observation from Hoeksema (1988) (mentioned in Abraham 1995:362) provides additional support for this idea. Relative pronouns in Dutch can occur as w-forms or as d-forms, depending on whether the pronoun refers to an argument of a verb or to an argument of a preposition:

18.There are three possible ways in which this assumption can be made more precise. The reference object of the particle could be (i) generated in SpecVP and moved to SpecvP (ii) generated inside the PP and moved to SpecvP, and (iii) directly merged into SpecvP. As I pointed out in note 4, the only thematically relevant position is the position in which an argument is interpreted at the conceptual interface level of syntactic structure. Since under all three alternatives, this position will be SpecvP, we do not have to decide here which of the three options (i)–(iii) is correct. For convenience, I represent all arguments of particle verbs as being generated in SpecVP from where they undergo movement to SpecvP. 19.Some particles realize their arguments as full FPPreps, which gives rise to the existence of pleonastic particle constructions like (i) which are discussed in Olsen (1996, 1997b): (i)

Er schüttete das Wasser aus dem Eimer aus he poured the water out.of the bucket part(out) ‘He poured the water out of the bucket’

It assume, following Olsen (1997b), that the prepositional phrase aus dem Eimer in (i) is adjoined to the particle phrase or to VP. I will not be further concerned with constructions like (i) in the following chapters.


(71) a.

de vrouw aan wie/*die ik nu denk the woman at who/that I now think ‘the woman about whom I now think’ b. de vrouw die/*wie deze brief leest the woman that/who the letter reads ‘the woman who reads the letter’

argument of P: Æ w-form argument of V: Æ d-form

If a relative pronoun refers to the argument of a particle verb, the d-form is chosen: (72) de vreemdeling *wie/die wij aanspraken the stranger *who/that we part-spoke ‘the stranger to whom we spoke’

If the relative pronoun in (72) was generated as the argument of the particle aan, we might expect that it has the same form as the relative pronoun in (71a), which is an argument of the preposition aan. However, since the relative pronoun in (72) is die, we can take this to be additional support for the assumption that DP-arguments of particle verbs are syntactically generated outside of the particle phrase. 3.2.3 Adjectival particles and resultatives I have argued that nominal and prepositional particles are represented as phrasal complements of the verb that lack a functional shell. Furthermore, I have shown that the lack of functional structure is semantically reflected in the non-referentiality of the particle phrase. This property distinguishes nominal particles from full DP-complements and prepositional particles from functional prepositional phrases. In the following, I will argue that adjectival particles follow the same pattern; their phrasal projection lacks functional structure. The question is whether the difference between adjectival particles and functional adjectival phrases can also be captured in terms of the referentiality/nonreferentiality difference that was discussed in the previous sections. This question is hard to answer, since it is not clear in what sense an adjective does or does not refer. Adjectives typically express Properties, and, as Jackendoff (1990) notes, Properties do not clearly differentiate tokens and types. What I suggest here is that phrasal adjectival complements in resultative constructions are functional projections, and I take the absence of a resultative interpretation as evidence for the lack of functional structure.


Recall from Section 3.1.3 that adjectival agreement in languages like French was taken as evidence for the presence of a functional head above the projection of the adjective: (73) Marie est petite M.-sg-fem is small-sg-fem ‘Marie is small’ (74) [AgrP Mariei [Agr¢ petitej [AP DPi A0j]]]

The agreement morpheme which is visible on the adjective in (73) is associated with Agr0. Agreement between the adjective and its argument is determined by this functional projection. The copula verb être, ‘be’, in (73) is unaccusative. Therefore, the DP-argument of the adjective in (73) must raise to SpecTP in order to receive case; it becomes the subject of the sentence. In contrast, in resultative constructions, v0 may be present to assign an Agent-role. The argument of the adjective then raises to SpecvP to receive accusative case (cf. Levin & Rappaport 1995): (75) a.

Peter redet sich heiser P. talks refl hoarse ‘Peter talks himself hoarse’ b. Peter tritt seine Tür kaputt P. kicks his door broken ‘Peter kicks his door to pieces’

(76) [vP sichi [VP [AgrP DPi¢ [AP DPi A0j] heiserj]]]

The assumption that adjectival resultatives are represented by functional Agrphrases gains strong support from Norwegian, which shows overt agreement between adjectives and their arguments in resultative constructions: (77) a.

Vi vaska golvet rein-t we washed floor-sg-neut clean-neut ‘We washed the floor clean’ b. *Vi vaska golvet rein (Åfarli 1985: note 8)

As (77) illustrates, the suffix -t which indicates (neutral) gender agreement between golvet and the adjective is obligatory. This shows that the verbal


complement in a resultative construction is a full functional adjectival projection AgrP.20 How can we show that functional structure is absent with adjectival particle verbs in German? Since postnominal adjectives do not agree overtly, visible agreement cannot be used as a test. What I will suggest in the following is that resultative constructions require their predicates to have an extended projection. Consequently, the absence of a resultative interpretation in a V-Adj construction suggests the presence of an adjectival particle verb. The meaning of (75a) can be paraphrased as “Peter causes himself to become hoarse by talking” or “As a result of his talking, Peter becomes hoarse”. It is an open question how this particular resultative (or causative) interpretation is licensed in these constructions. Jackendoff (1990) postulates a correspondence rule that links the particular syntax of this construction to its characteristic semantics. This correspondence rule expresses the fact that a syntactic construction with a verb and an adjectival complement corresponds to an interpretation that can roughly be paraphrased as follows: the state expressed by the adjective holds for the direct object as a result of the event expressed by the verb. Similarly, von Stechow (1995) argues that the meaning of a resultative construction is determined by a CAUSE and a BECOME operator (cf. Dowty 1979) that are interpolated by a semantic strategy when the verb phrase is interpreted. He also postulates a specific rule that stipulates that the relevant causative interpretation is automatically associated with a syntactic structure V + adjectival phrasal complement. In order to capture the fact that a resultative interpretation can only be associated with a structure when a functional adjectival projection is present (as the data from Norwegian tell us), we could simply stipulate that this information is part of (the syntactic part of) the correspondence rules suggested by Jackendoff (1990) and von Stechow (1995). However, I will try rather to derive the correspondence between a resultative interpretation and functional structure

20.The word order of the resultative construction in (77) is V-Obj-Adj. Interestingly, if the order of adjective and object is reversed, agreement is optional: (i)

a.

b.

Vi vaska golvet reint/*rein we washed (the) floor clean ‘We washed the floor clean’ Vi vaska rein/reint golvet (Åfarli 1985: note 8)

Æ

normal order, cf. (77a)

Æ

reversed order

I will discuss the absence of agreement in (i-b) in Chapter 7, Section 7.2.

145


from speculations about what it could mean for an adjective to be referential. Suppose that an adjective is referential if it can be predicated of an appropriate argument that has to be syntactically realized. Intuitively, such an assumption makes sense: The concept expressed by an adjective is a token and can be mentally projected only if there is a mental representation of a Thingconcept that this property can be applied to. The state expressed by an adjective can be interpreted as a token only in the sense that this adjective is predicated of a particular individual. In other words, only in ascribing a property to an individual can we interpret it as a specific Property-concept. If the claim made above is correct, and resultative constructions really require a referential resultative phrase and hence an AgrP, then we predict that the argument of a resultative predicate must always be expressed overtly. Interestingly, this prediction seems to be borne out. It is a general requirement that in resultative constructions, the Theme-argument of the adjective always has to be represented syntactically, either as a full DP or as a reflexive coindexed with the subject DP, as in (75a). It is very odd to omit the direct object in a resultative construction, even if the base verb can be used intransitively: (78) a. Harry hammered the metal b. Harry hammered (79) a. Harry hammered the metal flat b. *?Harry hammered flat

If a transitive verb is used intransitively, as in (78b), it still can be assigned an interpretation (“Harry hammered something”). Although this should be possible in resultative constructions as well (“Harry hammered something flat”), (79b) is odd. I assume that the argument of an adjectival resultative predicate must be present in order to license the referential interpretation of the adjective. If the Theme-argument is obligatory in resultatives, and its presence corresponds indeed to a referential interpretation of the adjectival phrase, then we can take this as further support for the claim that resultatives require referential resultative predicates, as already suggested by the Norwegian data in (77). Since referentiality is expressed by functional structure, this means that the AP in resultative constructions must be dominated by an Agr-projection. This in turn implies that a bare AP-complement of the verb can never trigger a resultative interpretation. Therefore, if we find a verb-adjective construction without a resultative meaning, it is likely that this situation is the result of the lack of functional structure, i.e. that we are dealing with an adjectival particle verb:


(80) a.

Peter hat krank gefeiert P. has part(sick) celebrated ‘Peter played hooky’ b. Peter hat kurz getreten P. has part(short) kicked ‘Peter went easy’

VP

(81)

AP

V0 feier-

A0 krank

Adjectival particles are bare APs, but bare APs do not yield a resultative interpretation. Notice that the examples in (80) do not include the conceptual arguments of the adjectives krank and kurz. If we add internal arguments that can function as possible arguments of these adjectives, the constructions in (80) immediately receive a resultative interpretation: (82) a.

Peter hat [seine Nachbarn krank] gefeiert P. has [his neighbors sick celebrated ‘Peter had so many parties that his neighbors finally became sick’ b. Peter hat [die Tür kurz und klein] getreten P. has [the door short and small kicked ‘Peter kicked the door to pieces’

In sum, if an adjective does not project functional structure, it cannot refer, which means that it is impossible to derive a resultative. Therefore, adjectival particle verbs can never have a resultative interpretation.

3.3 Conclusion In this chapter I have discussed the assumption that regular complements of the verb are extended projections of lexical heads, and I have provided evidence for the claim that particle phrases are exceptional in this respect, since they are bare PPs, NPs, or APs. The main claim of this chapter, defined as in (40) above, is repeated in (83):


(83) Particle definition Particles are heads of non-functional phrasal complements of the verb and do not leave their base position in overt syntax

(83) yields (84) as a definition of the notion “particle verb”: (84) Particle verb definition A particle verb consists of a verb and the head of its non-functional phrasal complement

The definition in (83) provides a solution to what Lüdeling (1998a) calls the “delimitation problem”: How can particle verbs be distinguished from “similar constructions” like e.g. resultatives? Lüdeling suggests that they cannot; she argues that particle verbs are “similar” to other constructions with a verb and a phrasal complement. In Chapter 2 I have shown that the verb-particle construction is in fact similar with respect to the phrasal status of particles. Particle phrases and “regular” phrasal complements are different, however, because particle phrases are non-functional. Therefore, and in contrast to what is assumed by Lüdeling, (83) shows that a clear definition for the class of particles can be given in purely syntactic terms. However, it can only be given on the basis of an articulated syntactic structure that takes into account the fact that “regular” phrasal complements of the verb are functional. (84) goes one step further and shifts the focus from the phrasal representation of the particle to the relation between the head of this phrase and the head that selects it. (84) defines a particle verb as consisting of two heads, not as a verb plus complement. According to the definition of “structural adjacency” that I introduced in Chapter 1, Section 1.3.1, the verb and a particle are structurally adjacent. This means that the verb-particle construction is characterized by a specific local relation between two lexical heads: VP

(85)

PrtP

V0 Structural adjacency

Prt0

Usually, a verb is structurally adjacent to a functional head, since regular complements of V0 are DPs, CPs, or FPPreps etc. One exception to this generalization


has already been discussed in Chapter 1, Section 1.3.3. There I discussed the derivation of deadjectival verbs in the syntax, where V0 is an affix which selects a bare AP. The affixal character of V0 then forces the head of its lexical complement to incorporate; consequently, the two lexical heads in these examples are part of the same word at the interface level of syntactic structure. The second exception are verb-particle constructions. Here we also have a verb and a non-functional complement, but since the verb is not an affix, no incorporation is required. Therefore, the two lexical heads do not form a morphological object, but are merely structurally adjacent. Particle verbs hence occupy an intermediate position; they are not words, but the relation between the two terminal nodes that they consist of is nevertheless more local than is the case with other verb-complement constructions. The “classical” local domain of two lexical terminal nodes is defined morphologically; i.e. the terminal nodes form a complex word. Crucially, particle verbs establish a local domain which is non-morphological. This aspect will become relevant in the following chapters of this study where I argue that the “word-like” properties of particle verbs follow from the two definitions given in (83) and (84).

Chapter 4

The lexical representation of particle verbs

In Chapter 3, I provided evidence for the claim that particle verbs are syntactically represented as verbs with non-functional complements, and I argued that particle phrases differ in this respect from “regular” syntactic complements of V0, which are functional. This of course raises the question of how this “defective” status of particle phrases is licensed. Why does syntax create a structure with a non-functional complement of the verb at all? In this chapter I will provide an answer to this question by looking at the semantics of a subset of particles, i.e. those of category P.1 My starting point is the observation that the semantics of prepositional particles differs in various ways from the semantics of the corresponding prepositions. I will try to combine this observation with the syntactic definition of a particle that I gave in Chapter 3. I suggest that the specific meaning of particles can only be associated with a terminal node in syntactic structure if this terminal node is structurally adjacent to a verb. In other words, I argue that the exceptional syntactic status of particle verbs is a requirement listed as part of their lexical representation. Before I make this claim explicit in Section 4.2, I first discuss the various ways in which the semantics of a particle can combine with the semantics of a verb in Section 4.1. I illustrate how the meaning of transparent particle verbs can be derived on the basis of the meaning of their parts, and I show that the idiosyncratic properties of non-transparent particle verbs are captured if they are analyzed as phrasal idioms and hence treated along the lines of the idioms discussed in Chapter 1. In Section 4.2, I then discuss the contextual condition that makes a lexical element a particle, and I develop what I have labeled a classbased theory of particle meanings. I argue that the meaning of a particle depends on properties of the verbal head that it is structurally adjacent to, and

1.The meaning of prepositional particles has received the most attention in the literature. Therefore, I will focus on these elements when I investigate the lexical and morphological properties of particle verbs.

152


in Section 4.3, I extend this analysis to “regular” prepositions. The key conclusion that I develop is that a prepositional element is represented through one single lexical entry which lists its syntactic category and its phonological form. Its meaning, however, depends on its syntactic environment; i.e. on whether it is structurally adjacent to a verb or to a functional head.

4.1 The semantics of particle verbs In this section I discuss the different ways in which particles combine with (the meaning of) their base verbs. Consider again the syntactic structure of particle verbs that I motivated in Chapters 2 and 3: VP

(1)

Spec

V′ PP

V0

P0

(1) is the representation of particle verbs at the interface level of syntactic structure that is accessed by the semantic component. In this process, the lexical entries of the particle and the verb are unified with the terminal nodes P0 and V0, respectively. They contribute their lexical indices, and the terminal nodes are linked to the meaning of the base verb and the particle. In the case of semantically transparent particle verbs, the combination of the meanings of these two elements yields a complex semantic representation that corresponds to the meaning of the particle verb at the level of V¢. Studies in the lexical semantics of German particle verbs (cf. e.g. Olsen 1996, 1997a; Stiebels 1996; McIntyre 2001) reveal that for many particle verbs a transparent semantic analysis is indeed available. In Section 4.1.1 I discuss particles that saturate argument positions in the LCS of their base verbs. In Section 4.1.2 I turn to particles that do not saturate argument positions of the base verb, but that combine with the verb to form a complex predicate, and in Section 4.1.3 I look at aspectual particles. Non-transparent particle verbs are


discussed in Section 4.1.4, where I show how the theory of lexical licensing discussed in Chapter 1 straightforwardly accounts for these verbs. 4.1.1 Particles as arguments The following examples show that particles can sometimes be substituted for regular functional prepositional phrases: (2) a.

Peter stellt den Topf [FP auf den Boden] P. puts the pot onto the floor ‘Peter puts the pot onto the floor’ b. Peter stellt den Topf [PP ab] P. puts the pot part ‘Peter puts the pot down’

(3) a.

Peter lädt die Koffer [FP auf den Wagen] P. loads the suitcases onto the cart ‘Peter loads the suitcases onto the cart’ b. Peter lädt die Koffer [PP auf] P. loads the suitcases part ‘Peter loads the suitcases onto something’ die Tür] Peter lehnt sich [FP an P. leans refl against the door ‘Peter leans against the door’ b. Peter lehnt sich [PP an] P. leans refl part ‘Peter leans against something’

(4) a.

(5) a.

Peter klebt ein Pflaster [FP auf die Wunde] P. glues a plaster [[[[on the wound ‘Peter applies a plaster to the wound’ b. Peter klebt ein Pflaster [PP auf] P. glues a plaster part ‘Peter applies a plaster’

(6) a.

Peter geht [FP aus dem Haus] P. goes out.of the house ‘Peter leaves the house’ b. Peter geht [PP aus] P. goes part ‘Peter goes out’

153

154


Following Jackendoff (1983, 1990), I assume that in each of the verbs in (2)–(6), the LCS includes a Path-argument slot which is linked to a regular functional prepositional phrase in the (a)-examples. The (b)-examples illustrate that the same argument position can be filled by a particle. (2)–(6) therefore exhibit examples of particles that function as arguments of the verb (cf. Stiebels & Wunderlich 1994; Stiebels 1996; Olsen 1996, 1997a). (7) further illustrates this point by giving the conceptual structures associated with the relevant parts of the VPs in (3) (I have given a simplified LCS of the verb laden, ‘load’): (7) a. laden: [Event CAUSE ([Thing]A, [Event GO ([Thing]A, [Path]A)])] b. auf den Wagen: [Path TO ([Place ON ([Thing INSTANT OF TYPE: CART])])] c. auf: [Path TO ([Place ON ([Thing])])] d. auf den Wagen laden: [Event CAUSE ([Thing]A, [Event GO ([Thing]A, [Path TO ([Place ON ([Thing INSTANT OF TYPE: CART])])])])] e. aufladen: [Event CAUSE ([Thing]A, [Event GO ([Thing]A, [Path TO ([Place ON ([Thing])])])])]

As (7a) shows, the verb laden requires an obligatory Path-argument. This argument position can be filled by either a functional prepositional phrase or by a particle. In (7d) and (7e), the Path-concepts expressed by the prepositional phrase and the particle are simply merged into the respective slot in the LCS of the base verb. As was noted in Chapter 3, Section 3.2.2, the Path expressed by the particle is interpreted non-referentially, which follows from the absence of functional structure. The meaning of the particle in (7c) is clearly based on the meaning of the corresponding preposition. The same holds for the particle meanings in (4)–(6); the only difference between the (a)- and (b)-examples is that the reference object of the preposition is syntactically represented in the former cases, whereas it remains implicit in the case of particle verbs. The observation that many particles differ from prepositions only with respect to the absence of an overt reference object has led researchers to label particles “intransitive prepositions” (cf. Emonds 1972; van Riemsdijk 1978; den Dikken 1995). However, not all particles leave their reference object implicit. The example in (8) illustrates that the particle zu saturates the predicative argument slot of the verb werfen, ‘throw’, but the reference object of the preposition is realized as the dative argument of the particle verb zuwerfen (cf. Olsen 1997a): (8) a.

Peter wirft den Ball [FP zu dem Kind] P. throws the ball-acc to the child


b. Peter wirft dem Kind den Ball [PP zu] P. throws the child-dat the ball-acc part ‘Peter throws the ball to the child’

In (8a), the reference object (the Goal-DP das Kind, ‘the child’) is the complement of the preposition; it receives case from P0. Since particle phrases lack functional structure and hence cannot assign case to their arguments, the GoalDP in (8b) must receive case in SpecvP; it is hence realized syntactically as an argument of the complex predicate zuwerfen. However, the resulting LCSs of the VPs in (8) are identical.2 In both examples, the DP dem Kind is interpreted as the reference object of the prepositional element zu. In the next section, I will discuss more examples where semantic arguments of prepositional particles are realized as internal arguments of the particle verb.3 4.1.2 Semantic adjunction and eventive particles A particle can also combine with a verb if this verb does not provide the appropriate argument position for prepositional phrases. I will call the semantic combination of a base verb and a particle in these cases “semantic adjunction”.4 Consider the examples in (9)–(11): (9) a.

Peter arbeitet P. works ‘Peter works’ (intransitive base verb) b. Peter arbeitet seine Schulden ab P. works his debts part ‘Peter works off his debts’ (transitive particle verb)

(10) a.

Peter spricht P. speaks ‘Peter speaks’ (intransitive base verb)

2.This is not quite true; there is a slight meaning difference between (8a) and (8b) that will be discussed in Section 4.2.3. 3.In a Montague-based semantic framework, the operation that combines the particle and the verbs in (2)–(6) is formally different from the operation that forms the particle verb in (8). The particles in (2b)–(6b) combine with the verb via Function Application. However, since the particle in (8b) has one more unsaturated argument than the particles in (2b)–(6b), the operation that combines zu and the verb in (8b) is Function Composition. 4.Stiebels (1996) calls this process “lexical adjunction”.

155


b. Peter spricht dem Mädchen nach Peter speaks the girl part ‘Peter speaks after the girl’ (transitive base verb) (11) a.

Peter hustete P. coughed ‘Peter coughed’ (intransitive base verb) b. Peter hustete ihr eine dicke Erkältung an Peter coughed her a thick cold part ‘She caught a strong cold from Peter’s coughing’ (double-object particle verb)

The base verbs arbeiten, ‘work’, sprechen, ‘speak’, and husten, ‘cough’, in (9)–(11) are intransitive and do not provide argument positions for prepositional phrases. Nevertheless, they can combine with particles. In this process, the particle adds its semantic argument(s) (cf. Stiebels & Wunderlich 1994; Stiebels 1996). For example, the particles ab and nach in (9) and (10) introduce one argument, and an in (11) introduces two arguments. Semantic adjunction also explains the examples in (12)–(14): (12) a.

Peter schreibt einen Brief P. writes a letter ‘Peter writes a letter’ (direct object = Theme) b. Peter schreibt das Finanzamt an P. writes the inland revenue office part ‘Peter writes to the inland revenue office’ (direct object = Goal)

(13) a.

Peter trinkt ein Bier P. drinks a beer ‘Peter drinks a beer’ (direct object = Theme) b. Peter trinkt sein Glas aus P. drinks his glass part ‘Peter empties his glass’ (direct object = Source)

(14) a.

Die Polizei sucht den Mörder von Jones the police search the murderer of J. ‘The police are looking for the murderer of Jones’ (intensional transitive verb) b. Die Polizei sucht das Feld ab the police search the field part ‘The police scour the field’ (extensional transitive verb)


In (12)–(14), the particle does not change the valency of the verb; both the particle verb and the respective base verb are transitive. However, as the (b)examples show, the particle has changed the selectional properties of the verb. This can be explained by assuming that schreiben, ‘write’, trinken, ‘drink’, and suchen, ‘search’, are used intransitively in the (b)-examples, and that the internal argument is introduced by the particle, as in (9)–(11). The derivation of the particle verbs in (9)–(14) is an instance of complex predicate formation. The particle does not saturate an argument position, but rather merges with the verb’s meaning to derive a complex LCS which incorporates semantic aspects of both the particle and the verb. The internal arguments are semantically arguments of the particle; the main event expressed by the particle verbs is characterized by the base verb. The question is, how can we formally account for the fact that in (9)–(14), the particle has combined with a verb, although the verb does not provide an appropriate argument position? In most theories on particle verbs that are concerned with comparable examples, the particles in (9)–(14) are taken to differ from prepositions in that they do not express spatial concepts, but Events. For example, van Hout (1998) assumes that particles, like verbs, express “event types”, and she proposes an operation called Event Type Composition that merges the two eventive predicates into one. This principle is similar to Kratzer’s (1994) principle of Event Identification, which allows the combination of two eventive predicates, even though neither of them provides an appropriate argument position for the other.5 If it is assumed that particles are also eventive predicates, the combination of V0 and the particle can be treated as another instance of Event Identification (cf. Zeller 1996 for a proposal along these lines). Stiebels (1996) adopts Wunderlich’s (1995) operation of Argument Extension, a lexical rule that simply modifies the lexical entry of the base verb and adds a predicative argument position which can be saturated by eventive predicates. Thereby, the particle contributes its argument(s) to the argument structure of the particle verb. All these proposals have in common the idea that in the process of combining the verb and the particle, two eventive predicates are unified.6

5.In Kratzer (1994), Event Identification is introduced in order to account for the semantic combination of the main verb V0 and Voice0 (= the light verb v0); both heads are semantically interpreted as eventive predicates in Kratzer’s theory. 6.The representation of particles as expressing events makes their meaning similar to that of verbs. This fact is particularly interesting in the light of results in computational linguistics. As Gamon & Reutter (1996) show, the difficult properties of German particle verbs can

157

158


In adopting Jackendoff’s (1983, 1990) framework, one is also forced to assume that certain particles express Events. Recall that according to Jackendoff, prepositions do not express relations, but spatial concepts with only one argument, the reference object. However, there are examples where the particle also introduces the Theme that traverses the Path expressed by a particle as an argument of the particle verb. For example, in order to capture the ditransitivity of the particle verb anhusten, ‘(give X a cold by) coughing’, in (11), it is necessary to assume that an has introduced both arguments, because the base verb husten, ‘cough’, is intransitive. However, the conceptual constituent that includes both the Theme and the reference object of a local relation is an Event in Jackendoff’s framework. Therefore, the derivation of particle verbs like anhusten requires the assumption that the LCS of the particle looks like (15b), since (15a) alone does not represent the Theme: (15) a. [Path TO ([Place AT ([Thing ]A)])] b. [EventGO ([Thing ]A, [Path TO ([Place AT ([Thing ]A)])])]

Finally, the assumption that particles express Events is reminiscent of Lieber & Baayen’s (1993) analysis of verbal prefixes in Dutch. Lieber & Baayen adopt Jackendoff’s (1983, 1990) formal system and represent verbal prefixes like verand be- as Motion-Events, arguing that this representation allows it to reduce all possible uses of these prefixes to one single lexical entry. From now on, I will assume that whenever a particle combines with a verb via semantic adjunction, its LCS corresponds to an Event. 4.1.3 Aspectual particles and meaning variation As the examples in (16)–(18) show, some prepositional particles function as aspectual markers: (16) a.

Peter brät das Fleisch P. fries the meat ‘Peter fries the meat’

best be implemented in the Microsoft Natural Language Processing System by treating them as elements of the category VERB that can expand into phrases. The rule that performs the lookup of the complex verb (consisting of the base verb and the particle phrase) can then be constructed parallel to rules that combine auxiliaries with participial VPs.

The lexical representation of particle verbs 159

b. Peter brät das Fleisch an P. fries the meat part ‘Peter fries the meat lightly’ (17) a.

Der Zug rollt the train rolls ‘The train is rolling’ b. Der Zug rollt an the train rolls part ‘The train starts to roll’

(18) a.

Peter schreit P. screams ‘Peter is screaming’ b. Peter schreit auf P. screams part ‘Peter gives a scream’

The particle an in (16) indicates that the event expressed by the base verb is only carried out “partially” or “lightly”; in (17), the same particle expresses ingressive aspect. The particle auf in (18) gives rise to a punctual reading of the event expressed by the verb. Apart from the fact that the particles in (16)–(18) are homophonous with prepositions, there does not seem to be any synchronic relation between the aspectual meaning of these particles and the meaning of the corresponding prepositions. Aspectual meanings are just the most drastic way in which particles differ semantically from their underlying prepositions. However, there are also other differences. Consider the particle verbs in (19): (19) a.

Peter reist ab P. travels part ‘Peter leaves to travel’ Æ Theme moving away from implicit reference object b. Peter fegt den Staub ab P. sweeps the dust part ‘Peter sweeps off the dust’ Æ Agent removes Theme from implicit reference object c. Peter fegt den Fußboden ab P. sweeps the floor part ‘Peter sweeps the floor’ Æ Agent clears the Source of implicit Theme


d. Peter schwatzt mir die CD ab P. chatters me the CD part ‘Peter talks me into giving him the CD’ Æ Theme changes possession from Source to Agent e. Peter büßt seine Strafe ab P. atones his sentence part ‘Peter serves his sentence’ Æ metaphorical (decremental) “from” (the sentence is reduced by removing time from it)

The relation between the meanings of the particle ab and the meaning expressed by the corresponding preposition varies in (19). For example, the particle ab in (19a) and (19b) roughly corresponds to the preposition ab (which is similar to von, ‘from’) in expressing a Path which leads “away from” somewhere.7 The reference object is implicit. The prepositional meaning of ab is also present in (19c), but here, the direct object denotes the reference object, and the Theme is implicit. In (19d), ab expresses a possessive relation, and both arguments are realized (cf. (11) in Section 4.1.2). Finally, ab in (19e) denotes events in which individuals reduce their size, time, length etc. (cf. (9) in Section 4.1.2); here, the meaning of the underlying preposition is present at most in a metaphorical sense. Nevertheless, I will argue below that the verbs in (19) are all based on one single lexical representation of the element ab. I will assume that, regardless of their semantic differences, particles and prepositions with the same phonological form are always represented through the same lexical entry, with their different meanings licensed by different syntactic environments. Since it is possible to associate all meanings in (19) with one lexical item ab, there is nothing problematic about extending this approach to aspectual meanings as well.8

7.The use of ab as a spatial preposition is restricted to a few expressions like ab hier, ‘from here’. However, ab still occurs as a preposition in temporal uses as in ab heute, ‘from today on’. The relation between the particle ab and the preposition von is discussed in Stiebels (1996, Section 6.1.3). 8.Notice that it is likely that prepositions and aspectual particles have the same historical root. For example, both the spatial and the aspectual use of auf might be diachronically related to the Old German adverb uf (cf. Stiebels 1996: 62).


4.1.4 Non-transparent particle verbs as phrasal idioms As is well-known, the meaning of particle verbs is not always as transparent as was the case with the verbs discussed in the preceding sections. In (20), I give a few examples of particle verbs with idiomatic interpretations: (20) a.

anfangen, ‘start’ part-catch, lit.: on-catch b. aufhören, ‘stop’ part-hear, lit.: up-hear c. (jemanden) einwickeln, ‘twist somebody round one’s little finger’ (somebody) part(in)-wrap; lit: wrap somebody into (something) d. auftauchen, ‘arise, appear suddenly’ part(up)-dive, lit.: ‘surface’

The particle verbs anfangen, ‘start’, and aufhören, ‘stop’, in (20a) and (20b) are the “classical” examples of irregular particle verbs in German. Their meaning seems to be entirely unrelated to their parts. The two examples in (20c) and (20d) are (more or less) metaphorically related to existing regular formations. The use of einwickeln in (20c) is an idiomatic version of the regular particle verb (in Papier) einwickeln, ‘wrap up (in paper)’, and the metaphorical use of (20d) is based on the literal meaning of auftauchen, ‘surface’, as a combination of the meaning of auf as ‘upwards’ and the meaning of the base verb. The conclusion that the constructions in (20) are lexically listed is unavoidable. Since the lexicon precedes the syntax under a lexicalist view, lexicalists have taken this observation as evidence for the claim that particle verbs are syntactic atoms, i.e. X0s. However, in Chapter 1 I have argued against this conclusion. The alleged conflict between semantic and syntactic properties of particle verbs results from the misconception of lexical insertion; the semantic irregularities of particle verbs pose a problem for a syntactic analysis only from the lexicalist’s point of view, according to which lexical items with special meanings are always X0s. However, as I have also shown in Chapter 1, a vast number of lexical items (or “listemes” in Di Sciullo & Williams’ (1987) terminology) associate idiosyncratic meaning with syntactic structures larger than words. Phrasal idioms and other fixed expressions are characterized by a special meaning that corresponds to a non-minimal syntactic structure. It was also shown in Chapter 1 that phrasal idioms do not raise problems for the theory of lexical licensing. Therefore, lexical licensing is the appropriate method to capture the idiosyncratic properties of particle verbs like those in

161


(20). Although particle verbs are syntactic constructions, they may be associated with special meanings listed in the lexicon. I hence assume that the lexical representation of idiomatic particle verbs like those in (20) mirrors that of phrasal idioms. To my knowledge, Emonds (1972) was the first who pointed out that idiomatic verb-particle constructions in English behave essentially like idioms consisting of a verb and a full prepositional phrase (e.g. take X for a ride, put X to the test etc.) and therefore should also be analyzed as syntactic V + prepositional phrase-constructions. The same reasoning of course applies to particle verbs in German and Dutch (cf. Lüdeling 1998a; Wurmbrand 1998). Idiomatic particle verbs like aufhören, ‘stop’, and idiomatic expressions like in der Tinte sitzen (lit. sit in the ink), ‘to be in the soup’, can be analyzed along similar lines. In both examples, a verb (hören, ‘hear’, and sitzen, ‘sit’, respectively) combines with a prepositional phrase, and the whole complex V + phrasal complement is linked to a special meaning. The lexical entry for a particle verb like aufhören is given in (21): (21) a. LPS b. (Links to the lexical entries of aufa and hörenb)

PP

LSS

c. LCS [Event STOP ([Event ])]x

n, n>0

V

x

V0b

P0a

The syntactic structure that the meaning of the particle verb aufhören is linked to is a non-minimal projection of V consisting of the verb and the particle phrase. The meaning of the particle verb that I provided in (21c) is associated with the syntactic node that dominates the particle and the verb. If V¢ is merged with an argument, this argument fills the conceptual argument position in (21c). Importantly, the lemma that links (21b) and (21c) corresponds to two lexical forms; i.e. the linkups between the terminal nodes in (21b) and the phonological realizations of the verb hören and the particle/preposition auf, respectively. Importantly, this means that the lexical indices a and b establish a link to two other lexical entries; in a sense, (21) can be regarded as a special branch of the entry for the base verb and as a special branch of the entry for auf. (21) provides the information that the “regular” meanings of these two entries are overwritten by (21c) when the two occur together in the syntactic context specified in (21b). Of course, we do not want that the entry of an idiomatic expression can


only be unified with a syntactic structure when the verb is in its base position. Taking it as it stands, in order for the lexical entry in (21) to be unified with a syntactic structure, the phonological representations of the terminal nodes must be associated with the positions inside the VP. The sentence in (22), however, creates an immediate problem for lexical licensing: (22) Peter hörte endlich auf P. heard finally part ‘Peter finally stopped’

In (22), the finite verb has moved to Comp0. Although the VP generated by the syntactic module still looks as in (23) (because the moved verb has left a copy inside its base position), a strict interpretation of the lexical entry in (21) would always require that the copy of the verb be spelled-out. VP

(23)

PP

V0i

verb movement

P0

Notice that the same problem is raised by phrasal idioms in general, which also permit verb movement: (24) Peter sitzt in der Tinte P. sits in the ink ‘Peter is in the soup’

In both (22) and (24), movement of the verb to Comp0 does not affect the acceptability of an idiomatic reading. This observation is interesting, because normally, there are stringent restrictions on the syntactic accessibility of parts of an idiom. In (25) and (26), this situation is illustrated with respect to passive, adverbial modification, and wh-movement (cf. Abraham 1989: 15; Jackendoff 1997: 166):9

9.Jackendoff (1997), following proposals by Bresnan (1978), Wasow, Nunberg & Sag (1984), Ruwet (1991), and Nunberg, Sag & Wasow (1994), suggests that “fixed” idioms like the ones in (25) cannot be decomposed into smaller irregular readings, whereas idioms with mobile chunks are composed of “subidiomatic” parts.


(25) a. *The bucket was kicked by John b. *The towel was thrown in by Bill (26) a. *weil Peter in der dicken Tinte sitzt because P. in the thick ink sits ‘because Peter is in the soup’ b. *Wo sitzt Peter? — In der Tinte ‘Where does Peter sit?’ — ‘In the soup’

In contrast to the syntactic operations illustrated in (25) and (26), verb movement seems to be a syntactic operation that may access the syntactic structure of idioms. It follows from the copy theory of movement that the configuration that is necessary for lexical licensing of an idiom is not destroyed; however, we still have to allow for the LPS of the verb to be associated with the head of the chain created by verb movement. I hence postulate the following correspondence rule: (27) Condition on unification of V0 Add the lexical index that links the LPS of a verb to V0 in syntactic structure to the highest position of the V0-chain

(27) is a descriptive statement that captures the fact that the Vocabulary item that corresponds to a verb is associated with the head of the verbal chain. (27) does not specify the syntactic position that the LCS of the verb is associated with. It is still possible to interpret the verb inside the VP even though it is spelled out in a higher syntactic position. Notice that the stipulative character of (27) is not a consequence of the theory proposed here; rather, something like (27) has to be postulated in any theory that wants to account for the fact that verbs are not always spelled-out in the position where they are semantically interpreted. For example, it is an implicit assumption in most theories that incorporate semantics into the Principles-and-Parameters approach that verbs are interpreted in their base position, even if their LPSs are associated with a different link of the chain (cf. e.g. Kratzer 1994; Büring 1996). In other words, the assumption that verbs “reconstruct” into their base positions is generally accepted; the principle in (27) merely makes this assumption explicit.10

10.(27) rules out the option of “covert” verb movement. In fact, I assume that the effects attributed to covert head movement can also be captured by referring to local domains established between two heads that do not form words (see Chapter 3, Section 3.1.4, and Chapter 6, Section 6.2.2). However, even if covert movement existed, it could be incorporated into the proposal made here by formulating conditions that must be fulfilled in order for (27) to be overruled.


(28) illustrates how (27) accounts for the interpretation of sentences like (22) (I have given a simplified syntax for (22), with v0/Voice0 omitted for ease of exposition): CP

(28)

C′

Spec DP C0 Tns0i/j V0i

b

TP C0 Spec

Tns0j

T′ VP

V0i

PP P0

Tns0i/j

x

V0i

Tns0j

a

The structure in (28) is generated by the syntax. The syntactic VP in (28) fulfills the requirements that are stipulated by the LSS of aufhören; hence, the lexical entry in (21) can be unified with (28) and can add its lexical indices. Since (21) associates the lexical index x that links the particle verb to its LCS with the verbal node immediately dominating the verb and the particle phrase, the meaning of the particle verb is associated with the VP in (28). Unification of (21) further adds the index a to P0 and links the LPS of the prepositional element auf to this terminal node. The unification of the lexical entry of hören is now subject to (27), which states that the lexical index b must be added to the highest position of the verbal chain. The highest position of the verbal chain in (28) is in Comp0. Consequently, the LPS-entry of the verb must be linked to the verb in this position. The idiosyncratic meaning of aufhören is assigned to the structure in (28), even though not all of the Vocabulary items of the terminal nodes are associated with heads inside VP, as is “literally” required by the context specified in (21). To conclude this section, the discussion has shown that the semantic properties of particle verbs are well compatible with the syntactic approach that I defended in Chapters 2 and 3. Transparent particle verbs are derived by interpreting the conceptual interface level of syntax by associating special


meanings with the terminal nodes P0 and V0; non-transparent particle verbs are phrasal idioms. If we simply analyzed every particle verb with irregular semantic properties as a phrasal idiom, we could end the discussion here. However, recall that there is still the open question of why the particle does not project functional structure in syntax, although other lexical elements always have an extended projection. I will try to answer this question in the next section by looking at the lexical representation of particle verbs. As it turns out, the proposal that I develop has interesting consequences for the analysis of nontransparent particle verbs.

4.2 Local domains and class-based meaning In this section I am concerned with a claim made by Marantz (1997, in prep.) with respect to the “Encyclopedia”; i.e. the list of special meanings (or LCS-entries in the present terminology; see Chapter 1, Section 1.2.3). According to Marantz (in prep., 4), encyclopedic knowledge is associated with terminal nodes, but conditioned by the syntactic environment of these nodes. On the basis of this observation, I argue in Section 4.2.1 that the special meaning of particles can only be assigned to a lexical node if P0 is structurally adjacent to the verb. I argue that this contextual condition is part of the lexical entry of a particle element. In Section 4.2.2 I further suggest that the particular meaning of a particle depends on the semantic class of its base verb. In Section 4.2.3 I discuss some examples that are problematic for this approach, and I draw the conclusion that membership in a class of possible base verbs is determined by semiproductive rules. Finally, in Section 4.2.4, I extend this analysis to account for properties of certain idiomatic particle verbs. In so doing, I illustrate that many apparently non-transparent particle verbs can be analyzed compositionally; their irregular meaning can be derived regularly on the basis of the special meaning of their parts. 4.2.1 Special meaning and syntactic contexts As noted in Chapter 1, both Marantz (1997, in prep.) and Jackendoff (1997) emphasize that phrasal idioms provide a strong argument against the traditional idea behind lexical insertion. When special meaning is associated with syntactic structures larger than X0, this information must be listed somehow in the lexicon. However, the actual treatment of idioms suggested in Marantz (1997,


in prep.) is slightly different from Jackendoff’s analysis that I illustrated in Section 1.1 and that I adopted in Section 4.1.4 in order to capture particle verbs with a non-transparent semantic structure. Recall that in Jackendoff’s theory, a phrasal idiom is a lexical entry that connects a constituent at conceptual structure with a non-minimal syntactic structure (whereas the terminal nodes of this structure are regularly linked to the Vocabulary items specified by these lexical entries). For example, the idiom kick the bucket links the terminal nodes V0, D0, and N0 to the LPSs of the entries for kick, the, and bucket, but the idiomatic meaning ‘die’ is directly linked to the whole VP. A whole phrase is linked to a particular LCS; the complex syntax of the idiom is not reflected in its meaning. In contrast, Marantz (1997, in prep.) shows that certain aspects of the meaning of kick the bucket are in fact the result of its transitive syntax; these aspects are therefore not part of the meaning of the intransitive verb die. For example, Marantz observes that the punctual, active reading of a transitive clause is preserved in the idiom, leading to the contrast in (29): (29) a. He was dying for three weeks before the end b. *He was kicking the bucket for three weeks before the end (Marantz 1997: 212)

(29) shows that the verb die, which is an achievement, can be used in the progressive; die then changes its aspectual character and turns into an activity. This aspectual shift is not possible with kick the bucket; as Marantz points out, kick the bucket is aspectually more like pass away, and therefore does not allow the progressive. The difference between (29a) and (29b) cannot be captured if the VP in (29b) is simply linked to the same conceptual structure as die. It follows, however, if one assumes that the (aspectual) meaning of kick the bucket is still the meaning of a transitive verb with a definite DP-argument. The syntax in (29b) has a direct impact on the interpretation, just as the syntax of regular VPs does. Therefore, Marantz (1997, in prep.) suggests that in kick the bucket, it is not the whole VP that receives a special interpretation, but rather the terminal nodes that this particular syntactic structure is built of. V0, D0, and N0 are not only linked to their regular LPS-entries, but are also individually associated with LCS-meanings. However, and most importantly, in the specific syntactic environment defined by the idiom, the terminal nodes are associated with a special meaning that differs from the regular meaning of the three words kick, the, and bucket. The regular combination of these irregular meanings of the parts of the idiom then leads to the idiomatic meaning of kick the bucket; therefore, the


compositional meaning determined by syntactic structure is generally preserved. In other words, one derives the meaning of kick the bucket on the basis of encyclopedic knowledge that states that the verb kick (a terminal node with a particular phonological form) receives a special interpretation in the context of a DP-complement whose terminal nodes are interpreted by the LPSs associated with the and bucket. It must be considered part of the lexical entry for kick that its regular meaning can be “bleached” or negated in the syntactic context of the DP the bucket. Similarly, the noun bucket may receive a different meaning in the context of the determiner the if the whole DP is the complement of the verb kick etc. Each terminal node in an idiom receives an interpretation which is different from its regular use, but the interpretation of the whole idiom proceeds compositionally on the basis of these special meanings. Unfortunately, Marantz is not very explicit about what these special meanings of the parts of kick the bucket could be. If I understand his proposal correctly, he would have to assume that kick is associated with the meaning of something like ‘to end’, and the bucket with the meaning of ‘one’s life’ etc. Such an analysis might look a little bit far-fetched, although it would explain why kick the bucket has different aspectual properties than die. However, there are examples that clearly illustrate that at least some idioms can be analyzed compositionally. Jackendoff (1997: 166) notes that the idiom “not playing with a full deck can be paraphrased roughly as ‘not acting with a full set of brains’, where playing means ‘acting’, full means ‘full’, and deck (of cards) means ‘set of brains’. That is, this fixed expression is a minimetaphor in which each of its parts makes metaphorical sense”. In order to capture this “minimetaphorical” character of the parts of the idiom, each of its basic constituents must be linked to a special meaning that is only licensed in the relevant context. Although it may be hard to determine whether idioms uniformly preserve the compositionality that is associated with their LSS-parts, for some idioms at least, this assumption seems quite straightforward (see Marantz 1997, in prep., Ruwet 1991, and Jackendoff 1997 for more discussion of this point). On the basis of these observations, I will argue in the following that the interpretation of particles is context-dependent. Recall from the discussion in Sections 4.1.1–4.1.3 that the semantics of most prepositional particles is different from the meaning of the underlying preposition. A particle is an element that is homophonous to a preposition, but with a different LCS-representation. We can say that semantically, the term “particle” refers to a special meaning of a prepositional element P0. In combination with the syntactic definition of “particle” that I have given at the end of Chapter 3, according to


which particles are heads of non-functional complements of the verb, we arrive at the following picture. The special “particle”-meaning of a prepositional element is associated with the terminal node P0 if P0 does not project functional structure and if it is the head of the complement of the verb. In other words, particle meanings are only licensed by specific contexts, and these contexts are defined in syntactic terms. Since conditions on syntactic environments are represented through LSS-entries, the LSS of a particle element looks like (30): V

(30)

PrtP

n, n>0

V0

Prt0a

The particle element itself is linked to its meaning and to its phonological form through the lexical linking index a. The LPS-entry that the particle is linked to is that of the regular preposition, while the LCS-entry may be different. Importantly, the LSS-representation in (30) specifies the syntactic context which is required for the lexical entry to be unified with syntactic, phonological, and conceptual structure. The meaning of a particle can only be associated with a terminal node in syntactic structure if this node is in the same locality domain as the verb, and this locality domain is defined by the LSS in (30). In a sense, then, particles are similar to affixes whose morphological subcategorization frame also specifies the contextual conditions on lexical licensing in their LSS-entries (see Chapter 1, Section 1.3.2). This similarity between particles and affixes will be the topic of Section 5.1 in Chapter 5. The lexical representation of a particle hence includes information about the particle’s syntactic context. In presenting this information by giving the structure of the whole syntactic tree, (30) includes the particle’s maximal projection as well as the non-minimal level of the verbal projection that immediately dominates the verb and the PP. However, I think that the crucial part of (30) is that particle verbs are characterized on the basis of a particular local relation between two heads, i.e. V0 and P0. The information provided by an LSS-entry of the form in (30) may therefore be reformulated as in (31): (31) Contextual condition on particles The meaning of a particle in conceptual structure can only be associated with X0 in syntactic structure if X0 is structurally adjacent to V0


The condition in (31) is a little less specific than the structure in (30) because the information that the verb has to properly head-govern the particle (and not the other way around) is only given in (30); it is not provided through the definition of structural adjacency in (31) (cf. the remarks in Chapter 1, Section 1.3.1). However, I assume that there is a general syntactic rule that states that whenever a verb and another non-complex11 lexical head are structurally adjacent, the verb is the designated governor. Therefore, it follows that a syntactic structure like (30) is the only way in which the contextual condition in (31) can be fulfilled. Most importantly, (31) emphasizes that particle verbs are characterized by the local relation between two terminal nodes. This is in line with Marantz’s (1997, in prep) claim discussed above. The syntactic properties of particle verbs now follow directly from (31). Because of the definition of structural adjacency that I have given in Chapter 1, the condition in (31) requires the particle to be the head of a non-functional complement of the verb in syntactic structure. Neither the PP nor the V¢/VP-node need to be mentioned in the lexical representation of a particle; these syntactic nodes are generated by syntactic rules in order to derive a structure that meets the requirement in (31). In other words, a particle verb consists of a particle and a verb in a particular locality domain; the fact that the particle is syntactically represented as a phrasal construction follows from the particularities of this local relation. Therefore, the particle verb is not just a normal verb + phrasal complement-construction. The fact that syntactic structure includes a particle phrase is an epiphenomenon of the particular licensing conditions stated in (31). There is empirical evidence for (31). As noted in van Riemsdijk (1978), prepositional phrases can occur as complements of nouns. This is illustrated for German in (32) and (33): (32) a.

in die Stadt fahren in the city drive ‘drive into the city’

11.The additional requirement that the second lexical node must be non-complex still leaves open the possibility that VPs may occur as sisters of lexical heads that correspond to derivational affixes (see, for example, the structure of process nominals in Chapter 1, Section 1.3.3). As far as I can see, derivational affixes are the only lexical nodes that may properly head-govern a verb. This observation might be relevant for the discussion of Verb Raising in Dutch in Chapter 7, Section 7.1.


b. die Fahrt in die Stadt the drive in the city ‘the drive into the city’ (33) a.

auf die Mauer springen on the wall jump ‘jump on the wall’ b. der Sprung auf die Mauer the jump on the wall ‘the jump on the wall’

The (b)-examples show that prepositional phrases can occur as complements of nouns. The structure of the NPs in (32) and in (33) is as in (34): NP

(34)

N0

FPPrep

In contrast, as van Riemsdijk (1978) observes, particles are not licensed as complements of N0: (35) a.

in den Hafen einfahren into the harbor part-drive ‘sail into the harbor’ b. *die Fahrt in den Hafen ein

(36) a.

aufspringen part-jump ‘jump up’ b. *der Sprung auf

The impossibility of (35b) and (36b) follows from (31). The structure of the NPs in (35b) and (36b) is given in (37):

171


*NP

(37)

N0 (Sprung)

PP P0 (auf)

In (37), the prepositional element is not structurally adjacent to a verb. (35b) and (36b) are hence excluded as a violation of (31). As noted above, (31) is a lexical condition on the particle’s syntactic environment. This implies that the exceptional status of particle phrases as nonfunctional categories is simply stipulated as part of the particle’s lexical entry. However, notice that every syntactic theory about particle verbs has to say something about the syntactic relation between a verb and a particle phrase. For example, on the basis of Pesetsky’s (1995) notion of L-selection (lexical selection), Svenonius (1996) argues that particles are L-selected by their base verbs. This means, according to Svenonius, that the syntactic relation between particles and verbs is listed as part of the verb’s lexical entry. In contrast, I assume that it is listed as part of the particle’s lexical representation.12 I assume that this peculiarity of the lexical representation of particles is a reflex of their diachronic history. In order to explain why particle verbs have maintained certain syntactic properties, although they are (wrongly) analyzed as morphological objects, proponents of the morphological approach often point out that particle verbs were regular syntactic constructions at earlier stages. I assume that particle verbs have never lost their status as phrasal constructions. However, they have lost their ability to project functional structure, and since constructions like these are normally not licensed by syntactic rules, this idiosyncratic property has to be listed in the lexicon. In the following sections I will try to provide more evidence for the claim that particle meanings are only licensed on the basis of the presence of a verb.

12.Interestingly, the definition in (31) is also not unrelated to the proposal made by Stiebels & Wunderlich (1994) and Stiebels (1996). Stiebels and Wunderlich argue that particle verbs are “formed in the lexicon”. In their theory, particles are “special” morphological elements, since they receive a feature [+max] (see Chapter 2, Section 2.1.2). According to Stiebels and Wunderlich, this feature [+max] can only be assigned to an element in the context of a verb. As in my account, the particle status of an element is contingent on the context of a verb.


I show that many semantic and thematic properties of particles depend on semantic properties of the respective base verb. This situation can be captured directly on the basis of a lexical representation of a particle that explicitly mentions its local relation to the verb. 4.2.2 Class-based semantics and possible hosts One example of a particle element which derives transparent particle verbs is the particle an with the meaning ‘the event is directed-towards some person/ object’. As Stiebels (1996) shows, this meaning of an is very productive. Some examples are given in (38): (38) a.

jemanden anschreien, ‘shout at somebody’ somebody part-shout b. jemanden anbetteln, ‘ask somebody for money’ somebody part-beg c. jemanden ansprechen, ‘address somebody’ somebody part-speak d. jemanden ansingen, ‘sing towards/to somebody’ somebody part-sing e. jemanden anspucken, ‘spit at somebody’ somebody part-spit

In (38), an combines with intransitive base verbs; the resulting particle verbs denote complex events that combine the meaning of the base verb and the meaning of the particle. (39) gives the lexical representation of the particle an with this “directed-towards”-meaning: LPS

(39) a.

b.

Worda

a

n

LSS P0a

c.

LCS [Event DIRECTED TOWARDS ([Event]A, [Thing ]A)]a if P0 is structurally adjacent to V0

In (39), I have added the contextual condition on particles that was given in (31) as an amendment to the LCS of an. Recall that this is just a notational alternative to a representation that uses an LSS like (30) above to specify the necessary context of a particle. The meaning of the particle that is given in (39c) can only be unified with a syntactic structure if the particle is structurally adjacent to a verb. If the syntax creates such a structure, the lexical entry of a verb is unified with V0; the LCS of the verb is merged into the Event-slot in

173


(39c), and the result is the conceptual structure of a transitive particle verb. However, despite the productivity of an, its use is not unrestricted. Stiebels & Wunderlich (1994: 950) point out that the “directed-towards”-meaning of an in (39) is typically associated only with agentive intransitive base verbs. Therefore, an cannot combine with the verbs in (40): (40) a.

lieben love

b. träumen dream

–

–

*jemanden anlieben somebody part-love intended reading: ‘direct the loving towards somebody’ *jemanden anträumen somebody part-dream intended reading: ‘direct the dreaming towards somebody’

The possibility of combining a particle with a particular base verb seems to be restricted by the semantic properties of the base verb. As another example for this observation, consider (41): (41) a.

(ein Buch) anlesen, ‘start reading a book/read a book partially’ (a book) part-read b. (ein Lied) anspielen, ‘start playing a song/play a song partially’ (a song) part-play c. (das Brot) anschneiden, ‘start cutting the bread, cut into the bread’ (the bread) part-cut

The aspectual meaning of an in (41) was already discussed in Section 4.1.3. Stiebels & Wunderlich (1994) note that the semantics of an in (41) is only licensed in combination with transitive base verbs with an incremental Theme. Consequently, the particle verbs in (42) do not exist: (42) a.

arbeiten work

b. lachen laugh

–

–

*anarbeiten part-work intended reading: ‘start working, work partially’ *anlachen part-laugh ungrammatical under the intended reading: ‘start laughing, laugh a little’

Notice that semantically, there is nothing wrong with the situations expressed in (42). This can be illustrated by a third example of a particle that only combines with a particular class of verbs. The semantics of the particle los is


similar to that of an in (41); los determines an ingressive reading of the event (in contrast to an, los does not express partitivity). However, los can only combine with intransitive base verbs: (43) a.

losarbeiten, ‘start working’ part-work b. loslachen, ‘start laughing’ part-laugh

(44) a. *ein Buch loslesen a book part-read intended reading: ‘start reading a book’ b. *ein Lied losspielen a song part-play intended reading: ‘start playing a song’ c. *das Brot losschneiden the bread part-cut intended reading: ‘start cutting the bread’

The contrast between the examples in (41) and (42) on the one hand, and the examples in (43) and (44) on the other, suggests that the meaning of the particles an and los is licensed only in combination with a particular class of verbs. In Stiebels & Wunderlich (1994), this observation is generalized through the following claim (see also Lüdeling 1998a): (45) Stiebels and Wunderlich’s Generalization (cf. Stiebels & Wunderlich 1994: 950) All productive uses of particles […] are sensitive to the semantic class of the base verb

I will henceforth assume that Stiebels and Wunderlich’s generalization in (45) is correct. A particular meaning of a particle is licensed only if the base verb is part of a particular semantic class. Fortunately, the lexical representation of particles that I suggested in (31) makes it possible to represent the respective requirement in the particle’s lexical entry. The lexical entry for an in (39) already includes the information that the terminal node P0 that is linked to the particle’s LCS must be structurally adjacent to the verb. All that is needed now is that we add the information that a particular particle meaning is only licensed with a particular class of verbs. This allows us to represent the aspectual and the “directed-towards”-meaning as different variants of the lexical entry of an:

175


LPS

(46) a.

b.

Worda

a

LSS P0a

c.

n

LCS [Event DIRECTED TOWARDS ([Event ]A, [Thing ]A)]a if P0 is structurally adjacent to a class-1 verb [Event PART ([Event ]A)]a if P0 is structurally adjacent to a class-2 verb

Class 1: agentive intransitive verbs Class 2: transitive verbs with an incremental Theme

(further meanings of an) if P0 is structurally adjacent to a class-3 verb (…)

The lexical representation in (46) is an example of what I will call the class-based analysis of particle verbs. It implements the generalization in (45) with respect to the particle an. The lexical entry of an may link the terminal node P0 to various LCSs; the choice of an LCS depends on the verb class. The classes mentioned in (46) are defined by semantic and thematic properties of verbs. The verb classes in (46) are rather large, due to the semantic criteria by which they are defined. However, as pointed out by Lüdeling (1998a), the semantic class that defines possible base verbs for a particular particle meaning can also be quite small. For example, as noted by Lüdeling, the particle ein licenses its particular aspectual reading mainly with verbs of sleeping; a class that might include not more than three or four base verbs: (47) einschlafen, einpennen, einknacken, ?eindösen part-sleep, part-doze etc.; ‘fall asleep’

The fact that the meaning of a particle depends on the semantic class of a verb is of course not a consequence of the syntactic approach defended in Chapter 2. The morphological approach to particle verbs would have to assume something similar to (46) in order to capture Stiebels and Wunderlich’s generalization. Only if lexical dependencies like those stated in (46) are (mis-)taken to be exclusively properties of morphological objects will one arrive at the conclusion that particle verbs must be words. However, as was shown in Chapter 1, the “lexicalist” view of grammar, according to which the lexicon is identified with the domain of morphology, is empirically incorrect. In light of the alternative conception of the lexicon as a set of correspondence rules that are part of the interface component, the non-morphological relation between the particle and


verb does not contradict its lexical status. One might object that the possibility of combining a particle with a particular base verb is restricted by general conceptual constraints that govern the combination of any phrasal complement and a verb. According to this view, the rules of conceptual structure simply cannot generate structures that would correspond to non-existing particle verbs like those in (44). Then it would be unnecessary to list possible verb classes as parts of the particle’s entry. However, there is evidence that the combinatorial properties of particles are in fact more restricted than those of “regular” syntactic complements. Consider (48): (48) a.

Peter lehnt das Bild an die Wand P. leans the picture at the wall ‘Peter leans the picture against the wall’ b. Peter lehnt das Bild an P. leans the picture part ‘Peter leans the picture (against an implicit non-referential reference object)’

In (48b), we have combined a verb that takes a predicative argument with the particle an. In line with the observations made in Section 4.1.1, the particle’s semantics is based on the local concept of the preposition in (48a). However, its reference object is not linked to syntax. The LCS of an in (48b) can be represented as in (49): (49) [Path TO ([Place AT ([Thing ])])]

Notice that (49) differs from the representation of the corresponding preposition in (48a), which obligatorily links its reference object to syntax: (50) [Path TO ([Place AT ([Thing ]A)])]

Now consider (51): (51) a.

Peter hängt das Bild an die Wand P. hangs the picture at the wall ‘Peter hangs the picture on the wall’ b. *Peter hängt das Bild an Peter hangs the picture part intended reading: ‘Peter hangs the picture (against an implicit nonreferential reference object)’

Curiously, although a particle verb anlehnen exists, no corresponding particle verb *anhängen can be derived. However, the base verbs lehnen and hängen


both combine happily with a full prepositional phrase headed by an. The contrast between (48b) and (51b) shows that the conditions that restrict the possibility of combining a particle with a particular verb are stricter than the conditions that govern the combination of a verb and a functional complement. Except with respect to linking, the LCSs in (49) and (50) are identical. If the combination of the particle an and a verb was excluded by a general conceptual principle, we would expect that the same principle would also rule out the grammatical example (51a) with a full prepositional phrase. Since this expectation is not borne out, I conclude that the combination of a particle and a verb must be lexically conditioned. The same conclusion must be drawn when particle verbs are compared to h-verbs (= combinations of verbs and intransitive h-postpositions): (52) a.

Die Milch läuft heraus the milk runs H-out ‘The milk runs out (of some contextually specified reference object)’ b. Die Milch läuft aus the milk runs part ‘The milk runs out’ (non-referential implicit reference object)

(53) a.

Die Feder springt heraus the spring jumps H-out ‘The spring jumps out (of some contextually specified reference object)’ b. *Die Feder springt aus the spring jumps part intended reading: ‘The spring jumps out’ (non-referential implicit reference object)

(52a) and (52b) illustrate the semantic difference between h-verbs and particle verbs with implicit reference objects that was already discussed in Chapter 3, Section 3.2.2. Both the h-postposition heraus and the particle aus can be combined with the base verb laufen, ‘run’. The pronominal adverb, which is a functional prepositional phrase, combines with other verbs of motion as well, as shown by (53a) (cf. also heraus kommen, ‘come out (of sth.)’, heraus fahren, ‘drive out (of sth.)’ etc.). In contrast, however, not every verb of motion can also combine with the particle aus, as shown by (53b). (53) supports the assumption that the combination of a verb and a functional complement is much freer than the derivation of a particle verb.


4.2.3 Particle verbs derived by semiproductive rules Although the generalization in (45) is confirmed by the examples discussed in the previous section, it is thus far based only on a rather poor data base. However, testing the class-based analysis with an “acceptable” amount of particles would certainly go beyond the scope of this study. Moreover, in order to see that there are problems for this analysis, one does not even have to consider particles different to the ones discussed above. Consider the particle verbs in (54): (54) a. *einen Brief anschreiben a letter part-write intended reading: ‘start writing a letter; partially write a letter’ b. *ein Haus anbauen a house part-build ungrammatical under the intended reading: ‘start building a house, partially build a house’

Although the base verbs bauen, ‘build’, and schreiben, ‘write’, are transitive and take incremental Themes, the use of an as a partitive or ingressive marker is not possible with these verbs. This observation raises doubts about whether the class of all possible base verbs for a particular particle can really be defined on purely semantic grounds. Stiebels & Wunderlich (1994) point out that the class of possible base verbs for a particular particle verb cannot be defined on the basis of semantic criteria like (in-)transitivity or agentivity alone. Rather, they assume that, “a more finegrained semantics is needed to delimit the possible input of complex verb formation” (1994: 951). One might therefore assume that the ungrammaticality of the verbs in (54) can be accounted for on the basis of a more careful investigation of the semantic conditions that determine membership in the class of possible base verbs for aspectual an. For example, notice that there is a difference between the incremental Themes of schreiben, ‘write’, and bauen, ‘build’, on the one hand, and those of lesen, ‘read’, and braten, ‘fry’, which permit the use of an, on the other. Only the Themes of the latter verbs exist independently of the event, whereas the incremental Themes of the former only come into existence through the event. For example, the same book can be read over and over again, but the same letter cannot be written more than once (see Krifka 1989, 1992 for a formal analysis of this difference). One could argue that the criteria that define the semantic class that includes lesen and braten, but excludes schreiben and bauen, must take this difference into account.


However, it still seems unlikely that it is possible to define verb classes on purely semantic grounds, such that every member of this class is a possible base verb of a particle with a particular meaning. For example, consider again the contrast between the possible particle verb anlehnen in (48b) and the impossible verb *anhängen in (51b). This contrast cannot be explained by general conceptual restrictions (as was confirmed by the fact that both base verbs can take the same functional prepositional phrase as a complement). However, it seems equally unlikely that a class of base verbs can be defined on purely semantic grounds that includes lehnen, ‘lean’, but excludes hängen, ‘hang’. In fact, I suspect that in defining the semantic criteria for such a class, one ends up defining precisely the semantic difference between these two verbs. But then, the class-based account would be nothing more than a way of disguising the fact that in many cases, the possible base verb for a particular particle verb must be individually listed in the lexical entry of the particle (cf. Risch (1994) for a proposal along these lines). Let me illustrate this problem with another example. Sometimes, two uses of a particle only differ with respect to linking. Compare the derivations in (55): (55) a.

Peter lädt die Koffer vom Gepäckwagen P. loads the suitcases off.the luggage-cart ‘Peter loads the suitcases off the luggage-cart’ b. Peter lädt die Koffer ab (Theme = direct object) P. loads the suitcases part ‘Peter loads the suitcases down’ c. Peter lädt den Gepäckwagen ab (Source = direct object) P. loads the luggage-cart part ‘Peter unloads the luggage-cart’

As (55b) and (55c) show, the particle verb abladen may realize either the Theme or the reference object of the particle as its direct object; (55) illustrates the so-called landmark flexibility (McIntyre 2001) that is attested with certain particles.13 The alternation in (55) can be explained on the basis of two different semantic representations of the particle ab (cf. Stiebels 1996). In

13.German researchers (cf. e.g. Kühnhold 1973; Hundsnurscher 1968) refer to the alternation between (55b) and (55c) as “Objektvertauschung” or “Objektumsprung”. Landmark Flexibility is also licensed with the particles an and aus, and a few instances are also attested with ein, auf, über and durch (cf. McIntyre 2001). A discussion of similar alternations in English is provided by Svenonius (1996).


(55b), ab saturates a predicative argument position of the base verb, and the reference object is left implicit. It can hence be assumed that ab in (55b) expresses a Path whose reference object is not linked to syntax, cf. (56a). In contrast, ab in (55c) combines with the (intransitive form of the) base verb via semantic adjunction and introduces its reference object to syntax. On the basis of the discussion in Section 4.1.2, this can be captured by representing ab as an eventive predicate, (56b): (56) a. die Koffer abladen: b. den Gepäckwagen abladen:

[Path FROM ([Thing])] [EventGO ([Thing], [Path FROM ([Thing]A)])]

Now compare (55) to (57): (57) a.

Peter stellt den Topf vom Herd Peter puts the pot from.the stove ‘Peter takes the pot from the stove’ b. Peter stellt den Topf ab (Theme = direct object) P. puts the pot part ‘Peter puts the pot down’ c. *Peter stellt den Herd ab (Source = direct object) P. put the stove part ungrammatical under the intended reading: ‘Peter takes something from the stove’

As the (a)-examples show, both stellen, ‘put’, and laden, ‘load’, take an obligatory prepositional phrase as an argument which may realize the Source. Like abstellen, the particle verb abladen can represent the Theme of the Event as a direct object, cf. (55b) and (57b). Interestingly, abladen also allows for the Source to be linked to syntax, (55c). In contrast, abstellen does not license landmark flexibility, (57c). Although the argument structures of the base verbs laden and stellen are the same, the realization of the reference object of ab is not possible with the particle verb abstellen in (57c). In terms of the class-based analysis, the conclusion must be drawn that laden is a member of the class of possible verbs that ab in (56b) can combine with, whereas stellen is not. Again, the similarity between both verbs makes it unlikely that such a class can be defined on semantic grounds. Nevertheless, I do not think that we have to give up the idea that the meaning of a particular particle is class-based. Neither do we have to develop more and more fine-grained semantic criteria until we have defined a class

181


which only includes all and only those verbs that a particular particle can combine with. Instead, I think that we can continue to define a possible base verb for a particle on the basis of semantic criteria, as long as we acknowledge that not every verb that fulfills these respective semantic conditions is a possible base verb. What I have in mind is that the relation between the semantic properties of a verb and the possibility of combining it with a particular particle is not entirely regular, but rather characterized by what Jackendoff (1997) calls a semiproductive rule. Jackendoff (1997) defines productive processes as those that “predict the existence and form of all derived forms”, whereas “the defining characteristic of semiproductive rules is that one needs to know whether each particular form exists, as well as (in many cases) particularities of its meaning and pronunciation” (1997: 121). As an example, Jackendoff (1997) discusses denominal verbs in English, like to saddle, to shelve, or to butter. He notes that the rule that derives these verbs from nouns like saddle, shelf, or butter is not fully productive. Crucially, this semiproductive rule does not yield outputs for every input (for example, there is no verb *to mustard, whose existence would be predicted by a productive conversion rule). Jackendoff concludes that even though there is a (semiregular) relation between e.g. the noun saddle and the verb saddle, the output of this semiproductive rule must be listed as an independent lexical item. With respect to the present issue, this means that whether or not a verb qualifies as a possible base verb for a particular particle is only semiregularly determined on the basis of the semantic properties of this verb. Although both mustard and butter are nouns that in principle would qualify as inputs to the conversion rule, only to butter is a grammatical output. In the same way, both laden and stellen are members of the class of verbs that ab in (49b) combines with, but factually, only laden is a possible base verb. This view has an important consequence which (at first sight) intuitively contradicts the claim made in this section. As in the case of denominal verbs in English, the view that possible base verbs for particles are defined by semiproductive rules implies that particle verbs like anlehnen, ‘lean on’, or abladen, ‘load off’, must be listed as lexical items. For example, consider again the contrast between the particle verb anlehnen and the non-existing particle verb *anhängen. A speaker of German must know that only the former particle verb exists; therefore, the lexicon must include a representation like (58):


n, n>0

V

(58)

PP

V0b

a: link to lexical entry for an b: link to lexical entry for lehnen

P0a

(58) provides a link between the particle an and the base verb lehnen; essentially, (58) represents the lexical information that there is a particle verb anlehnen. This seems to be a rather unwelcome result. The idea behind the class-based analysis is to reduce the size of the lexicon by only having to list one lexical entry for the particle that can productively be combined with the base verbs of a particular class, such that the resulting particle verb does not require a separate entry. One may wonder why a particle should be represented as a lexical entry at all if the verbs derived from it have to be listed anyway. However, it is important to consider the kind of information that exactly is provided by (58). Importantly, (58) does not give any information that is not present in the lexical entries of an and lehnen, except for the information that these two entries can be linked; this is the only independently needed information that must be stored. When a syntactic structure is unified with (58), this lexical entry establishes a connection between the lexical representation of the particle and the entry of its base verb. Since (58) links V0 to lehnen, the LPS and the LCS of this verb are associated with V0. The index a links P0 to the entry of an, whose LPS is unified with the structure as well. The contextual requirement for particles is fulfilled; therefore, P0 is a particle. Furthermore, the class-based analysis also accounts for the fact that the correct meaning of the particle is associated with P0; since lehnen belongs to a particular semantic class, a particular particle meaning is associated with the respective terminal node. This means that, in comparison to lexical entries that provide fully idiosyncratic information, the informational contribution of (58) is rather small. If we define the “size” of the lexicon in terms of informational “cost”, (58) does not increase the lexicon significantly. Jackendoff (1975) suggests that the informational cost of the lexicon should be measured in terms of “independent information content”. The lexicon includes devices that capture regularities between lexical entries and thereby reduce storage space. The listing of a lexical entry that is related to one or more existing entries requires less effort (is “cheaper”) than listing an entry whose content is completely idiosyncratic. It is hence “cheaper” to link the lexical entries of particle verbs to the lexical

183


representation of the particle and the verb than it is to list all particle verbs as unrelated lexical items. Therefore, I assume that even if an actual particle verb must be lexically listed, only those aspects of its meaning that cannot be captured through the lexical entries of its parts take up lexical storage space. The semiproductive relation between semantic properties of verbs and their ability to occur as base verbs of particular particles only requires the listing of the information that the entry of a verb is linked to the entry of a particle. All other information is contributed by the lexical entries of these two lexical elements. As a consequence of this view, the lexicon can be seen as a complex net, consisting partly of lexical entries whose only purpose is to specify links between the various sounds and meanings that may be associated with other lexical items. This idea is also useful to account for examples where a derived particle verb exhibits certain semantic properties that are not entirely captured through the meaning of its parts. For example, consider the particle verb in (59b) again, which has already been discussed in Section 4.1.1: (59) a.

Peter wirft den Ball [FP zu dem Kind] P. throws the ball to the child b. Peter wirft dem Kind den Ball [PP zu] P. throws the child-dat the ball-acc part ‘Peter throws the ball to the child’

At first sight, the meaning of the double-object particle verb in (59b) is identical to the meaning of the construction in (59a). However, (59a) and (59b) are not completely synonymous. Olsen (1997a) argues that the DP dem Kind, ‘the child’, in (59a) is the Goal of the preposition zu, whereas this DP is interpreted as benefactive in (59b). Olsen assumes that the particle verb zuwerfen in (59b) has incorporated an additional semantic component that indicates a possessive relation between the Theme and the Goal in (59b). Importantly, she assumes that this component is not part of the particle’s meaning, but has been added in the process of deriving the particle verb. Suppose that Olsen’s assumption is correct and that the meaning of the particle zu in (59b) indeed corresponds to the spatial use of the preposition zu in (59a). In that case, not all aspects of the semantics of the particle verb zuwerfen follow from the combination of the particle and the verb. But then, the particle verb must be listed as a complex lexical entry that provides this additional information. The lexicon must include a representation like (60) which connects V¢ with the (possessive) meaning of zuwerfen:


n, n>0

Vx

(60)

PP

V0b

b: a: x:

link to lexical entry for werfen link to lexical entry for zu link to special meaning of zuwerfen

P0a

If V¢ in (60) is associated with a special meaning, we in fact suggest an analysis of zuwerfen that is similar to the analysis of idiomatic particle verbs that were discussed in Section 4.1.4. However, what we also want is that it is still “cheaper” to list a verb like zuwerfen than to list a completely idiosyncratic entry, because large parts of the meaning of zuwerfen are still licensed through the lexical entries of its parts. Again, this intuition is captured through the notion of informational cost outlined above. I assume that although V¢ must be associated with the complex meaning of zuwerfen, its terminal nodes are also linked to the semantics of werfen and zu. Notice that the terminal nodes have to be linked to the lexical entries of these two elements anyway in order to guarantee that their regular LPS-representations are associated with the terminal nodes P0 and V0 in (60). The particle zu with the meaning in (59b) combines freely with verbs that belong to the same semantic class as werfen (cf. zuschieben, ‘push towards’, zuschicken, ‘send to’ etc.).14 Therefore, if werfen is associated with V0, P0 is linked to the right meaning of zu. Since part of the meaning that is associated with V¢ is therefore already derived compositionally, the only part of the information associated with V¢ that takes up storage place in the lexicon is the possessive part of zuwerfen. Even though this particle verb is listed, its entry only increases the size of the lexicon insofar as it adds special information to the particle’s class-based lexical representation. In that sense, zuwerfen can be compared to a denominal verb like shelve. The latter is clearly derived from the noun shelf, but it must be listed in the lexicon. Similarly, a particle verb like zuwerfen must be listed. Furthermore, although there is an obvious relation between the LPS of shelve and the LPS of shelf, one has to know that the denominal verb ends with a voiced consonant.

14.As far as I can see, the possessive aspect of zu is licensed with these verbs as well. This suggests that this semantic particularity is part of the particle meaning of zu; it is associated with P0 if P0 is structurally adjacent to V0. In that case, (59b) does not require a representation like (60), but rather is in full accordance with the claims made in the preceding paragraphs. However, I have adopted Olsen’s (1997a) proposal in the text for illustrative purposes.

185


This is additional phonological information that must be included in the entry for shelve. In the same way, the semantics of zuwerfen is based on the combination of zu and werfen; however, the additional possessive component must be listed with zuwerfen. In both examples, the size of the lexicon is only increased by this small piece of extra information, whereas the semantic and phonological information that comes with the semiproductive rule is “free”. It is therefore still cheaper to list a particle verb like zuwerfen and link its parts to the entries for zu and werfen than to store such particle verbs as “real” phrasal idioms. Let me summarize the results of this section. The lexical representation of a prepositional element may include various meanings that can be associated with a terminal node P0 only if this terminal node is structurally adjacent to a verb. If the structure created by syntax meets this condition, lexical licensing yields a construction with the syntactic and semantic characteristics of a particle verb. Furthermore, the lexical meanings of particles are enriched with information about classes of possible base verbs which are defined by certain semantic or thematic criteria. These class-based meanings define the range of possible particle verbs. In addition, the lexicon includes entries that correspond to existing particle verbs. The sole information that is provided by these entries is the link between a particular lexical representation of P0 and a particular lexical representation of V0. With this information provided, all other information associated with a particular particle verb comes for free; it is provided by the lexical entries of P0 and V0. If the factual base verb is determined, the grammar of a speaker checks the lexical entry of the respective particle that this verb is linked to in order to establish the right LCS on the basis of the semantic class that the verb belongs to. The LCS of the verb and the so-determined LCS of the particle are then associated with P0 and V0 in the syntax. Furthermore, the LPS-entries of V0 and P0 are associated with these terminal nodes (or with some higher position of the verbal chain in the case of verb movement). If the meaning of the particle verb includes certain semantic features that are not derived on the basis of the LCS of V0 and the LCS of P0, these features may be stored in addition to the lexical entry that links the representation of the particle and that of the verb. 4.2.4 More irregularities I will now show that the class-based analysis of particles reveals a new perspective on a particular class of particle verbs that are usually analyzed as irregular formations. Consider the particle verbs in (61):


(61) a.

anhalten, ‘stop’ part-stop b. ausprobieren, ‘test’ part-test

The events expressed by the particle verbs in (61) are related to the events expressed by the base verb; (61a) is a stopping, and (61b) is a testing. Therefore, it seems that on the one hand, the base verbs in (61) regularly contribute their semantics. On the other hand, the particle verbs in (61) have idiosyncratic properties, because the semantic contribution of the particles is unclear. For example, it seems impossible to relate the meaning of an in (61a) to any of the meanings of this particle that have been discussed in the previous sections. Similarly, it is hard to tell whether a particular meaning can be assigned to aus in (61b) that is also attested with other particle verbs formed with aus. Although the meaning of the base verbs is clearly reflected in the meaning of the particle verbs in (61), it seems that these examples have to be considered irregular formations. However, the class-based analysis offers a new way of looking at examples like (61). Consider the verbs in (62): (62) a.

jemanden ansingen “directed-towards” an somebody part-sing ‘sing towards/to somebody’ b. (ein Lied) anspielen aspectual an (a song part-play ‘start playing a song’

(62) repeats the productive uses of the particle an that I discussed above. Recall that both the “directed towards”-, and the aspectual, meanings of an are licensed in the context of base verbs that belong to a particular class. Now suppose that in the same way that the special meaning of the particle an in (62a) and (62b) is restricted to structurally adjacent verbs that belong to a particular class, the special meaning of an that derives the verb anhalten from halten in (61a) is also restricted to a particular class of verbs — but this class perhaps includes only one member.15 Recall that verb classes may in fact be quite small; for example,

15.The fact that both halten and anhalten are translated as ‘stop’ of course raises the question of whether there is any meaning difference that is caused by the particle. I suspect that the crucial semantic contribution of the particle an in anhalten is that it derives a verb that can be used transitively. As (ii) shows, this option does not exist with the base verb halten: (i)

a.

Das Auto hielt an


the aspectual use of ein was licensed only by a few verbs of sleeping (cf. (47)); accordingly, the meaning of an in (61a) is licensed by an extremely small class which only includes the verb halten. On this assumption, the single difference between “regular” an in the examples in (62) and “irregular” an in (61a) lies in the number of possible verbs with which the respective meaning is licensed. However, this suggests that the line between “regular” and “irregular” meanings is in fact thinner than one might think. In fact, it is probably inappropriate to call the meaning of an in (61) “irregular”. It seems that all meanings of an are equally regular, they are only unequally productive.16 The consequence of this view is that many particle verbs that have been classified as idiomatic can now be analyzed as being derived compositionally on the basis of the special meanings of their parts. This idea is compatible with the proposal made in Marantz (1997, in prep.) that I discussed in Section 4.2.1. As in the case of the transparent particle verbs discussed in Sections 4.1.1–4.1.3, certain idiosyncratic particle verbs can be analyzed on the basis of a special meaning of a particle that is combined with the special meaning of the verb. As emphasized by Marantz (1997, in prep.), special meanings are assigned in particular contexts. With respect to the meaning of particles, the relevant contexts are defined by structurally adjacent verbs.

b.

(ii)

a.

b.

the car stopped part Das Auto hielt the car stopped ‘The car stopped’ Die Polizisten hielten das Auto an the policemen stopped the car part ‘The policemen stopped the car’ *Die Polizisten hielten das Auto the policemen stopped the car

There are more, quite subtle, meaning differences between halten and anhalten that do not have to concern us here. 16.This statement only holds under a view of productivity that at least partially takes into account the absolute number of verbs that can be formed with a particular element. If the productivity of a particle is defined solely on the basis of the number of possible derivations within the pool of possible base verbs, this would lead to counterintuitive results: a particle like an in (61a) that can only combine with one possible base verb would be 100 percent productive by definition, whereas a particle that can combine with a whole class of base verbs, but does not fully exhaust its possible base, would be less than fully productive. See Lieber (1992) for a discussion and definition of productivity that captures this intuition.


A similar point can be made about the semantics of base verbs. It can be observed that in the same way that a particle’s meaning is licensed by a verb, the special meaning of a verb may be licensed by a particle. Consider again the particle verb anhalten. The base verb halten in (61) was translated as ‘to stop’, but in fact, halten has other meanings as well, like e.g. ‘to hold’. Crucially, these meanings are not attested with the particle verb anhalten. This means that although the lexical entry for halten may list various meanings, only one of these meanings can be assigned to V0 if V0 is structurally adjacent to P0. In Chapter 5, I will discuss more examples that illustrate the context-dependency of special verb meanings. These examples show that the local domain defined by structural adjacency of a particle and a verb is relevant for the assignment of special meanings to both terminal nodes. The question that is raised by this observation is whether all idiomatic particle verbs can be analyzed compositionally by assigning special meanings to the two terminal nodes V0 and P0. Although this would be a highly welcome result, I think this question must be answered in the negative. As already pointed out in the preceding sections, there are certainly particle verbs whose meanings cannot be derived on the basis of the meaning of their parts alone. Some cases were already discussed in Section 4.1.4. For particle verbs like auftauchen, ‘appear’, whose idiosyncratic meanings are metaphorically based on the meaning of transparent particle verbs, a compositional analysis can hardly be given. I therefore assume that at least some particle verbs must be lexically stored as phrasal idioms; I will return to this point in Chapter 5, Section 5.1.

4.3 The lexical licensing of prepositions I will now extend my proposal to regular prepositions in German which are dominated by functional structure in syntax. Consider again the example of an. This preposition can have both a directional and a locative meaning which corresponds to different case-assigning properties. As was shown in Chapter 3, these case properties are determined on the basis of grammatical properties of the functional head F0Prep which is characterized by a single specification [±directional]. If the functional head of the extended projection of the preposition an is marked as [−directional], an is interpreted as a locative preposition and assigns dative case to its reference object, (64). If F0Prep is [+directional], an expresses a Path and its complement bears accusative case, (65):


(63) a.

Das Kleid hängt an [dem Haken the dress hangs at [the hook]-dat ‘The dress hangs at the hook’ b. Peter hängt das Kleid an [den Haken P. hangs the dress at [the hook]-acc ‘Peter hangs the dress at the hook’

(64) a.

[FP [F′ [PP an [DP dem Haken]] F 0 [-directional]]] dative

b. LCS of preposition: (65) a.

[Place AT ([Thing ]A)]

[FP [F′ [PP an [DP den Haken]] F 0 [+directional] ]] accusative

b. LCS of preposition:

[Path TO ([Place AT ([Thing ]A)])]

In Chapter 3 I pointed out that the meaning of FPPrep corresponds to the feature specification of its functional head. I will now make this claim more precise. I assume that the functional head F0Prep does not specify the meaning of the whole prepositional phrase, but the meaning of its lexical head, i.e. the preposition. What I suggest is that the actual LCS that is associated with the preposition an in a particular syntactic environment is established in the same way as was argued to be the case with the particle an. The relevant factor that determines the meaning of a preposition is again the head that it is structurally adjacent to. The important difference between particles and prepositions is that in the case of the latter, the relevant head is not a verb, but the functional head F0Prep. The meaning of the preposition an is given in (66): (66) a.

LPS Worda a

n

b.

LSS P0a

c.

LCS [Place AT ([Thing ]A)]a if P0 is structurally adjacent to [−directional] F0 [Path TO ([Place AT ([Thing ]A)])]a if P0 is structurally adjacent to [+directional] F0

The lexical entry in (66) links a locative meaning to a syntactic P0-node in the syntactic context of a [−directional] F0Prep-head, whereas the directional meaning of an is associated with P0 in the context of a structurally adjacent [+directional] F0Prep. At the same time, the reference object of the preposition


receives the case that is determined by the respective feature of the functional head (I have not represented the case properties of P0 in (66), but recall that whether P0 assigns dative or accusative case depends on the specification of F0Prep). This captures the fact that an as a directional preposition assigns accusative case, while the DP-complement receives dative case if an is locative. In my proposal, the correspondence between case-assigning properties and the meaning of a preposition is mediated through the properties of the functional head F0Prep which is structurally adjacent to the preposition. Notice that the label of the relevant feature (“directional”) only intends to make this correspondence clearer. We might as well call the feature dichotomy of F0Prep “weak” versus “strong” or “high” versus “low”; it does not make a difference, as long as it is clear that this feature specification determines both the case-assignment properties and the meaning of the preposition. The important advantage of a lexical representation like (66) is that it enables us to combine the lexical representations of particles and their corresponding prepositions. The lexical entry of the prepositional element an is given in (67). It states that an can be used as a particle and as a preposition, depending on its syntactic context: (67) a.

LPS Worda a

n

b.

LSS P0a

c.

LCS [Event DIRECTED TOWARDS ([Event ]A, [Thing ]A)]a if P0 is structurally adjacent to a class-1 verb [Event PART ([Event ]A)]a if P0 is structurally adjacent to a class-2 verb (further meanings of an) if P0 is structurally adjacent to a class-3 verb (…) [Place AT ([Thing ]A)]a if P0 is structurally adjacent to [−directional] F0 [Path TO ([Place AT ([Thing]A)])]a if P0 is structurally adjacent to [+directional] F0

191


According to (67), the various meanings that can be associated with a prepositional element like an are a case of polysemy, not of homonymy. Particles are not separate lexical entries that have to be distinguished from their corresponding prepositions. The term “particle”, understood semantically, may be used to refer to a particular LCS of a prepositional element; alternatively, it can be understood syntactically in the sense of the definition given in Chapter 3; i.e. as referring to a lexical P-head that is structurally adjacent to a verb. The lexical representation given in (67) has an important implication. (67) claims that the meaning of a prepositional element can only be determined once the status of the head to which it is structurally adjacent is taken into account. A P0-head that is linked to a particular phonological form reveals its special meaning only in a particular syntactic environment. (67) may be taken as an implementation of the core idea behind Marantz’s (1997, in prep.) theory: special meanings of terminal nodes are always assigned on the basis of the syntactic context of the node.

4.4 Conclusion In this chapter I have discussed the semantic properties of particle verbs, focussing in particular on the way their meanings can be derived on the basis of the meanings of the particle and the verb. The semantics of many particle verbs supports the conclusion drawn at the end of Chapter 3: particle verbs are complex expressions that consist of two lexical heads in a local domain: (68)

VP

Spec

V′ PP P0

V0

Particle verb syntactic context for special meaning of P 0

This result is no surprise for proponents of a morphological approach to particle verbs. If it is assumed that particle verbs are morphological objects, the


context-dependency of the particle’s meaning can be captured through the local relation between two parts of a word. However, I have shown in Chapter 2 that particle verbs are not represented as complex words. Fortunately, on the basis of the alternative model of lexical licensing that I outlined in Chapter 1, the lexical properties of particles prove to be compatible with a syntactic approach as well, contrary to what is assumed by proponents of lexicalist models of grammar. Looking at the lexicon as a set of rules operative on the interface between syntax and conceptual structure (and phonology), we can incorporate the idea that the special meaning of a terminal node is determined by its structural environment, even if this environment is not determined by the local domain that is characteristic of morphology. Rather, in the light of the syntactic properties of particle verbs discussed in Chapters 2 and 3, their lexical properties can be taken as evidence for the claim that locality domains may also be defined by non-morphological syntactic structures.

Chapter 5

Local domains and morphology

In Chapter 4 I argued that particle verbs are characterized by a specific local relation between two terminal nodes (the particle and the verb). As I argued in Chapter 1, Section 1.3.1, the terminal nodes that form a complex morphological object also constitute a particular local domain. The domain that characterizes particle verbs (i.e. structural adjacency) is of course different from the domain defined by complex words. However, in this chapter I am concerned with the parallels between the two local domains, and I argue that certain word-like properties of particle verbs follow from this parallel. A first property of particle verbs that is also known from morphology is the context-dependency of terminal nodes. It is well-recognized that the phonological properties of an affix may depend on properties of its host; the choice of an allomorph of an affixal element depends on the phonological or syntactic features of the terminal node that it combines with inside a word. In the same way, as I showed in Chapter 4, the semantic properties of a lexical head that qualifies as a particle depend on properties of the base verb. On the basis of this parallel, I argue in Section 5.1 that particles are “semantic affixes”. In Section 5.2 I extend the comparison of the local domains that distinguish particle verbs from words; i.e. structural adjacency and the complex X0-head. I show that the verb-particle construction exhibits properties that are otherwise only attested with words. In line with the proposal defended in Chapters 2–4, I argue that these properties do not require an analysis that treats particle verbs as words. Rather, I suggest that the respective phenomena are licensed in particular local domains; since these domains may be defined by both complex words and by structural adjacency, they are expected to be observed with particle verbs as well. Finally, in Section 5.3, I address the properties of h-postpositions and pleonastic circumpositional constructions in German. It will be shown that these constructions provide interesting support for the claims made in Sections 5.1 and 5.2.


5.1 Particles as semantic affixes A morpheme whose realization depends on the presence of a particular terminal element within a particular locality domain is typically called an affix. I argued in Chapter 4, Section 4.2.1, that it is a lexical property of particles that they must be structurally adjacent to a verb. Hence, a particle is a kind of “semantic” affix (cf. Zeller 1997b); it requires a “host” (the verb) within a particular local domain. In this section, I therefore compare particles to phonological affixes with respect to phenomena that determine the choice of a particular allomorph. The discussion will reveal interesting parallels between the licensing of the semantics of particle verbs and morphophonological operations like conditioned allomorphy (Section 5.1.1) and suppletion (Section 5.1.2). 5.1.1

“Semantic” allomorphy

I concluded Section 4.3 of the preceding chapter with a lexical representation of the prepositional element an that looks like (1): (1) Lexical representation of an a. LPS b. LSS P0a Worda

a

n

c.

LCS [Event DIRECTED TOWARDS ([Event ]A, [Thing ]A)]a if P0 is structurally adjacent to a class-1 verb [Event PART ([Event ]A)]a if P0 is structurally adjacent to a class-2 verb (further meanings of an) if P0 is structurally adjacent to a class-3 verb (…) [Place AT ([Thing ]A)]a if P0 is structurally adjacent to [−directional] F0 [Path TO ([Place AT ([Thing ]A)])]a if P0 is structurally adjacent to [+directional] F0

According to (1), the lexical entry of an links a particular phonological form to the terminal node P0 in syntax. However, the respective meaning that is associated

Local domains and morphology 197

with P0 depends on this node’s environment; i.e. on the semantic class or the functional specification of the head that P0 is structurally adjacent to. The representation in (1) has an interesting parallel in the phonological domain. It is well-known that many morphemes exhibit conditioned allomorphy; i.e. the phonological realization of a particular morpheme depends on properties of the root that it combines with. Let me illustrate this situation on the basis of the past tense-morpheme in English. The choice of a particular past tense allomorph (i.e. the Vocabulary entry that realizes the syntactic node Tns0 with the grammatical feature [+past]) depends on the choice of the verb with which the morpheme combines. For concreteness, consider the alternation between the allomorphs Ø (the zero-morpheme), /t/, and /d/ in English. All three allomorphs correspond to [+past]-Tns0 in syntactic structure and to the past tense meaning in conceptual structure. However, the choice of a Vocabulary item depends on the class of the verb. The suffix Ø attaches to so-called strong verbs like put and beat (class 1), the suffix /t/ is selected by a second set of verbs (e.g. send, dwell etc.; class 2), and the suffix /d/ is inserted elsewhere. The lexical entry for past tense must include both the phonological forms of the three allomorphs and the contextual information that specifies which allomorph corresponds to which context. (2) is the lexical entry for the past tense affix (cf. Halle & Marantz 1993; Jackendoff 1997): (2) a.

LPS Ø if Tns0 is part of the same syntactic word as a class 1-verb /t/ if Tns0 is part of the same syntactic word as a class 2-verb /d/ if Tns0 is part of the same syntactic word as V0

b. LSS Tns0b

c.

LCS [Situation PAST ([Situation ])]b

(2c) specifies an unambiguous meaning of the past tense morpheme. Syntactically, tense is represented through the node Tns0. This node must be part of the same word as V0 in order to license the insertion of a Vocabulary item. In (2), I have listed this requirement as a condition on the licensing of the LPS. (I could also have incorporated it into the LSS of the affix; cf. the representation of the English plural suffix /-z/ that was given in Chapter 1, Section 1.3.2.) Importantly, the choice of the respective phonological realizations depends on properties of this syntactic V0-node. If the base verb belongs to class 1 or class 2, the respective morphemes Ø or /t/ are inserted. The allomorph /d/ is a default affix; as with the other affixes, it can only be inserted if Tns0 and V0 are part of the same word, but in addition, this is only possible if V0 neither qualifies as a class 1- nor a class


2-verb.1 Therefore, as in the case of the semantics of a particle, the phonology of the past tense morpheme is conditioned by properties of a second terminal node within the same locality domain. In contrast to particle verbs, the relevant locality domain of tense inflection is the domain of morphological objects. In the same way that the meaning of a particle depends on the class of the base verb, the choice of a Vocabulary item in (2) is class-based. I have argued in Chapter 4 that the class of possible base verbs can be defined on the basis of semantic properties of the verb. Correspondingly, allomorphy may be phonologically conditioned. For example, it seems likely that the classes of verbs that are relevant for the choice of a past tense allomorph are determined by phonological criteria. A clearer example for phonologically conditioned allomorphy, which is discussed by Spencer (1991: 121) and Jackendoff (1997: 139), is provided by Turkish. In this language, the phonological realization of the passive morpheme depends on phonological properties of the verb stem. If the root ends on /l/, the Vocabulary item is /In/; after vowels, it is /n/, and the phonological form /Il/ is inserted elsewhere: (3) Root al(‘take’) oku- (‘read’) yap(‘make’) sev(‘love’) (Spencer 1991: 121)

Passive alîn okun yapîl sevil

As (3) shows, the lexical representation of the passive morpheme refers to three possible classes of roots; membership in one of the classes is defined phonologically.2

1.Since all three morphemes share the contextual condition for Vocabulary Insertion that requires the presence of a verb, /-d/, /-t/, and Ø compete for insertion. In Distributed Morphology, the allomorph that appears in the most highly specified context always “wins”; therefore, if V0 is specified as e.g. a weak verb of class 2, /-t/ is automatically inserted, although the contextual conditions of /-d/ are met as well (cf. Halle & Marantz 1993 and Chapter 1, Section 1.2.3). The same result is achieved through Jackendoff’s (1997: 141) rule of “Allomorphic blocking” that defines the allomorph with the most highly specified context as the “designated” allomorph. 2.Of course, there is one crucial difference between class-based allomorphy and class-based particle meanings. In the former, but not in the latter, case there is a one-to-one relationship between allomorphs and their hosts. For example, every verb in Turkish can only combine with one possible passive morpheme, whereas it is always possible that a verb qualifies as a


Phonological and semantic properties of the host are not the only features that may determine the choice of a Vocabulary item or the interpretation of a terminal node. Phonological and “semantic” allomorphy may also be governed by syntactic properties of the host. For example, I have argued that the meaning of directional or locative prepositions depends on the grammatical feature of F0Prep. In the same way, the choice of an allomorph may depend on syntactic properties of its host. Aronoff (1994) discusses examples from Spanish and Russian where the choice of a particular inflectional morpheme (a gender marking suffix in Spanish, case endings in Russian) depends on the inflectional class of the noun. This inflectional class in turn is pre-determined by the syntactic gender of the noun. Therefore, the inflectional class of a noun can be represented as a syntactic feature; the phonological realization of a suffix depends on this syntactic property of its host noun (cf. Jackendoff 1997: 142). However, the examples from Spanish and Russian discussed by Aronoff (1994) are also interesting for another reason. They show that class membership cannot always be determined regularly on the basis of syntactic or phonological properties. For example, class membership of Spanish nouns is not exhaustively determined by the rules in (4) (cf. Aronoff 1994: 69): (4) a. Masculine b. Feminine

Æ Æ

class 1 class 2

Class 1-nouns are suffixed with o, whereas a is the class 2-suffix. However, although the rules in (4) determine the relation between the syntactic gender of a noun and its inflectional class in most cases, there are also exceptions. For example, the feminine noun mano, ‘hand’, belongs to class 1; the masculine noun dia, ‘day’, belongs to class 2. As Jackendoff (1997) notes, these exceptions show that the relation between the syntactic property of N0 (gender) and membership in a particular inflectional class is determined by a semiproductive rule. Although rules like (4) hold for the majority of cases, the exceptions must be listed in the lexicon; for example, Aronoff (1994) assumes that the inflectional

member of two different semantic classes and hence triggers different meanings of a particular particle. The realization of particle meanings through base verbs is not disjunctive; for example, verbs that trigger landmark flexibility can combine with two different LCSs of the same particle (cf. abladen in den Wagen abladen, ‘unload the wagon’, and das Heu abladen, ‘load down the hay’, which were discussed in Chapter 4, Section 4.2.3). Nevertheless, both the phonological realization and the semantic interpretation of a terminal node may depend on its syntactic environment.


class of exceptional nouns like mano and dia must be listed as part of their lexical entry. This echoes the discussion of similar irregularities that were attested with respect to the possible base verbs for particles. I have argued in Chapter 4 that a particular meaning of a particle is licensed by a particular semantic class of a verb, but I have also shown that a particle cannot always be combined with every possible member of this class. In the same way as membership in an inflectional class of a noun in Spanish and Russian is not defined entirely reliably on the basis of its syntactic gender, the possibility of having a particular verb as the base verb for a particle is only semiproductively determined by its semantic properties. Therefore, the additional information that a particular verb combines with a particular particle has to be listed in the lexicon (see Section 4.2.3 for details). The similarities between the lexical representation of particles and the lexical representation of affixes that exhibit allomorphy suggest that it is actually appropriate to consider particles as affixal elements. The different meanings that correspond to one particular particle are comparable to the different allomorphs that correspond to a particular morpheme. The various LCS-entries that are listed as part of a particle’s lexical entry could hence be called “allosemes”. One may object that morphological notions are not appropriate for particles, because particle verbs are not words. The locality domain in which the choice of a e.g. a Tns-morpheme is determined is the complex X0-element (the syntactic word), whereas the locality domain of the base verb and the particle is defined by structural adjacency (which requires a syntactic structure larger than the word). However, the notion “semantic affix” emphasizes the fact that both domains, despite different structural properties, are essentially characterized by strictly local relations between two terminal nodes. I will make this point even stronger in Section 5.2. 5.1.2 “Semantic” suppletion Halle & Marantz (1993) discuss examples of what they call partial suppletion. Consider the derivation of the irregular past tense form of go, i.e. went. As was shown in (2), the past tense morpheme /-t/ is licensed in the context of class-2 verbs like send, build etc. Halle & Marantz (1993) now argue that the phonetic realization /wεnt/ that corresponds to the past tense form of go is derived via affixation of /-t/ to an irregular stem /wεnd/:


Tns 0

(5) V0 Vocabulary Insertion of /w7nd/, because of a [+past]-Tns 0

Tns 0 Vocabulary Insertion of /-t/, because /w7nd/ is a class-2 verb

The idea behind partial suppletion is that the lexical entry of a verb like go includes the information that a [+past]-context triggers Vocabulary Insertion of an irregular verbal stem. The form /wεnd/ in (5) is an allomorph of the verb go. Since /wεnd/ belongs to class 2, this in turn produces the right environment for the association of the entry /-t/ with the Tns0-node, and at phonological structure, we derive /wεnt/. Thus, we have a kind of “mutual dependency”; a [+past]-context triggers insertion of an extra verbal stem, and in turn, this stem triggers the choice of a particular past tense allomorph. As argued above, different particle meanings that are listed in the lexical entry of a prepositional element are the semantic equivalents of allomorphs. A parallel situation to (5) could be constructed in which the semantics of the verb (the semantic “host” of the particle) is determined by a particular particle, just like the phonological form of the verb in (5) is determined by the Tns-affix. Interestingly, such cases of what could be called partial “semantic suppletion” (cf. Zeller 1997b) can indeed be found. Consider the particle verbs in (6): (6) a.

(jemanden) zurichten, ‘make a mess of someone’ (somebody) part-judge b. (ein Fest) ausrichten, ‘give a party’ (a party part-judge c. (einen Hund) abrichten (a dog) part-judge, ‘train a dog’ d. (die Wohnung) einrichten (the flat) part(in)-judge, ‘furnish’

The particle verbs in (6) consist of a particle and the base verb richten. Importantly, the meaning of this verb in (6) is not related to the meaning that richten has in isolation. In Modern German, the simple verb richten has the meanings ‘to judge’ or ‘to point, direct’. In contrast, the base verb richten in the examples in (6) does not have this interpretation. Rather, its meaning might be paraphrased


as ‘arrange something, do something properly, make something right’.3 Aside from a few idiomatic exceptions4, this meaning of richten is only licensed in the context of a particle; it is never attested with the normal use of the simple verb. We therefore have found an example where the meaning of a particular verb in isolation is different from the meaning that is assigned to it if it is structurally adjacent to a particle. I assume that both this special meaning and the “regular” meaning of the verb richten are associated with one lexical entry. Both meanings are semantic alternants of the same verb; they are “allosemes”. In fact, the meaning of richten in (6) is as regular as the meaning of the simple verb richten — just different. The assumption that verbs may have various meanings all stored as part of their lexical entry is of course not new. For example, Jackendoff (1990, 1997) notes that the verb go has at least two readings, one as in (7) and one as in (8): (7) a. Peter went to New York b. [Event GO ([Thing ]A, [Path ])] (8) a. The road goes to LA b. [State EXTEND ([Thing ]A, [Path ])]

Whereas both meanings of go are licensed regardless of their syntactic environment, richten only adopts its special “make right”-meaning in particular contexts (i.e. with particles and with the idiomatic expressions mentioned in note 4). Its lexical entry must include the information that the particular phonological form of the verb richten and the syntactic node V0 are associated with this meaning if V0 is structurally adjacent to a terminal element P0. The representation of the particle verb einrichten, ‘furnish’, in (6d) looks like (9):

3.Etymological considerations are relevant here. The verb richten is based on the old German rihten, ‘make right’ (cf. Kluge 1960). It is possible that this meaning leads to the development of the specific meaning ‘judge’, but has maintained its more general semantics in combination with particles. 4.The relevant meaning of richten also appears in idioms like er wird’s schon richten, ‘he will set it right’, and in fixed expressions like die Krawatte richten, ‘arrange the tie’.


n, n>0

V

(9) PP

V0

P0

LCS-entry of ein; licensed in the context of the verb class that includes richten with the special meaning in (6)

special LCS-entry of richten; licensed because V0 is structurally adjacent to a particle

For both the verb and the particle, their structural adjacency-relation licenses a special meaning. If the syntax provides a structure like (9), both terminal nodes fulfill each other’s contextual conditions, and the respective special meanings can be associated with the terminal nodes P0 and V0. The semantics of the particle verb is then derived on the basis of the meaning of its parts. Recall from Chapter 4, Section 4.2.1, that this is the analysis of phrasal idioms proposed by Marantz (in prep., 1997). According to Marantz, the special meaning of a phrasal idiom is not linked to a phrase, but is derived compositionally from the meanings associated with the terminal nodes that the idiom consists of. As (9) shows, the semantic properties of many particle verbs are well compatible with Marantz’s proposal. The verb richten is not the only example of a verb that adopts a special meaning in the context of a particle. A similar point can be made about the verb stehen, ‘stand’, in (10): (10) a.

(jemandem) beistehen, ‘assist, help somebody’ (somebody) part-stand (lit. stand with somebody) b. (jemandem) nahestehen, ‘be closely associated with somebody’ (somebody) part-stand c. (jemandem) zustehen, ‘belong to somebody’ (somebody) part-stand

The base verb stehen, ‘stand’, in (10) does not have its basic locational meaning. Rather, the “standing” in (10) is metaphorical, but all three particle verbs are derived from the same metaphorical use. The meaning of the particle verbs in (10) is hence based on a special meaning of stehen that is licensed in combination


with particles. Interestingly, the particles in (10) are metaphorical versions of the prepositions bei, ‘at’, nahe, ‘close’, and zu, ‘to’. These metaphorical particle meanings are also licensed by other members of the class of locational verbs that are used metaphorically: (11) a.

(jemandem) beikommen, ‘get at somebody’ (somebody) part-come b. (jemandem) nahekommen, ‘get closer to somebody’ (somebody) part-come c. (jemandem) zukommen, ‘to be received by somebody’ (somebody) part-come

I assume that the meaning of the particles in (11) is the same as in (10). These metaphorical meanings are licensed in the context of certain metaphorically used verbs. In turn, the metaphorical meaning of the base verbs is licensed only in the context of these particles. Again, we have mutual semantic dependencies between structurally adjacent lexical nodes. Consider next the examples in (12): (12) a.

sich den Pullover anziehen, ‘put on the sweater’ refl the sweater part-pull b. sich den Pullover ausziehen, ‘take off the sweater’ refl the sweater part-pull c. sich umziehen, ‘change clothes’ refl part-pull

The base verb ziehen in (12) normally means ‘pull, drag’.5 Interestingly, in (12), it adopts a special meaning that seems to impose conceptual restrictions on the (implicit or explicit) internal argument — the Things that the objects of the verbs in (12a) and (12b) refer to must be clothes (cf. also sich einen Hut aufziehen, ‘put on a hat’). Again, the special meaning of ziehen that incorporates this conceptual condition is only licensed in the context of the particles. Now consider the following contrast, already discussed in Chapter 2, Section 2.4.2: (13) a. *An hat er den Pullover gezogen part has he the sweater pulled ‘He put on the sweater’

5.The verb ziehen in fact has more meanings in isolation. However, none of them seems to be related to the meaning that it has in (12).


b. *Aus hat er den Pullover gezogen part has he the sweater pulled ‘He took off his sweater’ c. Zu hat er die Tür gezogen part(close) has he the door pulled ‘He pulled the door close’

(13) suggests that topicalization of the particle is not permitted if the particle licenses a special meaning of the verb, even if the particle verbs as such can be contrasted. It seems that in (13a) and (13b), topicalization moves the particle out of the local domain of the verb; therefore, the special meaning of ziehen illustrated in (12) cannot be assigned to V0 in (13a) and (13b). In contrast, (13c) is grammatical, because the “regular” meaning of ziehen does not require a structurally adjacent particle element.6 Interestingly, there are also verbs whose only meaning is licensed in the context of a particle. (14) illustrates an example of a particle verb derived from an otherwise non-existing base verb: (14) (die Glut) anfachen, ‘blow into a flame, glow’ (the glow part-?

In (14), we have the meaning of an that also occurs in formations like anschalten, ‘switch on’, and ankurbeln, ‘turn on’. However, a simple verb fachen does not exist. Therefore, instead of analyzing anfachen as an independent lexical entry, we can assume that it is derived compositionally on the basis of the particle an and the verb fachen. Importantly, the meaning of the verb is only licensed in the context of this particle. Many particle verbs that in this respect are similar to (14) are derived from non-existing denominal or deadjectival base verbs; I come back to these verbs in Section 5.2.1. Consider the particle verb anhalten, ‘stop’, that has already been discussed in Section 4.2.4: (15) a. halten, ‘stop’, ‘hold’ b. anhalten, ‘stop’, *‘hold’

6.Notice that strictly speaking, V and Prt are still structurally adjacent in (13a) and (13b), since the particle phrase leaves behind a copy. The data in (13) hence require an additional assumption, namely that copies of elements that have been moved for semantic reasons are invisible for lexical licensing. This would yield the result that special verb meanings are not licensed by topicalized particles, whereas verb movement (which has no direct semantic effect) does not prevent the verb from licensing special meanings of the particle.


The base verb halten is ambiguous. However, only one meaning (‘stop’) is licensed if halten is structurally adjacent to the particle an. (15b) might hence be considered an example where a particular meaning of a verb (‘hold’) is licensed only if the verb is not structurally adjacent to a particle. The idea that the meaning of a verb is conditioned by a particle may further be used to explain another subclass of “irregular” verb-particle combinations, where a productive particle combines with a base verb whose meaning is drained or modified (cf. Stiebels 1996): (16) a.

auftauchen, ‘surface’ part-dive b. abbauen, ‘demolish, deconstruct’ part-build

The meaning of the particles in (16) can be defined in isolation. The particle auf in (16a) expresses upward movement (as in the examples in (17)); the particle ab in (16b) is the decremental ab that has been discussed already in Section 4.1.3 (cf. the data in (18)): (17) a.

aufspringen, ‘jump up’ part-jump b. (einen Stein) aufheben, ‘lift up (a stone)’ (a stone) part-lift

(18) a.

(seine Strafe) abbüßen, ‘serve a sentence’ (his sentence) part-atone b. (seine Schulden) abarbeiten, ‘work off one’s debts’ (his debts) part-work

The particles in (16) contribute the same meaning as they do in (17) and (18), but the meaning of the base verbs in (16) is different from the meaning that these verbs have in isolation. However, it is not a completely different meaning; it rather seems that certain aspects of the original verb meaning have become lost in the process of combining the verb with a particle. The base verb tauchen literally means ‘dive’ and therefore implies a downward movement under water. This aspect is contradicted by the particle’s meaning and hence is suppressed when the verb combines with auf. Similarly, not all aspects of the “full” meaning of bauen are preserved in the particle verb in (16b). For example, it is part of the meaning of the verb bauen that something is constructed; bauen takes an incremental Theme. However, this aspect is cancelled out by the contradictory meaning of ab in (16b); therefore, if a German speaker uses abbauen, she refers


to an activity similar to building, but not with the implication that something is built. In accordance with the proposal outlined in this section, one possibility of accounting for the examples in (16) would be to assume that the “reduced” meanings of the base verbs are listed with the respective lexical entry, together with the information that these meanings are only licensed in the context of the particular particles auf and ab, respectively. However, there is another interesting option of accounting for the semantic shift exhibited in (16). Recall that in phonology, it is possible that readjustment rules, which operate at the interface between syntactic and phonological structure, modify the Vocabulary item of a stem such that it “fits” together with an affix. For example, if the past tense affix /-t/ combines with a stem like /wεnd/ or /sεnd/, the stem-final /-d/ is deleted before the phonological forms /wεnt/ and /sεnt/ are derived. We could now assume that similar readjustment rules are operative at the interface between syntax and conceptual structure. If a particle like ab and a verb like bauen combine, these readjustment rules delete those aspects of the meaning of the base verb that are incompatible with the meaning of the semantic affix, just like the phonological part of a stem that is incompatible with the phonological form of the affix is deleted. This means that the base verb bauen with its “regular” meaning is a member of the class of verbs that decremental ab combines with. The reduction process that changes the meaning of bauen then results from the application of a semantic readjustment rule at the interface. Finally, even a particle verb like anfangen in (19), which is the classical example of a particle verb with a completely non-transparent semantics, could now be analyzed as a case of partial semantic suppletion: (19) anfangen, lit. part(on)-catch, ‘begin’

The event expressed by the particle verb anfangen in (19) has nothing to do with catching. The meaning of an, however, seems to be related to its meaning in (20): (20) a.

Der Zug rollt an the train rolls part ‘The train starts to move’ b. Das Auto fährt an the car drives part ‘The car starts driving’

The particle an in (20) marks ingressive aspect of an event. The verb anfangen seems to have inherited this meaning:


(21) a.

Peter fängt an zu lesen P. catches part to read ‘Peter starts to read’ b. Peter fängt seine Hausaufgaben an P. catches his homework part ‘Peter starts doing his homework’

Conceptually, anfangen takes an argument which expresses an Event (this Event is expressed through a sentence in (21a) and through a DP in (21b)). The particle verb expresses the start of this event, just like an in (20) does. This suggests that the meaning of anfangen is determined by the meaning of an, and that the meaning of the base verb fangen in the context of an is semantically empty. The lexical knowledge of a particle verb anfangen is knowledge that an means ‘start to’ in the context of fangen, and that the meaning of fangen is simply zero in the context of an.7 To conclude, in partial semantic suppletion, the lexical entry of the verb must include the information that a special meaning (a meaning which is different from the LCS that the base verb has in isolation) is associated with V0 if V0 is syntactically adjacent to a particular particle. In turn, the respective verb with this special interpretation is part of the class that licenses the respective particle meaning. In principle, it is possible that a particular verb meaning is only licensed by one single particle. Now recall that I argued in Chapter 4, Section 4.2.4, that there are certain particles that only combine with one single possible base verb; i.e. their meaning is only licensed by one particular element. Theoretically, these assumptions could now be combined to argue that every particle verb can be analyzed compositionally. For example, it could be argued that even the idiomatic particle verb aufhören (lit. up-hear, ‘stop’) is semantically transparent. It consists of a base verb hören with a special meaning which is licensed only in the context of auf, and of a particle auf with a special meaning which is licensed only in the context of hören. If we combine these two special meanings, we derive the meaning of aufhören. Recall again that this is precisely the reasoning behind Marantz’s (1997, in prep.) analysis of idioms.

7.If the interpretation of fangen in the context of an is really zero, then this is a further curious parallel between phonology and semantics. In the same way as certain syntactic contexts license the occurrence of phonological null morphemes (for example, the past tense affix Ø in the context of strong verbs), certain contexts seem to license the occurrence of semantically empty morphemes.


I do not want deny that the (conceptually attractive) possibility to derive every particle verb compositionally may indeed exist. However, at the end of Section 4.2.4 I have already pointed out that the idiomatic use of a particle verb like auftauchen in (16a) (which can be used metaphorically and mean ‘appear’) may turn out problematic for this analysis. What seems to be required here is a “genuine” phrasal idiom analysis, i.e. the idiomatic meaning of the particle verb is associated with V¢ or VP, as was proposed in Chapter 4, Section 4.1.4. Notice that the phonological realization of a particular concatenation of morphemes may also be completely idiosyncratic. This phenomenon is what Halle & Marantz (1993) call “full” suppletion. For example, if the English copula verb be appears in the context of a [+past]-Tns0, the whole complex Tns0-node is fully supplied with the phonological features of an idiosyncratic entry like were or was (see also Chapter 1, Section 1.2.3). In the semantic domain, I will therefore tentatively adopt the same assumption for idiomatic particle verbs like auftauchen, ‘appear’. I will analyze these verbs as cases of “full” semantic suppletion; in these examples, the combination of P0 and V0 is associated with a special meaning in the lexicon.

5.2 Morphological properties of particle verbs I argued that the meaning of a particle can be associated with a lexical P0element only if it is the head of the verb’s sister, (23b). Particles are hence structurally adjacent to the verb, with structural adjacency defined as in (22): (22) Structural adjacency A head X and the head Y of its complement YP are structurally adjacent.

XP

(23) a.

YP

Y

0

VP

b.

X0

PP

V0

P0

The specific contextual requirement expressed by structural adjacency is listed in the particle’s lexical entry. Furthermore, I showed in Section 5.1 that lexical


properties of verbs may also depend on the presence of a structurally adjacent particle. In contrast to particle verbs, the local domain which is relevant for licensing operations between stems and affixes is the domain of the syntactic word. For example, in order for Vocabulary Insertion to apply to a Tns0-morpheme, Tns0 and its host V0 must be part of the same word, as shown in (25b). I will henceforth assume that (the syntactic nodes corresponding to) a phonological affix and its host must be morphologically adjacent: (24) Morphological adjacency If a head X and a head Y merge to derive a complex X0, then X and Y are morphologically adjacent.

X 02

(25) a.

Y0

Tns 0

b.

X 01

V0

Tns 0

The difference between the configurations expressed by (23) and (25) is obvious. Morphological adjacency describes the domain of words and hence of morphology; as I showed in Chapter 1, Section 1.3.1, two elements that form a complex X0 form a morphological object. In contrast, structural adjacency is perceived as a “syntactic” relation, because the definition in (22) and the principles of X-bar theory require Y0 to project a phrase in order to be structurally adjacent to X0. However, what I want to emphasize here is the parallelism between both structures. As I have already noted in Chapter 1, Section 1.3.1, the structural adjacency-relation is traditionally described as head-government: in a structure like (23), X0 properly head-governs Y0. Interestingly, the notion of government has also been applied to morphological objects. Roberts (1985: 28) notes that the relation between an affix and a stem can be considered “a kind of headcomplement relation”, and he suggests that affixes morphologically govern stems inside words. This means that in (25b), the Tns-affix governs V0, and more generally, that X10 governs Y0 in a structure like (25a). Roberts concludes that “there exists both syntactic government and morphological government. These are instances of the same relation” (1985: ibid.). To make this point even stronger: in both (23a) and (25a), two heads X0 and Y0 stand in a particular local configuration. It is well-known that certain operations and dependencies between


terminal elements are licensed in morphological domains like (25); therefore, it has been argued that whenever the respective phenomena are observed, a structure like (25a) must be present. However, I showed in Section 5.1 that there are certain parallels between the structures in (23) and the structures in (25) with respect to context-dependencies. I show now that objects of the form in (23a) may also have other characteristics that are otherwise only attested within morphological domains, i.e. inside (complex) X0-categories. I discuss three phenomena that illustrate that the verb-particle construction shares certain properties with morphologically complex verbs. In Section 5.2.1 I show that both particles and prefixes license the occurrence of verbs that do not exist in isolation (i.e. without the prefix or particle). In Section 5.2.2 I discuss examples of Dutch particle verbs which show that particles, like prefixes, may change the auxiliary chosen by the verb in periphrastic constructions. Finally, Section 5.2.3 shows that particles may determine the case-assigning properties of the verb; a transfer that is only expected to be found in complex words. 5.2.1 Particle verbs with “non-existing” base verbs A property of particle verbs that makes them look like morphological objects is discussed in Booij (1990) for Dutch and in Zeller (1997b) and in McIntyre (2001) for German. Consider the German particle verbs in the right column of (26) and (27): (26) input A: a. leer, ‘empty’ b. schwarz, ‘black’

– –

particle verb: ausleeren, ‘empty’ anschwärzen, ‘blacken someone’s name’

(27) input N: a. Sattel, ‘saddle’ b. Farbe, ‘color’

– –

particle verb: aufsatteln, ‘put a saddle on a horse’ abfärben, ‘rub off’

Many particle verbs are derived from adjectives or nouns. At first glance, it looks as if the particles in (26) and (27) had category-changing power. However, this view would require an analysis according to which the particles in (26) and (27) were of category V and could pass these features up to the next node that dominates the adjective/noun and the particle. This analysis has been criticized for a number of reasons that I do not want to discuss here (see Booij (1990), Lüdeling (1998a), McIntyre (2001)). An alternative proposal is made by Booij (1990) and has been adopted by various other authors (cf. Stiebels 1996; Zeller 1997b; Lüdeling 1998a). It has been argued that the particles in (26) and (27) do

211


not attach directly to the respective adjectives or nouns, but to verbs that are derived from these elements by zero-conversion. This analysis works straightforwardly for the cases in (26) and (27), since the adjectives leer and schwarz as well as the nouns Sattel and Farbe can in fact be zero-converted into verbs: (28) input A: a. leer, ‘empty’ b. schwarz, ‘black’

– –

zero-converted verb: leeren, ‘empty’ schwärzen, ‘blacken’

(29) input N: a. Sattel, ‘saddle’ b. Farbe, ‘color’

– –

zero-converted verb: satteln, ‘saddle’ färben, ‘dye’

The problem with this analysis is that it implies that the particle verbs in (30) and (31) are derived from “non-existing” base verbs: (30) input A: a. reich, ‘rich’ b. mager, ‘skinny’

– –

(31) input N: a. Freund, ‘friend’

–

b. Kanzel, ‘pulpit’

–

particle verb: anreichern, ‘enrich’, but: *reichern abmagern, ‘grow thin’, but: *magern particle verb: anfreunden, ‘become friendly with’, but: *freunden abkanzeln, ‘dress down, tick off’, but: *kanzeln

The adjectives in (30) and the nouns in (31) cannot be zero-converted into verbs if the verb is not also combined with a particle. In other words, the verbs from which the particle verbs in (30) and (31) are derived do not occur in isolated form; they only exist as base verbs of these particle verbs. It should be clear how the examples in (30) and (31) are analyzed in the light of the theory that I advocate. In German (as in English), zero-conversion is a semi-productive rule which requires the output to be listed (cf. Chapter 4, Section 4.2.3). Therefore, I assume that the converted nouns and adjectives in the examples (28)–(31) exist as independent lexical entries. However, there is an important contrast between the verbs in (28) and (29) and those in (30) and (31). Whereas the former behave like normal verbs, the lexical entries of the latter must include the additional information that these verbs are only licensed in the context of a particle. In this respect, denominal verbs like reichern and freunden have the same status as non-existing base verbs like fachen that were discussed in (14) in Section 5.1, repeated here in (32):


(32) (die Glut) anfachen, ‘blow into a flame, glow’

I have argued above that a verb like fachen is only licensed in the context of a particle; its lexical entry specifies that its meaning can only be unified with V0 if this terminal node is structurally adjacent to a particle. The analysis of the particle verbs in (30) and (31) is similar. In both examples, the zero-derived base verb is only licensed if it is syntactically represented within the same locality domain as the particle: *V 0

(33) a.

A0

b.

–

V′ V0

PP

V0 Ø

P0 an

reicher-

A0

V0 Ø

reicher-

Interestingly, non-existing verbs are also licensed by verbal prefixes (cf. Booij 1990; Zeller 1997b; Ackerman & Webelhuth 1998), as (34) and (35) show: (34) a.

input A langsam, ‘slow’ b. input N: Herberge, ‘inn’

– –

prefix verb: verlangsamen, ‘slow down’, but: *langsamen prefix verb: beherbergen, ‘put up’, but: *herbergen

(35) (ein Feuer) entfachen, ‘light a fire’ (a fire pref – ?

(34) shows that the non-existing deadjectival verb langsamen and the nonexisting denominal herbergen are licensed only if they occur together with the prefixes ver- and be-, respectively. Similarly, the prefix verb in (35) is formed by combining the prefix ent- with the non-existing verb fachen that occurs with the particle an in (32). The non-existing verbs in (34) and (35) are only licensed if they occur in the same domain as a prefix. As I argued in Chapter 2, prefix verbs are complex V0s; this suggests that the relevant locality domain is defined by morphological adjacency:

213


*V 0

(36) a.

A0

langsam-

b.

V0 Ø

–

V0 V0

Pref 0

ver-

A0

V0 Ø

langsam-

The zero-converted V0 in (36a) is morphologically adjacent to a prefix in (36b), which yields the grammatical verb verlangsamen. This suggests that it is part of the lexical information associated with the verb langsamen that this verb is only licensed if it is morphologically adjacent to a lexical head like a prefix. The similarity between prefix verbs and particle verbs depicted in (30)–(32) and (34)–(35) seems to pose a problem for the syntactic approach. As I have argued in Chapter 2, the syntactic structure of particle verbs is similar to that of other verb-complement constructions and different from the structure of complex words. But if this is so, then one may wonder why particle verbs behave like prefix verbs with respect to the licensing of otherwise non-existing base verbs. The answer to this question follows if we turn our attention to the similarities that we find in comparing prefix verbs and particle verbs. Both the structure in (33b) and the structure in (36b) represent the base verb in a local relation to a lexical head with affixal properties. In the case of prefix verbs, the verb and the prefix establish this local relation because they form a word. In the case of particle verbs, it has been one of my major concerns in this book to show that the particle and the verb also form a local domain because particle phrases lack an extended projection. The heads of regular phrasal complements of the verb are functional heads. Therefore, a verb is never structurally adjacent to a lexical head inside its regular complement. It is only because particles do not project functional structure that they stand in a local relation to the verb and hence are relevant with respect to contextually conditioned lexical licensing. What I therefore suggest is that context-dependencies as they are attested with both particles and prefixes should not be captured with reference to specific structures, but rather with reference to local domains. In (37), I formulate a generalized contextual licensing rule that is part of the lexical entries of verbs like freunden, fachen, and langsamen: (37) *V, if V is not in the same locality domain D as some lexical head L


(38) Definition of locality domain D Two heads X and Y are within the same locality domain iff a. X and Y are morphologically adjacent b. X and Y are structurally adjacent

Something like (38) holds for the licensing of the base verbs discussed in this section; it states that these verbs must stand in a local relation to a prefix or a particle. (38) gives two options of how this local relation may be defined. (38a) is fulfilled by the prefix verbs in (34) and (35); (38b) is fulfilled by the particle verbs in (30)–(32). Of course, not every particle or prefix licenses a particular non-existing base verb. Further restrictions might have to be listed together with the general condition in (37) in the verb’s lexical entry; alternatively, these restrictions may follow from contextual conditions that are associated with the entry of the respective particles. For example, the fact that the verb reichern is licensed by the particle an, but not by e.g. the particle aus or the prefix ver- may simply follow from the fact that this verb is not part of the verb class that licenses the particle or prefix. At any rate, regardless of how one captures these additional restrictions, it should be clear that in the approach represented by (37) and (38), the observed similarities between prefixes and particles do not raise problems. Since local domains can be established both morphologically and syntactically, context-dependent lexical licensing is expected to be attested with different structural configurations, as long as these structures establish the right locality domains. In sum, the proposal in (37) and (38) accounts for the similarities of particle verbs and prefix verbs without having to deny their structural differences. The important thing about particle verbs is not so much that they are phrasal constructions, but that they consist of two non-functional heads that occur within the same locality domain. Locality is a structural property that is not restricted to the domain of words; with respect to certain morphological or lexical licensing conditions, it is irrelevant whether locality domains are defined by structural or by morphological adjacency. In the next two sections, I will discuss two more phenomena that support this claim. 5.2.2 Auxiliary selection Booij (1990) makes another observation which shows that particle verbs are similar to prefix verbs. There are a number of particle verbs in Dutch that select the auxiliary zijn, ‘be’, although the respective base verb selects hebben, ‘have’,

215


as its auxiliary in the intransitive use. The Dutch examples in (39) are taken from Booij (1990: 54): (39) base verb (selects hebben): a. glijden, ‘glide’ b. vliegen, ‘fly’ c. lopen, ‘walk’

particle verb (selects zijn): uitglijden, ‘slide’ uitvliegen, ‘fly away’ weglopen, ‘walk away’

I do not want to discuss the question whether auxiliary selection is an appropriate test for unaccusativity (but see Burzio 1986; Grewendorf 1989; Hoekstra & Mulder 1990; Levin & Rappaport 1992, 1995). It is irrelevant whether or not the particle verbs in the right column in (39) are unaccusatives. The question that concerns me here is under which structural conditions the change of auxiliary in (39) is licensed. At first sight, it seems that functional phrasal complements can trigger auxiliary change in the same way as particles: Consider the data in (40) and (41): (40) a.

Jan heeft gesprongen J. has jumped ‘Jan jumped’ b. Jan is in de sloot gesprongen J. is in the ditch jumped ‘Jan jumped into the ditch’ (Hoekstra & Mulder 1990: 4)

(41) a.

Ik heb gevlogen I have flown ‘I have flown’ b. Ik ben naar Amsterdam gevlogen I am to Amsterdam flown ‘I have flown to Amsterdam’ (Booij 1990: 55)

In (40) and (41), we find that verbs of motion select the auxiliary hebben when they occur without a complement, but select zijn when they combine with a full prepositional phrase FPPrep. It seems that the phrasal complement of the verbs in (40b) and (41b) has the same effect on auxiliary selection as the particles in (39). The observed pattern follows if it is assumed that the choice of the auxiliary depends on the meaning of the whole VP. For example, Hoekstra & Mulder (1990) assume that the relevant aspectual property that triggers zijn-selection is perfectivity; Booij (1990) argues that a predicate selects the auxiliary zijn when it denotes a change of state. Both particles and functional prepositional phrases can change the meaning of the VP in the relevant way that causes the auxiliary


zijn to be selected. Then, of course, the observation that particles may change the choice of auxiliary does not require any specific assumptions. Like the functional prepositional phrases in (40b) and (41b), particles are phrasal complements of the verb; it seems that both functional and non-functional phrases have auxiliary-changing power. However, the story is not quite over. Notice that it is only with verbs of motion that the addition of an FPPrep-complement causes a change in the choice of the auxiliary. With non-motional verbs that select hebben, it is impossible to find examples where the addition of a full prepositional phrase leads to the choice of a different auxiliary. Curiously, however, combining a particle with a non-motional verb may again have precisely this effect: (42) base verb (selects hebben): a. bijten, ‘bite’ b. studeren, ‘study’ c. koelen, ‘cool’ (Booij 1990: 54)

particle verb (selects zijn): uitbijten, ‘erode away’ afstuderen, ‘finish one’s studies’ afkoelen, ‘cool down’

The particle verbs in (42) select zijn, although the base verbs select hebben and do not express movement. These examples show that particles do exhibit a syntactic effect that is not attested with regular (functional) phrasal complements. Importantly, the same effect is observed with prefixes: (43) base verb (selects hebben): a. branden, ‘burn’ b. groeien, ‘grow’ c. slapen, ‘sleep’

prefix verb (selects zijn): verbranden, ‘burn up/away’ vergroeien, ‘grow wrong’ ontslapen, ‘die’

The base verbs in (43) are not verbs of motion and select hebben. The derived prefix verb expresses a change of state and selects zijn. In this respect, particle verbs clearly pattern with the complex words in (43). The change in the choice of an auxiliary with non-motional verbs can only be triggered by a non-functional element within the same locality domain as the verb. This observation can be taken as further evidence for the claim that particle verbs are essentially complex elements that consist of two lexical nodes that stand in a particular local relation, and that the syntactic configuration that defines this relation establishes a local domain that shares properties of the


morphological domain. Consequently, similar effects occur with both prefix and particle verbs.8 5.2.3 Case assignment Consider the following particle verbs: (44) a. b. c. d.

einem Zugdat nachschauen jemandemdat nachlaufen jemandemdat nachsprechen jemandemdat nachspüren

– – – –

‘gaze after a train’ ‘run after somebody’ ‘speak after somebody’ ‘spy on somebody’

(45) a. b. c. d.

jemandemdat zuhören einem Plandat zustimmen jemandemdat zunicken einem Ortdat zueilen

– – – –

‘listen to somebody’ ‘agree with a plan’ ‘nod to somebody’ ‘rush towards a place’

The particle verbs in (44) and (45) are derived from the prepositional particles zu and nach that correspond to the prepositions zu, ‘to, towards’, and nach, ‘after, past’. The meaning of these particles is more or less related to the meaning of the corresponding prepositions. The internal arguments of the particle verbs in all examples in (44) and (45) have been introduced by the particle and correspond to the reference objects of the spatial or temporal relations expressed by nach and zu (cf. Fanselow 1990). For example, the particle verb einem Ort zueilen can be paraphrased as zu einem Ort eilen, ‘run towards a place’, where the object of the particle verb is realized as the internal argument of a preposition inside the FPPrep-complement. All these properties of the particle verbs in (44) and (45) are straightforward and can be analyzed on the basis of the discussion in Chapter 4, Section 4.1.1 and 4.1.2. The interesting thing about (44) and (45) is the case on the object. All particle verbs in (44) and (45) assign dative case. This situation does not seem to be caused by properties of the respective base verbs; for example, schauen, ‘look’, in (44a), laufen, ‘run’, in (44b), and eilen, ‘rush’, in (45d) are intransitive, whereas stimmen, lit. ‘tune’, in (45b) and hören, ‘hear’, in (45a) assign accusative case to their internal arguments. The particle verbs in (44) and (45) hence

8.I have chosen Dutch to illustrate this argument because the number of inchoative particle verbs in German that pattern with the Dutch verbs in (42) tends towards zero. The only verb that I could find is the inchoative verb einschlafen, ‘fall asleep’, which selects sein, ‘be’, whereas schlafen, ‘sleep’, selects haben, ‘have’.


contrast with particle verbs like (46), which are derived from the same base verbs plus a different particle: (46) a.

jemanden anschauen somebody-acc part-look ‘look at somebody’ b. jemanden aufspüren somebody-acc part-track ‘track down somebody’ c. jemanden anhören somebody-acc part-hear ‘listen to somebody’

The direct objects in (46) bear accusative case. Hence, the case-assigning properties of the verbs in (44) and (45) are not likely to be determined by the verb. Rather, the contrast between the verbs in (44)–(45) and (46) suggests that the particles nach and zu cause the assignment of dative case. This assumption is strongly supported by the following observation. As shown in (47), the prepositions that correspond to the particles in (44) and (45) in fact assign dative case to their complements: (47) a.

Peter sucht[FP [PP nach [einem Mann]dat]] P. seeks after [a man ‘Peter is looking for a man’ b. Peter geht[FP [PP zu [seiner Freundin]dat]] P. goes to [his girl-friend ‘Peter goes to his girl-friend’

In contrast to those prepositions whose case-assigning properties depend on whether they express a Path or a Place (as was the case with prepositions like an or auf etc.; see Chapter 3, Section 3.1.4), the respective properties of the prepositions zu and nach do not change. The assignment of dative case to the reference object is an inherent lexical property of the prepositions nach and zu; if a functional head is present that triggers the assignment of case, dative case is assigned automatically. This observation strongly supports the view that dative case in (44) and (45) is determined by the case-assigning properties of the particle. Apparently, despite the lack of functional structure, the case-assigning properties of nach and zu are licensed in a verb-particle construction as well. One might object to this conclusion by pointing out that not all particle verbs with zu and nach select arguments bearing dative case:


(48) a.

ein Wort nachschauen (im Wörterbuch) a word-acc part-look (in.the dictionary ‘look up a word (in the dictionary)’ b. jemanden nachahmen somebody-acc part-imitate ‘imitate somebody’

(49) a.

jemanden zudecken somebody-acc part-cover ‘cover somebody’ b. eine Mahlzeit zubereiten a meal-acc part-prepare ‘prepare a meal’

(48) and (49) show that particle verbs with nach and zu that take accusative objects also exist. However, it would be precipitous to conclude that dative case on the objects in (44) and (45) is therefore a lexical accident. Particle verbs with nach that take accusative objects are very rare to begin with; the few examples that can be found really seem exceptional. Particle verbs with zu and an accusative object occur more frequently, but it seems that these examples cannot be treated on a par with the examples in (45). Notice that the internal argument in (49) is not the reference object of the particle; the object in (49) rather seems to be the Theme of the verbs decken, ‘cover’, and bereiten, ‘prepare’. If we restrict our attention to examples with direct objects that are the reference objects introduced by zu, we find that particle verbs with zu and an accusative object are as exceptional as accusative objects of particle verbs with nach. I hence assume that (48) and (49) do not provide a serious challenge for the claim that dative case assignment in (44) and (45) is determined by the lexical properties of the prepositional particles nach and zu.9

9.Not all particle verbs with particles corresponding to dative case-assigning prepositions can assign dative case. The preposition aus, like zu and nach, always assigns dative case to a complement, (i), but particle verbs derived from aus always take accusative complements if the base verb is not itself a dative case assigner, (ii-c): (i) (ii)

Peter kommt [aus [dem Haus]dat] a. den Eimer-acc ausschütten, ‘empty the bucket’ b. einen Plan-acc ausarbeiten, ‘work out a plan’ c. jemandem-dat aushelfen, ‘help somebody out’ (helfen assigns dative case)


This conclusion supports the analysis of Chapter 4, where I have argued that prepositions and their corresponding particles are represented through a single lexical entry which determines different meanings of a single element with respect to its syntactic context. If dative case assignment is a lexical property of the prepositions nach and zu, it is not surprising that it appears with particle verbs derived from these elements as well. However, the data in (44) and (45) now raise the question of how the particles can determine dative case on the object. The first option that must be excluded is the idea that the particles in (44) and (45) assign dative case directly to their DP-arguments. I have argued in Chapter 3 that prepositions can only assign case to their complements inside the PP if they are structurally adjacent to a functional head that licenses the assignment of their case properties. According to this view, it is impossible that particles assign case to their arguments, because they lack functional structure by definition. Although the elements zu and nach are lexically specified for assigning dative case, they cannot assign this case without a functional licenser.10 But even if we drop for a moment the idea that case properties require functional structure to be licensed, it can be shown that the particles in (44) and (45) do not assign case to their objects directly. Notice that prepositions assign case to the right. This means that in a structural configuration like (50a), the reference object of P0 cannot receive case from P0: *V′

(50) a.

V0 eilen

PrtP DP einem Ort

Prt 0 zu

no case

10.Notice that the same assumption is made in Chomsky (1995) with respect to the verbal feature [(assign) accusative case], which is a lexical property of V0, but only activated by a functional head v0 (see Chapter 3, Section 3.1.1).


V′

b.

PrtP Prt 0 zu

V0 eilen DP einem Ort

case

The only way the prepositional particle could assign case directly to a DP would be through a configuration like (50b). This structure, however, does not correspond to the linear order DP-Prt-Verb found with particle verbs, but yields the wrong order *Prt-DP-Verb. (50a), however, is excluded, because no case can be assigned in this structural configuration. The impossibility of (50a) is further illustrated by the examples in (51) and (52): (51) a.

weil Peter nicht zu dem Laden geht because P. not to the shop goes b. *weil Peter zu nicht dem Laden geht because P. to not the shop goes ‘because Peter does not go to the shop’

As (51b) shows, nicht cannot intervene between a preposition and its complement, but rather has to precede the whole prepositional phrase. However, (52) shows that in the unmarked order, the negation intervenes between the dative DP and the particle: (52) weil Peter dem Mädchen nicht zuhört because P. the girl-dat not part-listens ‘because Peter does not listen to the girl’

If (50a) was the correct structural representation for (45d), the position of nicht in (52) could not be explained. Negation is an adverb right-adjoined to VP. If we assume that dative case, like accusative case, is assigned (or checked) by the verb in the second specifier of vP, and the DP must move to SpecvP in order to receive case from the verb, the linear order in (52) follows directly. Provided that dative case assignment in (44) and (45) is determined by the respective properties of the prepositions nach and zu, the latter assumption


leads to the conclusion that the base verb in these examples must have inherited the case-assigning properties from the particle. This observation of course raises the question of why a base verb inherits the case-assigning property of the particles nach and zu, but not those of a particle like aus (which also assigns dative case when used as a preposition inside FPPrep, see note 9), but I do not intend to go into this matter here. The question that I am concerned with is how the case-assigning property of the particle can be passed on to the verb. Consider the structure in (53): vP

(53)

Spec DP

v′

dative case assignment/checking v 0i/j

VP PP

V 0i

V 0i

v 0j

P0

If the main verb adjoins to v0, the resulting complex head can assign the case that is lexically specified by the verb to the DP in SpecvP. In order to see how the case-assigning property of the particle is transferred to V0 and then to v0, let me first explain how the complex v0 in (53) receives the ability to assign case to the DP in SpecvP. Recall that head movement is a case of word formation by Move. In Chapter 1, Section 1.3.2, I have argued that morphological operations are licensed in words derived by Move as well. The relevant morphological operation that is at work in (53) is feature percolation. According to Selkirk’s (1982) Percolation Conventions, features of non-heads may percolate if the head of a word is underspecified for the feature of the non-head. Therefore, since v0 does not assign case by itself, we can assume that in (53), the lexical case-assigning property of V0 percolates up to the highest v-node (indicated by the shared index of v0 in (53)). But how does the verb receive its case-assigning feature from the particle? The answer to this question is again provided by the central claim made in


this chapter. A particle verb should not so much be seen as a verb with its phrasal complement, but rather as an element that consists of two lexical heads within a particular locality domain. Case-assignment properties are determined by heads; and since the particle and the verb are heads that are structurally adjacent, the observed percolation effects follow from the parallel between the two locality domains discussed above. I now suggest that the kind of index or feature percolation that is illustrated in (53) is not restricted to the morphological domain, but is licensed also when two heads are structurally adjacent. This means that the verb and the particle are in the right structural configuration in order for the verb to inherit the case-assigning property of the particle. The appropriate representation of case assignment in the examples in (44) and (45) therefore looks like (54): vP

(54)

Spec DP

v′ VP

PP

P 0k

dative case assignment/checking v 0i/j/k

V 0i/k V 0i/k

v 0j

dative case-assignment property

In (54), the case-assigning property of the prepositional element can percolate to the node that it is structurally adjacent to. This node is V0, a lexical head. Therefore, V0 can inherit this property, and incorporation into v0 explains that the DP-argument of the particle verbs in (44) and (45) may receive the case that is normally assigned by the corresponding preposition. I have argued in Chapter 3, Section 3.1.4, that the case-assignment property of a lexical head is licensed by the functional head of this head’s extended projection. Recall that prepositions do not incorporate into F0Prep, but since P0 and F0Prep are structurally adjacent, case assignment is possible. In (54), no extended projection of P0 is present; therefore, P0 cannot assign case on its own. However, it can transfer this property to the lexical head that it is structurally adjacent to. Dative case assignment in (54) is then licensed through the


functional head of the verb’s extended projection; i.e. through v0. This shows that a particular property of a lexical head L is licensed either through identification by a functional head or by percolation of this property to another lexical head. Crucially, both alternatives are possible even if the licensing head and L do not form a word. It is sufficient that both heads are structurally adjacent. The case-assignment properties of P0 in the extended projection of P are licensed by a structurally adjacent functional head. The respective property of the verb is triggered by a morphologically adjacent functional head v0 (on the assumption that V0 incorporates into v0, as in (54)). This shows that not only with respect to the relation between two lexical heads, but also with respect to the relation between a lexical and a functional head, the two domains defined by structural and morphological adjacency behave alike. In the next section, I will illustrate this latter point in more detail through the discussion of h-postpositions in German.

5.3 Local domains and the licensing of P0 The phenomena discussed in Section 5.2 have shown that certain relations between two lexical heads (a verb V0 and a prepositional prefix or particle P0) can be licensed by both structural and morphological adjacency. In this section, I discuss these locality domains with respect to relations that hold between a lexical and a functional head. Therefore, I return to prepositional elements, which I argued are licensed in the context of a functional head F0Prep (see Chapter 3, Section 3.1.4). In Section 5.3.1 I take a look at the status of h-postpositions, and in Section 5.3.2, I show that a lexical prepositional element is licensed by a local functional head regardless of whether the local relation is defined by morphological or structural adjacency. In Section 5.3.3, I discuss the semi-lexical status of h-postpositions that follows from the analysis presented in Sections 5.3.1 and 5.3.2. 5.3.1 The morphological structure of postpositions The discussion of h-verbs (= combinations of a verb and an h-postposition) in Section 3.2.2 has shown that certain h-postpositions can function as referential prepositional proforms. I have argued in Chapter 3 that this observation follows from the assumption that postpositions are realizations of the functional head F0Prep, but I have not been very explicit about the internal morphological


structure of these functional h-postpositions. Recall that German h-postpositions are morphologically complex; they consist of a prepositional element and a deictic element hin or her. In pleonastic circumpositional phrases like (55), the prepositional part of the h-postposition is homophonous with the head of its PP-complement: (55) a.

Peter kommt aus dem Haus heraus P. comes out the house H-out ‘Peter comes out of the house’ b. Peter steigt auf den Berg hinauf P. climbs on the mountain H-on ‘Peter climbs up the mountain’

What is the morphological status of the h-postpositional elements in (55)? According to the Right-hand Head Rule, their prepositional parts (aus in heraus; auf in hinauf) must be their morphological heads. But since the complex h-postposition is a functional head, and since the categorial status of a word is determined by its morphological head, we have to conclude that the rightmost prepositional elements aus and auf inside the h-postpositions in (55) are not of category P0, but functional elements of category FPrep: F 0Prep-2

(56)

F 0Prep hin

F 0Prep-1 auf

(56) shows that an h-postposition is derived by attaching the h-prefix (which I take to be of category F0Prep as well)11 to another functional node which is homophonous with a normal lexical preposition. The question is, what is the relation between the functional use of this prepositional element and its occurrence as a lexical preposition or particle? What I want to suggest here is that auf in (56) is a postpositional element that is morphologically derived from a simple prepositional element through the suffixation of a functional head which is phonologically zero:

11.The h-elements hin and her can also occur as prepositional proforms, which suggests that they are of category F0Prep (see Zeller forthcoming a for details).


F 0Prep = F 0Prep-1 in (56)

(57)

P0 auf

F 0Prep Ø

According to this proposal, the apparently simple postpositional heads of h-postpositions are themselves morphologically complex. They consist of a bare preposition of category P0 that is combined with a functional head F0Prep; this head is the morphological head of the postposition. Combining (57) with the structure in (56), we hence derive the following representation for German h-postpositions like hinauf: F 0Prep-2

(58)

F 0Prep hin

F 0Prep-1 P0 auf

F 0Prep Ø

According to (58), the derivation of h-postpositions in German is a two-step process. The first step derives a functional postposition from a lexical preposition; the second step merges the h-prefix with this postposition. The resulting h-postposition is a functional element. Independent evidence for the claim that h-postpositions are derived from simple postpositional elements comes from Dutch. Notice that topological postpositions like aus, auf, ein etc. do not occur in their simple form in German; they require a prefix. However, the corresponding postpositions in Dutch are in fact homophonous with the respective prepositions (cf. Chapter 3, Section 3.1.4): (59) a.

het huis uit the house out b. uit het huis out the house ‘out (of) the house’

(60) a.

de gevangenis in the jail in


b. in de gevangenis in the jail ‘into the jail’ (61) a.

de berg op the mountain up b. op de berg up the mountain ‘up the mountain’

Because of the similarities between Dutch and German, it is of course feasible that postpositions in Dutch are also functional elements. This welcome result follows naturally from the structural analysis of complex postpositions in German that I suggest in (58). Whereas the prepositions uit, in, and op in the (b)-examples in (59)–(61) are lexical P0-heads, the respective postpositions in the (a)-examples are morphologically complex. They have the structure in (57); the only difference between German and Dutch is that in German, these postpositions must be further prefixed with a deictic element her or hin (these elements do not exist in Dutch). Although the zero-operator in (57) and (58) is phonologically invisible, it has a specific semantic function. In Zeller (forthcoming a), I argue that it maps the meaning of a preposition to the meaning of a postposition. Olsen (1999) assumes that the semantics of a postpositional element incorporates the semantics of the bare preposition; for example, the postposition aus in (57) expresses the same spatial concept as the preposition aus. However, Olsen shows that the argument structure of a postposition differs from the linking properties of the corresponding preposition in an important respect. Whereas the reference object of a simple transitive preposition is obligatorily linked to syntax, (62), the reference object of a postposition can be omitted. Postpositions can be used as intransitive elements; as I have shown in Chapter 3, Section 3.2.2, an intransitive h-postposition is a prepositional proform whose implicit internal argument is interpreted referentially, (63): (62) a.

Peter springt aus *(dem Fenster) P. jumps out *(the window ‘Peter jumps out of the window’ b. Peter kommt in *(mein Zimmer) P. comes in *(my room ‘Peter comes into my room’


(63) a.

Peter springt hinaus P. jumps H-out ‘Peter jumps out’ b. Peter kommt herein P. comes H-in ‘Peter comes in’

I suggest that the functional zero-head in (57) and (58) is an operator that derives a postpositional element with an implicit reference object from a preposition with an obligatory argument (see Zeller (forthcoming a) for a more detailed semantic analysis): (64) a. preposition; category P0: obligatory reference object b. postposition; category F0Prep (P0 + zero-operator): implicit reference object

Simple prepositional elements link their internal arguments obligatorily to syntax. The attachment of a functional zero-element derives a postposition with an implicit internal argument. Due to the functional status of the empty operator, this implicit reference object of a postposition is always interpreted referentially. This follows from the relation between functional structure and referentiality discussed in Chapter 3, Section 3.2. An intransitive h-postposition is hence always interpreted as a referential prepositional proform However, recall that h-postpositions can also occur with PP-complements. Olsen (1999) suggests that the optional PP-complement of an h-postposition that we find in circumpositional phrases is used to identify the implicit internal argument of an h-postposition. The implicit argument of a postposition is most easily identified in pleonastic circumpositional phrases as in (65), where the postposition and its PP-complement are derived from the same preposition:12 (65) a.

Peter springt aus dem Fenster hinaus P. jumps out the window H-out ‘Peter jumps out of the window’ b. Peter kommt in mein Zimmer herein P. comes in my room H-in ‘Peter comes into my room’

12.McIntyre (2001) suggests that the P0-head of the PP-complement is simply a cognate copy of the prepositional part of the h-postposition. The “cognate” PP-complement of hinaus in (65) may thus be compared to the cognate object of the verb dream in a construction like dream a dream.


Since the P0-part that the h-postposition is derived from and the head of the PP in (65) have the same semantics, the PP-complement of a pleonastic circumpositional phrase provides the ideal means to identify the implicit reference object of an h-postposition. Now recall that I argued in Chapter 4, Section 4.3, that the semantics of a lexical preposition is licensed through a structurally adjacent functional head. The P0-heads inside postpositions, however, are not structurally adjacent to a functional head. How is the assignment of an LCS to the terminal P0-nodes in (57) and (58) licensed? This is the question to which I now turn. 5.3.2 Functional heads and local domains An important fact about h-postpositions in German, and postpositions in Dutch, is that they are always directional (cf. Abraham 1995). Since the meaning of a postposition is based on the meaning of a lexical P0-element, it must be concluded that the prepositional part of a postposition (the P0-node in (57) and (58) above) is directional as well. For example, the bare prepositions auf and op that derive the postpositions hinauf and op must express Paths. The question is how the directional meaning is licensed if P0 is not structurally adjacent to a functional head. The answer is provided by the assumptions about locality domains that I motivated in this chapter. Although the prepositional element P0 and the ØF0Prep-head in (57) and (58) are not structurally adjacent, they are morphologically adjacent; i.e. part of the same word. Apparently, the meaning of a preposition is licensed whenever P0 bears a local relation to a functional head F0Prep, regardless of whether this relation is defined by the word or by structural adjacency. The analysis I present here therefore requires a restatement of the syntactic licensing conditions of prepositional elements. A functional head does not have to be structurally adjacent to a preposition in order to license its meaning; it is sufficient that both heads are part of the same locality domain, with “locality domain” defined as in (38) in Section 5.2 above: (66) Definition of locality domain D Two heads X and Y are within the same locality domain iff a. X and Y are morphologically adjacent b. X and Y are structurally adjacent

Again, we find that a relation between two nodes that is licensed by structural


adjacency can also be satisfied by morphological adjacency (i.e. if both terminal elements are part of the same word). In contrast to the examples discussed in Sections 5.1 and 5.2, the relation that is relevant here is that between a lexical and a functional head. The structural representation of a pleonastic circumpositional phrase like auf den Berg hinauf in German hence looks like (67): FPPrep

(67)

F 0Prep [+directional]

PP P 01 auf

DP den Berg

F 0Prep hin

F 0Prep P 02 auf

F 0Prep [+directional]

In (67), the directional preposition auf appears twice. Its first occurrence (P01 inside the PP) is licensed through the syntactic head of the FPPrep by structural adjacency.13 Its second occurrence (P02) is licensed by the morphological head of F0Prep (the zero-operator) via morphological adjacency. Consequently, the head of the whole circumpositional phrase is interpreted as directional, and its implicit reference object is simultaneously identified through a directional PP-complement.14

13.One could assume that the directional meaning of P0 in a pleonastic circumpositional phrase like (67) is determined by the syntactic feature [+directional] of the postposition. However, this assumption encounters problems with postpositional phrases like (i): (i)

unter [der Brücke durch under [the bridge]-dat through ‘(pass) under the bridge’

Although the postposition durch in (i) is directional, its complement-PP is locative (the preposition unter assigns dative case to its DP-complement die Brücke). This is due to the non-topological character of durch that requires the identification of the internal argument through a locative PP-complement (cf. Olsen 1999). It is therefore not the syntactic feature of the postposition, but its semantics, that determines the interpretation of the head of its complement. 14.It might be possible to analyze pleonastic constructions in terms of chain formation, such that the head of the PP and the P0-part of a postposition form a chain. See Zeller (forthcoming a) for some speculations.

231


5.3.3 Postpositions as semi-lexical elements In order to complete the discussion of h-postpositions, let me finally turn to an important implication of the analysis proposed here. The structure in (58) shows that postpositions are hybrids. On the one hand, they are functional elements; they are heads of fully functional prepositional phrases. Therefore, postpositional and circumpositional phrases pattern with simple prepositional phrases with respect to a number of phenomena that distinguish functional prepositional phrases from particle phrases. For example, the functional head of a circumpositional phrase enables the head of its PP-complement to assign case to a DP, whereas arguments of particles must receive case outside the PP; postpositional and circumpositional phrases express referential Path-tokens, whereas particle phrases express types (see Chapter 3, Section 3.2.2) etc. On the other hand, however, postpositions are “less” functional than genuine functional elements like e.g. determiners, inflectional morphemes, or zero-F0Prep-heads. Postpositions have thematic properties like lexical elements, because they incorporate the semantics of their corresponding prepositions. They are morphologically complex functional elements derived from a lexical element, whose semantic properties they inherit. Therefore, h-postpositions are “disguised” lexical categories; they are functional categories, but with properties of items from the non-functional domain. In Zeller (forthcoming a), I therefore propose to consider postpositions as “semi-lexical” heads. Note that postpositions are not the only elements which are difficult to classify as either functional or lexical. Emonds (1985) argues that items with semi-lexical properties are also found in the nominal domain (e.g. certain classifiers in pseudopartitive constructions as in “three pieces of cattle”) and in the verbal domain (e.g. auxiliaries). Although these elements are semantically comparable to fully lexical nouns and verbs, and are therefore “less grammatical” than e.g. determiners or inflectional morphemes, they are syntactically associated with a functional position. It seems that “semi-lexicality” plays more than a marginal role in grammar (see also van Riemsdijk 1998b for a discussion of various phenomena). However, at first sight, this notion seems incompatible with the dichotomy that is proposed by Grimshaw (1991) in her theory of extended projections (see Chapter 3, Section 3.1). Whereas Grimshaw assumes that an element is either lexical or functional (it has a functional value 0 or >0), semi-lexical elements now seem to constitute a third possible class. However, the account of postpositions that I have proposed in Section 5.3.1 offers a possibility to analyze semi-lexical elements on the basis of a morpholog-


ical interaction between a lexical and a functional part. A semi-lexical element is a complex word derived by combining a lexical and a functional head. The lexical head contributes its semantics, the functional head, as the morphological head of the derived word, determines the functional status of the derived element. According to this view, semi-lexical elements should not be compared to fully functional elements like e.g. determiners or complementizers, but rather to inflected verbs or nouns. This means that h-postpositions are semi-lexical elements; like particles, they have semantic content. If we now take a closer look at the local environment of an h-postposition, we notice another interesting parallel between particles and h-postpositions. Like particles, h-postpositions are structurally adjacent to the verb: V′

(68)

V0

FPPrep F 0Prep

PP F 0Prep

structural adjacency

F 0Prep P0

F 0Prep

(68) illustrates that the maximal projection of an h-postposition is a full FPPrep, which explains that postpositional and circumpositional phrases pattern with regular prepositional phrases. However, (68) also shows that h-postpositions pattern with particles with respect to their syntactic environment. Whereas a particle and a verb are structurally adjacent because the particle phrase lacks functional structure, a postposition is structurally adjacent to the verb because it is the head of the verb’s functional complement. The observation that h-postpositions are semi-lexical elements that are structurally adjacent to the verb explains why postpositions share many properties of particles. For example, it has been noted that the semantics of many h-postpositions is not as compositional as their transparent morphological structure suggests. The detailed analysis of the lexical representation of h-postpositions offered by McIntyre (2001) shows that h-postpositions are not related to their basic prepositional elements in a systematic manner. McIntyre notes


that although the meaning of h-postpositions like herein or hinaus comes close to the semantics of the respective prepositions ein and aus, other h-postpositions are not as clearly related to the semantics of their parts. For example, herab always expresses downward movement, a semantic aspect that is not necessarily present in the meaning of the preposition (or the particle) ab. The h-postposition hervor expresses emergence from concealment, whereas the simple preposition vor refers to (the Path to) the front of a Thing. McIntyre concludes that, although h-postpositions might be related to the lexical entries of prepositions by semi-productive rules, they are lexically listed as individual items. Furthermore, it can also be observed that certain h-postpositions, like particles, show a great variety of meanings. Some examples from McIntyre (2001) are given in (69) and (70): (69) herum: a. postpositional meaning, based on the meaning of the preposition um: movement around as seen from the center (um etwas) herumgehen, -fahren, -hüpfen, ‘go, drive, jump around (something)’ b. action without immediate accomplishment of its purpose herumdoktern, ‘try one’s hand at’, herumwirtschaften, ‘potter around’ c. inappropriate or inept activity herumschreien, ‘shout’, herumalbern, ‘fool around’ d. idiosyncratic meaning herumreiten auf etw., ‘harp upon something’ (70) heran: a. postpositional meaning, based on the meaning of the preposition an: movement towards a place (an etwas) heranfahren, -schleichen, -rollen, ‘drive, creep, roll towards something’ b. to come close to verge on herankommen, heranreichen, ‘come close to, approach, border on’ c. temporal approach heranrücken, heranbrechen, heranziehen, ‘draw near, approach, advance’

Again, it is clear that the different special meanings of the h-postpositions in (69) and (70) must be listed in the lexicon. A question that is raised by this observation is whether lexically listed h-postpositions depend on the presence of a verb in a similar way as particles.


Are h-postpositions restricted to verbal contexts, or is their meaning also licensed in the context of a structurally adjacent, non-verbal head? The data are ambiguous. Whereas h-postpositions with a transparent semantics that adopt the spatial meaning of the underlying preposition can occur as the complement of N0, (71), h-postpositions with more specialized meanings cannot, (72): (71) a.

die Fahrt hinaus ins Grüne the drive H-out into.the green ‘the drive out into the green’ b. der Gang um die Kirche herum the walk around the church H-around ‘the walk around the church’ c. der Weg hinunter the way H-under ‘the way down’

(72) a. *das Wirtschaften herum (cf. (69b)) b. *der Bruch heran (cf. (70c))

(72) might be taken as evidence for the assumption that at least some h-postpositions require the presence of a structurally adjacent verb, as was argued to be the case with prepositional particles. The elaboration of a fully developed account of h-postpositions would go beyond the scope of this study. What I have intended to show is that the differences between these elements and particles follow from the functional status of the h-postposition’s maximal projection, whereas the similarities follow from the semi-lexical status of h-postpositions and the fact that their local relation to the verb is the same as the one between the verb and a particle. As a result of the ambiguous status of postpositions, a speaker may have at her disposal two different mental representations of a construction with a verb and an FPPrep-complement headed by a postposition. For example, a clause like (73) might be analyzed as a combination of the verb gehen, ‘go’, and a pleonastic circumpositional phrase, (74): (73) in das Haus hineingehen in the house H-in-go (74) [in das Haus hinein] gehen

In that case, the PP-complement is semantically combined with the postposition, and the whole circumpositional FPPrep is interpreted as the complement of the


verb. However, there is an alternative. Recall that a particle verb consists of a verb and the head of its complement; although both nodes do not form a word in the morphological sense, they form a semantic unit. Now, if h-postpositions are in fact similar to particles, a speaker may analyze the combination of hinein and gehen in (73) as one semantic unit, with the PP in das Haus as an argument of the h-verb, (75): (75) [in das Haus] [hineingehen]

The reader should not be confused by the brackets in (75). The verb and the h-postposition do not form a word. (74) and (75) do not represent a morphosyntactic ambiguity (as is argued, for example, by Olsen (1996, 1997b)). An h-postposition never forms a constituent with the verb; structurally, it is always the head of the verb’s complement. Nevertheless, I argued that two structurally adjacent terminal nodes also form a construction, because structural adjacency defines a particular local syntactic domain. Therefore, an h-postposition and a verb do not have to be part of the same word in order to form a semantic unit. Moreover, h-postpositions are not “normal” functional, but semi-lexical, elements. These two observations combined make certain h-postpositions comparable to particles and suggest that the respective h-verbs and particle verbs may be treated along similar lines. In Chapters 6 and 7, I show that this possibility explains why postpositions and particles behave alike with respect to word formation in German and Verb Raising in Dutch. To summarize, there are differences between h-postpositions and particles that follow from the fact that an h-postposition is of category F0Prep, whereas particles are lexical categories. However, there are also similarities between particle verbs and h-verbs that are the result of (i) the structural relation that both elements bear to the verb and (ii) the semi-lexical status of h-postpositions which follows from the fact that postpositions are functional heads derived from lexical elements.

5.4 Conclusion In this chapter I discussed some of the morphological properties of particle verbs. I argued that they follow from the proposal made in Chapter 4, according to which a single lexical entry of a prepositional element includes different LCSs that are licensed in specific syntactic contexts. These contexts are determined by the terminal node that the element is structurally adjacent to; if it is a verb, the


prepositional element is a particle. The properties of a particle hence depend on properties of its “host”. I showed in Section 5.1 that a similar phenomenon is also attested in the phonological domain, where it is known as conditioned allomorphy. Due to the similarities between allomorphs and particles, I called particles “semantic affixes”. In Section 5.2 I showed that particle verbs have properties that are usually attested only in the morphological domain. I accounted for these properties by arguing that the local domain that is necessary for the licensing of zero-conversion, auxiliary selection, or the transfer of case-assignment properties, may be defined both by structural adjacency and by morphological adjacency. Finally, the discussion of the structural and lexical properties of h-postpositions and pleonastic circumpositional phrases in Section 5.3 confirmed the parallels between these two local domains. These parallels will be relevant in the next chapter as well when I look at the word formation properties of particle verbs. In sum, particle verbs share with genuine morphological objects the property that they consist of two terminal nodes. The requirement that particles must be structurally adjacent to verbs is listed in the lexical entry of a particle. In the next chapter, I show that this situation may also have another consequence. The local domain which is characteristic of particle verbs may be reanalyzed as the domain of morphological objects. In other words, the structural adjacency-requirement may also be satisfied by a syntactic structure in which the particle and the verb are morphologically adjacent.

Chapter 6

Particle verbs and word formation

The results of the analysis of particle verbs that I proposed thus far in this study can be summarized as follows. Like a complex morphological element, the verbparticle construction consists of two terminal nodes in a strictly local relation and therefore exhibits properties that are also attested in the morphological domain. Furthermore, I argued that the particle is a semantic affix, which means that it is a lexical peculiarity of particles that they must stand in a particular syntactic relation to a verb. However, in contrast to phonological affixes, the local domain in which a particle is licensed is not defined by the syntactic word, but by structural adjacency. Structural adjacency by definition means that the particle projects a phrase which is the verb’s complement. Therefore, the particle verb exhibits syntactic properties which are similar to other verb-complement constructions. For example, the particle verb cannot undergo movement as a whole, because the next node that dominates the particle and the verb is a V¢- or VP-projection, which cannot be accessed by head movement rules. In this chapter I look at some structural consequences of the ambivalent character of the particle verb. I show that in certain contexts, the local relation between the particle and its host may be reanalyzed as morphological adjacency. I propose that, although the syntactic representation of the particle verb as a verb plus phrasal complement follows from the lexical licensing conditions of the particle, the particle verb may be represented syntactically as a V0 in morphological contexts. I argue in Section 6.1 that the particle verb may be reanalyzed as a morphological object whenever it participates in operations of derivational word formation. In Section 6.2 I capture this situation through a principle of reanalysis, which operates on lexical unification. This principle states that LSS-entries with structurally adjacent terminal nodes can, under certain conditions, be unified with a syntactic structure in which these two nodes are morphologically adjacent. In Section 6.3 I briefly discuss the consequences of reanalysis with respect to h-postpositions and adjectival resultative constructions.


6.1 Particle verbs in words In this section I focus on the observation that particle verbs can productively provide the input to further word formation processes. This property is illustrated for German in (1) and (2) (cf. Booij 1990; Neeleman & Weerman 1993; Neeleman 1994 for Dutch): (1) particle verb a. einführen, ‘introduce’ b. ausgraben, ‘excavate’ c. einbrechen, ‘break in’

– – –

derived nominal Einführung, ‘introduction’ Ausgrabung, ‘excavation’ Einbrecher, ‘burglar’

(2) particle verb a. aufblasen, ‘inflate’ b. ausweichen, ‘evade’ c. annehmen, ‘accept’

– – –

derived adjective aufblasbar, ‘inflatable’ unausweichlich, ‘inevitable’ unannehmbar, ‘unacceptable’

The right columns in (1) and (2) show adjectives and nominals derived from particle verbs. These adjectives and nominals are syntactic words, i.e. they are represented as (complex) A0- or N0- categories in the syntax and project their own phrases. In this section, I address the question of what the internal structure of these words looks like. Section 6.1.1 provides evidence that the words in (1) and (2) are not formed by attaching a particle to a derived adjective or noun, but are truly derived from particle verbs. Section 6.1.2 shows that derivational affixes usually do not attach to phrases in German, which consequently creates a problem with respect to the derived words in (1) and (2). Finally, Section 6.1.3 provides empirical evidence that particle verbs inside derived -ung-nominals and in un-prefixed adjectives are represented as complex V0s. 6.1.1 The morphological bracketing of words derived from particle verbs In this section I show that the derived words in (1) and (2) above are derived from particle verbs by affixation; the respective derivational affixes (-bar, -ung etc.) attach to the particle verb. An alternative option to analyze word formation from particle verbs which I will briefly discuss and reject here has been suggested in Stiebels & Wunderlich (1994). Stiebels and Wunderlich argue that nouns such as e.g. Einführung, ‘introduction’, in (1a) are derived by combining a particle-like element and a noun derived from the base verb. In other words, according to Stiebels & Wunderlich (1994), Einführung does not have the

Particle verbs and word formation 241

expected morphological structure in (3a), but rather the structure in (3b):1 (3) a. [[ein-führ-]V0 -ung]N0 b. [ein [führ-ung]N0]N0

Notice that (3b) is a bracketing paradox. The semantics does not correspond to the proposed structure; einführen, means ‘introduce’, and accordingly, Einführung means ‘introduction’. But the bare noun Führung means ‘leadership’, and there is no obvious way to derive the meaning ‘introduction’ from the combination of this noun and the element ein. To solve this problem, Stiebels and Wunderlich adopt an interpretative convention called “lexical relatedness” (cf. Williams 1981; Di Sciullo & Williams 1987). This convention allows for the possibility of interpreting a noun with a structure like (3b) as if it had the structure in (3a). A number of arguments have been provided against this analysis of nouns derived from particle verbs. One argument is given by the fact that the approach suggested by Stiebels and Wunderlich predicts that words like the ones in (4) are derived from adjectives and nouns that do not exist independently (see Neeleman & Weerman 1993 who make the same point for Dutch): (4) a.

aufmerksam but: *merksam part-noticable, ‘attentive’ b. Ausschreibung but: *Schreibung part-writing, ‘announcement’

This argument is not very strong in light of the observation made in Chapter 5, where I showed that the lexicon may include verbs which are only licensed in the context of a particular particle or prefix. Stiebels and Wunderlich could therefore respond by suggesting that the adjective *merksam and the noun

1.Stiebels & Wunderlich (1994) are forced to analyze words derived from particle verbs as in (3b), since according to their proposal, a particle receives the feature [+max] in the context of a verb and hence remains visible in syntax. (This is how they account for syntactic separability, cf. the discussion in Chapter 2, Section 2.1.2.) However, since nouns derived from particle verbs are never syntactically transparent, Stiebels and Wunderlich have to stipulate that words like Einführung do not include an element marked as [+max]. This excludes the possibility to derive nouns from particle verbs as in (3a), since here the verb makes the prepositional element ein a particle, and ein is marked as [+max]. Instead, Stiebels and Wunderlich assume the structure in (3b), where the verb is nominalized before it combines with the element ein. Therefore, there is no verb in (3b) which makes ein a particle, and the [+max]-feature is not assigned.


*Schreibung in (4) are only licensed in the context of auf and aus, respectively. However, a related, but stronger argument against a structure like (3b) is given in Groos (1989) with respect to Dutch nouns derived from particle verbs. Groos points out that the Dutch nominal affix -ing is sensitive to the aspectual specification of the base verb to which it attaches. The requirement is that the base verb has terminative aspect. Certain particles can turn non-terminative verbs into terminative ones. Now (5) shows that -ing can attach to a terminative particle verb even if the base verb is non-terminative: (5) a. *drijving, ‘driving’ but: aandrijving, ‘drive, power’; uitdrijving, ‘expulsion’ b. *schrijving, ‘writing’ but: inschrijving, ‘registration’; anschrijving, ‘instruction’

(5) clearly proves that what has been nominalized is the terminative particle verb, not the non-terminative base verb. If it was always the base verb that a derivational morpheme like -ing combined with before the particle is attached, then the non-terminativity of the base verbs drijv- and schrijv- should exclude affixation of -ing in (5). Since these derivations are possible, (5) provides a strong argument against a structure as in (3b). I thus conclude that nominals and adjectives like the ones discussed above are really derived from the underlying particle verbs. (For more discussion and more arguments against Stiebels & Wunderlich’s analysis, see Lüdeling (1998a) and McIntyre (in prep.).) In the next section, I address the question of how a particle verb is represented inside these words.2

2.So-called root nominalizations with particle verbs may be derived by combining a prepositional element with a root noun: (i)

a. b.

durchsehenV part-see, ‘look through’ aufgehenV part-go, ‘rise’

– –

DurchsichtN ‘inspection’ Aufgang ‘rise’

As noted in McIntyre (in prep.), the Ablaut-process that led to the creation of nouns like Sicht, ‘view’, derived from the verb sehen, ‘see’, is truly extinct. It is hence problematic to assume that a root noun like Durchsicht is derived from a particle verb like durchsehen through the application of this process. This means that the structure of these elements might in fact look like (3b); i.e. the noun is modified with a preposition. However, as McIntyre (in prep.) shows, this conclusion does not carry over to examples of words derived from particle


6.1.2 The “No Phrase Constraint”, Compounding and Affixation I argued that particles are heads of phrasal complements of the verb. Consequently, the next node dominating both the particle and the verb is a nonminimal verbal projection, i.e. V¢ or VP. However, particle verbs can appear inside words as well, and Section 6.1.1 showed that these words are derived by attaching derivational affixes to the particle verb. Combining these claims, we arrive at the unwelcome conclusion that words derived from particle verbs include phrases. However, it is traditionally assumed that phrases are not allowed in words, a restriction that is sometimes dubbed the No Phrase Constraint (a term coined in Botha 19833). For example, this constraint explains why the VPs in (6) cannot undergo word formation: (6) a.

durchs Museum führen – through.the museum guide ‘guide through the museum’ b. ein rotes Auto kaufen – a red car buy ‘buy a red car’ c. auf den Tisch stellen – on the table put ‘put on the table’

*die Durchs-Museum-Führung

*der Ein-rotes-Auto-Kauf

*auf-den-Tisch-stellbar

(6) shows that word formation is not possible with constructions that consist of a verb plus a full phrasal complement. But if the unacceptability of (6) is due to a general ban on phrases in words, the syntactic approach that I argued for in this study fails to account for the word formation properties of particle verbs. A detailed investigation of the behavior of particle verbs in word formation is provided by Lüdeling (1998a). Lüdeling also advocates a syntactic approach to particle verbs. Despite the ungrammaticality of examples like (6), she argues that the structures in (7) are the correct representations for words derived from particle verbs:

verbs by affixation. I refer the reader to McIntyre (in prep.) for more discussion. 3.Notice that Botha (1983) himself argues against this constraint.


N0

(7) a.

N0 -ung

VP PP P0 aus

V0 grab-

A0

b.

VP PP P0 auf

V0 blas-

A0 -bar

Since nouns like Ausgrabung, ‘excavation’, or adjectives like aufblasbar, ‘inflatable’, are derived from particle verbs like ausgraben, ‘excavate’, and aufblasen, ‘inflate’, respectively, Lüdeling assumes that these words include full phrases; i.e. a V¢- or a VP-projection that dominates the verb and its phrasal complement, the particle phrase. Lüdeling’s main argument is based on the observation that the No Phrase Constraint is not uncontroversial and has been confronted with numerous counterexamples in the literature. The structure of phrasal compounds, which exist in various languages, is one of the most frequently noted violations of the No Phrase Constraint (cf. Toman 1983; Lieber 1992: 11ff.): (8) a. b. c. d.

(the) Charles and Di syndrome liewer vas mense (Afrikaans; ‘preferably together people’) lach of ik schiet humor (Dutch; ‘laugh or I shoot humor’) (die) Wer-war-das-Frage (German; ‘the who was it question’)

As (8) shows, in languages like English, Afrikaans, Dutch, and German, phrases may productively be the input to compounding. Clearly, in a compound like (8d), a full CP (Wer war das?, ‘Who was that?’) combines with the N0 Frage, ‘question’. The result is a word; i.e. a complex N0 which includes a full phrase. The examples in (8) therefore show that the No Phrase Constraint does not hold for compounds. But does this observation justify the conclusion that the No Phrase Constraint need not be obeyed inside other complex words either? Recall from Chapter 1, Section 1.3.2, that I do not assume a structural difference between words derived by compounding and words derived by affixation. A compound like Scherzfrage, ‘jocular question’, consists of X0-elements, just like the structure of a complex word like Lösung, ‘solution’, which is derived


from the verb lösen, ‘solve’, via affixation:4 N0

(9) a.

N0 Scherz

N0

b.

N0 Frage

V0 Lös-

N0 -ung

I assume that both compounds and derived words are formed syntactically by Merging two heads. A proposal in the same spirit is made in Höhle (1982), who argues that there is no formal difference between compounds and complex words formed by affixation. However, Höhle also emphasizes that there is one important property that still distinguishes affixes from morphemes that take part in compounding. Affixes are bound to a particular context that is specified as part of the lexical entry of the affix. In discussing the plural suffix /z/ in Section 1.3.2, I showed that this boundedness of an affix may be captured through its morphological subcategorization frame (cf. Lieber 1980, 1992). The affix /z/ requires a particular syntactic context; it must attach to an N0-category. In the system of lexical licensing adopted here, morphological subcategorization is represented through LSS-entries. For example, the LSS of an affix like -ung in (9b) looks like (10), with the lexical index a marking the actual element, and the rest of the structure specifying the contextual conditions: (10) LSS of -ung:

N0 V0

N 0a

The lower N0-element is linked to the LPS and the LCS of -ung through the

4.(9a) is a simplified structure, since it represents the compound as being derived from the free morphemes Scherz and Frage. However, as noted in Grewendorf & Hamm & Sternefeld (1987), there is evidence that compounding in German should rather be defined as the combination of stems, which are bound morphemes. According to this view, the nominal compound Scherzfrage would have the structure in (i): (i)

[[Scherz- [frag-]] -e]

I will ignore this qualification here. Stems still differ from affixes in that they do not put any constraints on the status of the element that they combine with. Therefore, it is immaterial for the argument that I make in the text whether compounds are derived from stems or free morphemes.


lexical index a. Crucially, (10) requires N0 to be morphologically adjacent to a verbal head. Consequently, we expect that examples with -ung attaching to a full phrase are ill-formed. The LSS in (10) cannot be unified with a structure in which the morphological subcategorization requirement of the affix -ung is not satisfied. Since the LSS of every affix specifies a contextual condition, the No Phrase Constraint is expected to hold for other derivational affixes as well. In contrast to affixes, elements involved in compounding do not put any constraints on the syntactic status of the element with which they combine. Therefore, in contrast to derivational affixes, a stem or a free morpheme may in principle combine with a full phrase and derive a complex X0 which violates the No Phrase Constraint, even though the existence of such constructions seems to be the exception rather than the rule (cf. Toman 1983; Di Sciullo & Williams 1987; and Chapter 1, Section 1.3.1). I assume that because of this important difference between derivation and compounding, it is not possible to take the existence of phrasal compounds as evidence for the claim that words derived by affixation may also include phrases. Derivation and Compounding are two different kinds of word formation. Affixes are specified for the syntactic context in which they must occur, and because of these contextual licensing conditions, it is rather unlikely that the No Phrase Constraint can be violated in words derived by affixation.5

5.Lieber (1992) discusses more examples that are supposed to show that syntactic words can be formed by attaching an affix to a phrase. For instance, the possessive affix -s in English attaches to complex NPs: (i)

a. b.

[a friend of mine]DP’s book [a man I know]DP’s hat

(Lieber 1992: 14)

However, the examples in (i) do not show what they are supposed to show, because the possessive affix -s and the preceding phrase do not form a syntactic word. Rather, it has been assumed that the phonological realization of the possessive affix is associated with the head of the DP, i.e. D0, whereas its phonological host in (i) is part of a syntactic phrase which is located in SpecDP. Therefore, (i) does not provide evidence that phrases can occur in syntactic words. Furthermore, Lieber (1992) argues that in Tagalog, phrases provide the input to verbal derivations with the prefix mag-: (ii)

a. b.

sa ayos in order sa dula in a play

magsaayos ‘put in order’ magsadula ‘put into a dramatic form’

(Lieber 1992: 17)

Finally, Lieber (1992) discusses examples from Tamil where the derivational affix -tal is used to derive nouns from VPs. Lieber notes that -tal can only attach to transitive verbs if their


Consequently, deriving words by attaching an affix to a phrase is not possible in German. The (b)-examples in (11)–(14) are excluded, because the prepositional phrases force the verb to project a VP, which is not a legitimate input for the affixes -ung or -bar. In contrast, derivational affixes do attach to particle verbs, as shown in the (a)-examples: (11) a.

ausstellen – part-put ‘exhibit’ b. aus dem Haus stellen out.of the house put ‘take out of the house’

Ausstellung ‘exhibition’ –

*aus dem Haus-Stellung

(12) a.

umleiten – Umleitung part-lead ‘re-routing’ ‘re-route’ b. um die Stadt herum leiten – *Um-die-Stadt-herum-Leitung around the city H-around guide ‘guide around the city’

(13) a.

einführen part-lead ‘introduce’ b. in den Wald führen into the forest lead ‘lead into the forest’

–

–

Einführung ‘introduction’ *In-den-Wald-Führung

internal subcategorized argument is present, (iii-a); furthermore, derived -tal-nominals may even include VP-adverbs, (iii-b): (iii)

a. b.

nilatt-ai uqu land-acc plow kazinamaaka uqai hard work

nilatt-ai uqu-tal ‘plowing the land’ kazinamaaka uqai-ttal ‘working hard’ (Lieber 1992: 18)

(ii) and (iii) may indeed show that in some languages, affixes may be morphologically subcategorized for phrases. However, this still does not provide sufficient evidence for a similar assumption about derivational morphemes in German. Furthermore, it is not completely clear whether the derived verbs and nouns in (ii) and (iii) really are V0- and N0heads, or whether these affixes only select PP- and VP-complements and project in a regular fashion, while their phonological affixhood is satisfied at phonological structure.


(14) a.

durchführen – part-lead ‘carry out’ b. durch den Wald führen – through the forest lead ‘lead through the forest’

durchführbar ‘practicable’ *durch-den-Wald-führbar

If VPs could really productively undergo word formation, we would expect nouns and adjectives to be derived from all kinds of verb+complement constructions. However, none of the VPs in the (b)-examples can function as the input for further word formation processes. Again, the ungrammaticality of these examples is the result of the morphological subcategorization frame of affixes like -ung and -bar. In contrast, the particle verbs in the (a)-examples can combine with derivational affixes. This suggests that they meet the contextual requirement of the derivational affixes. However, this in turn supports the conclusion that the particle verbs in (11a)–(14a) are V0s. 6.1.3 The particle verb as a V0 and word formation by Move Although affixes do not adjoin to phrases, it is possible that the interface level of syntactic structure includes a full phrasal projection of the affix’s host. This is what we find if we have word formation by Move. The potential host projects a phrase, but undergoes head movement and adjoins to the affix. (15) illustrates this option with Y0 being a derivational affix that requires a verbal host: YP

(15)

VP XP

Y0 V0

V0

Y0

word; domain where morphological subcategorization frame of Y 0 is fulfilled

X0

Due to the possibility of verb movement, the affix Y0 forms a word with its verbal host, which at the same time projects a full phrase. We may now wonder whether (15) is a possible structure for words derived from particle verbs. Notice that in (15), the head of the verb’s XP-complement and the (copy of the moved) verb are structurally adjacent. Therefore, one may suggest that XP = PrtP, and that (15) represents the way derivational affixes


combine with particle verbs. For example, we could assume that the derivational affix -ung in e.g. (13a) selects a VP where the verb and the particle are structurally adjacent. Morphological conditions then force V0 to move and to adjoin to the affix, as illustrated in (15). However, I will show that (15) cannot be the structural representation of nominals derived from particle verbs. Recall that the difference between process and result nominals is represented structurally through the difference between word formation by Move and word formation by Merge. If word formation by Move takes place, the verb projects a VP and is now structurally adjacent to (the complex) N0. This situation is responsible for the process nominal-interpretation (see (46) in Chapter 1, Section 1.3.3). Now, if word formation with particle verbs always proceeded as in (15), it would be predicted that nouns derived from particle verbs were always process nominals. However, (16) and (17) illustrate that particle verbs occur both inside result nominals (word formation via Merge) and inside process nominals (word formation via Move): (16) a.

Die Anweisung wurde ihm zugeschickt the part-order-ing was him sent ‘The instruction was sent to him’ b. Die Ausarbeitung lag auf dem Schreibtisch the part-work-ing lay on the desk ‘The elaboration lay on the desk’ c. Die mathematische Ableitung stand an der Tafel the mathematical part-direct-ing stood at the blackboard ‘The mathematical deduction was on the blackboard’ d. Die Auszahlung bestand nur aus 100-Dollar Scheinen the part-pay-ing consisted only out 100-$ bills ‘The disbursement consisted only of 100-dollar bills’

(17) a.

Die Auszählung der Stimmen dauerte vier Tage the part-count-ing the votes-gen lasted four days ‘The counting of the votes lasted four days’ b. Die Auswertung der Ergebnisse wird bald abgeschlossen sein the part-evaluate-ing the results-gen will soon finished be ‘The evaluation of the results will be finished soon’ c. Die Beisetzung des Toten wird am Sonntag stattfinden the part-sit-ing the dead-gen will at.the Sunday take.place ‘The burial of the dead will take place on Sunday’


d. Die Auszahlung der Löhne wird pünktlich beginnen the part-pay-ing the wages-gen will punctually start ‘The disbursement of the wages will start punctually’

The process nominals in (17) are derived by Move, but we still need a structure without a VP in order to represent the result nominals in (16). If particle verbs in words were always VPs, the existence of these nominals would remain unaccounted for, since Borer’s (1993) analysis requires that the verb in a result nominal is a V0 that does not project. The existence of both process and result nominals derived from particle verbs shows that at least in some cases of word formation, the particle verb must be a V0 which does not project a phrase. But can we maintain something like (15) in order to account for process nominals? The following discussion will show that we cannot. According to Borer (1993), the VP-complement of the derivational affix -tion in English appears to the right of N0. Notice that in German, nominals also take their complements to the right: (18) a.

der Bruder von Maria the brother of Mary ‘Mary’s brother’ b. die Schwester des Lehrers the sister the teacher-gen ‘the teacher’s sister’

Therefore, it seems plausible that the projection of V is also located to the right of N0 if N0 is a derivational affix (see also Rivet (1999) for a similar assumption about -er-nominals derived via Move). This means that a process nominal like Ausgrabung, ‘excavation’, would have a structure like (19): (19)

*N′ N0 V0 grab-

VP N0 -ung

PrtP

Prt 0 aus

V0

Particle verbs and word formation

In (19), N0 combines with VP; verb movement creates the right context for the nominal affix, and at the same time, the particle and the (copy of the moved) verb are structurally adjacent. However, as a result of N0 selecting its VP-complement to the right, the particle is now predicted to follow the nominal rather than preceding it. If the particle verb is represented as a VP, only the verb moves and combines with the affix. Due to the order N-complement, this would yield the wrong order *V-ung-Prt at phonological structure. Consequently, we cannot represent particle verbs inside derived nominals as VPs. This means, however, that a particle verb inside a derived nominal must be a complex V0. As for process nominals, I assume that the particle verb is a V0 which projects a VP; the complex verb incorporates into the nominal head corresponding to the affix -ung. Since the (copy of the) particle verb is structurally adjacent to N0, the nominal receives an eventive interpretation. This situation is illustrated in (20a). In (20b), we have the structure of result nominals derived from particle verbs. The particle verb is a V0 that does not project; the nominalizing suffix is merged with this V0, and the nominal receives a result interpretation: NP

(20) a.

N0 V0 Prt 0

VP

N0 V0

Prt 0

NP

b.

N0 N0

V0 Prt 0

V0

V0 V0

251


Let me turn to adjective formation with particle verbs. Consider the adjectival passive constructions in (21): (21) a.

Der Schrank ist schlecht aufgestellt the closet is badly part-put ‘The closet is badly assembled’ b. Das Geschirr ist schlampig abgewaschen the dishes are sloppily part-washed ‘The dishes are sloppily washed’

The adjectival participles in (21) are compatible with the VP-adverbs schlecht, ‘badly’, and schlampig, ‘sloppily’. Kratzer (1994) and, following Kratzer, Rapp (1997), argue that the adjectival participles in (21) are derived by combining the verbal participle with an adjectival zero-morpheme (semantically, this morpheme is an operator that maps the eventive reading of the verbal participle into its stative reading). The verb projects a full VP (which licenses the occurrence of adverbs) and incorporates into the phonologically empty A0-head. The adjectival participles in (21) are hence an instance of adjective formation via Move; the derivation is comparable to the syntactic derivation of derived verbs like widen discussed in Chapter 1, Section 1.3.3. Again, we have to wonder whether a structural representation for (21) is possible in which the particle projects a full phrase. Kratzer (1994) argues that the zero-adjectival head selects its VP-complement to the left. Hence, (22) might be a possible structure for the derivation of the participles in (21): AP

(22)

A0

VP PrtP

V0

V0 gestellt-

A0 Ø

Prt 0 auf

In (22), A0 combines with a VP; this projection licenses the occurrence of VP-adverbs. Verb movement combines the participle with the zero-affix, and at the same time, the particle and the (copy of the moved) verb are structurally adjacent. Phonologically, the particle precedes the verb; therefore, the data in


(21) do not tell us whether the whole particle verb in adjectival passives is a V0, or whether (22) is a possible structure. However, only the former assumption can explain the following data: (23) a.

Der Schrank ist unaufgestellt the closet is un-part-put ‘The closet is unassembled’ b. Das Geschirr ist unabgewaschen the dishes are un-part-washed ‘The dishes are not washed’

In (23), the negative adjectival prefix un- attaches to the adjectival participle. If the participle was derived as in (22), we would have to assume that un- selects the AP-projection in (22), and that the participle moves and adjoins to un-. However, (23) shows that un- attaches to the particle, not to the base verb. If the adjectival participle aufgestellt was derived as in (22), i.e. by moving only the verbal part and by stranding the particle inside the VP, we could not derive a structure where un- precedes the particle verb. Rather, we would derive the incorrect order *auf-ungestellt. In contrast, if it is assumed that the particle verb is a V0 from which the adjectival participle is derived, the whole particle verb would be the input to un-prefixation, and the right order would follow. Furthermore, (24) shows that VP-adverbs are not licensed if un- is prefixed to the adjectival participle: (24) a. *Der Schrank ist schlecht unaufgestellt the closet is badly un-part-built b. *Das Geschirr ist schlampig unabgewaschen the dishes are sloppily un-part-washed

The impossibility of having VP-adverbs shows that there is no VP-projection in (24). This in turn suggests that the adjectives in (23) are derived by Merge. In fact, Kratzer (1994) argues that the prefix un- can never combine with phrasal categories. This means that the particle verbs in (23) are complex V0s that attach directly to the adjectival zero-morpheme. The resulting complex A0s can then be prefixed with un-:


AP

(25)

A0 A0 un-

A0 V0

Prt 0 ab

V0 gewischt

A0 Ø

The particle verbs inside the un-prefixed adjectival participles in (23) are V0s; their structural representation in (25) corresponds to the structure of result nominals in (20b). With respect to the data in (21), where the presence of an adverb indicates that V0 projects a VP and that the adjectival participles are derived by Move, I will assume that two options exist. The particle verb may be a complex V0 that projects a VP and moves as a whole; this option corresponds to the derivation of process nominals in (20a): AP

(26)

A0

VP V0 Prt 0

V0 V0

Prt 0 ab

V0 gewischt

A0 Ø

Alternatively, the adjectives in (21) may have the structure in (22), where the particle and the verb are structurally adjacent. In this latter case, only the base verb moves and combines with the adjectival zero-morpheme; since A0 takes its VP-complement to the left, the particle still precedes the participle at phonological structure. (I will come back to adjectival passives in Section 6.3.) To summarize, the discussion has shown that with respect to certain instances of word formation, the conclusion that particle verbs are V0s cannot be avoided. The properties of words derived from particle verbs suggest that whenever particle verbs in German undergo ung-nominalization or occur in derived adjectives prefixed with un-, they must be represented as V0s.


6.2 Structural adjacency and morphological structures The idea that particle verbs are V0s when they appear in morphological contexts contradicts the basic claim made in this study; viz. that the particle and the verb are structurally adjacent. Clearly, we do not want to abandon this claim; the discussion in Chapter 2 has provided a good amount of empirical evidence in favor of the assumption that the particle is syntactically represented as a phrasal complement of the verb. However, as a result of the discussion in the previous section, we also have to acknowledge that certain structural environments (i.e. those required by the morphological subcategorization frame of affixes) force the particle verb to be represented as a morphological object. In Section 6.2.1, I account for this situation through a principle of reanalysis, and in Section 6.2.2, I compare this principle to similar proposals that can be found in the literature. 6.2.1 Reanalysis I argued that particle verbs, despite their syntactic representation as phrasal constructions, remain essentially elements that consist of two heads. What I propose now is that this situation is responsible for the possibility of reanalyzing the structural relationship between the particle and verb. Instead of having to represent particles as being structurally adjacent to the verb, I assume that they may exceptionally be realized as morphologically adjacent to their base verbs. What is the status of reanalysis in the framework of lexical licensing that I have adopted in this study? Recall that I argued in Chapter 4 that it is a lexical requirement for particles that they must be structurally adjacent to the verb. If we want to say that this structural relation may be reanalyzed as morphological adjacency, we are in fact saying that the LSS of a particle can be unified with a structure generated by syntax even if in this structure, the particle and the verb are morphologically adjacent. I capture this situation through the Principle of Reanalysis in (27): (27) Principle of Reanalysis Given two terminal nodes X, Y, and a lexical entry L that requires X and Y to be structurally adjacent.6 Then the lexical entries of X and Y can be unified with a syntactic structure in which X and Y are part of the same word X0.

6.L might be the lexical entry of the particle or the verb, or the lexical entry that includes both the particle and the verb.


(27) allows for a reinterpretation of the local domain which is specified as part of the lexical entry of an element. It expresses the possibility that the lexical entry of a particle may be unified with a syntactic structure if the particle and the verb are morphologically adjacent. (27) hence postulates a systematic relationship between a representation in which a terminal node is the head of the complement of another terminal node and a representation in which both terminal nodes are part of the same word. Recall that this systematic relationship has already been the topic of Chapter 5, where I showed that certain phenomena are licensed in local domains, with “locality domain” defined either by structural adjacency or by the word: (28) Definition of locality domain D Two heads X and Y are within the same locality domain if a. X and Y are morphologically adjacent b. X and Y are structurally adjacent

For example, I showed that the occurrence of otherwise non-existing verbs may be triggered by the presence of a morphologically adjacent prefix or a structurally adjacent particle. Another example was provided by the lexical licensing conditions of the semantics of a prepositional element. I argued in Chapter 5, Section 5.3.2, that the LCS of a topological preposition in German can be associated with a P0-head if this head is either structurally, or morphologically, adjacent to a functional head F0Prep. I therefore argued that the lexical licensing of these prepositions does not specifically require a structurally adjacent functional head, but rather, more generally, a functional head within a particular locality domain. This raises the question of whether it is also possible to generalize the contextual condition that holds for particles. This would mean that we would have to relax the structural condition on lexical licensing of the particle by stating it in terms of the locality domain (defined as in (28)). One could assume that the lexical semantics of a particle can be associated with a P0-head whenever this head is either structurally adjacent or morphologically adjacent to V0. The former case covers the data discussed in Chapter 2; the latter captures the behavior of particle verbs with respect to word formation. Notice that the principle in (27), as it stands, does nothing else than that — if we can freely reanalyze structural adjacency of a particle and a verb as morphological adjacency, then we may also say that the particle and verb must be represented within the same locality domain. In Chapter 7, I will discuss the possibility that (27) in fact applies in an


unconditioned way in languages like English and Norwegian. In German, however, morphological adjacency is not just an alternative structural option. Reanalysis is not possible in all syntactic contexts. For example, it is not possible to generate a particle verb as a V0 in a finite clause, and then move it as a complex to Comp0. Instead, the option expressed through (27) is restricted to cases where the particle verb undergoes an operation of derivational morphology. It seems that in German, the option that the particle and the verb can be combined inside a word is allowed if (and only if !) the particle verb is itself part of a derived word. I capture this observation in (29): (29) Amendment for (27) Condition: X0 must be adjoined to W0, W0 a lexical head.

(29) guarantees that a reanalyzed particle verb may only occur inside a larger morphological structure. One could say that the morphological combination of the particle and the verb derives a complex element which is itself a bound morpheme; it must combine with a derivational affix. Without the derivational affix, representing the particle verb as a V0 yields ungrammaticality, because the complex V0 created by syntax cannot be unified with the lexical entries of the particle and the verb.7 This means that the basic local relation between a particle and a verb in German is structural adjacency. However, since the domain of morphology is the complex X0-element, morphological operations can only apply to particle verbs if they are represented as V0s. This conflict can be solved through reanalysis, but the reanalysis-option does not exist when the particle verb is not part of a derived nominal or adjective. Interestingly, similar conclusions about the status of particle verbs have been reached independently in the literature. For example, Groos (1989) also assumes that the particle verb is a V0 when it is part of a word, whereas all other occurrences of particle verbs are syntactic (the details of her analysis are different from mine, but this is of no interest for the present point). She claims that there is a “general principle that forces the syntactic realization of [particle] verbs whenever possible” (1989: 56). Instead of postulating a “general principle”, I have suggested that structural adjacency of the particle and verb is lexically required. Therefore, this syntactic context must be created “whenever

7.This raises the question of why only derivational, and not inflectional, morphology triggers reanalysis. The lexical properties of a verb are influenced more strongly by derivational than by inflectional operations. I suspect that therefore, grammar prefers the input to this derivational operation to be a complete morphological object.


possible”. However, due to the specific requirements of morphology, it is impossible to create this syntactic configuration when the particle verb appears inside a word. A morphological context requires the particle verb to be a morphological object. The analysis of particle verbs proposed by Ackerman & Webelhuth (1998) also comes close to the proposal made in this section. They account for the properties of particle verbs through lexical entries that explicitly allow the particle and the verb alternately to be generated either as a word or as a phrasal construction: “The lexical entry of this kind of predicate specifies that it can be realized either as a morphological particle-verb compound … or as a particle and a verb which are projected into phrase structure as two independent syntactic atoms” (Ackerman & Webelhuth 1998: 340). The respective choice between these two options depends on the syntactic and morphological context in which the particle verb is realized. For example, Ackerman and Webelhuth assume that in morphological derivations, the particle verb is a compound, whereas in main clauses with V2, the particle verb is a “discontinuous predicate”. Although the framework in which they formulate their proposal, and many of their basic assumptions, are different from the theory I have presented here, the similarity between their solution and the one proposed in this section is striking. In sum, the fact that with respect to the word formation properties of particle verbs, many researchers have arrived at the same conclusion, albeit on the basis of different theoretical frameworks, provides evidence that the analysis presented here is on the right track. The word formation properties of particle verbs do not provide evidence that particle verbs are always generated as words — they only show that particle verbs, despite their basic syntactic character, can adopt the structure of words under certain conditions. 6.2.2 Morphological Merger and Parallel Morphology The principle in (27) accounts for the facts observed in Section 6.1; it is motivated on the basis of the claim that local relations between terminal nodes can be expressed by both morphological structures and structures that are larger than words. In this section, I want to show that similar ideas are also found in other theories on the syntax-morphology interface. Marantz (1984, 1988, 1989) formulates the idea that certain structural relations between elements can be expressed by combining the two elements inside a word through the principle of Morphological Merger:


(30) Morphological Merger (Marantz 1988: 261) (…) a relation between X and Y may be replaced by (expressed by) the affixation of the lexical head of X to the lexical head of Y.

Marantz (1984, 1988, 1989) proposes (30) in order to deal with examples where two heads show ambiguous behavior with respect to whether or not they are part of the same word. (30) states that whenever two terminal nodes stand in a particular (local) relation, this relation can be expressed by affixation. In Marantz’s earlier work, Merger takes place at a level between PF and syntactic structure which is syntactic in nature, but does not obey the rules of X-bartheory (Phonological Structure in Marantz (1988, 1989), Morphological Structure in Halle & Marantz (1993)). In terms of the theory presented in Chapter 1, and in accordance with the assumptions articulated explicitly in Marantz (in prep.), I assume that Morphological Merger is operative in “Phonology”, i.e. at the interface between syntactic structure and phonological structure. The application of Morphological Merger is not entirely unrestricted; not any relation between two elements X and Y can be expressed by affixation. Morphological Merger can only operate on heads within a particular locality domain which is defined syntactically. Importantly, Halle & Marantz (1993) define this locality domain as structural adjacency. This means that two heads which are syntactically represented as structurally adjacent can be reanalyzed as being part of the same word at phonological structure. One example is provided by auxiliary clitics in Papago. According to Marantz (1989), the auxiliary clitic ‘o is a suffix and needs a host to its left. Since ‘o occupies the Comp-position, the phonological condition imposed by the clitic is met whenever a major constituent is moved to SpecCP; ‘o then cliticizes to the right of this constituent and hence ends up in second position. However, the affixal requirement of ‘o is not met in a structure like (31), where no topicalization occurs. Therefore, without a full phrase in SpecCP, the auxiliary clitic must appear linearly after the first element of the syntactic constituent that follows Comp0 (Marantz 1989: 100): (31) a.

[C¢ ’o [pi iam-hu cikpan g Huan]] aux [neg there work the John b. *’o pi iam-hu cikpan g Huan aux neg there work the John c. pi ’o iam-hu cikpan g Huan neg aux there work the John ‘John is not working here’


Marantz (1988, 1989) suggests that the phonological string that yields the order in (31c) is derived via the application of Morphological Merger. At phonological structure, the linear order Aux–Neg, which is determined by syntactic structure alone, is reversed in order to fulfill the affixal conditions of the clitic. An auxiliary in Comp0 and negation in Papago are structurally adjacent. This relation is reanalyzed at phonological structure, where the two elements are part of the same word.8 According to Marantz (1988: 107), the principle of Morphological Merger “describes the only way in which a syntactically independent morpheme (or X0) may cease to project its syntactic relations and become part of another word morphologically”. This is exactly what my proposal says about the word formation properties of particle verbs. The particle normally is syntactically independent and projects a phrase, but it may exceptionally become part of a complex V0-element by virtue of (27). However, although both Morphological Merger and the Principle of Reanalysis that I proposed in (27) can be taken as representations of the possibility of expressing certain local structural relations between heads through word formation, they cannot be taken to be the same principle. Morphological Merger differs from reanalysis in that the former is an interface condition, whereas (27) operates on unification. In the example in (31), the syntactic relation between Neg0 and Comp0 remains unchanged; both heads are structurally adjacent. The phonological requirement of the clitic, however, overwrites the linear order determined by the syntactic structure and forces the application of Morphological Merger at phonological structure. Only in Phonology are the Vocabulary entries represented inside one phonological word, with the Auxclitic attached to the right of the negation. In contrast, what I suggest in (27) is that the syntactic relation between two heads itself may be reanalyzed. Instead of representing the particle and the verb as structurally adjacent, as required by the lexical entry of the particle, the syntax generates them as part of the same morphological object. Whereas (31) exhibits an application of Morphological Merger at phonological structure, which forms a phonological word out of two structurally adjacent heads, the word formation properties of particle verbs are an example of the application of (27), such that structural adjacency is replaced by the formation of a syntactic word. As a result of reanalysis, grammar licenses

8.Similar cases are discussed in Halpern (1995). Morphological Merger corresponds essentially to Halpern’s rule of Prosodic Inversion, which determines the reordering of terminal elements in the mapping from syntactic structure to phonological structure.


two syntactic representations of particle verbs, one which requires the particle to project a phrase, and one which represents the particle verb as a complex V0. This situation is reminiscent of the cases discussed by Borer (1988, 1991, 1993) in her framework of Parallel Morphology (see Chapter 1, Section 1.3.3). Recall that the core idea behind Parallel Morphology is the insight that words can be derived by Move and by Merge. According to the former option, a terminal node is the head of the phrasal complement of a second node and incorporates. According to the latter option, both nodes are directly combined and form a complex X0. One might therefore suggest that particle verbs can be analyzed along the lines of Parallel Morphology. According to the first option, the particle projects a phrase. The second option, word formation via Merge, covers the fact that particle verbs in words are V0s. Of course, the first option is still incomplete. If a Parallel Morphologyanalysis is adopted, one also has to assume that the head of the particle phrase incorporates into the verb. However, I showed in Chapter 2 that the particle does not incorporate into the verb overtly. Rather, the properties of the verbparticle construction all follow from the fact that at the interface level of syntactic structure that is interpreted by phonology, the particle is represented inside the particle phrase. Therefore, in order to make particle verbs compatible with a Parallel Morphology-approach, I have suggested in Zeller (1997a, b) that the particle incorporates into the verb covertly and forms a complex V0 with the base verb at LF. The resulting complex V0 is associated with the particle verb’s semantics, whereas the phonological representation of the particle is associated with the base position of the particle chain inside the particle phrase.9 This means that in sentences where the particle and the verb are visibly separated (as in V2, for example), it only looks as if the particle had stayed inside the particle phrase. In fact, the particle does move, but this movement is invisible, because the particle’s Vocabulary item is associated with the foot position of its chain. According to this Parallel Morphology-approach, if the particle is the head of an independent phrase in overt syntax, word formation takes place at LF. In

9.The fact that the LPS of the particle cannot be associated with the head of the particle chain can be derived from a proposal made in Roberts (1997). Roberts suggests the general rule (i): (i)

*[X0 W1 W2], where Wn are morphological words.

(Roberts 1997: 426)

Since particles are phonologically free morphemes, they count as morphological words and cannot be spelled out together with the verb as part of the same X0.


morphological contexts, however, the particle does not project a phrase, and word formation takes place via Merge. The general idea behind this proposal is similar to the proposal that I have made in this section. Both analyses account for the observation that particle verbs are syntactic constructions in “phrasal” contexts, but are morphological elements in morphological contexts. Furthermore, notice that the locality domain that I take to be characteristic of the verb-particle construction (i.e. the domain defined by structural adjacency) is exactly the domain which usually licenses head movement. Therefore, the assumption that word formation may proceed via abstract incorporation can be regarded as a metaphorical means to express the claim that particle verbs consist of two heads in this particular locality domain. However, a problem arises with respect to a particular prediction that is made by the theory of Parallel Morphology. It predicts that both means of word formation, i.e. Merge and Move, are equally available. Parallel Morphology explains why the particle verb can be represented both as a syntactic and a morphological construction, but it does not give an explanation for why the particle verb cannot be represented as a morphological object in all contexts. For example, why does V2 not permit word formation of the particle verb via Merge? The Parallel Morphology-account does not explain why the Mergealternative is excluded in non-morphological contexts. This is where the reanalysis-approach is superior. According to my analysis, the syntactic representation of the particle verb that explains separability in V2 is more basic than the morphological alternative. Particle verbs in German are fundamentally non-morphological constructions; the particle is the head of the verb’s phrasal complement. Their occurrence as complex V0s is a marked exception — “marked” in the sense that this option is not available in all contexts, but only if external requirements (= affixation of a derivational morpheme) fulfill the condition stated in (29) above. The present account meets the spirit, but not the letter, of Parallel Morphology. It shares with the Parallel Morphology approach the insight that particle verbs can be represented both as phrasal constructions and as V0s, but it differs from this approach in that it does not claim that both options are available in all contexts. Instead, it is assumed that the possibility of reanalyzing structural adjacency as morphological adjacency is restricted to structural configurations in which the particle verb is attached to a lexical head. (For more discussion of a Parallel Morphology- and a reanalysis-approach to particle verbs, see Zeller forthcoming b.)


6.3 Word formation with resultatives and h-verbs The crucial difference between particle verbs and other verb-complement constructions is that the particle only projects a phrase because its lexical representation requires it to be structurally adjacent to the verb. Essentially, the particle verb remains a complex element consisting of two terminal elements. Therefore, (27) can be applied to particle verbs, but not to “genuine” verbcomplement constructions that do not consist of only two heads. This proposal raises the question of whether reanalysis in word formation is also licensed with constructions that are not particle verbs, but that also consist of a verb and a phrasal complement which is only represented through a single word. We know of two such constructions, i.e. adjectival resultatives and verbs that combine with intransitive h-postpositions (h-verbs). In this section, I will consider each in turn. Let me start with resultatives. The reanalysis proposal outlined in the previous section predicts that word formation with adjectival resultatives is impossible, because the adjective and the verb are not structurally adjacent. As I argued in Chapter 3, the adjectival predicate in a resultative construction is dominated by a functional (agreement-) layer. The resultative adjective is forced to adjoin to the head of its extended projection. This means that what is structurally adjacent to the verb is not the bare adjective, but the complex Agr-head: VP

(32)

AgrP

Agr 0 = Y in (27)

AP 0

A

V 0 = X in (27)

0

A

Agr 0

If we apply (27) to (32), X is V0, but the head that corresponds to Y in (27) is not the adjective A0, but the complex Agr-head. Therefore, reanalysis of A0 and V0 is impossible. Reanalysis of V0 and the functional Agr0-head is impossible, too, because the latter head does not correspond to any lexical entry that can be unified with syntactic structure. This, however, is required by (27). Therefore, resultatives are not expected to appear in words.


The following data confirm this expectation. As has been pointed out on numerous occasions in the literature (cf. Stiebels & Wunderlich (1994), Kratzer (1994), Zeller (1997b) for German; Neeleman & Weerman (1993) and Neeleman (1994) for Dutch), the possibility of deriving words from resultative constructions is much more restricted than word formation with particle verbs. The resultatives in the (b)-examples in (33)–(35) do not allow word formation, while the particle verbs in the (a)-examples do: (33) a.

das Haus anstreichen the house part-paint ‘paint the house’ Anstreicher part-painter ‘painter’ b. das Haus rot streichen the house red paint ‘paint the house red’ ? * Rotstreicher redpainter

(34) a.

die Stelle ausschreiben the job part-write ‘advertise the job’ Ausschreibung part-write-ing ‘advertisement’ b. den Brief fertig schreiben the letter ready write ‘finish writing the letter’ *Fertigschreibung ready-write-ing

(35) a.

(verb-particle construction)

(derived nominal)

(verb-resultative construction)

(derived nominal) (verb-particle construction)

(derived nominal)

(verb-resultative construction)

(derived nominal)

die Schuhe abwaschen (verb-particle construction) the shoes part-wash ‘wash the shoes’ die Schuhe sind abwaschbar (derived adjective) the shoes are part-wash-able ‘the shoes can be washed’


b. die Schuhe sauber waschen (verb-resultative construction) the shoes clean wash ‘clean the shoes’ *die Schuhe sind sauber waschbar the shoes are clean wash-able (derived adjective)

Notice that the following example does not contradict this result: (36) Die Schuhe sind saubergewaschen the shoes are clean washed ‘The shoes are cleaned’

As (36) shows, adjectival participle formation with resultatives is possible. However, recall from Section 6.1.3 that particle verbs allow for two options to derive adjectival participles. As I showed above, one option is to derive the adjective from a full VP (which includes the verb and its phrasal complement) via incorporation of the base verb into the adjectival zero-head (cf. structure (22) in Section 6.1.3). We can now account for the grammaticality of (36) by assuming that this is how adjectival participles are derived from resultatives. According to this assumption, the VP in (32) is the complement of the adjectival zero-head. The verbal participle (gewaschen in (36)) incorporates into A0, and the resultative predicate is stranded. In that case, the adjective is still adjacent to the participle. The second option to derive adjectival participles requires reanalysis of the particle verb (cf. (26) in Section 6.1.3); this option is therefore predicted to be excluded with resultatives. One possibility to test this prediction is through un-prefixation. It was argued above that un-prefixed adjectival participles can only be derived from complex verbs if these verbs are V0s. We therefore expect that adjectival participles derived from resultatives cannot be prefixed with un-. (37b) shows that this expectation is borne out: (37) a.

Die Schuhe sind unabgewaschen the shoes are un-part-washed ‘The shoes are not washed’ b. *Die Schuhe sind unsaubergewaschen the shoes are un-clean-washed

The particle verb in (37a) must be a V0, because the participle is prefixed with un-. In contrast, un-prefixation is impossible if the participle is derived from a resultative, which follows from my claim that a resultative predicate and V0 cannot be reanalyzed.


However, some resultative constructions exceptionally seem to be licensed in derived words (cf. Lüdeling (1998a): (38) a.

die Zeit totschlagen the time dead-beat ‘kill time’ b. Es gibt Zeiten, die sind untotschlagbar ‘There are times that are un-death-beatable’ (from the German translation of Martin Amis’ novel “Money”)

(39) a.

die Wäsche weich spülen the laundry soft wash ‘wash the laundry with the result of making it soft’ b. Weichspüler, ‘fabric softener’

If the adjective untotschlagbar in (38b) is really acceptable, and not just a free literal word creation, it is only possible in the context of the resultative construction totschlagen in the phrasal idiom die Zeit totschlagen, ‘kill time’. This phrasal idiom is a lexical entry that associates special meaning with a structure like (32). This means that in (38), structural adjacency of the Agr0- and the verbal node are in fact required by a lexical item. Consequently, the resultative in (38a) may be subject to the reanalysis principle in (27). The same reasoning explains the possibility of a nominal like Weichspüler. The resultative weich spülen is a fixed expression (its use is restricted to the washing of laundry). Therefore, it may allow for the possibility to be reanalyzed as a complex V0 that can undergo further processes of word formation.10 The data in (38) and (39) can be taken as illustrations of an observation made by Lüdeling (1998a). If a construction is lexicalized, it becomes more likely that it may (exceptionally) provide the input to word formation.11 Let me now turn to h-postpositions. The structure in (40) shows the relation between an h-postposition and a verb: 10.If V0 and Agr0 are reanalyzed, the resulting V0 is a lexical element that includes functional material. This might further restrict the possibility to reanalyze resultative constructions as V0s. 11.This observation seems to suggest that there is a correspondence between morphology and lexical listedness; an idea that I explicitly rejected in Chapter 1 as being part of the lexicalist view of grammar. However, the observation mentioned in the text does not imply that special meanings can only be associated with structures of the size of words. It only implies that certain phrasal constructions can exceptionally occur in words only if they are lexicalized.


V′

(40)

V0

FPPrep PP

F 0Prep

I argued that h-postpositions are (semi-lexical) realizations of the head F0Prep. This means that, in contrast to (intransitive or transitive) prepositions, h-postpositions share with particles the property that they are always structurally adjacent to the verb. Furthermore, I showed in Chapter 5, Section 5.3, that many h-postpositions, despite their transparent morphosyntactic structure, are lexically listed as complex F0Prep-elements with a special meaning. Furthermore, I showed that at least some of these h-postpositions are not licensed as complements of nouns, which suggests that their meaning is lexically licensed only if they are structurally adjacent to a verb. I concluded that these properties make some h-postpositions look very much like particles. Therefore, we expect that some h-verbs behave like particle verbs also with respect to word formation. In fact, the principle in (27) predicts that reanalysis is structurally possible at least with those h-postpositions that are lexically stored and that require a structurally adjacent verb. This prediction is borne out. The h-verbs in (41)–(44) participate in operations of derivational morphology: (41) a.

Peter fordert seinen Gegner heraus P. demands his opponent H-out ‘Peter challenges his opponent’ b. Herausforderung, ‘challenge’

(42) a.

Peter zieht einen Anwalt hinzu P. pulls a lawyer H-to ‘Peter consults with a lawyer’ b. Hinzuziehung, ‘consultation’

(43) a.

Peter fügt etwas Milch hinzu P. directs some milk H-to ‘Peter adds some milk’ b. Hinzufügung, ‘addition’

(44) a.

Peter nimmt das Futter heraus P. takes the lining H-out b. herausnehmbar, lit. out-takeable


The semantics of the h-postpositions in (41)–(44) must be assigned to the whole h-postposition and cannot be derived from the combination of a P-element and a functional head. Therefore, the whole complex head F0Prep is a lexical entry, and is required to be structurally adjacent to V0. It can hence be reanalyzed with the verb. Interestingly, if an h-postposition has maintained a spatial reading, which is derived compositionally on the basis of the underlying preposition, the combination of the h-postposition and the verb cannot be reanalyzed. The (a)examples in (45) and (46) show an h-verb, and the (b)-examples show that the respective h-postposition does not have to be structurally adjacent to a verb, but can also project a phrasal complement of N0. Consequently, as the (c)-examples show, reanalysis with these h-postpositions is impossible: (45) a.

(auf den Berg) hinaufsteigen (*aufsteigen) (on the mountain) H-up-climb (*part-climb ‘climb up the mountain’ b. der Weg hinauf the way H-up ‘the way up’ c. *Hinaufstieg d. Aufstieg, ‘ascent’

(46) a.

(aus dem Zimmer) hinausgehen (out the room) H-out-go ‘leave the room’ b. der Gang hinaus the walk H-out ‘the walk outside’ c. *Hinausgang d. Ausgang, ‘exit’

(*ausgehen) (*part-go

Interestingly, the (d)-examples show that there exist nominals that correspond to the h-verbs in (45a) and (46a). However, the h-postpositions are not licensed inside these nominals. Instead, the h-postpositions are replaced by simple prepositional elements. In fact, the element that has combined with the verb inside these derived nominals looks like a particle. I do not intend to explain the precise mechanism that apparently turns h-postpositions into particles in these examples, but notice that these data illustrate in a curious way that not all h-postpositions can be reanalyzed with the verb. The only problem with the derived words in the (b)-examples of (41)–(44)


is that the h-postposition was argued to be a functional element, and functional elements are usually not found in derived words. However, recall from Section 5.3 that the classification of h-postpositions as “genuine” functional heads has proved to be not quite correct. Since h-postpositions incorporate the representation of a lexical (prepositional) node, I classified them as semi-lexical elements. This explains why word formation with these elements is possible, although normally, functional heads inside derived words are ill-formed.12 I conclude that the reanalysis proposal also captures the observation that word formation with h-verbs is possible and occurs more frequently than word formation with productively derived phrasal constructions like resultatives. The definition of reanalysis refers to the lexical requirement that two heads X and Y must be structurally adjacent. Only then can their lexical entries exceptionally be unified with a syntactic structure in which X and Y are represented as part of the same word.

6.4 Conclusion In this chapter, I extended my claim that the strictly local relation between particles and verbs is responsible for their word-like properties. I argued that structural adjacency can be reanalyzed as morphological adjacency if this local configuration is independently required by morphological principles. I showed in Section 6.1 that particle verbs inside derived nouns and adjectives must be analyzed as morphological compounds, and I argued in Section 6.2 that this possibility is the result of a Principle of Reanalysis. Two terminal heads that are required to be structurally adjacent may exceptionally be represented inside a word. In Section 6.3, I discussed the word formation possibilities that exist with h-verbs and resultatives. This discussion provided further support for my analysis.

12.However, the functional properties of h-postpositions may have some effect on morphological operations. Word formation is not as productive as is the case with particle verbs. For example, I could not find any un-prefixed adjectives derived from h-verbs: (i)

*das Futter ist unherausnehmbar the lining is un-H-out-take-able ‘The lining cannot be taken out’

I suspect that the mixed properties of h-postpositions lead to the contrast between (41)–(44) and (i).


This chapter completes my analysis of the properties of particle verbs in German. In the final chapter of this book, I will address particle verbs in other Germanic languages in the light of the theory presented in this and the preceding chapters.

Chapter 7

Typological remarks and reanalysis

In this final chapter, I look at some consequences of my theory of particle verbs and local domains by discussing the behavior of particle verbs in Germanic languages other than German. The following remarks reveal some interesting cross-linguistic parallels, and they show how my analysis can account for some of the most interesting observations that derive from these parallels. I also show that analyzing the verb-particle construction in Dutch, English, and Mainland Scandinavian according to the results of the previous chapters raises a number of difficult questions that I will discuss and speculate on. In the previous chapters, I concentrated mainly on German, but I implicitly assumed that my analysis also accounts for the facts that are observed with respect to particle verbs in a language like Dutch. However, an interesting challenge to this assumption is provided by the properties of Dutch particle verbs in so-called Verb Raising constructions. In Section 7.1, I therefore discuss these properties in the light of the reanalysis principle proposed in Chapter 6. In Section 7.2, I turn to the so-called verb-particle alternation in English and in the Mainland Scandinavian languages. I suggest that reanalysis is also involved in this alternation. In contrast to what was observed in German, the discussion will show that reanalysis of particle verbs is also attested in contexts where the particle verb is not part of a derived word. The conclusion that I draw from this observation is that languages differ with respect to the degree to which they allow reanalysis.

7.1 On Verb Raising in Dutch In Dutch, certain verbs (like modals, auxiliaries, raising, perception and causative verbs) trigger raising of an embedded infinitive or participle into the matrix clause and adjunction to their right, an operation called Verb Raising (cf. Evers 1975; van Riemsdijk 1978; den Besten & Edmondson 1983):


(1) dat hij [een huis ti] wilde kopeni that he [a house wanted buy ‘that he wanted to buy a house’

(1) must have been derived via head movement of the embedded infinitive. As (2) shows, the modal willen does not permit extraposition of a full phrase (cf. Koster 1989): (2) *dat hij ti wilde [een huis kopen]i that he wanted [a house buy ‘that he wanted to buy a house’

Therefore, (1) cannot be derived by long scrambling of the embedded object, followed by extraposition of the emptied embedded clause (the so-called “Third construction”; cf. den Besten & Rutten 19891). Interestingly, if the embedded infinitive in a Verb Raising construction is a particle verb, two options exist. Compare the (a)- and (b)-examples of the following pairs: (3) a.

dat Jan zijn moeder op ti wil [bellen]i that J. his mother part wants [phone b. dat Jan zijn moeder ti wil [opbellen]i that J. his mother wants [part-phone ‘that Jan wants to call up his mother’ (Neeleman 1994: 24) omdat hij auto ti kan [rijden]i because he part(car) can [drive b. omdat hij ti kan [auto rijden]i because he can [part(car) drive ‘because he can drive a car’ (van Riemsdijk 1978: 102)

(4) a.

In (3), we have Verb Raising with the particle verb opbellen, ‘call up’; in (4), the (nominal) particle verb is autorijden, ‘drive a car’. In (3a) and (4a), the matrix

1.Den Besten and Rutten (1989) show that at least some of the cases where a matrix verb and an embedded verb undergo local string inversion are instances of the Third construction (see also Broekhuis et al. 1995). According to den Besten and Rutten, (i) is an example of extraposition of a full clause, with the direct object being scrambled into the matrix clause: (i)

dat Jan [Marie]j ti heeft geprobeerd [tj te kussen]i that J. [M. has tried to kiss ‘that Jan has tried to kiss Marie’ (den Besten & Rutten 1989: 45)

Typological remarks and reanalysis 273

verbs willen and kunnen have triggered movement of only the base verb, stranding the particle. This is what we expect from a phrasal analysis, according to which the particle is the head of a phrasal complement of the verb. However, in both (3b) and (4b), the whole particle verb has undergone Verb Raising. Since an extraposition analysis is not available, (3) and (4) strongly suggest that the particle verbs in these examples are represented as complex V0s that can undergo Verb Raising. Whereas the particle and the (copy of the) verb are structurally adjacent in the examples (3a) and (4a), which causes the particle to be stranded when the base verb undergoes Verb Raising, the particle and the verb are morphologically adjacent in the (b)-examples and therefore move together.2 Following the analysis proposed in Chapter 6, I assume that the possibility of moving the particle verb as a complex V0 is the result of the Principle of Reanalysis introduced in Section 6.2: (5) Principle of Reanalysis Given two terminal nodes X, Y, and a lexical entry L that requires X and Y to be structurally adjacent. Then the lexical entries of X and Y can be unified with a syntactic structure in which X and Y are part of the same word X0. (6) Amendment for (5) Condition: X0 must be adjoined to W0, W0 a lexical head.

(5) licenses unification of the lexical entries of the particle and the verb with (the terminal nodes of) the complex V0 that has undergone Verb Raising in (3b) and (4b). The amendment in (6), which I assume holds in both German and Dutch, formulates the condition for reanalysis. According to (6), the particle verb can only be represented as a V0 if this V0 is attached to a lexical head in syntactic structure. It is clear that the relevant lexical head in (3) and (4) that licenses reanalysis is the matrix verb. Since (6) is fulfilled, reanalysis can apply. This explanation depends on a particular view about the structure of bare infinitival complements. I will assume that the infinitives in constructions like (3) and (4) are bare VPs which are selected as non-functional complements of

2.Raising of the base verb and raising of the particle verb are equally acceptable in Dutch. However, according to le Roux (1988), separation of the particle verb in Verb Raising is a highly marked option in Afrikaans. Since this difference between two otherwise so closely related languages like Dutch and Afrikaans requires a more detailed comparison than I can offer in this chapter, I have to ignore her observation here. However, see the concluding chapter for some remarks.


matrix verbs that trigger Verb Raising (cf. e.g. Emonds 1985; Wurmbrand 2000a). Therefore, the structure of a sentence like (3b) looks like (7): V′

(7)

V0

VP

DP

zijn moeder Prt

V 0

V0 wil

0

V

0

V0 Prt 0 op

V0 bellen

Since the embedded verb lacks an extended projection, its internal argument cannot receive case in its base position. Therefore, it must move into a case position in the matrix clause (alternatively, it may be directly merged in the matrix clause as an argument of the complex verb derived by Verb Raising). The verb cluster wil opbellen in (7) consists of the particle verb attached to a lexical head V0; the matrix verb. Therefore, the derivation of (3b) and (4b) is possible for the same reason as word formation with particle verbs. The fact that particle verbs can undergo operations of derivational morphology, and the possibility of moving a particle verb as a whole in Verb Raising, are now explained by the same principle.3

3.The question of whether German has Verb Raising in infinitives is hard to settle, since the embedded verb never appears to the right of the matrix V0 if this verb is a modal, a causative, or a raising verb: (i)

dass ich meine Mutter anrufen will/*will anrufen that I my mother part-call want/want part-call ‘that I want to call my mother’

However, auxiliaries allow Verb Raising in German in certain constructions: (ii)

dass ich meine Mutter ti habe [anrufen wollen]i that I my mother have [part-call want ‘that I wanted to call my mother’

With respect to (i), one could assume that infinitives raise in German as well and attach to the left of the matrix verb. In that case, the analysis proposed for Dutch can be carried over to German. Alternatively, one could assume that Verb Raising in German takes place covertly; i.e. what is associated with the LPS of the infinitive in (i) is the copy of the moved verb. Another alternative would be to assume that infinitives in German clauses like (i) do not move at all. (Notice that this would make bare infinitives verbal particles, according to


It looks like an attractive possibility to assume that Verb Raising in Dutch is in fact some kind of verbal word formation; this assumption would treat modal verbs like willen and kunnen in Verb Raising constructions as derivational morphemes. This idea is in accordance with a proposal made in Bierwisch (1990), where it is suggested that modal verbs can function as pseudo-affixes that combine with infinitival forms of verbs, deriving single, but internally complex, verbs. However, a problem for this assumption is raised by the fact that the complex verb cluster does not obey the Right-hand Head rule. The matrix verb that triggers Verb Raising is the head of the verbal complex, but occurs on the left. Consequently, since the Right-hand Head Rule also holds for words derived by Move, the verbal complex does not qualify as a well-formed morphological object. It seems that not every instance of syntactic head movement into lexical positions can be analyzed as a genuine morphological operation. The alternative option illustrated by the (a)-examples in (3) and (4) is explained as follows. The particle projects a phrase, as required by its lexical representation, and only the base verb moves to combine with the matrix verb. The structure of (3a) is given in (8):

the definition of particles that I have given in Chapter 3.) Regarding (ii), it has to be noted that the verb wollen may not surface as a past participle in a verb cluster; it has to appear in its infinitival form (the Infinitivus Pro Participio, IPP). This so-called IPP-effect may be taken as evidence that (ii) involves Verb Raising (cf. den Besten & Edmondson 1983). However, this does not explain why in German, phrasal constituents can occur inside the verb cluster as well: (iii)

dass ich meine Mutter ti habe [zu Hause anrufen wollen]i that I my mother have [at home part-call want ‘that I wanted to call my mother at home’

Therefore, it is not clear that (ii) is indeed an instance of Verb Raising in German.


V′

(8)

V0

VP

DP

V0 wil

V′

zijn moeder PrtP

V 0i bellen

V 0i

Prt 0 op

Recall that I argued in Chapter 6, Section 6.1.3, that there are also two ways of deriving adjectival participles from particle verbs. According to the first option, the particle verb is a complex V0 and combines as such with the adjectival zeroaffix (see (26) in Chapter 6). According to the second option, the particle projects a phrase, and only the verb moves (see (22) in Chapter 6). These two alternatives correspond to the two possibilities illustrated in (7) and (8). Let me provide more evidence in favor of the claim that (3b) and (4b) are the result of reanalysis. First, notice that prepositions in Dutch can never occur inside the verb cluster: (9) *dat zij de jas ti de stoel heen hebben over gelegd that they the coat the chair H have over put ‘that they have put the coat over the chair’ (Koopman 1993: 33)

In (9), we have a circumpositional phrase over de stoel heen, ‘over the chair (over)’, with the preposition over having moved together with the participle. However, this kind of reanalysis is impossible. The preposition and the verb are not structurally adjacent, due to the intervention of a functional head F0Prep which is filled with the postposition heen in (9). Since structural adjacency of two terminal nodes is a necessary condition for reanalysis to be licensed, (9) is correctly ruled out by my analysis. Second, my proposal makes the correct predictions about intransitive prepositions like boven in (10): (10) Jan woont boven J. lives upstairs ‘Jan lives upstairs’


As I argued in Chapter 3, Section 3.1.4, intransitive prepositions are represented as regular prepositional phrases and project functional structure. The head F0Prep intervenes between the verb and P0 in (10). Therefore, intransitive prepositions, like their transitive counterparts, are never structurally adjacent to the verb. F0Prep hence prevents intransitive prepositions from occurring inside the verb cluster (cf. van Riemsdijk 1978; den Dikken 1995): (11) a.

dat Jan boven wil wonen that J. upstairs wants live b. *dat Jan wil boven wonen that J. wants upstairs live ‘that Jan wants to live upstairs’ (den Dikken 1995: 30)

The assumption that reanalysis in (11b) is blocked because the verb and the intransitive preposition are not structurally adjacent is further supported by the following data. Some prepositional elements can be used as particles and as intransitive prepositions. The following example (12) illustrates the difference between these two uses (cf. van Riemsdijk 1978: 55): (12) omdat hij [voor ti] schijnt te staani because he (it) [part/in.front seems to stand ‘because it (the team) seems to be leading’ (particle reading) ‘because it seems to stand in front’ (intransitive preposition-reading)

In (12), we have a Verb Raising-construction in which only the base verb staan has moved into the matrix clause. The combination of this verb and the prepositional element voor in (12) is semantically and structurally ambiguous. The stranded element voor may be located inside a bare PP-complement of the (copy of the moved) verb. In that case, voor is structurally adjacent to the verb and hence is a particle, and the special meaning of voorstaan can be derived from the combination of the lexical P0-element and the verb. Alternatively, voor in (12) may be an intransitive preposition inside a functional FPPrep. According to the theory that I outlined in Chapter 4, the prepositional meaning is now associated with voor in the syntactic context of this functional head F0Prep. This analysis makes the following prediction. If voor raises together with the verb as a result of reanalysis, we expect that the sentence is disambiguated. A Verb Raising-structure can only be unified with the lexical entry of the particle verb, since only as a particle is voor structurally adjacent to the verb staan. The prepositional meaning of voor is only licensed under structural adjacency with F0Prep, but this configuration does not permit reanalysis of P0 and V0.


This expectation is borne out. If the whole complex voorstaan appears inside the verb cluster, the preposition unambiguously receives the particleinterpretation (cf. van Riemsdijk 1978: 55): (13) omdat hij ti schijnt [voor te staan]i because he (it) seems [part to stand ‘because it (the team) seems to be leading’ (only particle reading)

According to my proposal, voor and the verb in (13) form a complex V0 which has undergone Verb Raising as a whole. This option is only available for two elements for which structural adjacency is a lexical requirement. Therefore, the possibility of having voor inside the verb cluster implies that it is a particle; the interpretation shows that this implication is realized in (13).4 Finally, the reanalysis proposal is confirmed by the behavior of Dutch postpositions, which contrast heavily with intransitive and transitive prepositions with respect to Verb Raising. For German, it was argued in Chapter 6, Section 6.3, that postpositions may be reanalyzed with the verb, since they are structurally adjacent to the verb. If the same option also exists for Dutch postpositions, we predict that postpositions in Dutch allow for reanalysis and may appear in the verb cluster. This prediction is also borne out. As observed by a number of authors (van

4.It must be noted that the example in (13) also raises a serious problem. Since the matrix verb in (13) is a raising verb, and raising verbs take te-infinitives in Dutch, what has undergone Verb Raising in (13) is the whole complex [particle-te-verb]. But according to the analysis of zu-infinitives in German that I proposed in Chapter 2, Section 2.2.1, this constituent is not a head. I have argued that the infinitival prefix is associated with the Infl0(Tns0)-position; the fact that te intervenes between the particle and the verb follows from the idea that the particle is the head of a phrase, whereas only the verb has moved to combine with the inflectional element. On the one hand, this seems to imply that the bracketed constituent in (13) cannot be a verb, but must be a full TP. On the other hand, however, this result would be incompatible with the observation that raising verbs, like modals, do not allow extraposition (cf. Hinterhölzl 1996). Furthermore, the fact that (13) disambiguates the sentence in (12) only follows from the reanalysis proposal, according to which the moved constituent in (13) is a V0. If (13) was derived by rightward movement of a TP, voor with the meaning of an intransitive preposition should also be possible inside the verb cluster. I cannot present a straightforward solution for this problem here, but I think that it might be found in a reinvestigation of the mechanisms that determine the inflectional properties of verbs and of other lexical categories. In Zeller (forthcoming b), I therefore develop a different view on inflection and its consequences for Verb Raising. For now, I have to note that the possibility of having particle verbs as te-infinitives inside the verb cluster is a non-trivial problem for my analysis.


Riemsdijk 1978; Groos 1989; den Dikken 1995; Koopman 1993, 1999b), postpositions in Dutch can undergo Verb Raising with the verb: (14) a.

omdat zij de boom in ti is [geklommen]i because she the tree in is [climbed b. omdat zij de boom ti is [in geklommen]i because she the tree is [in climbed ‘because she has climbed into the tree’ (Koopman 1993: 27) dat ik haar net nog de trap op ti zag [lopen]i that I her just now the stairs up saw [walk b. dat ik haar net nog de trap ti zag [op lopen]i that I her just now the stairs saw [up walk ‘that I saw her walking up the stairs just now’

(15) a.

(Groos 1989: 57)

(16) a.

dat Jan de bal over ti heeft [geschoten]i that Jan the ball over has [shot b. dat Jan de bal ti heeft [over geschoten]i that Jan the ball has [over shot ‘that Jan has shot the ball over’ (den Dikken 1995: 108)

Although the occurrence of some postpositions in the verb cluster is certainly predicted, there is a problem. The postpositions that are reanalyzed with the verb in (14)–(16) are not required to be structurally adjacent to the verb: (17) a.

de weg de stad in the way the city in ‘the way into the city’ b. die tippel de berg op that hike the mountain up ‘that hike up the mountain’ (van Riemsdijk 1978: 96)

The data in (17) show that the postpositions in and op can occur as complements of N0 and hence do not have to be structurally adjacent to the verb for their meaning to be licensed. In fact, all the postpositions in (14)–(16) are semantically closely related to the meaning of their underlying prepositions. In Chapters 5 and 6, I assumed that postpositions that show this behavior are not “semantic affixes” that need a verbal host. But if there is no lexical requirement for postpositions and verbs to be structurally adjacent, then we do not expect reanalysis to be allowed. For German, I showed that h-postpositions that have maintained a strict spatial meaning cannot be reanalyzed with the verb in word formation. Rather,


these h-postpositions are replaced by particles. (18) and (19) repeat examples (45) and (46) from Chapter 6, Section 6.3: (18) a.

(auf den Berg) hinaufsteigen (*aufsteigen) (on the mountain) H-up-climb (*part-climb ‘climb up the mountain’ b. der Weg hinauf the way H-up ‘the way up’ c. *Hinaufstieg d. Aufstieg, ‘ascent’

(19) a.

(aus dem Zimmer) hinausgehen (out the room) H-out-go ‘leave the room’ b. der Gang hinaus the walk H-out ‘the walk outside’ c. *Hinausgang d. Ausgang, ‘exit’

(*ausgehen) (*part-go

The simple particle elements auf and aus in (18d) and (19d) replace the complex h-postpositions hinauf and hinaus in the (a)-examples of (18) and (19). I will now follow a proposal made in Groos (1989) and assume that a similar process takes place in (14)–(16). Groos argues that postpositions inside verb clusters are reanalyzed as particles. Therefore, what appears inside the verb cluster in (14)–(16) is not the (functional) postposition, but a (lexical) particle that is substituted for the postposition. Since most postpositions and particles in Dutch are homophonous, the replacement is invisible in this language (see also Abraham 1995 for similar assumptions). This claim is confirmed by an observation made by Groos (1989) and Koopman (1993). Both Groos and Koopman point out that not all postpositions can easily appear inside the verb cluster. Both authors mention the postposition vandaan as an example of this situation: (20) a.

dat Jan bij de buren vandaan ti wilde [bellen]i that J. from the neighbors away ( wanted [call b. *dat Jan bij de buren ti wilde [vandaan bellen]i that J. from the neighbors wanted [away call ‘that Jan wanted to call (somebody) from the neighbors’ (Groos 1989: 58)


(21) ??dat het book (van) onder het bed ti is [vandaan gekommen]i that this book (from under the bed is [from come ‘that this book came from under the bed’ (Koopman 1993: 33)

As noted by Koopman (1993), whereas the postpositions in and op in (14) and (15) also occur as prepositions, the postposition vandaan in (20) and (21) is not homophonous with an existing prepositional element. The assumption that postpositional elements with a spatial meaning are replaced by homophonous particles when they are reanalyzed explains why reanalysis (and hence movement in Verb Raising) is not possible for those postpositions that do not have a corresponding particle form. Finally, Groos (1989) points out that judgements about postpositions inside the verb cluster are hazy, and Koopman (1993: 33) notes that the acceptability of postpositions inside the verb cluster “varies across speakers and within speakers judgements may vary depending on individual lexical items”. These comments may be taken as indications that the possibility of reanalyzing the structural relation between postpositions and verbs is lexically governed. It depends on highly idiosyncratic aspects of a speaker’s idiolect whether a postposition can be reanalyzed with a verb to undergo Verb Raising. Postpositions inside the verb cluster provide another test for the internal structure of complex verbs that undergo Verb Raising. In Dutch, as in German, postpositions can be modified with adverbials, which I take to be adjoined to the postpositional phrase (cf. Olsen 1999). Importantly, if a postposition appears inside the verb cluster, an adverbial modifier can neither be stranded in its base position, (22b), nor can it occur inside the verb cluster, (22c): (22) a.

dat Jan de bal [pal over]FP ti heeft [geschoten]i that J. the ball [right over has [shot b. *dat Jan de bal [pal tj]FP ti heeft [overj geschoten]i that J. the ball [right has [over shot c. *dat Jan de bal ti heeft [pal over geschoten]i that J. the ball has [right over shot ‘that Jan has shot the ball right over’ (Den Dikken 1995: 108)

(22b) is interesting, because it shows that the complex verb that underwent Verb Raising could not be derived from (22a) by overt incorporation, for in that case we would expect that the modifier can be stranded. This shows that complex verbs that appear inside the verb cluster in Dutch are derived by Merge, as argued by the reanalysis proposal. More importantly, (22c) illustrates that the


adverb cannot modify the postposition if the postposition occurs in the Verb Cluster. This provides strong evidence for the claim that the complex verb that has undergone Verb Raising is a genuine V0 and does not include any phrasal projections that could function as possible adjunction sites for an adverb.5 Postpositions in Dutch hence fit into the general analysis proposed in Chapter 6. What about resultatives? It was argued in Chapter 6 that resultatives cannot be reanalyzed with the verb. This claim is confirmed by the observation that in Dutch Verb Raising, phrasal resultative predicates cannot occur inside the verb cluster: (23) a.

dat Jan de fiets uit elkaar ti wil [halen]i that J. the bicycle to pieces wants [take b. *dat Jan de fiets ti wil [uit elkaar halen]i that J. the bicycle wants [to pieces take ‘that Jan wants to take the bicycle to pieces’

(24) a.

dat Jan de deur violet ti wil [verven]i that J. the door violet wants [paint b. *dat Jan de deur wil ti [violet verven]i that J. the door wants [violet paint ‘that Jan wants to paint the door violet’ (Neeleman 1994: 23)

In (23), the resultative predicate is clearly a prepositional phrase that cannot be reanalyzed with the verb for structural reasons. (24) shows that adjectival heads that function as resultative predicates cannot be reanalyzed either. This follows from the fact that, due to the intervening Agr-layer, the adjective and the verb in a construction like (24a) are not structurally adjacent. However, (25) shows that some resultatives may exceptionally undergo Verb Raising in Dutch: (25) a.

dat Jan de deur [groen]AP ti wil [verven]i that J. the door [green wants [paint

5.In Chapter 6 and above, I argued that particle verbs are V0s if they undergo Verb Raising or are subject to operations of derivational morphology. A possibility that I did not consider was that the particle inside this verbal head is itself a phrase. This would mean that the particle verb was a “phrasal compound”, i.e. a V0 that includes a PrtP. (22c) provides evidence that contradicts this idea. If a complex verb that undergoes Verb Raising included a phrasal representation of the verb’s complement, we would expect that adverbials that modify this complement are licensed in the verb cluster. The ungrammaticality of (22c) shows that reanalyzed structures exclusively consist of minimal elements.


b. dat Jan de deur wil ti [groenA verven]i that J. the door wants [green paint ‘that Jan wants to paint the door green’ (Neeleman 1994: 23)

In Section 6.3 I have assumed that reanalysis of resultatives is possible if the combination of the resultative predicate and verb is lexicalized. This would mean that the resultative groen verven in (25) is a lexical element. This assumption is certainly not uncontroversial; I will postpone the discussion to the next section, where I address resultative constructions in English and Norwegian. The ungrammaticality of (26b) and (26c) shows that as soon as a resultative predicate occurs inside the verb cluster, modifiers of this predicate are no longer licensed: (26) a.

dat Jan de deur [heel groen]AgrP ti wil [verven]i that J. the door [very green wants [paint b. *dat Jan de deur [heel tj]AgrP ti wil [groenj verven]i that J. the door [very wants [green paint c. *dat Jan de deur ti wil [heel groen verven]i that J. the door wants [very green paint ‘that Jan wants to paint the door very green’ (Neeleman 1994: 23)

As expected, adverbs modifying resultative adjectives are impossible if the verb is reanalyzed with the adjective to undergo Verb Raising. Like the example in (22b), (26b) shows that the complex V0 that moves is not derived by incorporation of the adjective. Like (22c), (26c) shows that the complex V0 that undergoes Verb Raising does not include any phrasal projections. I conclude that the crucial behavior of particle verbs in an SOV-language like Dutch is also captured by the analysis that I suggest in this study. The problematic properties of particle verbs with respect to Verb Raising can be accounted for through the Principle of Reanalysis that I proposed in Chapter 6. In the next section, I will show that this principle is also relevant for the analysis of particle verbs in SVO-languages like English and Norwegian.

7.2 On the verb-particle alternation in English and Mainland Scandinavian In this section I investigate how particle verbs in languages like English and Norwegian are analyzed in the light of the theory advocated in this study. I


argue that the so-called verb-particle alternation can be accounted for on the assumption that the particle and the verb can be reanalyzed. In Chapter 2, Section 2.7, I looked at particle verbs in Germanic SVO-languages. I argued that the syntactic representation of particle verbs in these languages is similar to their representation in German and Dutch. The particle is the head of a (non-functional) phrasal complement of the verb: (27) a.

German/Dutch: SOV das Bier austrinken v′

v0 trink-

VP V′

Spec das Bier

V0

PP

P0 aus b. English/Norwegian etc.: SVO drink the beer up v′

v0 drink

VP Spec the beer

V′ V0

PP P0 up

(27) illustrates why, in contrast to German and Dutch, the particle and the verb are separated by the direct object in English, Norwegian etc. Due to the


SVO-character of these languages, verb raising to v0 moves the base verb in front of the object in SpecVP.6 The typological difference with respect to particle verbs follows from different syntactic structures. However, the structure in (27) does not yet account for the so-called verbparticle alternation which is attested in Norwegian and English and illustrated in (28) and (29): (28) a. John drank his beer up b. John drank up his beer (29) a.

Mannen har drukket vinnen opp the.man has drunk wine part b. Mannen har drukket opp vinnen the.man has drunk part wine ‘The man has drunk up the wine’ (Norwegian; Svenonius 1996: 10)

In the (b)-examples, both the verb and the particle precede the direct object. It is not immediately obvious how these data can be derived from a structure like (27b). A vast number of analyses have been elaborated in order to explain the “characteristic paradigm” (Johnson 1991) of particle verbs illustrated in (28) and (29) (cf. Taraldsen 1983; Kayne 1985; Åfarli 1985; Johnson 1991; den Dikken 1995; Svenonius 1996).7 The proposal which comes closest to the theory that I have offered to account for word formation in German and Verb Raising in Dutch is the one made in Larson (1988b). Larson suggests that the V¢-node dominating the verb and the particle phrase in (27b) can be reanalyzed and can undergo movement as a complex V0:

6.This presupposes that the direct object does not have to move to SpecvP overtly in English for case reasons; cf. Chomsky (1995). See also Koster (1988) for independent evidence of the idea that the order V-NP-Prt is derived via movement of the verb from a base position which is to the right of its object. 7.The most detailed comparative study of the verb-particle alternation in English and the Scandinavian languages of which I am aware is provided by Svenonius (1996). Although Svenonius defends a syntactic approach, his analysis is based on a number of further assumptions that I do not share. For example, he argues that particle phrases are functional projections; a claim that is incompatible with the results that I reached in Chapter 3. I therefore do not adopt his proposal here.


(30) a. b.

[v′ drinki [VP his beer [V′ ti [PP up] ] ]] = (28a) [v′[v0drink up]i [VP his beer ti ] ]

= (28b)

(cf. Larson 1988b: note 11)

Larson’s proposal is attractive, because it allows for the possibility of explaining word formation with particle verbs, the behavior of Dutch particle verbs with respect to Verb Raising, and the verb-particle alternation in English and Norwegian by the same principle. However, in contrast to Dutch and German, where reanalysis is restricted to cases where the particle verb is attached to a lexical head, no such restriction is attested in English and Norwegian. The following data provide evidence that Larson’s proposal is correct: (31) a. John looked the information up b. John looked up the information (32) a. John looked the information right up b. *John looked right up the information (33) a. John flew the plane back b. John flew back the plane (34) a. John flew the plane straight back b. *John flew straight back the plane

(Neeleman 1994: 185ff.)

(35) a.

Jon sparka hunden langt ut J. kicked the.dog far part b. *Jon sparka langt ut hunden J. kicked far part the.dog ‘Jon kicked the dog far out’ (Åfarli 1985: 76)

(32a), (34a), and (35a) show that particles in English and Norwegian sometimes can be modified by adverbs (see Chapter 2, Section 2.7).8 However, if the verb and the particle are adjacent, these adverbs are not possible. This follows from a reanalysis account, according to which the verb and the particle in the (b)-examples in (31)–(35) form a complex V0. (31)–(35) could be regarded as the mirror image of the Verb Raising-data with postpositions discussed in

8.Some elements that are usually classified as particles may also be “postpositions”, i.e. semi-lexical representations of a functional head F0Prep. Since with respect to the alternation illustrated in (28) and (29), we expect these elements and particles to behave alike, I leave this point open.


Section 7.1. As in the Dutch examples, the reanalyzed form of the verb in the (b)-examples does not license any modifiers that require the prepositional element to project a phrase. The Mainland Scandinavian languages show interesting differences with respect to reanalysis. I showed in (29) that Norwegian exhibits the particle alternation. However, as illustrated in (36), Danish does not allow for the order verb-particle-object: (36) a.

Boris skrev kontrakten under Boris wrote the.contract part b. *Boris skrev under kontrakten Boris wrote part the.contract ‘Boris signed the contract’ (Danish; Svenonius 1996: 12)

(36) shows that while reanalysis applies freely in English and Norwegian, the grammar of Danish does not seem to allow for this option. Curiously, in Swedish we find the opposite situation. Reanalysis seems to be obligatory in this language; the order verb-object-particle is not permitted: (37) a.

Johan skrev upp numret J. wrote part the.number b. *Johan skrev numret upp J. wrote the.number part ‘Johan wrote down the number’

(Swedish; Svenonius 1996: 17)

The data in (36) and (37) raise the interesting question of what causes reanalysis to be restricted in this way in Danish and Swedish. As far as I can see, a satisfactory answer can only be provided through a detailed comparison of the grammatical systems of these languages (cf. Svenonius 1996 for a proposal). Since this would be the topic of a separate study, I have to leave this question open. A potential problem for the reanalysis approach to the verb-particle alternation is raised by the observation that English particle verbs do not productively undergo derivational word-formation processes (see den Dikken 1995: 23, note 21, but see Johnson 1991 for contrary assumptions).9 However,

9.As has been pointed out to me by Ian Roberts (p.c.), English has a quite productive pattern of nominalizations that forms nouns like pick-up, break-through, work-out, push-up or take-away. The linear order pre-determined by the syntax is preserved, but the main stress is shifted from the particle to the verb (cf. PICK up vs. pick UP). The latter nominals are probably derived by the same rule that creates the following nominals in French:


word formation with particle verbs was argued to be the main source of reanalysis in German. If the data in (28) really illustrate that English allows for reanalysis, it is unclear why this option cannot be used to derive nominals or adjectives from English particle verbs. In order to account for this curious gap in the English grammar, I suggest the following explanation. Suppose that in order to be unified with a certain lexical entry, the internal structure of a complex V0 that represents the result of reanalysis must preserve the linear order that is determined by the syntactic configuration of the two elements involved in reanalysis. In German and Dutch, which are SOV, the reanalyzed word is therefore a V0 with the base verb on the right. In contrast, reanalysis of the particle verb in English yields the order [verb-particle]: (38) a.

[VP [PrtP aus ] trink-]

b.

[VP drink [PrtP up ]]

Reanalysis

[V 0 austrink-] [V 0 drink up]

Now assume that the input to a morphological operation like derivation must be a well-formed morphological object. The reanalyzed V0 in German qualifies as such; its structure obeys Williams’ (1981) Right-hand Head Rule, according to which the head of a complex word is its rightmost element. Therefore, further word formation is possible. In contrast, the reanalyzed verb in English is not a legitimate morphological object; consequently, no further morphological processes can take place. Still, the V0 derived by reanalysis in English can undergo syntactic operations like head movement to v0. It is just that the linear order verb-particle in English prevents the reanalyzed verb from participating in further morphological derivations. The requirement that reanalysis must preserve the linear order pre-determined by structural adjacency seems to be less restrictive in Norwegian. As illustrated in (82), some Norwegian dialects prefer the order particle-verb when the particle verb is used in the participle form:

(i)

a. b. c.

trompe-l’oeil arc-en-ciel hors-la-loi

(lit. deceive-the-eye) (lit. arc-in-sky = rainbow) (lit. outside-the law = outlaw)

As illustrated in Chapter 1, sec. 1.3.1, Di Sciullo & Williams (1987: 82) argue that the examples in (i) are formed by a rule as in (ii) which assigns morphological status to certain lexicalized phrases: (ii)

N Æ XP

N Æ [VP pick up]


(39) a.

Trea vart nedhogne the.trees were part-chopped b. ??Trea vart hogne ned the.trees were chopped part ‘The trees were chopped down’ (Western Norwegian; Svenonius 1996: 19)

As noted in Svenonius (1996), the sentence with the particle on the left of the verb favors a stative interpretation of the participle, whereas the particle on the right favors an eventive reading. This suggests that (39a) is an adjectival passive, while (39b) seems to be a verbal passive. In Section 6.1.3, I analyzed participles in adjectival passives as A0s. Furthermore, I argued that for a particle verb to derive an adjectival participle, two options exist. According to the first option, the particle verb is reanalyzed as a V0, and this complex V0 combines with the adjectival zero-operator which yields the stative reading. In (Western) Norwegian, this option seems to be preferred. In (39a), the particle verb has been reanalyzed in order to combine with the derivational adjectival affix. However, as I argued above, this word formation operation requires the input (the particle verb) to be a well-formed morphological object. Therefore, reanalysis in Norwegian must change the order verb-particle to the order particle-verb. According to the second option, adjectival participles are derived by moving only the base verb to A0, stranding the particle. This option is required to derive adjectival passive constructions with particle verbs in a language that does not allow reanalysis to change the order of the terminal elements (as in English), or that does not allow reanalysis at all (as in Danish). (40) and (41) illustrate that the particle has to follow the participle in English and Danish, regardless of whether the passive receives a stative or an eventive interpretation: (40) a. The dog was thrown out b. *The dog was outthrown (41) a.

Hunden blev smedet ud the.dog was thrown part b. *Hunden blev udsmedet the.dog was part-thrown ‘The dog was thrown out’

(Danish; Svenonius 1996: 19)


Finally, recall that in Swedish reanalysis is obligatory. In this language, the particle must precede the participle:10 (42) a.

Mjölken blev aldrig inställd the.milk was never part-put b. *Mjölken blev aldrig ställd in the.milk was never put part ‘The milk never got put away’

(Swedish; Svenonius 1996: 19)

The example in (39a) is interesting, because it shows that reanalysis in Norwegian can yield different outputs. If the verb and the particle form a V0 that undergoes syntactic movement, as in (29b) above, the verb precedes the particle. This follows from the fact the particle phrase (in the non-reanalyzed structure) occurs to the right of the verb. However, if the V0-structure is subject to a morphological derivation, reanalysis shifts the order of the particle and verb, because the Right-hand Head Rule requires the verb to be on the right.11 This proposal implies that a particle verb in the order verb-particle-object is some kind of quasi-morphological object. It is a syntactic word; i.e. a V0, but it is morphologically not well-formed, since it is not right-headed. Åfarli (1985: 76) notes that in Norwegian, the particle can optionally form a wordaccent unit together with the verb in this order. This may be taken as further evidence that the particle verb in these constructions has some kind of intermediate status. Let me now turn to resultative constructions in English and Norwegian. As is predicted by the reanalysis proposal, resultative constructions do not enter into the characteristic paradigm attested with particle verbs in these SVO-languages: (43) a. John cuts the pear into pieces b. *John cuts into pieces the pear (Neeleman 1994: 195) (44) a. The doorman beat the drunks senseless b. *The doorman beat senseless the drunks

10.If (42) is interpreted as a verbal passive, the ungrammaticality of (42b) remains unexplained, since it is not clear why reanalysis is required if no derivational morpheme is present. 11.There are a few instances of nominals derived from particle verbs in English that behave similarly, i.e. they obey the Right-hand Head Rule and have shifted the particle to the left (cf. onlooker, bypass, offcuts, downfall, outlook).


(45) a. The firefighters hoisted the equipment high b. *The firefighters hoisted high the equipment Vi måla bilen fiolett we painted the.car violet b. *Vi måla fiolett bilen we painted violet the.car ‘We painted the car violet’

(Svenonius 1996: 4)

(46) a.

(Åfarli 1985: 90ff.)

The resultative predicate in (43) is a complex prepositional phrase which under no circumstances can form a verbal head together with the base verb. Reanalysis of a verb and the adjectival predicates in (44)–(46) is excluded, because the two terminal nodes are not structurally adjacent. However, there are some resultatives that do occur in the shifted order in English and Norwegian: (47) a. John cuts the melon open b. John cuts open the melon

(Neeleman 1994: 195)

(48) a. John made his intentions clear b. John made clear his intentions

(Neeleman 1994: 185)

(49) a. The activists set the lab rats free b. The activists set free the lab rats

(Svenonius 1996: 5)

(50) a.

Vi slo ormen ihjel we hit the.snake dead b. Vi slo ihjel ormen we hit dead the.snake ‘We beat the snake to death’

(51) a.

Vi gjorde bilen klar we made the.car ready b. Vi gjorde klar bilen we made ready the.car ‘We made the car ready’

(Åfarli 1985: 79)

In order to account for those resultatives in English that can be reanalyzed as V0s, Neeleman (1994) argues that these examples are lexicalized combinations of the verb and the resultative predicate. He assumes that in a resultative construction like cut open, the adjective morphologically subcategorizes for a verb (this explains why other resultatives with open may also appear in the reanalyzed variant, cf. kick open, break open etc.). This information must be stored as part of the lexical entry of the adjective.


Svenonius (1996) reaches a similar conclusion. In order to account for the exceptional behavior of the verb-adjective combinations set free and make clear, Svenonius argues that these specific resultative combinations must be listed in the lexicon. He adopts Pesetsky’s (1995) term of L-selection (lexical selection) and argues that the verb lexically selects for the specific resultative adjective. Again, the idea is that those resultative constructions which can be shifted must be lexically marked. Both proposals are compatible with the analysis that I suggested in Section 6.3, where I argued that adjectival resultatives in German can be reanalyzed if the combination of the adjective and the verb is lexically stored. The behavior of resultative constructions with respect to the alternation illustrated in (28) and (29), which is normally found only with particle verbs, supports the idea that this alternation should be explained in terms of the Principle of Reanalysis that I proposed in Chapter 6. In Section 7.1 I assumed that in those examples where adjectival resultatives can appear inside the verb cluster in Dutch Verb Raising constructions, reanalysis can apply, because the resultative construction is lexically listed. Therefore, the reanalysis principle allows the lexical entry that specifies the meaning of the adjective and the verb to be unified with a syntactic structure in which these two elements are morphologically adjacent (and have undergone Verb Raising as one complex head). A comparison of the examples from Dutch in (25) and (24) in Section 7.1 (repeated in (52) and (53)) and the examples from Norwegian in (54) and (55) confirms this assumption: (52) a.

dat Jan de deur groen wil verven that J. the door green wants paint b. dat Jan de deur wil groen verven that J. the door wants green paint ‘that Jan wants to paint the door green’

(53) a.

dat Jan de deur violet wil verven that J. the door violet wants paint b. *dat Jan de deur wil violet verven that J. the door wants violet paint ‘that Jan wants to paint the door violet’

(54) a.

Vi måla bilen blå we painted the.car blue b. Vi måla blå bilen we painted blue the.car ‘We painted the car blue’


(55) a.

Vi måla bilen fiolett we painted the.car violet b. *Vi måla fiolett bilen we painted violet the.car ‘We painted the car violet’

(Åfarli 1985: 90ff.)

Interestingly, the color adjectives groen and violet show the same surprising difference in Dutch Verb Raising contexts as the adjectives blå and fiolett do in Norwegian with respect to the “characteristic paradigm”. The resultative constructions ‘paint green/blue’ allow reanalysis of the adjective and the verb into a V0 in both Dutch and Norwegian, while the resultative construction ‘paint violet’ does not allow reanalysis in either of these two languages. The parallel between (52)–(53) on the one hand and (54)–(55) on the other supports my claim that the occurrence of Dutch resultatives in verb clusters and the adjacency of the verb and resultative predicate in examples like (54b) are manifestations of the same situation. If a resultative construction is lexicalized, reanalysis can apply. The resultative is then realized as a complex verb which can undergo Verb Raising in Dutch or enter into the alternation illustrated by examples like (54). One might object to this conclusion, pointing out that the meaning of resultatives like groen verven is entirely transparent. These resultative constructions seem to be derived productively, but why then should they be stored in the lexicon? Notice that this question presupposes a view of the lexicon as a “prison” that contains only the “lawless” (cf. Di Sciullo & Williams 1987: 3), i.e. expressions with irregular properties. However, this view has been challenged. As pointed out by Pinker & Prince (1991) and Jackendoff (1997), the lexicon not only includes irregular forms, but also some productively derived constructions. For example, Jackendoff discusses evidence from experimental work that suggests that the plural forms of high-frequency nouns are stored as separate lexical items, whereas the plural forms of low-frequency nouns are produced “on-line”. A resultative construction may therefore become an independent lexical item by virtue of the high-frequency of its use. Neeleman (1994) suggests that the possibility of undergoing Verb Raising together with the verb depends on the phonological weight of the resultative predicate. I do not want to exclude the possibility that phonological properties of an adjective might influence reanalysis, which is quite generally governed by various linguistic and extra-linguistic factors. Often, judgements vary from speaker to speaker or may be dependent on the particular context in which an expression is used. However, reanalysis does not depend on phonological


properties alone.12 I suspect that besides frequency of use, something like cognitive prominence (which may be determined partly by phonological weight) is responsible for the fact that a resultative construction like paint green is more “natural” and therefore more likely to be listed as a lexical item than the construction paint violet.13 I will therefore continue to assume that reanalyzing the verb with a resultative adjective is a lexically conditioned phenomenon. The behavior of resultatives in English, Norwegian, and Dutch, can hence be explained by the same principle. The exceptional cases with resultatives being reanalyzed as V0s provide further evidence in favor of the reanalysis proposal. For example, as was the case with resultatives that can undergo Verb Raising in Dutch, resultatives in English that allow for reanalysis do not permit adverbials: (56) a. John cut the melon right open b. *John cut right open the melon (Neeleman 1994: 263, note 6)

Finally, there is interesting evidence from Norwegian that shows that the reanalyzed structure of the shifted verb plus adjective is different from the nonshifted form. As argued in Chapter 3, in a non-shifted regular resultative construction, the projection of the resultative adjective is dominated by a functional Agr-projection whose head is structurally adjacent to the base verb. In Norwegian, this functional head is lexicalized through an agreementmorpheme which combines with the resultative predicate if A0 moves to Agr0 (see Chapter 3, Section 3.2.3): (57) a.

Vi vaska golvet reint Æ we washed the.floor-sg-neut clean-neut

functional structure

12.In discussing the contrast between (54) and (55), Åfarli (1985) also considers a phonological explanation, but he rejects this option after careful consideration and instead also opts for a lexical explanation. 13.Hornstein & Weinberg (1981) suggest that the possibility of stranding prepositions in English is contingent on reanalysis of the verb and the preposition (see also van Riemsdijk 1978). They show that this kind of reanalysis is restricted by non-syntactic factors like “naturalness” of the reanalyzed verb. For example, the contrast in (i) is accounted for by assuming that the reanalyzed V0 and P0 sleep in in (i-a) form a semantic unit (a “natural predicate”), whereas in (i-b), they do not: (i)

a. b.

That bed was slept in ti *New York was slept in ti


b. *Vi vaska golvet rein we washed the.floor clean ‘We washed the floor clean’ (Åfarli 1985: note 8)

I argued that the verb forms a complex V0 with the resultative adjective if reanalysis takes place and I pointed out that in reanalyzed structures, functional structure tends to be avoided. This assumption is confirmed by an observation made by Åfarli (1985) and Svenonius (1996: note 4). Both authors note that resultative adjectives in the shifted form typically appear in non-agreeing forms: (58) Vi vaska rein golvet Æ we washed clean the.floor ‘We washed the floor clean’

no functional structure

(58) shows that if reanalysis has taken place, no functional structure is present; the verb has combined with the bare adjective. I conclude that the reanalysis proposal, which was motivated by the word formation properties of particle verbs in German, and which also accounts for the properties of complex verbs in Dutch Verb Raising constructions, further explains the verb-particle alternation in English and Norwegian.14 The wordlike properties of the verb-particle construction in all Germanic languages follow from the same principle. For the strictly local representation of a head and a verb, specified as structural adjacency, grammar provides the possibility that these two heads are represented as parts of a complex V0.

14.Although the reanalysis approach reveals some interesting parallels and differences between SOV- and SVO-languages, it is not unproblematic. One obvious problem concerns inflectional morphology. Inflectional morphology attaches to the verbal part of the particle verb. As noted by Kayne (1985), this is unexpected if the verb and the particle really form a complex head: (i)

a. b.

John look-ed up the information *John look up-ed the information

This problem recalls the problem, discussed in note 4, that was raised by the observation that te-infinitives in Dutch can undergo Verb Raising. Again, it seems to be an unwelcome consequence of the reanalysis proposal that it requires something like the infixation of inflectional morphology. See Zells (forthcoming b) for an analysis of particle verbs that addresses this problem.


7.3 Conclusion In this chapter I have shown that some of the most surprising properties of the verb-particle construction in Germanic languages follow from the theory about German particle verbs that I developed in this book. The specific local relation that characterizes the verb-particle-construction in German (i.e. structural adjacency) also determines the status of particle verbs in languages like Dutch, English, and Norwegian as elements between syntax and morphology. From a cross-linguistic perspective, this is a welcome result. I argued that structural adjacency can be reanalyzed as morphological adjacency in both SOV-and SVO-languages. Importantly, the level of freedom to apply the Principle of Reanalysis that I suggested in Chapter 6 underlies language-specific variation. In German and Dutch, unification of a particle verb with a syntactic structure in which the particle and the verb form a V0 is only possible if this V0 is adjoined to another lexical head. Whereas the word formation data from German discussed in Chapter 6 illustrated this point for lexical heads of categories N and A, Dutch Verb Raising (Section 7.1) was argued to be a syntactic construction in which the reanalyzed particle verb attaches to a lexical head of category V. English and Norwegian (Section 7.2) allow unification even if the reanalyzed combination of the particle and the verb is not part of a larger-size word. The obvious question that is raised by this analysis is why languages should differ as they do above. For example, why do reanalyzed verbs in SOV-languages have to be part of a larger syntactic word, whereas no such requirement seems to hold for SVO-languages? Surely, parametric variation is to a certain extent inexplicable, but it would be interesting to see whether consideration of more data from Germanic languages reveals any other striking differences between these languages that can be traced back to the same parameterized property. However, I have to leave this and other issues as a topic for future research.

Conclusion

The debate concerning particle verbs is characterized by a controversy between the morphological approach and the syntactic approach. In its most radical form, the syntactic approach treats particle verbs on a par with regular verbcomplement constructions, while the morphological approach treats particle verbs as complex words. In this book, I tried to argue against the dichotomized character of this controversy and instead offered a third way of analyzing particle verbs. The alternative that I proposed combines aspects of both the morphological and the syntactic approach. My major claim is that a particle verb is a syntactic construction characterized by a local relation between the particle and the verb. As syntactic constructions, particle verbs share many characteristics of regular verb-complement constructions. At the same time, because no functional structure intervenes between the particle and the base verb, their relation is more local than the relation between lexical nodes in other syntactic constructions. Since morphology is the module of grammar that typically deals with terminal nodes in local domains, particle verbs therefore also resemble morphological objects in many respects. Moreover, I argued that the local domain established by the particle verb can even be reinterpreted as a morphological structure in certain contexts, and operations that otherwise only apply to real words can also apply to the particle verb. From this point, it is only a small step to reach an analysis that entirely gives up the idea that particle verbs are either words or phrases and instead assumes that they are words and phrases at the same time. In the following, I want to discuss briefly the major claims of my analysis in light of Chomsky’s (1994, 1995) bare phrase structure theory. The theoretical implications of this theory suggest that particle verbs are structurally ambiguous. I argued in Chapter 3 that particle phrases do not include arguments of the particle, due to the absence of functional structure. Therefore, the particle phrase is only represented by a single terminal node, its head:

298 Conclusion

n, n>0

V

(1)

V0

PrtP Prt0

According to (1), the particle projects, although it does not combine with a complement. But recall that Chomsky’s (1994, 1995) bare phrase structure theory (which I briefly discussed in Chapter 1) excludes non-branching projection levels. In this theory, a structure like (1) would be represented as (2): V

(2)

Prt

V

In the bare phrase structure theory, an element that does not project at all is a minimal projection; an element that does not project any further is a maximal projection. If the combination of a particle element and a verb is represented as in (2), the syntactic node Prt is simultaneously both minimal and a maximal projection. It is minimal, because it is not a projection, but it is also maximal, because it does not further project. If the maximal status of the particle in (2) is emphasized, (2) is a verbcomplement structure. As a maximal projection, Prt then requires that the verb projects. Consequently, the node that immediately dominates the verb and the particle must itself be non-minimal. This perspective corresponds to the structure in (1); the particle verb is taken to be a V¢ or a VP, which explains its syntactic properties. Importantly, however, the particle verb can at the same time be interpreted as a V0. From the perspective that Prt is minimal, (2) represents the structure of a verbal compound, in which case the dominating V-node is a complex verbal head. Accordingly, (2) behaves like a morphological object with respect to its ability to function as the input to affixation and like a V0 with respect to its ability to move in Dutch Verb Raising constructions. Furthermore, in languages like Norwegian or English, where Prt is merged to the right of the verb, the possibility of interpreting V in (2) as both a phrase and a complex head straightforwardly explains the “characteristic paradigm” discussed in Chapter 7. This suggests that both the syntactic and the word-like properties of particle verbs can be captured through the single bare phrase structure in (2).

Conclusion 299

The idea that particles are simultaneously both minimal and maximal might lead one to expect that combinations of verbs and other “single word”-complements (like intransitive prepositions, adjectival resultative predicates or bare plural nouns, for example) show the same ambiguous properties as particle verbs. However, recall that these intransitive complements still differ from particles in that they have extended projections. In Zeller (2000) I argue that lexical categories become maximal projections (= unambiguous phrasal complements; XPs in X-bar-theoretical terms) if they are selected by functional heads, i.e. by virtue of their extended projections. Therefore, since resultative predicates or bare plurals, for example, always have extended projections, they are always “genuine” complements, and the structure of resultatives or verbbare plural combinations is never ambiguous in a way that the structure of particle verbs is. Of course, we still have to ask under what conditions a structure like (2) counts as a verb-complement construction and under what conditions it represents a verbal compound. Again, the theory that I proposed in this study provides a possible answer to this question. I argued in Chapter 6, Section 6.2, that a particle verb in German can only be reanalyzed in the context of a derivational affix. Particle verbs are only words if they merge with a lexical head. In all other contexts, particle verbs maintain their “phrasal” properties. We could now assume that (2) is the structure of particle verbs, and that this structure qualifies as a verb-complement construction only if it is merged with a functional element. This is shown in (3a), where F is a head of the extended projection of the verb (either v or Tns). By virtue of functional structure, the verbal category that dominates V and Prt is a maximal projection, which in turn implies that the particle is interpreted as a phrase. In contrast, if (2) is merged with a lexical element, the particle verb is a compound. This is shown in (3b), where L is an adjectival or nominal suffix: F

(3) a.

V Prt

L

b.

V

F V

Prt

L V

According to this proposal, the answer to the question of whether the particle verb is a phrasal construct or a word depends on properties of the particle verb’s syntactic context. Notice that the Principle of Reanalysis and the amendment

300 Conclusion

that I proposed in Chapter 6, Section 6.2.1, state essentially the same thing, albeit in slightly different terms. However, an analysis which is based on a bare phrase structure like (2) has the crucial advantage of not having to postulate two different structural representations in order to account for the phrasal and the morphological properties of particle verbs. Therefore, expressing the results of my analysis in a bare phrase structure theory may provide a better understanding of the parallels between the syntactic structure of particle verbs and a complex head-structure which I have taken to be the result of reanalysis. Furthermore, such a theory may also provide an interesting starting point for a critical evaluation of bare phrase structure theory itself with respect to open questions about syntactic projection and the syntax-morphology interface. In Zeller (2000) I address some of these questions and extend the analysis presented here on the basis of a bare phrase structure theory about projection. The major part of my analysis has been developed by focusing on particle verbs in German. However, I also evaluated my proposal against the properties of particle verbs in other Germanic languages, like Dutch, English, or Norwegian. I looked at these latter languages in some detail in Chapter 7, and I showed that one major difference between particle verbs in English and the Scandinavian languages on the one hand and German and Dutch on the other hand is straightforwardly explained by the theory I have proposed in this book. Since particles are realized as sisters of the verb, they follow the verb in SVO-languages, but precede the verb in SOV-languages. However, it is clear that this explanation is only the tip of the iceberg — the typological perspective still raises a number of interesting questions that I could only partly address in this study. One general topic which is still worth further investigation concerns typological differences between languages within the SVO- and the SOV-paradigm. With respect to SVO-languages, I highlighted an interesting difference between Norwegian, Danish, and Swedish in Chapter 7. Whereas Norwegian licenses both the order V-NP-Prt and V-Prt-NP, Danish only allows for the first option, while in Swedish, only the latter order exists. According to the theory outlined in this book, the order V-Prt-NP is the result of reanalysis of the verb and the particle. The obvious question is why reanalysis is impossible in Danish, but obligatory in Swedish. The fact that not all SVO-languages show the “characteristic paradigm” with particle verbs is particularly interesting in light of an observation made by le Roux (1988) regarding Afrikaans, an SOV-language in which particle verbs pattern in most respects with their Dutch counterparts. Recall that in Dutch Verb Raising constructions, the particle verb can be separated by movement of

Conclusion 301

the verb, (4a). However, the particle verb can also move as a whole, (4b), an option that I argued to be the result of reanalysis: (4) a.

omdat Jan Marie op wil bellen because J. M. part wants ring b. omdat Jan Marie wil opbellen because J. M. wants part-ring ‘because Jan wants to ring up Mary’

(Dutch)

However, as noted by le Roux, the two options exhibited in (4) are not equally available in Verb Raising constructions in Afrikaans. Le Roux observes that “speakers of Afrikaans find sentences […] in which the verb is separated from the particle by the application of V-Raising only marginally acceptable, if not completely ill-formed” (1988: 21): (5) a. ?omdat Jan haar op wou bel because J. her part wanted ring b. omdat Jan haar wou opbel because J. her wanted part-ring ‘because Jan wanted to ring her up’ (Afrikaans, le Roux 1988: 21) (6) a. *Hy sal nie die antwoorde by my af kan kyk nie he will not the answers from me part can look not b. Hy sal nie die antwoorde by my kan af kyk nie he will not the answers from me can part look not ‘He will not be able to crib from me’ (Afrikaans, le Roux 1988: 241)

In terms of the theory I proposed in this book, it seems that many particle verbs in Afrikaans are obligatorily reanalyzed in Verb Raising constructions, whereas Dutch always allows for the non-reanalyzed structure. In other words, the relation between Dutch and Afrikaans particle verbs with respect to Verb Raising mirrors the relation between particle verbs in Norwegian and Swedish with respect to the characteristic paradigm. It seems worthwhile to examine further whether the two more “restricted” languages, Afrikaans and Swedish, have some independent property in common which is not found in Norwegian and Dutch and which explains the more restrictive character of these languages. Let me end with a few remarks on a subject which normally is not of particular interest to formal linguists, although they might sometimes be confronted with it in dinner table conversation, namely the spelling conventions in German.

302 Conclusion

In German, verbs and their nominal objects are normally written as two distinct words (I use the #-symbol to indicate incorrect spelling):1 (7) a. Bücher lesen, ‘read books’, not: #bücherlesen b. Geld sparen, ‘save money’, not: #geldsparen

The combination of a prefix and a verb in German is spelled as one word: (8) a. unterschreiben (lit. under-write), ‘sign’, not: #unter schreiben b. umsegeln, ‘sail around’, not: #um segeln

We know that particle verbs share properties of both the verb-object combinations in (7) and of the prefix verbs in (8). It is therefore interesting to see what the spelling conventions have to say about them. Prepositional particle verbs are treated like prefix verbs. When the particle and the verb are adjacent (e.g. in their infinitival form or in verb-last contexts), they are spelled as one word: (9) a. aufschreiben, ‘write up’, not: #auf schreiben b. absegeln, ‘sail away’, not: #ab segeln

Given (9), one might assume that in German dictionaries, particle verbs are generally treated as words. However, consider the following examples: (10) a. Maß nehmen (lit. measure-take), ‘aim’, not: #maßnehmen b. Auto fahren, ‘drive a car, be a car driver’, not: #autofahren

According to the theory that I have presented, Maß nehmen and Auto fahren are nominal particle verbs (since a singular count noun without a determiner combines with the verb; see Chapter 3, Section 3.2.1). (10) suggests that nominal particles are treated on a par with genuine phrasal complements like the ones in (7), while prepositional particles are analyzed as prefixes. The spelling conventions of German therefore treat particle verbs inconsistently. Unfortunately, not even all nominal particle verbs are treated in a uniform fashion. Consider the spelling of the next two examples: (11) a. teilnehmen, (lit. part-take), ‘take part’, not: #Teil nehmen b. heimfahren, ‘drive home’, not: #Heim fahren

(11) shows that some nominal particle verbs are spelled like the prepositional particle verbs in (9). The dictionary treats them as exceptions; the general rule

1.It is a further peculiarity of German spelling that nouns always start with a capital letter.

Conclusion 303

is that constructions with (separable) nouns and verbs in German are spelled as two separate words. Exceptions like those in (11) are captured through the following rule of thumb: if the meaning of a noun used as a particle is somehow different from its “regular” meaning, the particle and the verb are spelled as one word. Apart from the problematic character of the distinction between regular and irregular meanings (discussed in Chapters 4 and 5), it seems unlikely that a writer who does not know the correct spellings of, for example, the two verbs in (10a) and (11a) could actually derive this difference on the basis of such a rule. He would rather want to consult a dictionary.2 These inconsistencies suggest that the spelling conventions may have been formulated without a clear notion of what constitutes a particle. It might be that the constructions in (9)–(11) are not even viewed as being instances of the same grammatical phenomenon. In this respect, it would be worthwhile to use a rigorous theory of particle verbs to inform the formulation of spelling conventions, whatever the rules of this theory may be. In this book, I have defined particles as “phrases without functional structure”. It is certainly a non-trivial job to “translate” this definition into a terminology which can be easily operationalized. But acknowledging the existence of formal theories about grammatical phenomena can only help to deal with these phenomena also in domains other than formal linguistics. By the same token, I also think that formal linguists should at times be willing to question whether their work has any relevance outside academics — for example, does it facilitate learning a foreign language, understanding grammar, or writing correctly? More exchange between “formal” and “practical” linguists is surely fruitful for both.

2.The reform of the spelling conventions that was introduced in Germany on August 1st, 1998 at least removed even more dramatic inconsistencies from the dictionaries. For example, although the spelling of the nominal particle verb Auto fahren, ‘drive a car’, was as in (i) even before 1998, the particle verb Rad fahren, ‘ride a bicycle’, was spelled as in (ii) until 1998 (cf. Duden 1991): (i) (ii)

Auto fahren, not: #autofahren radfahren, not: #Rad fahren

Although the spelling of Rad fahren has now been adjusted to the spelling of Auto fahren (cf. Duden 1996), the contrast between (10) and (11) shows that the reform failed to deal successfully with all inconsistencies.

References

Abney, S. 1987. The English Noun Phrase in Its Sentential Aspect. Ph.D. dissertation, MIT, Cambridge. Abraham, W. 1989. “Idioms in contrastive and in universally based typological research: toward dintinctions and relevance”. In Proceedings of the First Tilburg Workshop on Idioms, M. Everaert and E.-J. van der Linden (eds), 1–22. Tilburg University. Abraham, W. 1995. Deutsche Syntax im Sprachenvergleich. Tübingen: Gunter Narr. Ackema, P. 1995. Syntax below Zero. Ph.D. dissertation, OTS, University of Utrecht. Ackerman, F. and Webelhuth, G. 1998. A Theory of Predicates. Stanford: CSLI Publications. Åfarli, T. 1985. “Norwegian Particle Constructions as Causative Constructions”. Nordic Journal of Linguistics 8: 75–98. Alexiadou, A. and Wilder, C. (eds). 1998. Possessors, Predicates, and Movement in the Determiner Phrase. Amsterdam and Philadelphia: John Benjamins. Anderson, S.-R. 1992. A-Morphous Morphology. Cambridge: Cambridge University Press. Archangeli, D. and Langendoen, D. T. (eds). 1997. Optimality Theory: An Overview. Oxford: Blackwell. Aronoff, M. 1994. Morphology by Itself. Stems and Inflectional Classes. Cambridge: MIT Press. Baker, M. 1988. Incorporation. A Theory of Grammatical Function Changing. Chicago and London: The University of Chicago Press. Baker, M., Johnson, K. and Roberts, I. 1989. “Passive Arguments Raised”. Linguistic Inquiry 20: 219–251. Bayer, J. 1993. V(P)-Topicalization and the Role of Traces. Unpublished manuscript, University of Stuttgart. Bennis, H. 1991. “Theoretische aspekten van partikelvooropplaatsing II”. TABU Bulletin voor Taalwetenschap 21: 89–95. Bennis, H. 1992. “Long Head Movement: The position of particles in the verbal cluster in Dutch”. In Linguistics in the Netherlands 1992, R. van Hout and R. Bok-Bennema (eds), 37–47. Amsterdam and Philadelphia: John Benjamins. Besten, H. d. 1983. “On the interaction of root transformations and lexical deletive rules”. In On the Formal Syntax of the Westgermania, W. Abraham (ed.), 47–131. Amsterdam and Philadelphia: John Benjamins. Besten, H. d. and Edmondson, J. A. 1983. “The verbal complex in continental West Germanic”. In On the Formal Syntax of the Westgermania, W. Abraham (ed.), 155–216. Amsterdam and Philadelphia: John Benjamins. Besten, H. d. and Rutten, J. 1989. “On Verb Raising, Extraposition and Free Word Order in Dutch”. In Sentential Complementation and the Lexicon, D. Jaspers et al. (eds), 41–56. Dordrecht: Foris.

306 References

Bhatt, C. 1990. Die syntaktische Struktur der Nominalphrase im Deutschen. Tübingen: Gunter Narr. Bierwisch, M. 1990. “Verb Cluster formation as a morphological process”. In Yearbook of morphology 3, G. Booij and J. v. Marle (eds), 173–199. Dordrecht: Foris. Bobaljik, D. 1995. Morphosyntax: The Syntax of Verbal Inflection. Ph. D. dissertation, MIT, Cambridge. Booij, G. 1989. “Complex verbs and the theory of level ordering”. In Yearbook of morphology 2, G. Booij and J. v. Marle (eds), 21–30. Dordrecht: Foris. Booij, G. 1990. “The boundary between morphology and syntax: separable complex verbs in Dutch”. In Yearbook of morphology 3, G. Booij and J. v. Marle (eds), 45–63. Dordrecht: Foris. Borer, H. 1988. “On the morphological parallelism between compounds and constructs”. In Yearbook of Morphology 1, G. Booij and J. v. Marle (eds), 45–65. Dordrecht: Foris. Borer, H. 1991. “The Causative-Inchoative Alternation: A Case Study in Parallel Morphology”. The Linguistic Review 8: 119–158. Borer, H. 1993. Derived Nominals. Unpublished manuscript, University of Massachusetts, Amherst. Borer, H. (forthcoming). “The Construct in Review”, to appear in: Studies in Afroasiatic Grammar. Borer, H. 1998. “Passive without Theta Grids”. In Morphology and Its Relation to Phonology and Syntax, S. Lapointe (ed.), 60–99. Stanford: CSLI Publications. Borer, H. 1999. “Deconstructing the Construct”. In Beyond Principles and Parameters, K. Johnson and I. Roberts (eds), 43–89. Dordrecht, Boston and London: Kluwer. Botha, R. 1983. Morphological Mechanisms. Oxford: Pergamon Press. Bresnan, J. 1978. “A realistic transformational grammar”. In Linguistic theory and psychological reality, M. Halle, J. Bresnan and G. Miller (eds), 1–59. Cambridge: MIT Press. Bresnan, J. and Mchombo, S.A. 1995. “The Lexical Integrity Principle: Evidence from Bantu” Natural Language and Linguistic Theory 13: 181-254. Brody, M. 1995. Lexico-Logical Form. Cambridge: MIT Press. Broekhuis, H., den Besten, H., Hoekstra, K. and Rutten, J. (1995), “Infinitival complementation in Dutch: On remnant extraposition”. The Linguistic Review 12: 93–122. Büring, D. 1991. “Semantische Transparenz und Linking”. Linguistische Berichte 135: 346–374. Büring, D. 1996. The 59th Street Bridge Accent. Ph.D. dissertation, University of Tübingen (SfS-Report 05–96). Büring, D. and Hartmann, K. 1997. “Doing the right thing”. Linguistic Review 14: 1–42. Burzio, L. 1986. Italian Syntax. A Government and Binding Approach. Dordrecht, Boston, Lancaster and Tokyo: Reidel. Carrier, J. and Randall, J. H. 1992. “The Argument Structure and Syntactic Structure of Resultatives”. Linguistic Inquiry 23: 173–234. Chomsky, N. 1970. “Remarks on Nominalization”. In Readings in English Transformational Grammar, R. Jacobs and P. Rosenbaum (eds), 184–221. Waltham, Mass.: Ginn. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986a. Knowledge of Language. New York: Praeger. Chomsky, N. 1986b. Barriers. Cambridge: MIT Press.

References 307

Chomsky, N. 1991. “Some notes on Economy of Derivation and Representation”. In Principles and Parameters in Comparative Syntax, R. Freidin (ed.), 417–454. Cambridge: MIT Press. Chomsky, N. 1994. “Bare Phrase Structure”. MIT Occasional Papers in Linguistics 5, MIT, Cambridge. Chomsky, N. 1995. The Minimalist Program. Cambridge: MIT Press. Chomsky, N. 1998. Minimalist Inquiries. Unpublished manuscript, MIT, Cambridge. Cinque, G. 1991. Types of A¢-Dependencies. Cambridge: MIT Press. Corver, N. 1990. The Syntax of Left Branch Extractions. Ph. D. dissertation, University of Tilburg. Corver, N. 1997. “The internal syntax of the Dutch extended adjectival projection”. Natural Language and Linguistic Theory 15: 289–368. Dewell, R. B. 1996. “The separability of German über-: A cognitive approach”. In The Construal of Space in Language and Thought, M. Pütz and R. Dirven (eds), 109–133. Berlin and New York: de Gruyter. Di Sciullo, A. M. and Williams, E. 1987. On the Definition of Word. Cambridge: MIT Press. Dikken, M. d. 1995. Particles. On the Syntax of Verb-Particle, Triadic, and Causative Constructions. New York and Oxford: Oxford University Press. Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht: Kluwer. Drach, E. 1963. Grundgedanken der deutschen Satzlehre. Darmstadt. (Reprint of 3rd ed., Frankfurt 1940). Duden20. 1991. Die deutsche Rechtschreibung. Duden21. 1996. Die deutsche Rechtschreibung. Emonds, J. 1972. “Evidence that Indirect Object Movement is a Structure-Preserving Rule”. Foundations of Language 8: 546–561. Emonds, J. 1985. A Unified Theory of Syntactic Categories. Dordrecht: Foris. Evers, A. 1975. The Transformational Cycle in Dutch and German. Ph.D. dissertation, University of Utrecht/Indiana University Linguistics Club. Fanselow, G. 1990. Über Partikelverben. Addendum to a talk given at the University of Cologne. Fraser, B. 1976. The Verb-Particle Combination in English. New York: Academic Press. Gamon, M. and Reutter, T. 1996. The Analysis of German Separable Prefix Verbs in the Microsoft Natural Language Processing System. Unpublished manuscript, Microsoft Research, Redmond. Gee, J. P. 1977. “Comments on the Paper by Akmajian”. In Formal Syntax, P. Cullicover, T. Wasow and A. Akmajian (eds), 461–493. New York: Academic Press. Giusti, G. 1989. Zu-Infinitivals and the Structure of IP in German. Unpublished manuscript, University of Geneva. Grewendorf, G. 1989. Ergativity in German. Dordrecht: Foris. Grewendorf, G. 1990. “Verbbewegung und Negation im Deutschen”. In Groninger Arbeiten zur Germanistischen Linguistik 30: 57–125. Grewendorf, G., Hamm, F. and Sternefeld, W. 1987. Sprachliches Wissen. Frankfurt: Suhrkamp. Grewendorf, G. and Sabel, J. 1994. “Long Scrambling and Incorporation”. Linguistic Inquiry 25: 263–308.

308 References

Grimshaw, J. 1990. Argument Structure. Cambridge: MIT Press. Grimshaw, J. 1991. Extended Projection. Unpublished manuscript, Brandeis University. Grimshaw, J and Mester, R.-A. 1988. “Light Verbs and θ-Marking”. Linguistic Inquiry 19: 205–232. Groos, A. 1989. “Particle-verbs and adjunction”. In Linguistics in the Netherlands 1989, H. Bennis and A. van Kemenade (eds), 51–60. Amsterdam and Philadelphia: John Benjamins. Guasti, M. T. 1991. “Incorporation, Excorporation and Lexical Properties of Causative Heads”. Linguistic Review 8: 209–232. Haiden, M. 1997. “Verbal inflection and the structure of IP in German”. Groninger Arbeiten zur germanistischen Linguistik 41: 77–106. Hale, K. and Keyser, S. 1993. “On Argument Structure and the Lexical Expression of Syntactic Relations”. In The View from Building 20, K. Hale and S. Keyser (eds), 53–110. Cambridge: MIT Press. Halle, M. and Marantz, A. 1993. “Distributed Morphology and the Pieces of Inflection”. In The View from Building 20, K. Hale and S. Keyser (eds), 111–176. Cambridge: MIT Press. Halpern, A. 1995. On the Placement and Morphology of Clitics. CSLI Publications: Stanford. Hankamer, J. 1973. “Unaccepted Ambiguity”. Linguistic Inquiry 4: 17–68. Hartmann, K. 2000. Right Node Raising and Gapping. Amsterdam and Philadelphia: John Benjamins. Haverkort, M. 1990. “Clitic Climbing and Barrierhood of VP”. In Current Approaches to African Linguistics 7, J. Hutchison and V. Manfredi (eds), 145–158. Dordrecht: Foris. Hazout, I. 1990. Verbal Nouns: Theta-Theoretical Studies in Hebrew and Arabic. Ph.D. dissertation, University of Massachusetts, Amherst. Hinterhölzl, R. 1996 Coherent Infinitives in German, Dutch and Westflemish. Unpublished manuscript, University of Southern California. Höhle, T. 1982. “Über Komposition und Derivation: zur Konstituentenstruktur von Wortbildungsprodukten im Deutschen”. Zeitschrift für Sprachwissenschaft 1: 76–112. Hoeksema, J. 1988. Dutch postposition incorporation. Paper read at the 1988 LSA, New Orleans. University of Pennsylvania. Hoeksema, J. 1991. “Theoretische aspekten van partikelvooropplaatsing”. TABU Bulletin voor Taalwetenschap 21: 18–26. Hoekstra, T. 1988. “Small Clause Results”. Lingua 74: 101–139. Hoekstra, T., Lansu, M. and Westerduin, M. 1987. “Complexe verba”. In GLOT 10, 61–79. Hoekstra, T. and Mulder, R. 1990. “Unergatives as Copular Verbs: Locational and Existential Predication”. Linguistic Review 7: 1–79. Horn, L. R. 1989. A natural history of negation. Chicago and London: The University of Chicago Press. Hornstein, N. and Weinberg, A. 1981. “Case Theory and Preposition Stranding”. Linguistic Inquiry 12: 55–92. Hout, A. v. 1998. Event Semantics of Verb Frame Alternations. A Case Study of Dutch and its Acquisition. New York and London: Garland. Hundsnurscher, F. 1968. Das System der Partikelverben mit AUS in der Gegenwartssprache. Göppingen: Kümmerle.

References 309

Jackendoff, R. 1975. “Morphological and Semantic Regularities in the Lexicon”. Language 51: 639–671. Jackendoff, R. 1983. Semantics and Cognition. Cambridge: MIT Press. Jackendoff, R. 1987. “The status of Thematic Relations in Linguistic Theory”. Linguistic Inquiry 18: 369–411. Jackendoff, R. 1990. Semantic Structures. Cambridge: MIT Press. Jackendoff, R. 1993. “The role of conceptual structure in argument selection: A reply to Emonds”. Natural Language and Linguistic Theory 11: 279–312. Jackendoff, R. 1997. The Architecture of the Language Faculty. Cambridge: MIT Press. Johnson, K. 1991. “Object Positions”. Natural Language and Linguistic Theory 9: 577–636. Kaufmann, I. 1995. Konzeptuelle Grundlagen semantischer Dekompositionsstrukturen. Tübingen: Niemeyer. Kayne, R. 1984. Connectedness and Binary Branching. Dordrecht: Foris. Kayne, R. 1985. “Principles of Particle Constructions”. In Grammatical Representation, J. Guéron, H.-G. Obenauer and J.-Y. Pollock (eds), 101–140. Dordrecht: Foris. Kluge, F. 1960. Etymologisches Wörterbuch der deutschen Sprache. Berlin: de Gruyter. Koopman, H. 1993. The structure of Dutch PPs. Unpublished manuscript, University of California at Los Angeles. Koopman, H. 1995. “On Verbs That Fail to Undergo V-Second”, Linguistic Inquiry 26: 137–163. Koopman, H. 1999a. “The internal and external distribution of pronominal DPs”. In Beyond Principles and Parameters, K. Johnson and I. Roberts (eds), 91–132. Dordrecht and Boston and London: Kluwer. Koopman, H. 1999b. Revised version of Koopman 1993. To appear in The syntax of Specifiers and Heads: collected essays by Hilda Koopman. London and New York: Routledge. Koster, J. 1975. “Dutch as an SOV-language”. In Linguistics in the Netherlands 1973, A. Kraak (ed.), 165–177. Assen: Van Gorcul. Koster, J. 1988. “The residual SOV structure of English”. Groningen Papers in Theoretical and Applied Linguistics TENK 5. Koster, J. 1989. “Left-Right Asymmetries in the Dutch Complementizer System”. In Sentential Complementation and the Lexicon, D. Jaspers et al. (eds), 271–282. Dordrecht: Foris. Kratzer, A. 1994. The Event Argument and the Semantics of Voice. Unpublished manuscript, University of Massachusetts, Amherst. Kratzer, A. 1995. Class lectures, Fall 1995, University of Massachusetts, Amherst. Krifka, M. 1989. “Nominal Reference, Temporal Constitution and Quantification in Event Semantics”. In Semantics and Contextual Expression, R. Bartsch, J. van Benthem and P. van Emde Boas (ed.), 75–115. Dordrecht: Foris. Krifka, M. 1992. “Thematic Relations as Links between Nominal Reference and Temporal Constitution”. In Lexical Matters, I. Sag and A. Szabolcsi (ed.), 29–53. CSLI Lecture Notes, Stanford, Cal. Kroch, A. and Santorini, B. 1991. “The Derived Constituent Structure of the West Germanic Verb-Raising Construction”. In Principles and Parameters in Comparative Grammar, R. Freidin (ed.), 269–338. Cambridge: MIT Press.

310 References

Kühnhold, I. 1973. “Präfixverben”. In Deutsche Wortbildung: Das Verb, I. Kühnhold and H. Wellmann (eds), 141–362. Düsseldorf: Schwann. Lapointe, S. 1980. A Theory of Grammatical Agreement. Ph.D. dissertation, University of Massachusetts, Amherst (published 1985, New York and London: Garland). Larson, R. 1988a. “On the double object construction”. Linguistic Inquiry 19: 335–391. Larson, R. 1988b. “Light Predicate Raising”. MIT Cognitive Science Center Working Papers. MIT, Cambridge. Le Roux, C. 1988. On the interface of morphology and syntax: Evidence from verb-particle combinations in Afrikaans. M.A. thesis, SPIL, University of Stellenbosch. Levelt, W. 1989. Speaking: From intention to articulation. Cambridge: MIT Press. Levin, B. and Rappaport, M. 1986. “The Formation of Adjectival Passives”. Linguistic Inquiry 17: 623–661. Levin, B. and Rappaport Hovav, M. 1992. “The lexical semantics of verbs of motion”. In Thematic Structure: Its Role in Grammar, I. M. Roca (ed.), 247–269. Berlin and New York: Foris. Levin, B. and Rappaport Hovav, M. 1995. Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge: MIT Press. Lieber, R. 1980. On the organization of the lexicon. Ph.D. dissertation, MIT, Cambridge. Lieber, R. 1992. Deconstructing Morphology. Chicago and London: The University of Chicago Press. Lieber, R. and Baayen, H. 1993. “Verbal prefixes in Dutch: a study in lexical conceptual structure”. Yearbook of Morphology, G. Booij and J. v. Marle (eds), 51–78. Dordrecht, Boston and London: Kluwer. Löbel, E. 1990. “D und Q als funktionale Kategorien in der Nominalphrase”. Linguistische Berichte 127: 232–264. Longobardi, G. 1994. “Reference and Proper Names: A Theory of N-Movement in Syntax and Logical Form”. Linguistic Inquiry 25: 609–665. Lüdeling, A. 1998a. On Particle Verbs and Similar Constructions in German. Ph.D. dissertation, University of Stuttgart. Lüdeling, A. 1998b. Strange Resultatives in German: New evidence for a semantic treatment. In The Syntax and Semantics of Predication. [Proceedings of the 1997 Texas Linguistics Society Conference], R. C. Blight and M. J. Moosally (eds), 223–233. Marantz, A. 1984. On the nature of grammatical relations. Cambridge: MIT Press. Marantz, A. 1988. “Clitics, Morphological Merger, and the Mapping to Phonological Structure”. In Theoretical Morphology: approaches in modern linguistics, M. Hammond and M. Noonan (eds), 253- 270. San Diego: Academic Press. Marantz, A. 1989. “Clitics and Phrase Structure”. In Alternative conceptions of phrase structure. Chicago, M. Baltin and A, Kroch (eds), 99–116. Chicago and London: The University of Chicago Press. Marantz, A. 1993. “A Late note on Late Insertion”. In Explorations in Generative Grammar: A Festschrift for Dong-Whee Yang, Y.-S. Kim et al. (eds), 396–413. Seoul: Hankuk. Marantz, A. 1997. “No escape from Syntax: Don’t try Morphological Analysis in the Privacy of Your Own Lexicon”. Penn Working Papers in Linguistics, Volume 4/2 [Proceedings of the 21st Annual Penn Linguistics Colloquium], 201–225, University of Pennsylvania.

References

Marantz, A. in prep. ‘Cat’ as a phrasal idiom: Consequences of late insertion in Distributed Morphology. Unpublished manuscript, MIT, Cambridge. McCarthy, J. and Prince, A. 1986. Prosodic Morphology. Unpublished manuscript, University of Massachusetts, Amherst, and Brandeis University. McIntyre, A. 2001. German double particles as preverbs: Morphology and Conceptual Semantics. Tübingen: Stauffenberg. (to appear) McIntyre, A. in prep. German particle verbs and morphology. Unpublished manuscript, University of Leipzig. Miller, D. G. 1993. Complex Verb Formation. Amsterdam and Philadelphia: John Benjamins. Mulder, R. 1992. The Aspectual Nature of Syntactic Complementation. Ph.D. dissertation, University of Leiden (HIL Dissertations 3). Neeleman, A. 1994. Complex predicates. Ph.D. dissertation, OTS, University of Utrecht. Neeleman, A. 1997. “PP-complements”. Natural Language and Linguistic Theory 15: 89–137. Neeleman, A. and Weerman, F. 1993. “The Balance between Syntax and Morphology: Dutch Particles and Resultatives”. Natural Language and Linguistic Theory 11: 433–475. Nespor, M. and Vogel, I. 1986. Prosodic Phonology. Dordrecht: Foris. Nunberg, G., Sag, I. and Wasow, T. 1994. “Idioms”. Language 70: 491–538. Olsen, S. 1996. “Pleonastische Direktionale”. In Wenn die Semantik arbeitet, G. Harras and M. Bierwisch (eds), 303–329. Tübingen: Niemeyer. Olsen, S. 1997a. “Der Dativ bei Partikelverben”. In Sprache im Fokus, C. Dürscheid, K. H. Ramers and M. Schwarz (eds), 307–328. Tübingen: Niemeyer. Olsen, S. 1997b. “Zur Kategorie Verbpartikel”. Beiträge zur Geschichte der deutschen Sprache und Literatur 119: 1–32. Olsen, S. 1999. “Durch den Park durch, zum Bahnhof hin: komplexe Präpositionalphrasen mit einfachem direktionalem Kopf”. In Deutsch kontrastiv, H. Wegener (ed.), 111–134. Tübingen: Stauffenberg. Ouhalla, J. 1991. Functional Categories and Parametric Variation. London: Routledge. Parsons, T. 1990. Events in the Semantics of English. Cambridge: MIT Press. Paulissen, D. and Zonneveld, W. 1988. “Compound Verbs and the Adequacy of Lexical Morphology”. In Morphology and Modularity. In honour of Henk Schultink, M. Everaert, A. Evers, R. Huybrechts & M. Trommelen (eds), 281–302. Dordrecht: Foris. Pesetsky, D. 1985. “Morphology and Logical Form”. Linguistic Inquiry 16: 193–246. Pesetsky, D. 1995. Zero Syntax. Cambridge: MIT Press. Pinker, S. and Prince, A. 1991. “Regular and irregular morphology and the psychological status of rules of grammar”. In Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society, L. A. Sutton, C. Johnson, and R. Shields (eds), 230–251. Berkeley Linguistics Society, University of California, Berkeley. Pollock, J.-Y. 1989. “Verb Movement, Universal Grammar, and the Structure of IP”. Linguistic Inquiry 20: 365–424. Postal, P. M. 1969. “On so-called ‘pronouns’ in English”. In Modern Studies in English, D. A. Reibel and S. A. Schane (eds), 201–224. Englewood Cliffs: Prentice-Hall. Prince, A. and Smolensky, P. 1993. Optimality Theory: Constraint interaction in generative grammar. Unpublished manuscript, Rutgers University and University of Colorado. Rapp, I. 1997. Partizipien und semantische Struktur: Zu passivischen Konstruktionen mit dem 3. Status. Tübingen: Stauffenberg.

311

312

References

Riemsdijk, H. v. 1978. A Case Study in Syntactic Markedness. Dordrecht: Foris. Riemsdijk, H. v. 1990. “Functional Prepositions”. In Unity in Diversity, H. Pinkster and I. Genée (eds), 229–241. Dordrecht: Foris. Riemsdijk, H. v. 1998a. “Head Movement and Adjacency”. Natural Language and Linguistic Theory 16: 633–678. Riemsdijk, H. v. 1998b. “Categorial fetaure magnetism: The endocentricity and distribution of projections”. Journal of Comparative Germanic Linguistics 2: 1–48. Risch, G. 1994. Verbpräfigierung des Deutschen: Skalierungsverben mit über- und unter-. Ph.D. dissertation, University of Stuttgart. Ritter, E. 1988. “A Head Movement Approach to Construct-State Noun Phrases”. Linguistics 26: 909–929. Ritter, E. 1991. “Two Functional Categories in Noun Phrases: Evidence from Modern Hebrew”. Syntax and Semantics 25: 37–62. Rivet, A. 1999. “Rektionskomposita und Inkorporationstheorie”. Linguistische Berichte 179: 307–342. Rizzi, L. 1990. Relativized Minimality. Cambridge: MIT Press. Roberts, I. 1985. “Agreement Parameters and the Development of English Modal Auxiliaries”. Natural Language and Linguistic Theory 3: 21–58. Roberts, I. 1991. “Excorporation and Minimality”. Linguistic Inquiry 22: 209–218. Roberts, I. 1997. “Restructuring, Head Movement, and Locality”. Linguistic Inquiry 28: 423–460. Ruwet, N. 1991. Syntax and human experience. Chicago and London: University of Chicago Press. Sabel, J. 1996a. Restrukturierung und Lokalität [studia grammatica 42]. Berlin: Akademie Verlag. Sabel, J. 1996b. “Vacuous V-to-Infl in German”. Unpublished manuscript, University of Frankfurt. Sabel, J. 2000. “Das Verbstellungsproblem im Deutschen: Synchronie und Diachronie”. To appear in Deutsche Sprache. Schoorlemmer, M. 1995. Participial Passive and Aspect in Russian. Ph.D. dissertation, OTS, University of Utrecht. Selkirk, L. 1982. The Syntax of Words. Cambridge: MIT Press. Simpson, J. 1983. “Resultatives”. In Papers in Lexical-Functional Grammar, B. Levin, M. Rappaport, and A. Zaenen (eds), 143–157. Bloomington, Ind.: Indiana University Linguistics Club. Spencer, A. 1991. Morphological Theory. Oxford: Cambridge University Press. Stechow, A. v. 1995. “Lexical Decomposition in Syntax”. In Lexical Knowledge in the Organization of Language, U. Egli et al. (eds), 81–117. Amsterdam and Philadelphia: John Benjamins. Sternefeld, W. 1990. “Scrambling and Minimality”. In Scrambling and Barriers, : G. Grewendorf and W. Sternefeld (eds), 143–198. Amsterdam and Philadelphia: John Benjamins. Stiebels, B. 1996. Lexikalische Argumente und Adjunkte [studia grammatica 39]. Berlin: Akademie Verlag. Stiebels, B. and Wunderlich, D. 1994. “Morphology feeds syntax: the case of particle verbs”. Linguistics 32: 913–968.

References

Stowell, T. 1981. Origins of Phrase Structure. Ph.D. dissertation, MIT, Cambridge. Stowell, T. 1989. “Subjects, Specifiers, and X-bar Theory”. In Alternative conceptions of phrase structure, M. Baltin and A. Kroch (eds), 232–262. Chicago and London: The University of Chicago Press. Stowell, T. 1991. “Determiners in NP and DP”. In Views on Phrase Structure, K. Leffel and D. Bouchard (eds), 37–56. Amsterdam: Kluwer. Svenonius, P. 1996. The Verb-Particle Alternation in the Scandinavian Languages. Unpublished manuscript, University of Tromsoe. Talmy, L. 1985. “Lexicalization Patterns: Semantic structure in lexical forms”. In Language Typology and Syntactic Description vol. III, T. Shopen (ed.), 57–149. Cambridge: Cambridge University Press. Taraldsen, K. T. 1983. Parametric Variation in Phrase Structure. Ph.D. Dissertation, University of Tromsoe. Taraldsen, K. T. 1990. “D-projections and N-projections in Norwegian”. In Grammar in progress, M. Nespor and J. Mascaró (eds), 419–431. Dordrecht: Foris. Toman, J. 1983. Wortsyntax. Tübingen: Niemeyer. Travis, L. 1984. Parametres and effects of word order variation. Ph.D. dissertation, MIT, Cambridge. Truckenbrodt, H. 1995. “Extraposition from NP and Prosodic Structure”, NELS 25: 503–517. Wasow, T., Nunberg, G. and Sag, I. 1984. “Idioms: An interim report”. In Proceedings of the 13th International Congress of Linguists, Tokyo, S. Hattori and K. Inoue (eds), 102–115. The Hague: CIPL. Williams, E. 1978. Notes on Lexical Theory. Unpublished manuscript, University of Massachusetts, Amherst. Williams, E. 1981. “On the Notions ‘Lexically Related’ and ‘Head of a Word’”. Linguistic Inquiry 12: 245–274. Wunderlich, D. 1995. “Argument Extension by Lexical Adjunction”. In The Blaubeuren papers, F. Hamm, J. Kolb, and A. von Stechow (eds), 434–474. Seminar für Sprachwissenschaft, University of Tübingen. Wurmbrand, S. 1998. “Heads or Phrases? Particles in particular”. In Phonology and Morphology of the Germanic Languages, W. Kehrein and R. Wiese (eds), 267–295. Tübingen: Niemeyer. Wurmbrand, S. 2000a. Infinitives. Restructuring and Clause Structure. Berlin and New York: de Gruyter. Wurmbrand, S. 2000b. The structure(s) of particle verbs. Unpublished manuscript, University of Montreal. Zeller, J. 1996. On verbal modifiers and their thematic properties. Unpublished manuscript, University of Massachusetts, Amherst. Zeller, J. 1997a. “Against overt particle incorporation”. Penn Working Papers in Linguistics 4/2 [Proceedings of the 21st Annual Penn Linguistics Colloquium], 291–307, University of Pennsylvania. Zeller, J. 1997b. Particle verbs and a theory of late lexical insertion. Unpublished manuscript, University of Frankfurt.

313

314 References

Zeller, J. 2000. Particle verbs and the conditions of projection. Unpublished manuscript, University of Frankfurt. Zeller, J. 2001. “Prefixes as Transitivizers”. In Structural aspects of semantically complex verbs, N. Dehé and A. Wanner (eds), 1–34. Frankfurt/Berlin/New York: Peter Lang. Zeller, J. (forthcoming a), “Lexical particles, semi-lexical postpositions”, to appear in Semilexical categories, N. Corver and H. van Riemsdijk (eds), Berlin and New York: de Gruyter. Zeller, J. (forthcoming b), “Particle verbs are heads and phrases”, to appear in Verb-particle explorations, N. Dehé, R. Jackendoff, A. McIntyre and S. Urban (eds) (working title; order of editors could change), Berlin and New York: De Gruyter. Zwart, J. W. 1994. “Dutch is head-initial”. The Linguistic Review 11: 377–406.

Name Index

A Abney 112, 113, 135 Abraham 57, 67, 118, 120, 142, 163, 230, 280 Ackema 41, 61 Ackerman 75, 213, 258 Åfarli 105, 144, 145, 285, 286, 290, 291, 293, 294, 295 Alexiadou 114 Anderson 35 Archangeli 98 Aronoff 199 B Baayen 158 Baker 41, 42, 44, 73, 75, 125 Bayer 92, 93 Bennis 82, 89 den Besten 271, 272, 275 Bhatt 114 Bierwisch 275 Bobaljik 29, 31 Booij 72, 82, 131, 132, 211, 213, 215, 216, 217, 240 Borer 15, 44, 45, 46, 47, 114, 123, 128, 250 Botha 243 Bresnan 43, 163 Brody 24, 25, 26, 28 Broekhuis 272 Büring 58, 81, 93, 164 Burzio 111, 216 C Carrier 59

Chomsky 17, 23, 24, 25, 26, 27, 36, 37, 43, 68, 109, 110, 111, 115, 117, 136, 221, 285, 297, 298 Cinque 91 Corver 114, 115 D Dewell 57, 58, 116, 138 Di Sciullo 22, 33, 34, 35, 42, 43, 105, 161, 241, 246, 288, 293 den Dikken 53, 100, 101, 105, 154, 277, 279, 281, 285, 287 Dowty 145 Drach 52 E Edmondson 271, 275 Emonds 105, 154, 162, 232, 274 Evers 55, 104, 271 F Fanselow 218 Fraser 86 G Gamon 157 Gee 70 Giusti 70 Grewendorf 52, 53, 70, 71, 74, 75, 89, 90, 95, 96, 216, 245 Grimshaw 47, 91, 92, 109, 123, 232 Groos 242, 257, 279, 280, 281 Guasti 44 H Haiden 52, 124, 141

316 Name Index

Hale 111 Halle 15, 24, 29, 30, 31, 36, 197, 198, 200, 209, 259 Halpern 260 Hamm 245 Hankamer 83 Hartmann 81, 83 Haverkort 99 Hazout 114 Hinterhölzl 92, 278 Höhle 39, 86, 245 Hoeksema 89, 99, 142 Hoekstra 52, 53, 59, 60, 117, 216 Horn 42 Hornstein 294 van Hout 157 Hundsnurscher 180 J Jackendoff 14, 15, 16, 17, 18, 20, 21, 22, 23, 24, 31, 33, 39, 40, 59, 116, 117, 121, 127, 137, 143, 145, 154, 158, 163, 166, 167, 168, 182, 183, 197, 198, 199, 202, 293 Johnson 51, 73, 86, 88, 105, 285, 287 K Kaufmann 59 Kayne 53, 74, 102, 285, 295 Keyser 111 Kluge 202 Koopman 47, 52, 55, 100, 114, 116, 117, 121, 122, 123, 126, 141, 276, 279, 280, 281 Koster 51, 272, 285 Kratzer 60, 73, 111, 124, 157, 164, 252, 253, 264 Krifka 179 Kroch 134 Kühnhold 180 L Langendoen 98 Lansu 52, 53 Lapointe 43

Larson 102, 111 285, 286 Le Roux 51, 86, 93, 273, 300, 301 Levelt 21 Levin 59, 73, 144, 216 Lieber 39, 41, 42, 61, 82, 158, 188, 244, 245, 246, 247 Löbel 114 Longobardi 114, 128, 129, 131, 132, 133, 135 Lüdeling 52, 59, 69, 89, 92, 95, 96, 162, 175, 176, 211, 242, 243, 266 M Marantz 14, 15, 18, 19, 21, 22, 23, 24, 29, 30, 31, 32, 33, 36, 40, 166, 167, 168, 188, 192, 197, 198, 200, 203, 208, 209, 258, 259, 260 McCarthy 34 McIntyre 51, 61, 65, 66, 67, 78, 116, 118, 120, 137, 138, 139, 140, 141, 152, 180, 211, 229, 233, 234, 242, 243 Mchombo 43 Mester 91, 92 Miller 127 Mulder 52, 102, 117, 216 N Neeleman 44, 51, 61, 62, 63, 64, 72, 82, 86, 88, 90, 93, 100, 101, 103, 117, 240, 264, 272, 282, 283, 286, 290, 291, 293, 294 Nespor 34 Nunberg 163 O Olsen 51, 118, 142, 152, 154, 184, 185, 228, 229, 231, 236, 281 Ouhalla 113 P Parsons 60 Paulissen 82 Pesetsky 18, 19, 28, 173, 272 Pinker 293

Name Index

Pollock 70, 110 Postal 135 Prince 34, 98, 293 R Randall 59 Rapp 252 Rappaport 59, 73, 144, 216 Reutter 157 van Riemsdijk 52, 55, 56, 68, 81, 104, 116, 117, 118, 119, 121, 123, 154, 170, 171, 232, 271, 272, 277, 278, 279, 294 Risch 180 Ritter 114 Rivet 250 Rizzi 36 Roberts 44, 73, 80, 91, 210, 261, 287 Rutten 272 Ruwet 163 S Sabel 70, 99 Sag 163 Santorini 134 Schoorlemmer 47, 74 Selkirk 42, 62, 223 Simpson 59 Smolensky 98 Spencer 125, 198 von Stechow 60, 145 Sternefeld 70, 245 Stiebels 42, 51, 64, 65, 69, 73, 76, 77, 78, 79, 82, 90, 92, 93, 94, 95, 101, 102, 103, 129, 152, 154, 155, 156, 157, 160, 172, 173, 174, 175, 176, 179, 180, 206, 211, 240, 241, 242, 264 Stowell 53, 114, 128, 131, 135 Svenonius 105, 106, 116, 172, 180, 285, 287, 289, 290, 291, 292, 295 T Talmy 116 Taraldsen 105, 114, 285 Toman 82, 244, 246

Travis 74 Truckenbrodt 82 V Vogel 34 W Wasow 163 Webelhuth 75, 213, 258 Weerman 51, 63, 64, 72, 90, 93, 103, 240, 264 Weinberg 294 Westerduin 52, 53 Wilder 114 Williams 22, 33, 34, 35, 42, 43, 105, 106, 161, 241, 246, 288, 293 Wunderlich 42, 51, 64, 65, 73, 76, 77, 78, 79, 82, 90, 101, 102, 103, 129, 154, 156, 157, 172, 174, 175, 176, 179, 240, 241, 242, 264 Wurmbrand 52, 53, 57, 58, 92, 93, 96, 98, 162, 274 Z Zeller 14, 52, 57, 58, 67, 72, 82, 92, 120, 157, 196, 201, 211, 213, 226, 228, 229, 231, 232, 261, 262, 264, 278, 295, 299, 300 Zonnefeld 82 Zwart 81

317

Subject Index

A accentuation –of certain prefixes 65, 67 –of the particle 58, 65, 67, 287 adjacency –condition 102f. morphological – 210ff., 231, 246, 255ff., 273, 292 structural – 36, 49, 109, 125f., 127, 148, 151, 166, 169f., 172f., 175, 186, 189ff., 202ff., 209ff., 230ff., 248f., 251f., 254, 255ff., 273, 276ff. adjective 45ff., 110, 114ff., 130f., 134, 143ff., 240f., 244, 248, 252ff., 263ff., 282f., 288ff. adverb 96, 99, 105, 132, 163f., 222, 252f., 281ff., 286, 294 affix 39f., 45, 47, 49, 62, 69, 149, 158, 169f., 196, 207f., 210, 240ff., 275, 276, 279, 289, 299 Afrikaans 244, 301 agreement 110, 113ff., 134, 144f., 294f. allomorphy 31, 33, 196ff.

lexical 109ff. causative 45 chain 25f., 38, 114, 125, 164f., 186, 231, 261,see also movement characteristic paradigm 285ff., 298, 300f. circumpositional phrase 118ff., 137, 229, 232, 276 pleonastic – 119f., 226, 229, 231, 235 clitic 40, 79, 136, 259f. complex predicate formation 157 complexity constraint 63ff., 103 compound 34, 39, 41, 76ff., 87f., 244ff., 282, 298 conceptual structure 15ff., 24, 33, 48, 107, 110, 154, 167, 169, 174, 177, 193, 197, 207 contrastive stress 67 conversion 212ff. copy theorysee movement correspondence rule 17ff., 27f., 38, 48f., 98, 145, 164, 176 covert movementsee movement

B backformations 76ff. binary branching 102 bracketing paradox 18f., 28, 241 Burzio’s generalization 111

D Danish 106, 287, 289 determiner phrase DP 113f., 123, 127ff., 167f., 191, 222ff., 246 Distributed Morphologysee morphology

C case 47, 111f., 115, 117, 119f., 124ff., 133f., 141, 144, 154f., 189ff., 218ff., 232, 274, 285 categories functional 27, 109ff.

E encyclopedia 32f., 166, 168 excorporation 44, 68 extended projection 109ff., 127, 145, 224f., 232, 274, 299

320 Subject Index

extraposition 80ff., 107, 272f., 278 F focus contrastive – 93ff. French 34, 113, 115, 136, 144, 288 functional categoriessee categories functional prepositional phrase FPPrep 116ff., 135ff., 153f., 178, 180, 184, 190, 216f., 218, 223, 224, 225ff., 277 functional projectionsee functional categories functional structure 47, 109ff., 127ff., 154f., 169, 189ff., 214f., 219, 221, 224ff., 263, 277, 294f., 297, 299,see also functional categories G Gapping 83ff. ground 116 H head –government 36, 170, 210 –movement 26, 30, 40ff., 45, 47, 70ff., 82f., 114f., 125, 149, 163ff., 186, 222, 248ff., 261f., 272ff. –constraint 74 –of a word 42, 63, 103, 226, 231, see also Right-hand head rule Hebrew 45 HMC, see Head Movement Constraint h-postposition 118ff., 135ff., 178, 225ff., 266ff., 279f. h-verb 138ff., 178, 225ff., 267ff. I idiom 21ff., 34, 98, 161ff., 166ff., 185f., 188f., 202f., 208f., 266 inchoative 45 incorporation 41, 45, 125, 149, 224f., 251, 261, 281, 283 abstract –see covert movement

–approach 52ff., 68f., 74f., 86, 101, see also head movement infinitive 70ff., 272ff. in situ approach 52ff., 72ff., 80ff., 88, 106 inflection 30f., 41, 82, 98, 125, 257, 278, 295 inseparable prefix verbs,see prefix verbs interface 16ff, 24ff. 29ff., 48, 176, 193, 207, 258, 260 –level 17ff, 24ff. 32f., 53, 117, 125, 142, 149, 152, 165, 248, –module 17, 20, 176 Italian 91 J Japanese 92 L landmark 116, 138ff.,see also reference object –flexibility 180f., 199 Landmark Referentiality Generalisation 138, 141 LCSsee lexical conceptual structure lemma 21, 23, 162 lexical –categoriessee categories –conceptual structure 8, 20, 33, 39, 152, 154, 157f., 164f., 166f., 169, 173, 175ff., 183, 186, 190, 192, 203, 208, 230, 245 –entrysee lexical item –form 21, 23, 32, 162 –index 20ff., 152, 162, 164f., 169, 183, 245f. –insertion 15, 22, 31, 161 late – 29 –integrity principle 43f., 46f., 58, 61, 80, 87 –item 31, 123, 160ff., 183, 190ff., 196ff., 212, 221, 255ff., 273 as correspondence rule 20

Subject Index

–licensing 20f., 29, 69, 161, 164, 169, 185f., 189ff., 193, 205, 208, 214, 245, 255 –logical formsee LLF –phonological structure 20, 26ff., 32, 39, 72, 82, 164f., 167ff., 183, 185, 197, 245, 261, 274 –selection 172, 292 –syntactic structure 20, 32, 39, 69, 165, 168f., 197, 245f., 255 lexicalist models of grammar 13, 22, 161, 176, 193, 266 lexicalist theoriessee lexicalist models of grammar LF 19, 27ff., 43, 52, 261 light verb (phrase) v(P) 111f., 221ff.,see also Voice listeme 34, 161 LLF 25 local domainsee locality locality 26, 33, 36, 49, 109, 125f., 148f., 164, 169ff., 189, 192f., 195ff., 210ff., 230ff., 256ff., 295, 297 locality domainsee locality definition of – 215, 230, 256 logical formsee LF LPSsee lexical phonological structure LSSsee lexical syntactic structure M Merge 32, 37ff., 45ff., 52, 58, 68, 73, 112, 117, 249, 253, 261f., 281 modifier 46f., 48, 100f., 105, 114, 130, 163f., 281ff., 286f., 294,see also adverb and adjective morpheme 31, 197, 242, 246 morphological –adjacencysee adjacency –approach 51ff., 61ff., 75, 86, 92, 101f., 172, 176, 192, 297 –merger 259 –object 33ff., 38, 41, 149, 172, 192, 198, 210, 255ff., 275, 288ff., 297f. –structure 29ff.

–subcategorization 40, 49, 169, 245f., 248, 255, 291 morphology –and lexical insertion 13, derivational – 41 distributed – 28ff., 41, 198 inflectional –see inflection parallel – 44ff., 261f. Move 37ff., 46ff., 52, 58, 68, 112, 117, 223, 248ff., 261f., 275 movement 42,see also chains,see also Move –and chains 25f. –and derivations 26f. copy thory of – 25f., 44, 98, 164, 205 covert – 24, 26, 27f., 125, 128, 164, 261f., 274 head –see head movement N negation 43, 91, 131, 132, 222, 253, 259f. nominals 87, 91f., 240ff., 264, 268, 287, 290 process – 47ff., 170, 249ff. result – 47ff., 249ff. No Phrase Constraint 243ff. Norwegian 106, 144, 145, 285f., 289, 291ff. O Oneida 75 Optimality Theory 98 P P-prefix 57f.,see also prefix verbs Papago 259 Parallel Morphologysee morphology participle adjectival – 73, 76, 252ff., 265, 276, 289f. verbal – 73, 265, 271, 275, 276, 289f. particle –as Event-concept 157f

321

322 Subject Index

–as intransitive preposition 154 –as preposition with special meaning 168ff. –as semantic affix 200 contextual condition on -s 169, 173, 183, 203 –definition 127, 148, 168 –shiftsee characteristic paradigm particle verb class-based analysis of -s 176 –definition 148 passive 73, 112, 134, 163f., 198, 252ff., 289f. Path 117, 121f., 124, 137f., 154, 158, 181, 189, 219, 230, 234 percolation 42, 223ff. PF 26, 29, 259,see also phonological structure phonological structure 15ff., 32, 66, 75, 78f., 106, 107, 169, 201, 207, 247, 251, 254, 259 phonologysee phonological structure Place 117, 121f., 124, 137, 219 pleonastic circumpositional phrasesee circumpositional phrase postposition 56, 117ff., 136ff., 226ff., 276, 278ff.,see also h-postposition prefixsee affix prefix verbs 56ff., 64, 72, 75, 82f., 84, 86, 213ff., 217f. prepositional phrase PP 22, 47, 59f., 67, 72f., 81, 83, 105, 116ff., 130, 135ff., 153, 156, 162, 170f., 178, 190, 222, 229, 231f., 277 directional – 121ff. , 189ff., 199 locative – 121ff., 189ff., 199 prepositional proform 137ff., 178, 225, 226, 228f. Procrastinate 27 productivity 188 projection 38f., 45f., 298ff. pronominal adverbsee prepositional proform pronoun 128, 135ff.

relative – 142f. R reanalysis 77f., 90, 255ff., 272ff., 300f. Principle of – 255, 272, 299 Amendment for Principle of – 257, 272, 299 reference object 116f., 137ff., 154f., 158, 160, 177, 180f., 189f., 218ff., 228ff., 231 referentiality 127ff., 154, 228f. representation –and derivation 26 level of – 25 mixed – 16 representational modularity 16, 20 resultatives 59ff., 72f., 80, 90, 101ff., 143ff., 148, 263ff., 282f., 290ff., 299 Right-hand Head Rule 42, 71, 106, 226, 275, 288, 290 Right Node Raising 85 S semi-lexicality 232ff., 269 semiproductive rule 182ff., 199f., 212, 234 Small Clause 53, 60, 74f., 117, 128 Spanish 199 spell-out 24 Stiebels and Wunderlich’s generalization 175 stress hypothesis 65ff. structural adjacencysee adjacency suffixsee affix suppletion 31, 200ff. Swedish 287, 290 syntactic approach 52ff., 102, 127, 129, 165, 176, 214, 243, 285, 297 syntactic atom 22, 33ff. syntactic structure 16, 28, 48, 53, 78f., 106, 107, 125, 142, 149, 152, 162f., 167ff., 183, 197, 207, 248, 255, 259, 273, 292

Subject Index 323

T Tagalog 246 Tamil 247 terminal nodes 30ff., 35 third construction 272 token 128, 137, 140f., 143, 146 topicalization 79, 81, 86, 88ff., 101, 119, 205, 259 tripartite parallel architecture 16, 28, 48, 78, 98, 107 Turkish 198 type 128, 130, 137, 140f., 143 U unaccusative prepositions particles as – 141 unaccusativity 111f., 136, 144, 216 unification 21, 69, 152, 165, 173, 246, 255f., 260, 273, 277, 292

V verb –Raising 54f., 104, 170, 271ff., 300ff. – -particle alternationsee characteristic paradigm –second 54ff., 77f., 79, 97f., 163, 261 vocabulary –insertion 29ff., 83, 110, 198, 201 –item 30f., 71, 125, 164f., 167, 197ff., 207, 260, 261 Voice 111, 124, 157 W word 33ff., 38ff., 149, 197, 200, 210, 214, 225, 240ff., –formation 53, 223, 240ff., 274f., 279, 287ff. phonological – 34, 260

In the series LINGUISTIK AKTUELL/LINGUISTICS TODAY (LA) the following titles have been published thus far, or are scheduled for publication: 1. KLAPPENBACH, Ruth (1911-1977): Studien zur Modernen Deutschen Lexikographie. Auswahl aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene Malige-Klappenbach. 1980. 2. EHLICH, Konrad & Jochen REHBEIN: Augenkommunikation. Methodenreflexion und Beispielanalyse. 1982. 3. ABRAHAM, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983. 4. ABRAHAM, Werner & Sjaak De MEIJ (eds): Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks, Groningen, 1984. 1986. 5. GREWENDORF, Günther and Wolfgang STERNEFELD (eds): Scrambling and Barriers. 1990. 6. BHATT, Christa, Elisabeth LÖBEL and Claudia SCHMIDT (eds): Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences. 1989. 7. ÅFARLI, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. 8. FANSELOW, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. 9. GELDEREN, Elly van: The Rise of Functional Categories. 1993. 10. CINQUE, Guglielmo and Guiliana GIUSTI (eds): Advances in Roumanian Linguistics. 1995. 11. LUTZ, Uli and Jürgen PAFEL (eds): On Extraction and Extraposition in German. 1995. 12. ABRAHAM, W., S. EPSTEIN, H. THRÁINSSON and C.J.W. ZWART (eds): Minimal Ideas. Linguistic studies in the minimalist framework. 1996. 13. ALEXIADOU Artemis and T. Alan HALL (eds): Studies on Universal Grammar and Typological Variation. 1997. 14. ANAGNOSTOPOULOU, Elena, Henk VAN RIEMSDIJK and Frans ZWARTS (eds): Materials on Left Dislocation. 1997. 15. ROHRBACHER, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and pro-drop. 1999. 16. LIU, FENG-HSI: Scope and Specificity. 1997. 17. BEERMAN, Dorothee, David LEBLANC and Henk van RIEMSDIJK (eds): Rightward Movement. 1997. 18. ALEXIADOU, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. 19. JOSEFSSON, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. 20. LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. 21. KLEIN, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. 22. ALEXIADOU, Artemis and Chris WILDER (eds): Possessors, Predicates and Movement in the Determiner Phrase. 1998. 23. GIANNAKIDOU, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. 24. REBUSCHI, Georges and Laurice TULLER (eds): The Grammar of Focus. 1999. 25. FELSER, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999. 26. ACKEMA, Peter: Issues in Morphosyntax. 1999.

° 27. RUZICKA, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. 28. HERMANS, Ben and Marc van OOSTENDORP (eds.): The Derivational Residue in Phonological Optimality Theory. 1999. 29. MIYAMOTO, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 1999. 30. BEUKEMA, Frits and Marcel den DIKKEN (eds.): Clitic Phenomena in European Languages. 2000. 31. SVENONIUS, Peter (ed.): The Derivation of VO and OV. 2000. 32. ALEXIADOU, Artemis, Paul LAW, André MEINUNGER and Chris WILDER (eds.): The Syntax of Relative Clauses. 2000. 33. PUSKÁS, Genoveva: Word Order in Hungarian. The syntax of È-positions. 2000. 34. REULAND, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. 35. HRÓARSDÓTTIR, Thorbjörg. Word Order Change in Icelandic. From OV to VO. 2000. 36. GERLACH, Birgit and Janet GRIJZENHOUT (eds.): Clitics in Phonology, Morphology and Syntax. 2000. 37. LUTZ, Uli, Gereon MÜLLER and Arnim von STECHOW (eds.): Wh-Scope Marking. 2000. 38. MEINUNGER, André: Syntactic Aspects of Topic and Comment. 2000. 39. GELDEREN, Elly van: A History of English Reflexive Pronouns. Person, ‘‘Self’’, and Interpretability. 2000. 40. HOEKSEMA, Jack, Hotze RULLMANN, Victor SANCHEZ-VALENCIA and Ton van der WOUDEN (eds.): Perspectives on Negation and Polarity Items. 2001. 41. ZELLER, Jochen : Particle Verbs and Local Domains. 2001. 42. ALEXIADOU, Artemis : Functional Structure in Nominals. Nominalization and ergativity. 2001. 43. FEATHERSTON, Sam: Empty Categories in Sentence Processing. 2001. 44. TAYLAN, Eser E. (ed.): The Verb in Turkish. n.y.p. 45. ABRAHAM, Werner and C. Jan-Wouter ZWART (eds.): Issues in Formal German(ic) Typology. n.y.p 46. PANAGIOTIDIS, Phoevos: Pronouns, clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. n.y.p. 47. BARBIERS, Sjef, Frits BEUKEMA and Wim van der WURFF (eds.): Modality and its Interaction with the Verbal System. n.y.p.

Particle Verbs and Local Domains (Linguistik Aktuell Linguistics Today, LA 41)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

Experimental Pragmatics Semantics (Linguistik Aktuell Linguistics Today)

Multiple Wh-Fronting (Linguistik Aktuell Linguistics Today)

Parentheticals (Linguistik Aktuell Linguistics Today, 106)

Argument Structure (Linguistik Aktuell Linguistics Today)

Phase Theory (Linguistik Aktuell Linguistics Today)

Causatives in Minimalism (Linguistik Aktuell Linguistics Today)

Cyclical Change (Linguistik Aktuell Linguistics Today)

Grammaticalization As Economy (Linguistik Aktuell Linguistics Today, LA 71)

Current Issues in Generative Hebrew Linguistics (Linguistik Aktuell Linguistics Today)

Balkan Syntax and Semantics (Linguistik Aktuell Linguistics Today)

Adverbials and the Phase Model (Linguistik Aktuell Linguistics Today, 177)

Perspectives on Negation and Polarity Items (Linguistik Aktuell Linguistics Today)

Copular Clauses: Specification, Predication And Equation (Linguistik Aktuell Linguistics Today)

Scrambling and the Survive Principle (Linguistik Aktuell Linguistics Today)

Syntatic Aspects of Topic and Comment (Linguistik Aktuell Linguistics Today)

Continuity and Change in Grammar (Linguistik Aktuell Linguistics Today)

The Minimalist Syntax of Defective Domains: Gerunds and Infinitives (Linguistik Aktuell Linguistics Today)

Quantifier Scope in German (Linguistik Aktuell Linguistics Today)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

The Derivation of Anaphoric Relations (Linguistik Aktuell Linguistics Today)

The Emergence of Order in Syntax (Linguistik Aktuell Linguistics Today)

Theoretical Approaches to Universals (Linguistik Aktuell Linguistics Today)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Movement Theory of Control (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Agreement Systems (Linguistik Aktuell Linguistics Today, Volume 92)

Particle Verbs and Local Domains (Linguistik Aktuell Linguistics Today, LA 41)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

Experimental Pragmatics Semantics (Linguistik Aktuell Linguistics Today)

Multiple Wh-Fronting (Linguistik Aktuell Linguistics Today)

Parentheticals (Linguistik Aktuell Linguistics Today, 106)

Argument Structure (Linguistik Aktuell Linguistics Today)

Phase Theory (Linguistik Aktuell Linguistics Today)

Causatives in Minimalism (Linguistik Aktuell Linguistics Today)

Cyclical Change (Linguistik Aktuell Linguistics Today)

Grammaticalization As Economy (Linguistik Aktuell Linguistics Today, LA 71)

Current Issues in Generative Hebrew Linguistics (Linguistik Aktuell Linguistics Today)

Balkan Syntax and Semantics (Linguistik Aktuell Linguistics Today)

Adverbials and the Phase Model (Linguistik Aktuell Linguistics Today, 177)

Perspectives on Negation and Polarity Items (Linguistik Aktuell Linguistics Today)

Copular Clauses: Specification, Predication And Equation (Linguistik Aktuell Linguistics Today)

Scrambling and the Survive Principle (Linguistik Aktuell Linguistics Today)

Syntatic Aspects of Topic and Comment (Linguistik Aktuell Linguistics Today)

Continuity and Change in Grammar (Linguistik Aktuell Linguistics Today)

The Minimalist Syntax of Defective Domains: Gerunds and Infinitives (Linguistik Aktuell Linguistics Today)

Quantifier Scope in German (Linguistik Aktuell Linguistics Today)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

The Derivation of Anaphoric Relations (Linguistik Aktuell Linguistics Today)

The Emergence of Order in Syntax (Linguistik Aktuell Linguistics Today)

Theoretical Approaches to Universals (Linguistik Aktuell Linguistics Today)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Movement Theory of Control (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Agreement Systems (Linguistik Aktuell Linguistics Today, Volume 92)

Recommend Documents