Ergativity in German [Reprint 2012 ed.] 9783110859256, 9783110130652

Citation preview

Ergativity in German

Studies in Generative Grammar The goal of this series is to publish those texts that are representative of recent advances in the theory of formal grammar. Too many studies do not reach the public they deserve because of the depth and detail that make them unsuitable for publication in article form. W e hope that the present series will make these studies available to a wider audience than has hitherto been possible. Editors: Jan Köster Henk van Riemsdijk Other books, still available in this series: 1. Wim Zonneveld A Formal Theory of Exceptions in Generative Phonology

23. S.J. Keyser/W. O'Neil Rule Generalization and Optionality in Language Change

3. Geert Booij Dutch Morphology

24. Julia Horvath FOCUS in the Theory of Grammar and the Syntax of Hungarian

7. Anneke Neijt Gapping 9. Noam Chomsky Lectures on Government and Binding 10. Robert May and Jan Koster (eds.) Levels of Syntactic Representation 13. Hagit Borer Parametric Syntax 14. Denis Bouchard On the Content of Empty Categories 15. Hilda Koopman The Syntax of Verbs 16. Richard S. Kayne Connectedness and Binary Branching 17. Jerzy Rubach Cyclic and Lexical Phonology: the structure of Polish 18. Sergio Scalise Generative Morphology 19. Joseph E. Emonds A Unified Theory of Syntactic Categories 20. Gabriella Hermon Syntactic Modularity 21. Jindfich Toman Studies on German Grammar

25. Pieter Muysken and Henk van Riemsdijk Features and Projections 26. Joseph Aoun Generalized Binding. The Syntax and Logical Form of Wh-interrogatives 27. Ivonne Bordelois, Heles Contreras and Karen Zagona Generative Studies in Spanish Syntax 28. Marina Nespor and Irene Vogel Prosodie Phonology 29. Takashi Imai and Mamoru Saito (eds.) Issues in Japanese Linguistics 30. Jan Köster Domains and Dynasties. The Radical Autonomy of Syntax. 31. Maria-Luisa Zubizarreta Levels of Representation in the Lexicon and the Syntax 32. Lars Hellan Anaphora in Norwegian and the Theory of Grammar 33. Jon Ortiz de Urbina Parameters in the Grammar of Basque 34. László Marácz and Pieter Muysken Configurationality

G ü n t h e r

G r e w e n d o r f

Ergativity in German

ν

1989 FORIS PUBLICATIONS Dordrecht - Holland/Providence - U.S.A.

Published

by:

Foris Publications Holland P.O. Box 509 3300 AM Dordrecht, The Netherlands Distributor

for the U.S.A. and

Canada:

Foris Publications USA, Inc. P.O. Box 5904 Providence RI 02903 U.S.A. Distributor

for

Japan:

Sanseido Book Store, Ltd. 1-1, Kanda-jimbocho-cho Chiyoda-ku Tokyo 101, Japan

ISBN 90 6765 426 4 ® 1989 Foris Publications - Dordrecht No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission from the copyright owner. Printed in The Netherlands by ICG Printing, Dordrecht.

For Dorle and Simon

Contents

Preface

ix

Chapter 1. The Ergative Hypothesis

1

Chapter 2. The Justification of the Ergative Hypothesis: Its Syntactic Predictions 2.1. #e-Cliticization and Auxiliary Selection in Italian 2.2. Auxiliary Selection in German 2.3. Participles as Attributes 2.4. Impersonal Passive - Agentive von (by)-PP 2.5. Derivational Morphology: -er and -ling Nominalizations 2.6. Topicalization of Subject + Participle II 2.7. Discontinuous Phrases 2.8. Extraction from NP Subjects 2.9. Extraction from Subject Sentences 2.10. Extraction and Extraposition 2.11. Extraction of Subjects 2.12. Dative 2.13. Control 2.14. Focus 2.15. Reflexives in A.c.I.-Constructions Chapter 3. The Inverted Subject Position and the Case Assignment Problem 3.1. The First Two Versions of the 'pro-drop' Parameter 3.2. pro and the Third and Fourth Versions of the pro-àrop Parameter 3.3. Case Assignment to "Inverted" Subjects in German 3.4. pro Subjects in German 3.5. The Definiteness Effect

5 5 8 18 21 21 23 26 34 48 65 87 91 94 96 97

109 109 123 134 145 164

Chapter 4. Ergative Verbs and "Theme Verbs" 4.1. The Ergativity Tests 4.2. "Subject Inversion" and Non-Ergative Verbs 4.3. "Subject Inversion" and Focus

177 177 188 203

References

221

Preface

This study is concerned with various descriptive and theoretical problems which arise in the study of German syntax. Its primary concern is to show that there are ergative structures in German the analysis of which leads to the conclusion that this language, like other Germanic languages, can to a certain extent be viewed as a pro-drop language. The main objectives of this book can be summarized as follows. Firstly, it will be shown that many descriptive generalizations of German sentence structures that have gone unnoticed up to now can be accounted for in terms of Burzio's well known "ergative hypothesis". Secondly, an investigation will be undertaken into what consequences the existence of subjectless constructions in German has for the theory of pro-drop and related phenomena which are usually associated empirically and conceptually with the phenomenon of pro-drop (as, for example, the absence of *[that-t] effects, free subject inversion, the definiteness effect). A crucial claim of the book will be that there exists in German an empty expletive pronominal subject. Proving German to be what has been called a "semi-pro-drop language", this study not only argues against a view that has been characterized as "pro pro-drop drop" it also differs quite substantially from approaches that abandon the extended pròjéction principle and argue for the existence of subjectless clauses. As should be clear from the previous remarks, the theoretical framework adopted in this study is the theory, developed in Chomsky (1981), (1982), (1986), that has become misleadingly known as the Government-BindingTheory. This label is a misnomer because, as Chomsky himself pointed out during his lectures in Brussels in 1986, the concepts of "Government" and "Binding" are given undue prominence in a modular theory of universal grammar that is thought to consist of more subsystems of principles than just the government and binding theories. Throughout this study, I will assume that the reader is familiar with the central concepts of this approach, such as the notion of Universal Grammar, the distinction between principles and parameters, the theories of government, binding, Case, and control, the theta-theory and the bounding theory. Within the relevant literature on the Government and Binding framework, there are two studies upon which the following investigations are primarily based. More than anything else, the content of the following

χ

Preface

Chapters was inspired by the important work of Luigi Burzio on intransitive verbs in Italian. Similarly, what I have to say on pro-drop owes much to Luigi Rizzi's work on empty subjects and objects in Italian. Chapter 1 introduces the so-called "unaccusative hypothesis" which was originally proposed by Perlmutter. Using the name "ergative hypothesis", Perlmutter's analysis has been elaborated upon by Burzio within the framework of the Government and Binding Theory. It will be shown in this Chapter how this hypothesis interacts with certain principles of Universal Grammar. Chapter 2 presents empirical generalizations that can be accounted for in terms of the ergative hypothesis. After briefly illustrating the (wellknown) pertinent generalizations in Italian, I will investigate fourteen German constructions, arriving at generalizations on such phenomena as auxiliary selection, impersonal passive, topicalization, extraction processes, control, focus, and reflexivization. I hope to demonstrate that the predictions vis-à-vis the ergative hypothesis arising from these generalizations are all borne out, thus giving strong support for this hypothesis as well as for the existence of ergative verbs in German. Chapter 3 addresses theoretical problems of the pro-drop analysis. After a short survey of several versions of pro-drop parameters, certain phenomena will be analyzed that are thought to strongly correlate with the pro-drop property, such as the occurrence of postpositioned subjects at S-structure, the problem of assigning nominative Case into the VP, the occurrence of empty objects, and the definiteness effect. Moreover, independent arguments are presented which show that German is endowed with an empty expletive pronominal subject that shows up in ergative configurations as well as in constructions such as the impersonal passive. In Chapter 4, a class of verbs will be examined that on the surface appears to behave much like ergative verbs. With regard to the Italian counterparts of a subclass of these verbs (the so-called psych-verbs), Belletti/ Rizzi (1986) have tried to show that they enter the same D-structural configurations as do ergative verbs. Relying on the ergativity tests presented in Chapter 2, I want to establish that contrary to the view of Belletti/ Rizzi, the corresponding German verbs have no derived subjects, i.e. their subject positions are theta-positions. The peculiar behavior of these verbs with respect to the surface order of their subjects will be attributed to their particular theta-grids ("theme"-subjects) as well as to focus properties of German. Most of the work on this book was completed in 1986. An original version appeared in German at the end of 1985 as material for lectures held at the University of Frankfurt in 1986 and 1987. Since then, many colleagues and institutions have given me the opportunity to lecture on parts of this study at their universities or during informal talks and meetings.

Preface

xi

I would like to express my gratitude to all of them for the critical, instructive, and encouraging way they contributed to the contents of this book. Special thanks go to my friends and colleagues from the GGS (Generative Grammar South) Group. The regular meetings of this group of linguists from different universities in South Germany have not only led to the most stimulating discussions I have ever experienced, but were also characterized by a friendly and encouraging atmosphere of which I have fondest memories. This book would probably not have been possible without the many discussions I have had in this group with Peter Staudacher, Hubert Haider, Gisbert Fanselow, Sascha Felix, Tilman Höhle, Martin Prinzhorn, and many others. I am grateful to Gabereil Drachmann for giving me the opportunity to present most of this book in lectures at Salzburg University during the spring of 1987. I benefited greatly from the discussions I had with him and Peter Hummer as well as with the other participants of the class. During the 1987 GLOW-conference in Venice, I benefited greatly from discussions of certain ideas in the book with Adriana Belletti and Richard Kayne. Furthermore, I would like to thank Guglielmo Cinque for his comments and suggestions, and Frans Plank for pointing out to me some interesting facts on ergativity and passivization in German. I would also like to thank Bob Freidin for having given me the opportunity to give a lecture at Princeton and for interesting discussion with members of The Program in Linguistics there. I am also grateful to Martin Prinzhorn for organizing the Fourth Vienna Round Table in the autumn of 1987, where I presented parts of this study and profited from the comments of Guglielmo Cinque, Jan Koster, Henk van Riemsdijk, Alessandra Tomaselli, and others. In addition, I have been able to discuss some of my ideas at the University of Cologne with Jürgen Lenerz, Thilo Tappe, Heinz Vater, and the participants of their seminar. As I have already indicated and as should be clear from several Chapters, the contents of this study were inspired to a great extent by the work of Luigi Burzio and Luigi Rizzi. I should also like to mention that I had my first discussions on the work of these two linguists with Angelika Kratzer as we were both teaching at the Universities in Berlin during 1982-1984. I would like to thank her for this. The English translation was done in cooperation with Ron Feemster. I am grateful to Paul Davies for his corrections. Susan Olsen read the whole text and suggested many improvements on both content and style. This book would probably not exist without the wonderful cooperation I experienced from her. I would like to thank her very much. In preparing the final text, I received excellent editorial assistance from Gereon Müller at the University of Frankfurt. I also owe a lot to my friend and Frankfurt colleague Fritz Hamm.

xii

Preface

My final and deepest thanks go to Dorle and Simon, who provide me with the right atmosphere in which to deal with these obscure subjects of linguistic desire. Günther Grewendorf Frankfurt a.M. September 1988

Chapter 1

The Ergative Hypothesis

Language typologists generally distinguish between ergative and accusative languages. Although scholars do not agree about what an adequate explication of ergativity is (cf. Plank (1979)), characterizations of the following type (cf. Sasse (1978)) have won a following. In ergative languages, the subject of an intransitive verb takes the same case (the "nominative" or "absolutive") as the object of a transitive verb, while the subject of a transitive verb is assigned another case referred to as "locative", "instrumental", "genitive", "agentive", or "ergative", depending on which ergative language one is dealing with. This conception has sometimes been replaced by a characterization (cf. Plank (1979)) in which reference is made to construction-specific realizations of thematic roles. According to this view, accusative constructions are those in which a transitive agent appears as a transitive subject, and a transitive patient as a direct object. In ergative constructions, on the other hand, the transitive patient appears as subject, and the transitive agent is assigned a special oblique case. Now it is not the case that languages are necessarily homogeneous with respect to an ergative or accusative syntactic organization. For this reason, the attempt has been undertaken to determine the degree of ergativity in languages depending on the number of ergative vs. accusative constructions they contain. It should thus come as no surprise that ergative constructions can be found in clearly accusative languages and vice versa. In German, for instance, the intransitive verb sterben (to die) (as opposed to töten (to kill)) occurs in constructions in which the patient role is realized as a grammatical subject. Intransitive verbs whose surface subjects' grammatical behavior is similar in many respects to that of objects of transitive verbs have recently been referred to as ergative verbs (cf. Burzio (1981); the term probably stems originally from Sapir). This terminology was criticized as "helpless" (cf. Wunderlich (1985)). I would like to retain it nonetheless for the following two reasons. First of all, we must consider the fact that the native speaker of an accusative language can only conceive of an ergative language from the perspective of an accusative language. Because, as I mentioned earlier, the

2 Ergativity in German characterizations of ergative languages are - equally helplessly and probably inadequately - marked by this accusative view, it does not quite indicate "ignorance of the classical concepts of their discipline" (Wunderlich, p. 222) when scholars describe as "ergative" those intransitive verbs whose subjects - once again regarded from the perspective of an accusative language - exhibit properties corresponding to the properties of transitive objects. Secondly, Wunderlich's own suggestion - of regarding these verbs as "genuine" intransitive verbs having not an agentive but a theme subject (and therefore calling them "theme verbs") - is no better. For it is not clear that all verbs with a theme subject exhibit exactly the same syntactic properties characteristic of the verbs which, following Burzio, are referred to as "ergative verbs". As will be seen, ignoring these differences leads to serious inadequacies. But let us return to our topic. In his investigation of intransitive verbs in Italian, Burzio observes that for verbs occurring transitively as well as intransitively, the selection rules for the direct object in the transitive case correspond to those for the subject in the intransitive case. Consider the following examples: (1)

a.

b.

(2)

a. b.

Die Sonne hat den Schneemann geschmolzen the sun has the snowman melted 'The sun has melted the snowman' Der Schneeman ist geschmolzen the snowman is melted 'The snowman has melted' Peter hängt das Bild an die Wand 'Peter is hanging the picture on the wall' Das Bild hängt an der Wand 'The picture is hanging on the wall'

Burzio calls the intransitive variants of these pairs "ergative verbs". His thesis is not only that the surface subjects of these verbs are D-structure direct objects, but that there exist many more ergative verbs than just the intransitive variants of pairs such as those introduced above. Following the idea that ergative forms occur in constructions of the following type (3)

[s [NP e] tvp V NP]]

and that they thus have the basic form of passive constructions (whereby the "type" of empty element in the subject position should remain open), Burzio returns to the point of view advocated in relational grammar (cf.

The Ergative Hypothesis

3

Perlmutter (1978)). In this tradition, the analysis given above is known as the "unaccusative hypothesis". Burzio's "ergative verbs" are referred to in this theory as "unaccusative verbs" and the "true" intransitive verbs as "unergative verbs". The correspondence to D-structure passive constructions as well as the correlation in technical terminology with relational grammar indicates how the S-structures of sentences containing ergative verbs are to be derived. It is considered characteristic of ergative verbs that they do not assign the accusative Case to their D-structure direct objects, and that in sentences with ergative verbs - as in the passive - the subject position is not assigned a thematic role; that is, the subject position is a non-theta position. The derivation of the S-Structure is exactly as in the passive. According to Case theory, a direct-object NP must abandon a position in which it receives no Case. Theta theory stipulates that such an object can only be moved to a non-theta subject position. The corresponding S-structure is as follows: (4)

ts NP¡ [vp V t j ]

Now Burzio shows that the "unaccusative" property of ergative verbs follows from a general hypothesis and therefore need not be stipulated as a lexical peculiarity of these verbs. This hypothesis, which need not be discussed further here, is known as (5)

Burzio's Generalization The subject position has no theta-role iff the object position has no Case.

In view of this generalization, the ergative hypothesis can be formulated as follows. There is a certain class of intransitive verbs (the ergative verbs) which has exactly one of the properties mentioned in (5) as a lexical peculiarity. If one assumes with Hoekstra (1984) that the selection of thematic properties is a lexical matter - since the assignment of thematic roles to grammatical functions is determined in the lexicon - then ergative verbs can be characterized by the following lexical property: in sentences in which ergative verbs occur, the subject position is not assigned a thematic role.

Chapter 2

The Justification of the Ergative Hypothesis: Its Syntactic Predictions

2.1. M?-CLITICIZATION AND AUXILIARY SELECTION IN ITALIAN

It has already been mentioned that according to Burzio, the domain of ergative verbs encompasses more than the intransitive variants of the type introduced in Chapter 1 (1) and (2). How can this claim be put to the test? And how can we determine which intransitive verbs are to be considered ergative in Burzio's sense? It is clear that the lexical peculiarity of ergative verbs makes syntactic predictions about the D-structural object nature of the surface-structure subject. Using phenomena specific to D-structural direct objects, one cannot only test the ergative hypothesis, but one can also determine which verbs are ergative in Burzio's understanding of the term. In Italian, direct objects share a fairly clear characteristic property. Burzio points out that the cliticization of a partitive phrase by the particle ne is only possible for direct objects. Consider the following examples. (1)

a. b.

(2)

Giovanni ha insultato due amici 'John has insulted two friends' Giovanni ne ha insultati due John of-them has insulted two 'John has insulted two of them'

a.

Giovanni ha parlato a due amici 'John has talked with two friends' b. •Giovanni ne ha parlato a due John of-them has talked with two 'John has talked with two of them'

In Italian, the subject can be freely inverted; that is, an Italian subject can be positioned in the underlying preverbal subject position or postverbally. For such inverted surface subjects we can observe that the subjects of transitive verbs do not allow ne-cliticization, but the subjects of verbs in passive form do:

6 Ergativity in German (3)

a. *Ne hanno fatto domanda molti of-them have submitted an application many 'Many of them have submitted an application' b. Ne furono arrestati molti of-them have been arrested many 'Many of them have been arrested'

This indicates that the D-structure position of the direct object allows necliticization, and that the passive subject in (3b) but not the inverted transitive subject in (3a) can occupy this position. Thus, Burzio assumes that the former is to be analyzed as an NP within the VP (generated there in the base) as in (4)

tvpVNP]]

while the inversion in the latter case has to be represented as adjunction of the NP from its D-structure subject position to the VP as in (5)

[ S [vp[vpV]NP]]

In such an analysis, one must obviously clarify how the surface subject in question can be assigned nominative Case. This is an important question, particularly for the ergative case (4), for in the analysis of ergative constructions given in Chapter 1, NP-movement has been triggered in the absence of Case assignment to the object position. The question of how an "ergative subject" can receive nominative Case in the object position, and whether this is only possible in Italian or also in German will be discussed at length in Chapter 3. If one assumes that the clitic pronoun in examples such as (1) was moved from the position of the partitive phrase to the verb (with the corresponding consequences for agreement), cf. (6)

a. b.

[ VP ha insultato [ NP molti ne]] [vp [y ne¡ ha insultati] [ NP molti t¡]]

then structures (4) and (5) suggest a reason why only direct objects permit ne-cliticization: apparently, the clitic must c-command its trace, a relation that obtains between ne and an NP in object position, but not between ne and an NP inverted into the adjunct position. As far as the ergative hypothesis is concerned, the crucial observation is that inverted subjects permit ne-cliticization for certain intransitive verbs, but not for others, cf.

The Justification of the Ergative Hypothesis (7)

a.

b.

(8)

7

Arrivano molti studenti arrive many students 'Many students arrive' Ne arrivano molti of-them arrive many 'Many of them arrive'

a.

Lavorano molti studenti work many students 'Many students work' b. *Ne lavorano molti of-them work many 'Many of them work'

However, it must be noticed that even in the case of such intransitive verbs as, for example, arrivare (to arrive), whose inverted subjects permit cliticization, ne-cliticization is not possible when the subject takes the preverbal - i.e. underlying - subject position, cf. (9)

a.

Molti studenti arrivano 'Many students are arriving' b. *Molti ne arrivano 'Many of them are arriving'

In view of these data, Burzio assumes that the postposed subject is in the object position in (7), and that it is inverted into the adjunct position in (8). (10)

a. b·

[VP arrivano [ NP molti studenti]] [vp [vp lavorano] [ NP molti studenti]]

So in (10a), the surface subject remains in its underlying object position. The fact that this option exists alongside option (9a) which is derived through NP-movement must be explained as an option of nominative Case assignment in Italian. Let us then say that ne-cliticization provides a test for ergative verbs in Italian: intransitive verbs whose inverted subjects allow this sort of cliticization are ergative. Before we come to the corresponding situation in German, let me point out a further criterion for ergative verbs in Italian: auxiliary selection. The ergative verbs form their present perfect tense with essere (to be), whereas the other intransitive and transitive verbs generally form it with avere (to have) (the exceptions are not relevant in this context), cf.

8 Ergativity in German (11)

a.

b.

Giovanni è arrivato John is arrived 'John has arrived' Giovanni ha lavorato 'John has worked'

This can also be seen from the intransitive/transitive variants mentioned in Chapter 1. (12)

a. b.

La marina americana ha affondato la nave 'The American navy has sunk the ship' La nave è affondata the ship is sunk 'The ship has sunk'

The present perfect of ergative verbs, as is also the case with the passive, makes use of essere as the auxiliary in Italian. This indicates that the choice of essere is related to the fact that in these cases only the object - not the subject - position receives a theta-role. How the auxiliary selection is related to the dethematized argument in the passive (it can be picked up again by a ¿^-phrase) and to the argument gap typical of ergative verbs (an agentive ¿y-phrase is not possible) will be discussed in connection with the corresponding German phenomenon. So far we have become acquainted with two tests for ergative verbs in Italian. One makes use of the object position (the NP within the VP) of the surface subject, the other refers to the argument structure typical of ergative verbs.

2.2. AUXILIARY SELECTION IN GERMAN

Are there ergative verbs in German? If so, it should be possible to analyze them in a way similar to those in Italian, and the analysis should be testable using syntactic predictions specific for German. At first sight, it appears as though there is no consistent test in German comparable to the ne-cliticization test in Italian. And, in fact, the ergative hypothesis is somewhat more difficult to test in German. Nonetheless, there are a few reliable tests even in this language: let us begin by considering auxiliary selection. In German, transitive verbs form their present perfect tense with haben (to have).

The Justification of the Ergative Hypothesis (13)

9

Jupp hat den Ball getroffen Joseph has the ball kicked 'Joseph has kicked the ball'

This is also true of inherently reflexive verbs. With respect to the latter, one must assume that the accusative reflexive pronoun sich does not really constitute an argument. Finally, it also holds true for the so-called impersonal verbs. (14)

a. b.

Jupp hat sich geschämt 'Joseph was ashamed' Es hat gedonnert 'It has thundered'

A few transitive verbs which form the present perfect active with sein (to be) do not readily fit into place, cf. (15)

a.

b.

c.

d.

e.

Ich bin die Arbeit durchgegangen I am the work gone through Ί have gone through the work' Er ist den Bund fürs Leben eingegangen he is the bond for life gone through 'He took the marriage vows' Er ist die ganze Stadt abgelaufen he is the whole town run through 'He has run through the whole town' Sie ist ihn geflohen she is himacc fled 'She has fled from him' Sie ist ihn angegangen she is himacc approached 'She has approached him'

That these accusatives are really instances of direct objects {fliehen (to flee) is an exception) is shown by the fact that they can undergo passivization. With intransitive verbs the case is less uniform. There is a class of intransitive verbs which generally forms the present perfect active with haben (to have). For example: (16)

a. b.

one place: arbeiten (to work), schlafen (to sleep), blühen (to bloom) with indirect object: zustimmen (to agree), zuhören (to listen to), helfen (to help)

10 Ergativity in German There is, however, another class of intransitive verbs which generally forms its present perfect tense with sein (to be). For example: (17)

a.

one place: ankommen (to arrive), verblühen (to wither), wachsen (to grow), fallen (to fall), ertrinken (to drown), ersticken (to suffocate), entstehen (to come into existence/arise/result) b. with indirect object: unterliegen (to succumb), unterlaufen (to happen, in the sense of a mistake), auffallen (to strike a p.), gelingen (to succeed), passieren (to happen), zustoßen (to befall s.o./to happen to s.o.), einfallen (to occur to a p.)

Let us ignore the few exceptions to the transitive rule for the moment and concentrate on the intransitive verbs. Here we would seem to find a situation similar to that in Italian. Moreover, it should be noticed that in German, the present perfect passive is also generally formed with sein. In traditional German grammars, one finds generalizations of the following type. The present perfect tense of the perfective intransitive verbs (those giving a result) is formed with sein\ that of the imperfective intransitive verbs is formed with haben (Paul (1920), Vol. IV, paragraph 359). However, Hermann Paul himself points out that this regularity is "darkened and distorted" in the course of developing analytical verb forms. He further notes that certain verbs, such as (18)

Er ist gelegen/geschwommen vs. Er hat gelegen/geschwommen he is lain/swum vs. he has lain/swum

cannot be clearly classified, and that (in this respect) regional differences between Northern and Southern Germany can be observed. According to Duden (1973), intransitive verbs signifying a change of state or place and thus a "newly arrived-at state" form the perfect tense with sein, while intransitive verbs signifying an uncompleted event or a continuing event select haben. Both of these aspects can affect one and the same verb differently, depending on whether directionality or locality is expressed. Some of the verbs of motion show this effect. For example: (19)

a.

b.

Hans John 'John Hans John 'John

hat in seinem Zimmer getanzt has in his room dat danced has danced in his room' ist in sein Zimmer getanzt is into his room acc danced has danced into his room'

The Justification of the Ergative Hypothesis

11

Imperfective verbs of motion, such as gehen (to walk, to go), laufen (to run), and rennen (to race), are exceptions to this generalization. These verbs, as well as the verbs sein and bleiben (to stay or to remain), which are clearly stative verbs - that is, verbs signifying an enduring state - form the present perfect tense with sein. The same is true, incidentally, for the Italian pendants essere and rimanere which also select essere in the present perfect tense, while the French être forms its perfect tense with avoir and not with être. This shows that the semantic diagnosis of traditional grammar obviously covers only a secondary aspect of auxiliary selection. Let us assume - despite all obvious confusion - that German verbs exhibiting sein selection are, as in Italian, ergative verbs. This assumption gets us into trouble with the verbs introduced in (15), for these are transitive verbs that form the present perfect tense with sein. And verbs taking an accusative object cannot be classified as ergative under the analysis given above. (Later on, I will discuss suggestions such as the one by Belletti (1988) according to which ergative verbs assign partitive Case, the accusatives in question therefore being "actually" partitives). One could parry the difficulty posed by (15) historically. Almost all of the verbs listed above were intransitive verbs which have become transitive by the addition of a prefix. (Fliehen is an exception, but this was the only case in which the accusative object could not be passivized, thus raising doubts as to whether the verb is "really" transitive.) In general, a change from sein to haben selection is associated with such cases. The problematic verbs could then be described as cases in which the analogy to the intransitive verbs is so strong that sein selection remains in effect. One could also counter the difficulties differently. Toman (1986), for instance, relates the accusative marking in such cases to the appearance of a separable prefix. The first solution seems to lead immediately to a second difficulty. The intransitive verbs selecting sein - from which the compound verbs are derived - belong to precisely that group of verbs which, according to the generalizations of traditional grammar, present an exception to the usual sein/haben distribution. Even in recent analyses (cf. Wunderlich (1985)), in which the thematic role of the subject is held responsible for auxiliary selection, these verbs are an exception to the hypothesized regularity according to which intransitive verbs with an agent-subject select haben. Should one say, then, that intransitive verbs of motion are also ergative? One point which could be made against this hypothesis is the fact that many of these verbs (like tanzen (to dance), laufen (to run) and schwimmen (to swim)) permit haben as well as sein selection (cf. (19)). Should one follow Wunderlich (1985) in seeking the solution in a relationship between "Aktionsarten" (action types) and thematic roles? Such a relationship

12 Ergativity in German might, for instance, be of the following kind: resultativity (terminativity) involves factors that cannot be influenced by an agent (that is, the absence of an agent), while the presence of an agent implies the continuity of influenced processes (that is, nonterminativity). Unfortunately, this hypothesis fails to explain the relationship between thematic roles and "Aktionsarten" as well as that between thematic roles and auxiliary selection. The following observation seems to me to point to a solution: there is a use of the past participles of imperfective verbs of motion with the verb kommen which was already common in Middle High German, cf. (20)

Er kommt gegangen/gelaufen/gerannt/gesprungen/geflogen/gefahren etc. he comes gone (going)/walked (walking)/run (running)/leaped (leaping)/flown (flying)/driven (driving) etc.

This construction expresses directionality per se. When verbs select haben with locality and sein with directionality, the use of the kommen form is only permitted with the sein variants. (21)

a.

Er he b. *Er he

(22)

a. b.

hat im Zimmer getanzt has in the room danced kommt im Zimmer getanzt comes in the room danced

Er ist ins Zimmer getanzt he is into the room danced Er kommt ins Zimmer getanzt he comes into the room danced 'He comes dancing into the room'

Hermann Paul points out ((1920), Vol. IV, paragraph 323) that with respect to these constructions, the use of such participles in the active voice "contradicts everything that we otherwise know". His only explanation is that these participles must be understood "passively". If it is correct that the directional reading of these participles has a passive character, then the preference for sein in these cases can be explained by the usual selection of sein in the passive. The historically younger haben variant can be interpreted as an adaptation to the situation we usually find with imperfective intransitives. That some of these verbs only exhibit sein selection - even for the local reading - might be due either to the fact that a directional aspect can be recognized here, or to the uniformity of the paradigm.

The Justification of the Ergative Hypothesis

13

The passive character of these participles is also shown - not insignificantly - by its conformity to the usual intuitions of the native speaker. In the Duden grammar, for example, reference is made to the "differences in meaning" associated with auxiliary selection. When, for instance, verbs such as fahren (approximately, to drive) and fliegen (to fly) form the present perfect tense with sein, the suggested meaning emerges according to which the subject is a passenger. The selection of haben, on the other hand, is more likely to indicate that the person concerned was a driver or pilot. These intuitions of traditional grammar receive an interesting theoretical confirmation in Hoekstra's (1984) "small-clause" analysis of intransitive verbs of motion. Taking Dutch as his example, Hoekstra uses independent arguments to show that prepositional phrases can be analyzed as so-called "small clauses" (categories with a subject), and that these PP small clauses are adjuncts in local cases and complements in directional cases. The sentence (23)

daß Hans im Zimmer tanzt that John in the room dances 'that John dances in the room'

is analyzed accordingly as S

(24)

VP

NP Hans

VP

PP NP PRO

\

Ρ im Zimmer

while the sentence (25)

daß Hans ins Zimmer tanzt that John into the room dances 'that John dances into the room'

has the following underlying structure:

V

I

tanzt

14 Ergativity in German

The subject position is a non-theta-position in the directional case, but a theta-position with a PRO as small clause subject in the local case. This analysis is confirmed by the following considerations. In the local case, haben is selected in the present perfect tense. This indicates a theta-subjectposition. In the directional case, sein is selected, indicating a non-theta subject position. In Dutch, the PP can occur postverbally in the local, but not in the directional case. In German, an analogous observation can be made with respect to the postverbal occurrence of PPs combined with the phrase und zwar. (27)

a.

Hans John b. *Hans John

hat getanzt, und zwar im Zimmer has danced and what is more in the room ist getanzt, und zwar ins Zimmer is danced and what is more into the room

The fact that (27b) is ungrammatical can be explained with reference to structure (26). The subject of the small clause must be governed by V, for it must either receive Case from V or, if this is not possible, must be moved into the subject position leaving behind a trace. This trace must be governed. But in Dutch, as in German, V governs only to the left. In the adjunct case, PRO is not and may not be governed. Thus postverbal position of the PP is possible in this case. The assumption that the directional PP with sein selection is indeed a complement, whereas the local PP with haben selection behaves as an adjunct receives further support from the following data: (28)

a.

Hans John b. *Hans John

hat getanzt has danced ist getanzt is danced

Finally, the "passive" structure of the directional case is further confirmed by the following observation. It is well known that the attributively used

The Justification of the Ergative Hypothesis

15

past participle of transitive verbs refers to the object, but not the subject, cf. (29)

der geküßte Student = der Student, the kissed student the student hat has Φ der Student, the student hat has

den jemand geküßt whoacc somebodynom kissed

der jemanden geküßt whonom somebodyacc kissed

One can therefore predict on the basis of the historical hypothesis that they are to be derived from underlying objects that the personal subjects of directionally interpreted participles of motion can occur with the past participle as attribute. In fact, the subject and the attributively used participle occur together on the directional but not on the local reading, cf. (30)

a.

der the 'the b. *der the 'the

ins Zimmer getanzte Ehemann into the room danced husband husband who danced into the room' im Zimmer getanzte Ehemann in the room danced husband husband who danced in the room'

If we extend these considerations to the class of intransitive verbs of motion whose perfect tense selects sein, we find all of the phenomena attested which were used to motivate the structure (26). That is, there are intransitive verbs of motion for which - cf. the constant auxiliary selection - only structure (26) can be assumed. This represents theoretically the intuitive hypothesis Hermann Paul posed for these verbs. However, this does not explain why the passive and the present perfect of ergative verbs select sein, while haben is selected in all other cases. What passive and ergative structures have in common is that the object is assigned a theta-role, but that the subject position is a non-theta position. It would thus seem plausible to posit a connection between auxiliary selection and theta-role assignment, i.e. to ask whether the representation of a certain argument structure requiring an auxiliary does not have to take into account the argument structure of the auxiliary itself. This would mean that, if a verb exhibiting an argument structure with a theta-subject position is to be represented in the present perfect tense, an auxiliary must be chosen which maintains this argument structure, or which allows it to "percolate".

16 Ergativity in German In my opinion, there are two ways to think of the percolation mechanism of an argument structure (i.e. of a theta-grid) as determined by the auxiliary. First of all, one could assume that the argument structure of the auxiliary and that of the participle must be isomorphic. One could thus say, for instance, that sein, which can also be considered an ergative verb (cf. for example Kayne (1985)) - that is, one characterized by a non-theta subject position - will be selected when the participle also displays a non-theta subject position (as in the passive and with ergative verbs). Haben, on the other hand, can allow two theta-roles to percolate, and thus is appropriate in the case of transitive verbs. This way of looking at things is not without problems, however. First of all, the argument structure of haben must be assigned an extremely high degree of variability, because haben is also selected in intransitive cases as well as with participles assigning more than two theta-roles. Secondly, further difficulties arise in view of the fact that haben/sein selection is the same under the predicative reading of a participle as under the present perfect tense reading, although the argument structure obviously differs in the two cases, cf. (31)

a.

b.

Die Wäsche ist getrocknet (in the sense of: ist eine getrocknete) (transitive case) the wash is dried (in the sense of: is one that has been dried) 'The wash is dry' Wir haben Flüchtlinge versteckt (in the sense of: Wir haben we have refugees hidden Flüchtlinge als Versteckte) 'We have refugees in hiding'

Obviously, the crucial factor for the auxiliary selection is played not by the argument structure as a whole, but by that of the subject position. I therefore prefer the second way of analyzing the percolation mechanism. This analysis goes back to a suggestion made by Hoekstra (1984) and Kayne (1985) according to which past participles in general (i.e. in the passive as well as in the active) are characterized - according to their own lexical representation - by the property of absorbing the theta-role assigned to the subject, and as a consequence of this by the further property of not assigning Case. Past participles thus do not assign Case to a governed NP. In other words, we can say that the participle is unable to further project the theta-role assigned to the subject position. One may thus regard as a specific characteristic of haben (cf. Hoekstra (1984), Haider (1984), Kayne (1985)) its ability to "deblock" (Haider) the subject theta-role absorbed by the participle, and to "transmit" (Kayne) it onto its own syntactic subject position, in short, to "revitalize" (Gre-

The Justification of the Ergative Hypothesis

17

wendorf) the subject theta-role. This ability lies in haben's own theta marking and in a percolation mechanism, as is found in word-formation rules (cf. Hoekstra (1984), Chapter 4.4.4). In the case of sein, whose subject position is itself a non-theta position, no theta-role is transported to the subject position. The advantage of the analysis above is that participles have the same characteristics in all of their uses (the present perfect tense as well as the passive). Furthermore, the selection of the auxiliary is dependent only on the theta marking of its subject position. In this way, disadvantages of the other solutions can be avoided. Let us take a brief look at how this analysis works. Case (1): Transitive Verbs Here there are two options. Either the subject theta-role that was "sedated" by the participle is reactivated by haben and a sentence in present perfect active ("( )" = absorption) is formed, (32)

hat has

(

I (Subj-Θ)

geküßt kissed

f Obj-Θ

or sein is chosen and the theta-role of the subject remains absorbed, with the consequence that the object of the participle receives no CaSe (cf. Burzio's Generalization) and therefore must be moved across the absorbed theta position of the participle into the non-theta position of the auxiliary in order to get Case. In other words, a passive structure is formed, cf. (disregarding werden) (33)

Objj I Subj-Θ'

ist is

t¡ f (Subj-Θ)

geküßt kissed

t¡ f Obj-Θ

Case (2): Non-ergative Intransitive Verbs Here there is a subject theta-role absorbed by the participle. It can be reactivated by haben so that the present perfect active is formed. Should sein be chosen, there is no subject theta position, and an impersonal passive is formed. Case (3): Ergative Intransitive Verbs Unlike passive participles, the participles of ergative verbs do not absorb or block a subject theta-role. There is simply no such theta-role to be absorbed, as can be seen from the fact that it cannot be picked up again in a vo«(èy)-phrase.

18 Ergativity in German There is therefore no "sedated" role that can percolate into the subject theta position of haben. Sein is then selected with the same effect as in the passive of transitive verbs: the object is moved into the non-theta subject position of the auxiliary in order to receive Case, cf. (34)

Obji I Subj-Θ'

ist is

t' f Subj-Θ'

gestorben died

t¡ I Obj-Θ

Intransitive verbs of motion such as tanzen (to dance), which can select haben as well as sein, are analyzed like the non-ergative intransitive verbs under the local reading and like the ergative intransitive verbs under the directional reading.

2.3. PARTICIPLES AS ATTRIBUTES

The fact that ge-participles (past participle, participle II) can occur as attributive predicates of the nouns corresponding to their direct objects has already been mentioned. (35)

a. b.

(36)

Maria hat den Studenten geküßt Mary has the student kissed der geküßte Student = der Student, the kissed student the student geküßt hat kissed has Φ der Student, the student geküßt hat kissed has

den jemand whoacc somebody,

der jemanden who nom somebody

a.

Maria hat dem Studenten das Buch gegeben Mary has the student da , the book acc given 'Mary has given the book to the student' b. ""der gegebene Student the given student c. das gegebene Buch the given book

It is therefore to be expected that only transitive verbs permit their participles to be used attributively. This prediction is confirmed when we consider non-ergative intransitive verbs.

The Justification of the Ergative Hypothesis (37)

19

a.

Der Student hat gearbeitet 'The student has worked' b. *der gearbeitete Student the worked student

However, it is interesting to note that ergative intransitive verbs permit attributive participles (cf. Grewendorf (1983)): (38)

a.

b.

(39)

a.

b.

Der Student ist eingeschlafen the student is fallen asleep 'The student has fallen asleep' der eingeschlafene Student the fallen-asleep student Dem Hans ist ein Schaden entstanden the John dat is a harm arisen 'John has suffered harm' der entstandene Schaden the arisen harm 'the resulting harm'

Recalling the analysis suggested for ergative verbs, we see that the fact that an ergative verb allows an attributive participle II is compatible with the generalization that the attributive participle II is predicated of an underlying direct object. The observation that the participles of ergative verbs can occur attributively may therefore be regarded as a confirmation of this analysis and as a further test for ergativity. Notice that an analysis such as that of Bresnan (1982), which is based on grammatical functions as well as thematic roles, does not adequately capture the phenomena. According to this analysis, attributive participles II should be possible in those cases where a subject with the theta-role theme occurs, that is, when an ergative or passivized transitive verb is present. It is thus assumed that the participle is predicated of a theme subject. This analysis is not only less simple, it fails to capture the data correctly. Even though the identification of thematic roles is difficult and extremely unclear, linguists are likely to agree that none of the following sentences has a theme subject even though the participle can be used as an attribute. (40)

a.

b.

Der Feuerwehrmann ist erstickt the fireman is suffocated 'The fireman has suffocated' der erstickte Feuerwehrmann the suffocated fireman

(Patient)

20 Ergativity in German (41)

a.

b.

(42)

a.

b.

Hans ist geschlagen worden John is beaten been 'John has been beaten' der geschlagene Hans the beaten John Hans ist gebeten worden... John is asked been... 'John has been asked...' der gebetene Hans the asked John

(Patient)

(Goal)

There should also be general agreement that the following examples contain theme subjects. Nevertheless, attributive participles are not possible. (43)

a.

Der Film hat dem Hans gefallen the film nom has the John dal pleased 'The film has pleased John' b. *der gefallene Film the pleased film

(44)

a.

(45)

a.

Hans hat in Berlin gewohnt John has in Berlin lived 'John has lived in Berlin' b. *der gewohnte Hans the lived John Dem Vortrag folgte eine Diskussion the lecturedat followed a discussion,,, 'The lecture was followed by a discussion' b. *die gefolgte Diskussion the followed discussion

These counterexamples to Bresnan's analysis also provide an argument against Wunderlich's (1985) strategy. As mentioned in Chapter 1, Wunderlich identifies ergative verbs with theme verbs. The examples above show that a further distinction is necessary. The characteristic aspect of ergative verbs is not that they have a theme subject on their syntactic surface, but that they have a non-theta subject position at the D-structure level. However, an object forced into NP-movement by Case theory must not always have the thematic role theme. This sort of objection to Wunderlich's approach will be raised again in the following sections.

The Justification of the Ergative Hypothesis

21

2.4. IMPERSONAL PASSIVE - AGENTIVE VON (BY)-PP

A further characteristic of ergative verbs is that they cannot take the passive voice. Unlike non-ergative intransitive verbs, they do not allow the impersonal passive form. (46)

a.

Es wurde getanzt/gearbeitet/geschrien/geschlafen it was danced/worked/screamed/slept b. *Es wurde angekommen/gewachsen/gefallen/eingeschlafen it was arrived/grown/fallen/fallen asleep c. Dem Hans wurde zugehört/geholfen the John dat was listened/helped 'John was listened to/was helped' d. *Dem Hans wurde unterlaufen/entstanden the John dat was (a mistake) happened/resulted

The suggested analysis of ergative verbs also provides an explanation of this phenomenon. If the crucial characteristic of the passive is the absorption of the subject theta-role, it is clear that there can be no such absorption in the case of ergative verbs. For it is characteristic of such verbs that unlike non-ergative intransitive verbs, they have no subject theta-role to be absorbed. A consequence of this fact is that ergative verbs, although they have an underlying structure not unlike that of the passive construction, do not allow the von-phrase permissible in the passive. In the passive, as mentioned earlier, a subject theta-role is absorbed by the participle, but is not thereby completely eliminated. The theta-role can be "reactivated" by a haben auxiliary or, in the case of a sein auxiliary, it can be picked up again by a von-phrase. In the case of ergative verbs, this is impossible. A theta-role was not absorbed; it simply did not exist in the first place. So according to my analysis, the fact that a prepositional phrase cannot pick up the theta-role again is explained in the same way that the impossibility of haben selection is explained.

2.5. DERIVATIONAL MORPHOLOGY: -ER AND -LING NOMINALIZATIONS

Burzio (1981) has drawn our attention to an additional phenomenon identifying ergative verbs that is indirectly related to a syntactic property. It stems from derivational morphology, and concerns the observation that unlike non-ergative intransitives, ergative verbs do not allow the derivation of nouns with the agentive suffix -er, cf.

22 Ergativity in German (47)

a.

b.

Arbeiter, Tänzer, Träumer, Denker, Helfer, Diener, worker, dancer, dreamer, thinker, helper, servant, Schläfer sleeper * Ankommer, *Faller, *Wachser, *Ersticker, *Unterlieger, •arriver, *faller, *grower, *suffocator, *succumber, •Einschläfer *faller-asleep

This observation concerns word-formation constraints, that is, not genuinely syntactic phenomena. The constraints in effect here obviously correlate with the theta framework of the verbal entries in the lexicon. Because the analysis offered for ergative verbs rests on a hypothesis about their theta frameworks, the difference illustrated in (47) can be accorded an indirect syntactic relevance. The explanation for this difference is found in the fact that with -er nominalizations the thematic role of the subject argument is always identified (although it is not necessarily always an agent, cf. Besitzer (owner), Empfänger (receiver)). An Ankommer is not one who arrives, for the same reason that a Schläger is not someone who someone else hits. Peter Hummer (Salzburg) has pointed out another argument to me in this connection. It concerns nominalizations with the -ling suffix. This sort of nominalization is no longer productive, but in the cases in which it occurs, it does provide an argument for the ergative hypothesis. Deverbative -ling derivatives are associated with transitive verbs, and are always identified with the object argument of the base verb (a Prüfling (examinee) is one who is examined). Consequently, they often have antonymic counterparts with the -er suffix (Prüfling (examinee) - Prüfer (examiner)). But they are also associated with a series of intransitive verbs. These are precisely (and only) those verbs which we have encountered as ergative verbs. The ergative hypothesis explains why it is precisely these intransitive verbs which exhibit the nominalization in question, cf. (47)

c.

transitive:

d.

ergative:

Lehrling (apprentice), Prüfling (examinee), Pflegling (foster-child), Findling (foundling), Liebling (darling), Mischling (half-caste), Setzling (young plant), Impfling (a vaccinated person), Säugling (suckling (baby)), Schützling (protégé), Sträfling (convict) Ankömmling (someone who has just arrived), Sprößling (sprout), Eindringling (intruder), Emporkömmling (upstart)

The Justification of the Ergative Hypothesis

23

2.6. TOPICALIZATION OF SUBJECT + PARTICIPLE II

The nominative subject and parts of the verbal complex cannot normally form a constituent. Thus a subject and a past participle cannot normally be topicalized together (placed into the "prefield" (CP-specifier position) of the sentence), although an object and a past participle can. (48)

a. *Der Kanzler gerettet hat gestern einen Pudel the chancellornom rescued has yesterday a poodleacc b. *Eine Frau vergessen hat wohl diesen schwarzen Hut a woman nom forgotten has probably this black hatac, c. Einen Pudel gerettet hat der Kanzler gestern a poodleacc rescued has the chancellorn0m yesterday 'Yesterday, the chancellor rescued a poodle' d. Diesen schwarzen Hut vergessen hat wohl eine Frau this black hat acc forgotten has probably a woman no Ά woman has forgotten this black hat'

Some verbs which form their present perfect tense with sein constitute an exception to the generalization illustrated by (48a/b). It is obvious that this is once again the class of ergative verbs. Their past participle can be topicalized in conjunction with the nominative subject, cf. (49)

a.

b.

c.

Ein Fehler unterlaufen ist dem Hans a mistake nom happened is the John dat schon lange nicht mehr not for a long time 'For a long time now, John has not made a mistake' Eine Concorde gelandet ist gestern in a Concorde nom landed is yesterday in München-Riem Munich-Riem Ά Concorde landed yesterday at the Munich-Riem airport' Ein Hattrick geglückt ist Dieter Hoeneß noch nie a hat trick succeeded is Dieter Hoeneß dat sofar never 'So far, Dieter Hoeneß has not succeeded in scoring a hat trick'

One notes, however, that such topicalizations of past participles and nominative subjects can become ungrammatical - or at least questionable - when the subject is definite.

24 Ergativity in German (50)

a. *Der Fehler unterlaufen ist dem Hans gestern schon the mistakenom happened is the John dat yesterday once wieder again b. *Der Kanzler gelandet ist gestern in the chancellornom landed is yesterday in München-Riem Munich-Riem c. *Der Sieg über Lendl geglückt ist Boris Becker [the victory over Lendl]nom succeeded is Boris Beckerdat zufallig by chance

Passive subjects constitute a second exception to the generalization under discussion. Here one can make observations similar to those made in the case of ergative verbs. Passive subjects can be topicalized in conjunction with a past participle when the subject is indefinite, but not - or at least not very well - when the subject is definite, cf. (51)

a.

b.

(52)

Ein Buch geschenkt wurde der Studentin zu a book nom given was the studente for Weihnachten Christmas Ά book was given to the student for Christmas' Eine Reise spendiert wurde der Gewinnerin des a journey nom paid for was the winnerdat of the Wettbewerbs contest Ά journey was paid for the winner of the contest'

a. *Canettis neuestes Buch geschenkt wurde der Studentin zu Canetti's latest book n0m given was the student dat for Weihnachten Christmas 'Canetti's latest book was given to the student for Christmas' b. *Der Europapokal spendiert wurde Juventus the European Cup nom given on a plate was Juventus Turin letztes Jahr Turin dat last year 'Last year, the European Cup was given Juventus Turin on a plate'

The Justification of the Ergative Hypothesis

25

One must bear in mind that topicalization data are not easy to evaluate. This might be due to the fact that there is no general agreement about the criteria for defíniteness (cf. section 3.5 as well as Belletti (1988)), or it might be that a variety of factors contribute to the acceptability of topicalizations. Thus Haider (1985) points out that in cases where no argument remains behind in the "middle field" of the German sentence, the acceptability of otherwise grammatical cases of topicalization is reduced. This is the case, for example, when monovalent ergative verbs or bivalent passive verbs are topicalized. (53)

a. ?Eine Tanne gewachsen ist früher an dieser Stelle a pine-tree grown is earlier at this spot 'Earlier, a pine-tree grew on this spot' b. ?Ein Weltmeister angekommen ist schon oft in a world champion arrived is already often in München-Riem Munich-Riem Ά world champion has often arrived before at the Munich-Riem airport'

(54)

a. ?Ein Professor ausgelacht wurde schon oft im a professor laughed at was already often in the Seminar seminar Ά professor has often been laughed at before in front of the class' b. ?Ein Tier gequält wurde schon häufig an animal tortured was already often 'An animal has been often tortured before'

Nevertheless, even in this case, it should be clear that the differences in grammaticality between the examples in (53/54) and the non-ergative structures are obvious: (55)

*Ein Linguist gearbeitet hat früher auf dieser Stelle a linguist worked has earlier in this position

It is not in itself surprising that indefinite ergative and passive subjects can occur in the "prefield" (CP-Spec) position with a past participle. Because both cases entail underlying objects, they naturally show a "basic" similarity to the acceptable object cases of (48c/d). Apart from this, however, the following problems arise. Postulating a non-theta subject position for passive as well as for ergative

26 Ergativity in German structures leads to the consequence (cf. Burzio's Generalization) that the direct object cannot be assigned Case and, therefore, that NP movement must result. Subsuming the topicalization of subject 4- past participle under the object case would, however, presuppose that the passive and ergative subjects would remain in their original positions within the VP. Only then would the required constellation of constituents be present. Explaining the problematical data with reference to the underlying objectcharacter of the topicalized nominative (and thus with respect to the ergative hypothesis) presupposes that the underlying objects have the option of remaining in the object position. And that means that they must have the option of receiving nominative Case in the object position. The fact that this option actually exists, and how this fact is to be explained within the framework of the government and binding theory will be the subject of section 3. Why topicalization is only possible in the case of indefinite NPs remains to be explained. This and other similar problems will be discussed in detail in sections 3 and 4. Assuming that the problems formulated above can in fact be solved as indicated, the observations in this section provide a further argument in favor of the ergative hypothesis. In addition, they can be used as a test for the identification of ergative verbs.

2.7. DISCONTINUOUS PHRASES

Discontinuous - that is, split - phrases have been analyzed in Fanselow (1985), Grewendorf (1988), Chapter 13, and, more recently, in van Riemsdijk (1987) and Fanselow (1988). As the following examples show (56)

a.

b.

Kleider hat er immer dreckige an clothes wears he always dirty-ones 'As for clothes, he always wears dirty ones' Geld hat er keines money has he none 'As for money, he has none'

in the case of discontinuous phrases, an NP (for example, dreckige Kleider) is split in such a way that Ν or N1 is topicalized, leaving behind the rest of the NP with the determiner in what is usally called "the source position". Contrary to the view expressed by Fanselow (1985), it does not seem obligatory to me that the second part of the NP necessarily has to be moved to COMP, cf.

The Justification of the Ergative Hypothesis (57)

a. b.

weil because weil because

er Kleider he clothes Kleider er clothes he

immer always immer always

dreckige dirty-ones dreckige dirty-ones

27 anhat wears anhat wears

Such an NP can be split across sentence boundaries (58a), and, surprisingly, both the topicalized part and the source position can contain a determiner (58b) (cf. van Riemsdijk (1987)) (58)

a.

b.

Kleider glaubt er hat Hans immer dreckige an clothes believes he wears John always dirty-ones 'As for clothes, he believes that John always wears dirty ones' Einen Wagen hat er sich noch keinen leisten können a car has he refi, yet none afford could 'As for a car, he has not yet been able to afford one'

NP splitting can also occur in conjunction with an element of the verbal complex, as in (59)

a.

b.

Kleider getragen hat er immer dreckige clothes worn has he always dirty-ones 'As for clothes, he has always worn dirty ones' Geld ausgegeben hat er noch keines money spent has he not yet any 'As for money, he has not spent any yet'

Now the morphological form of kein in (59b) appears to be appropriate for cases in which kein occurs alone, cf. (60)

a.

Er hat noch keines he has yet not-any 'He has not spent any b. *Er hat noch keines he has yet not-any

ausgegeben spent yet' Geld ausgegeben money spent

This observation led Haider (1985) to regard the sequence in front of the finite verb in (59) as a complex verb and to speak of "incorporated arguments". In Fanselow (1985), this consideration is worked out in such a way that NP-splitting itself is analyzed as verb topicalization. Such an analysis becomes possible if one assumes that the second part of a split NP (i.e., the nominal head) has been reanalyzed as part of the verbal complex, cf.

28 Ergativity in German (61)

VP

a.

V

NP

anhaben

dreckige Kleider VP

b. NP dreckige

VK e¡

Kleider¡

anhaben

The V-topicalization rule then operates in such a way that the first element of the verbal complex can be moved up into the topicalization position, and every continuous chain to the left and the right from this verbal element can be carried along (cf. Grewendorf (1988), Chapter (13.), Fanselow (1983)). The fact that "split topicalization" exhibits movement as well as nonmovement properties (see e.g. van Riemsdijk (1987)) has led to considerable disagreement on whether this construction has to be analyzed in terms of movement of a non-maximal projection along with some kind of "regeneration" compatible with X-bar theory (if operative at S-structure) of the missing structure (cf. van Riemsdijk (1987)), or as generation of two NPs with some adjustment concerning the application of the thetacriterion to such structures. Fanselow ((1985), Chapter 5.3) offers arguments against a discontinuous base generation and also against a wA-movement analysis. He shows that all essential properties of NP-splitting can be explained by reanalysis and V-topicalization. In Fanselow (1988), this analysis is spelled out in the following way. Fanselow starts out from the observation that there is a correlation between NP-splitting and NP's functioning as adjectives and determiners. As for split topicalization, he assumes base generation of two NPs: one consisting of an adjective/determiner and a pro-head (cf. Olsen (1987)), and another one (the one to be topicalized) adjoined to the verb, that is, located in a non-thematic position. Since the two NPs are coindexed, topicalization of the latter creates A'-binding of the prohead in a way roughly analogous to A'-binding of clitic pronouns. If an analysis of this kind is correct, then several relevant predictions can be made with reference to the ergative hypothesis. Because reanalysis is subject to the adjacency condition, and because we are concerned here with a reanalysis within the VP (where possible variations within the VP can be exploited, cf. Grewendorf (1988)), we can predict that constituents outside the VP are not subject to this reanalysis

The Justification of the Ergative Hypothesis

29

process. In other words, our prognosis would be that transitive and nonergative intransitive subjects do not permit NP splitting. This prediction is confirmed by the following examples. (62)

a. ""Studenten haben fleißige das Seminar students nom have hard-working ones the seminaracc besucht taken part in b. ""Eltern haben freigiebige dem Kindergarten eine parents nom have generous ones the nurserydat a Schaukel gespendet swingacc donated to c. ""Politiker glaube ich haben korrupte diesen Staat politiciansnom think I have corrupt ones this stateacc ruiniert ruined 'As for politicians, I think corrupt ones have ruined this state' (Because of the veracity of this example, I felt obliged to give an accurate English translation.)

Ergative and passive subjects, on the other hand, are again an exception. They allow NP splitting, cf.: (63)

a.

b.

c.

Fehler sind dem Hans vermeidbare unterlaufen mistakesnom are the John dat avoidable ones happened 'As for mistakes, John has made avoidable ones' Widersprüche sind dem Richter mehrere aufgefallen contradictions nom are the judge dat several been striking 'As for contradictions, the judge noticed several' Demonstranten wurden mehrere verletzt demonstrators were several injured 'As for demonstrators, several were injured'

Again, it is easy to subsume such exceptions under the theory of NPsplitting outlined above, provided that we assume that ergative and passive subjects are allowed to remain in their D-structure positions. With the same reservations as noted at the end of 2.6, subject NP-splitting can be used as an argument for the ergative hypothesis and thus also as a test for identifying ergative verbs. The fact that NP-splitting of ergative and passive subjects only seems to be acceptable in indefinite cases indicates that the solution outlined at the end of 2.6 points in the right direction, cf.

30 Ergativity in German (64)

a. *Gewinner wurden nur die ersten zehn mit einer winners were only the first ten with a Urkunde geehrt certificate honoured 'Of the winners, only the first ten were honoured with a certificate' b. Gewinner wurden nur weibliche mit einer Urkunde winners were only females with a certificate geehrt honoured O f the winners, only the females were honoured with a certificate' c. *Gäste sind nur die angemeldeten angekommen guests are only the announced ones arrived d. Gäste sind nur angemeldete angekommen guests are only announced ones arrived 'Only announced guests have arrived'

A remark on the grammaticality judgements made in this section might be in order here. It may be that the judgements "*" vs. "ok" are overstated. Nevertheless, there exists clear differential behavior of subject and object NPs with respect to split topicalization, and this difference is what the foregoing argument relies on. Note that the point of this argument can be further established using a type of example where the judgements "*" vs. "ok" are undoubtedly correct. These are examples of split NPs where a type of partitive relation between two nouns is involved, cf. (64)

e.

Wein wine f. *Wein wine g. Wein wine

One must bear are frequently often adduced not necessarily (65)

a.

habe ich drei Kisten gekauft have I three cases bought haben drei Kisten meinen Keller gefüllt have three cases my cellar filled-up sind drei Kisten angekommen are three cases arrived

in mind that examples of the type illustrated in (65), which introduced as illustrations of NP-splitting and which are as counterexamples to the analysis presented above, are to be analyzed as such.

Arbeiter haben viele die SPD gewählt workers have many the SPD voted 'Of the workers, many have voted for the SPD'

The Justification of the Ergative Hypothesis b.

31

Politiker glaube ich haben keine zum Wohl der politicians think I have none to the good of the Welt beigetragen world contributed

It might be that these cases are best subsumed under the phenomenon "quantifier floating" and that an analysis other than the one offered here for discontinuous NPs can be found. For a further example of how discontinuous constituent phenomena can become relevant to the discussion of the ergative hypothesis, consider the following cases of "subextraction" discussed for Dutch by den Besten (1981). Den Besten has observed that direct objects in Dutch have a property comparable to the property which allows «e-cliticization in Italian. He observed that the Dutch analogue to German was für (what kind of) constructions - wat voor - can only be split with a direct object and not with a subject or indirect object. Though degrees of acceptability partially vary among German speakers, den Besten's observation concerning "subextraction" can, in principle, also be applied to German, where was für is likewise a question quantifier, and where the following judgements seem, at least approximately, to correctly represent the intuitions of German speakers (see also van Riemsdijk (1987), Chapter 3.5.6). (66)

a.

b.

(67)

a.

(68)

a.

Was für Museen hast du in Italien besucht? what for museums have you in Italy visited 'What kind of museums have you visited in Italy?' Wasj hast du in Italien [ NP e¡ für Museen] besucht? whatj have you in Italy [ NP e¡ for museums] visited

Was für Leute haben dieses Bild gemalt? what for people have this picture painted 'What kind of people have painted this picture?' b. *Was, haben [ NP e¡ für Leute] dieses Bild gemalt? whati have [ NP e¡ for people] this picture painted Was für Leuten hast du deine Adresse gegeben? what for people dat have you your addressacc given 'What kind of people have you given your address?' b. *Was¡ hast du [NP e¡ für Leuten] deine Adresse what; have you [ NP e¡for people]dat your addressacc gegeben? given

32 Ergativity in German However, analyzing these data analogously to ne-cliticization is not possible because the extracted was c-commands its original position in every case. Subjacency cannot be referred to either, because it is either violated in all three cases or in none of them. The same seems to be true of the Empty Category Principle, if one assumes (cf. Fanselow (1985)) that the was/«r-position is properly governed by the adjacent noun. It therefore seems reasonable to follow the analysis suggested by Bennis (1983) for Dutch (cf. also Fanselow (1985)) according to which was fîirphrases can be restructured so that a phrase such as (69a) can be reanalyzed as (69b). (69)

a. b.

[ NP [ NP was] [ PP für Leute]] [np [np what] [pp for people]] [ NP was] [pp für Leute] [Np what] [pp for people]

This reanalysis has a certain plausibility, especially if one considers the historical origins of the was /«/--construction. The sequence was für ein takes the place of a genitivus partitivus linked to the pronoun was (cf. Paul (1920), III, paragraph 231). It originates from constructions such as Was hast du für einen Vater in the sense of "what do you have in place of a father". The restructuring suggested, which eliminates the outer NP-boundary, enables us to explain the subextraction data in terms of subjacency. Subextraction violates the subjacency condition in a structure like (69a), but not in a structure like (69b). Explaining the data (66) - (68) along these lines requires once again the additional hypothesis that reanalysis is limited to the domain of the VP. It is interesting - and relevant to the ergative hypothesis - to observe that subextraction is impossible with transitive (67b) and intransitive (70) but not with ergative (71) and passive (72) subjects (or that it is at least significantly less acceptable in the former than in the latter case). (70)

*Was haben für Leute gearbeitet? what have for people worked

(71)

Was sind für Leute angekommen? what are for people arrived 'What kind of people have arrived?'

(72)

Was sind in der Kantine für Leute gesehen worden? what are in the canteen for people seen been 'What kind of people have been seen in the canteen?'

The Justification of the Ergative Hypothesis

33

On the basis of Bennis' analysis - and on the assumption that the Case problem has been solved - the ergative hypothesis would also provide an explanation for this phenomenon. In (71) and (72), the surface subject is in its underlying object position, that is, within the VP. Because the reanalysis process illustrated in (69) is limited to the domain of the VP (cf. Fanselow (1985)), the data (66)-(68) can be explained by assuming that the outer NP node in (69a) cannot be eliminated when it is located outside of the VP. Assuming that this argument is sound, the data in (70) - (72) provide a further argument for the ergative hypothesis and a further test for ergative verbs. Regardless of whether or not this particular explanation turns out to be adequate, the ergative hypothesis seems to be involved at any rate since subextraction is impossible (or at least much less acceptable) even with ergative and passive verbs whenever the surface subject is not in its underlying position, cf. (73)

a.

Was sind dem Lehrer für Fehler aufgefallen? what are the teacherdat for mistakesnom been striking 'What kind of mistakes has the teacher noticed?' b. *Was sind für Fehler dem Lehrer aufgefallen? what are for mistakesnom the teacher^ been striking

(74)

a.

Was sind dem Kanzler für Geschenke überreicht what are the chancellordat for gifts nom presented worden? been 'What kind of gifts have been presented to the chancellor?' b. *Was sind für Geschenke dem Kanzler überreicht what are for giftsnom the chancellordat presented worden? been

(Once again, the asterisk is not meant to convey an absolute judgement.) This situation reflects precisely the situation we found with Italian necliticization. Even in the presence of an ergative or passive verb, necliticization is impossible when it originates from the subject position (cf. (9b)). Before closing this section, I would like to briefly mention that an explanation of the subextraction phenomenon need not necessarily appeal to restructuring and subjacency. As already pointed out above, the claim that the ECP is irrelevant to the explanation of subextraction rests on the assumption that the extraction position is properly governed by a noun:

34 Ergativity in German the "question quantifier" wasfür is thus conceived as an attribute. However, if one assumes a structure such as (69a), which lacks a governing N, the ECP might regain its relevance. If we assume that government of the complex NP percolates to the was (as head) position, the subject/directobject asymmetries could be interpreted as an ECP effect (see also van Riemsdijk (1987)). (The situation with indirect objects requires an independent explanation as is indicated by the fact that the NP-splitting described in the first part of this section is likewise impossible for dative objects (cf. Fanselow (1985), Chapter 5.3)). However, the fact that the unacceptable examples of subextraction are not "completely" ungrammatical provides evidence against an ECP account of this phenomenon.

2.8. EXTRACTION FROM NP SUBJECTS

Let us now examine the way in which extractions from NP subjects more exactly, the extraction of prepositional phrases from NP subjects can provide us with arguments for the ergative hypothesis. Our starting point is the observation that in German, extraction of PPs from object NPs seems to be possible in certain cases, cf. (75)

a.

b.

c.

d.

[Über Boris Becker]¡ hat Peter Handke [ein Buch e¡] [about Boris Becker]¡ has Peter Handke [a book e¡] geschrieben written 'Peter Handke has written a book about Boris Becker' [Von Flaubert], hat Fritz [zahlreiche Romane e¡] [by Flaubert]] has Fritz [numerous novels e¡] gelesen read 'Fritz has read numerous novels by Flaubert' [Zum ECP]¡ hat der Student [eine Frage e¡] [about the ECP]¡ has the student [a question e¡] gestellt asked 'The student asked a question about the ECP' [Mií Peter]¡ hat sie [das gemeinsame Leben e¡] beendet [with Peter]¡ has she [the life together e¡] terminated 'She has terminated the life together with Peter'

The following examples show that such an extraction is not always possible.

The Justification of the Ergative Hypothesis (76)

35

a. *[Über Wittgenstein]; hat Georg [ein Buch e¡] gestohlen [about Wittgenstein]; has George [a book e¡] stolen 'George has stolen a book about Wittgenstein' b. *[Aus Rom]; hat Hans [den Zug e¡] verpaßt [from Rome]; has John [the train e¡] missed 'John has missed the train from Rome' c. *[Mit dem Schwanz]; will Frauchen dem Hund [with the tail]; wants missy the dog