The Function of Function Words and Functional Categories
Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Ling...
63 downloads
779 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
The Function of Function Words and Functional Categories
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 Editors Werner Abraham
Elly van Gelderen
University of Vienna
Arizona State University
Advisory Editorial Board Cedric Boeckx
Ian Roberts
Harvard University
Cambridge University
Guglielmo Cinque
Ken Safir
University of Venice
Rutgers University, New Brunswick NJ
Günther Grewendorf
Lisa deMena Travis
J.W. Goethe-University, Frankfurt
McGill University
Liliane Haegeman
Sten Vikner
University of Lille, France
University of Aarhus
Hubert Haider
C. Jan-Wouter Zwart
University of Salzburg
University of Groningen
Christer Platzack University of Lund
Volume 78 The Function of Function Words and Functional Categories Edited by Marcel den Dikken and Christina M. Tortora
The Function of Function Words and Functional Categories Edited by
Marcel den Dikken The Graduate Center, The City University of New York
Christina M. Tortora The College of Staten Island and The Graduate Center, The City University of New York
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 The function of function words and functional categories / edited by Marcel den Dikken, Christina M. Tortora. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 78) Includes bibliographical references and index. 1. Grammar, Comparative and general--Function words. 2. Functionalism (Linguistics) 3. Grammar, Comparative and general-Grammatical categories. I. Dikken, Marcel den, 1965- II. Tortora, Christina. III. Linguistik aktuell ; Bd. 78. P283.F86 2005 415--dc22 isbn 90 272 2802 7 (Hb; alk. paper)
2005048395
© 2005 – 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 The function of function words and functional categories Marcel den Dikken and Christina M. Tortora Verb second as a function of Merge Jan-Wouter Zwart Nonnative acquisition of verb second: On the empirical underpinnings of universal L2 claims Ute Bohnacker
vii 1 11
41
Clause union and clausal position Josef Bayer, Tanja Schmid and Markus Bader
79
Explaining Expl Marc Richards and Theresa Biberauer
115
Reflexives in contexts of reduced valency: German vs. Dutch Marika Lekakou
155
Simple tense Guido Vanden Wyngaerd
187
Possessor licensing, definiteness and case in Scandinavian Marit Julien
217
Pronouns are determiners after all Dorian Roehrs
251
Index
287
Acknowledgements
We would like to thank several parties associated with the 19th Comparative Germanic Syntax Workshop (CGSW19) here. First, we are very grateful to Alison Gabriele, Shukhan Ng, and Erika Troseth for their help in organizing the conference. There were several other papers presented at this conference, beyond the contributions to this volume. We would like to thank the two invited speakers, Hans Bennis and Alison Henry, as well as Jonathan Bobaljik, Ellen Brandner, Hans Broekhuis, Siobhán Cottell, Vera Lee-Schoenfeld, Thomas Leu, Lisa Levinson, Erik Magnusson, and Eric Stenshoel for their contributions to the event. We thank the many colleagues who reviewed the abstracts submitted to the workshop, and would also like to express our gratitude to the reviewers of the papers submitted to this volume: Sjef Barbiers, Hans Bennis, Judy Bernstein, Jonathan Bobaljik, Cédric Boeckx, Anna Cardinaletti, Norbert Corver, Mürvet Enç, Gisbert Fanselow, Kleanthes Grohmann, Barbara Hemforth, Christer Platzack, Cecilia Poletto, Henk van Riemsdijk, Ian Roberts, Ken Safir, Cristina Schmitt, Bonnie Schwartz, Markus Steinbach, Peter Svenonius, Øystein Alexander Vangsnes, Susi Wurmbrand, and Martha Young-Scholten. Finally, we would like to thank Werner Abraham for inviting us to publish this collection of papers from CGSW19 in the Linguistik Aktuell / Linguistics Today series.
The function of function words and functional categories Marcel den Dikken and Christina M. Tortora
The papers contained in this volume were all presented at the 19th Comparative Germanic Syntax Workshop, held at The Graduate Center of The City University of New York, 3–5 June 2004. While the workshop itself had no special theme, the selected papers brought together here all address, in one way or another, the question raised by the title of the volume: what is the function of function words and functional categories? In these pages, we briefly introduce the contributions to this volume with this central theme as our guide, relating the papers to one another by presenting them in a particular order, and highlighting what we believe are their most significant theoretical and empirical results. By relating the various contributions to one another in this particular way, this introduction also serves to provide a rationale for the order in which the papers are presented in the volume. The syntax of function words and functional projections has dominated research in generative grammar in the last two decades, with perhaps the strongest impetus to this research having been given by Borer’s (1984) hypothesis that all parametric variation is confined to morpho-lexical properties of functional categories – a hypothesis that has since become the basis of work on parametric variation within various different paradigms of research, including that defined by Chomsky’s (1995 et passim) minimalist program. In his contribution to the volume, Jan-Wouter Zwart departs in an interesting way from this widely accepted hypothesis, denying that morphological properties of functional heads in the left periphery (their ‘richness’, in particular) could be responsible for the question of whether a language does or does not exhibit Verb Second. Zwart’s approach to Verb Second is profoundly different from those which take some morphological property (be it feature strength or the
Marcel den Dikken and Christina M. Tortora
‘EPP feature’) to drive the raising of the finite verb into a position in the left periphery. (It is worth highlighting in passing that while Zwart downplays the role of morphology in the domain of Verb Second, Richards and Biberauer’s contribution to this volume, reviewed below, makes a crucial appeal to morphological richness in determining the parametric differences between languages in the area of EPP-satisfaction.) Zwart’s approach also differs from ones that assume that Verb Second (like head movement phenomena in general, on Chomsky’s current assumptions) is a PF phenomenon. Rather, placing Verb Second squarely in the domain of (‘narrow’) syntax, Zwart provides a perspective which reanalyses the phenomenon as a positional dependency marking strategy. That is, the finite verb is placed at the left edge of the clause, functioning as a linker between the element in initial position (which may be a null element, or perhaps a discourse factor of sorts, as in some V1 constructions) and the remainder of the clause (the dependent). He pursues his hypothesis that Verb Second is a positional dependency marking strategy – in other words, a function of Merge, on the assumption that Merge must always create a dependency relationship – by reassessing the analysis of garden-variety V2, as well as deviations from the expected pattern, including not just V1, but also cases of V2 triggered by a conjunction, ‘repeated’ V2, and V3 word orders. Zwart discusses a variety of instantiations of V3, suggesting that the constituents ‘spoiling’ the V2 pattern either are ‘extradependent’ (i.e., the element between the sentence-initial constituent and the finite verb is an interpolation), or are ‘extracyclic’ (i.e., the element in sentence-initial position lies outside the cycle that is the locus of positional dependency marking); as such, they do not count in the computation of the dependency relationship mediated by the linker. Deviations from Verb Second are central to Ute Bohnacker’s paper as well. She thoroughly examines the oft-heard claim that V2 is hard to acquire for second language learners, focusing specifically on Swedish learners of German. While both Swedish and German are Verb Second languages, Swedish allows more deviations from the rigid V2 pattern than does German. Bohnacker demonstrates that the more flexible V2 system of Swedish results overall in more liberal deviations from V2 in the German produced by Swedish L2 learners, so that the Swedish learners’ spontaneously produced German contains V2 errors that are typical of their native Swedish structures (esp. ones involving discourse particles like så ‘so’ and sen ‘then’). This suggests that L1 transfer plays a significant role in the problem that V2 poses. Bohnacker also finds that L2 knowledge of English has a striking effect on Swedish learners’ production of V2 in German. Specifically, going beyond the V3-potential of their native Swedish, L3-learners of German with previous knowledge of English
The function of function words and functional categories
show V3 patterns in German which seem to be the result of transfer from their L2. Bohnacker’s study thus provides evidence for transfer from both L1 and L2 in the acquisition of Verb Second. Amidst all of these findings, however, Bohnacker emphasises that there is no overarching trouble with V2 in her Swedish learners of German; L1- and L2-induced errors aside, violations of the Verb Second requirement are extremely rare. This suggests that V2 per se is not difficult to acquire for second-language learners at all. This in turn suggests that UG-based hypotheses which take functional categories in the C-domain to be ‘vulnerable’ in second language learning (or which take L2-acquisition to be guided initially by a more basic SVX order) need to be reconsidered. While the first two papers in this volume are exclusively concerned with the left periphery of the root clause, Josef Bayer, Tanja Schmid and Markus Bader concentrate in their contribution on the functional superstructure of embedded control infinitives with zu, focusing on German (but also bringing Dutch and Bangla data to bear on the questions they address). The standard perspective on zu-infinitives is that when they exhibit ‘clause union’ effects, their structure is reduced such that there is no functional structure present between the matrix control verb and the VP of the infinitive; in contrast, when no ‘restructuring’ effects are exhibited, the control verb’s complement is taken to be a full-fledged infinitival CP. Bayer, Schmid and Bader start out by assessing the adequacy of the standard approach with a corpus- and questionnaire-based study. Their results show that ‘extraposed’ zu-infinitives are consistently preferred; when the zu-infinitive is not in extraposed position, it is preferably analysed as part of a mono-clausal construction. They argue that while ‘extraposed’ zu-infinitival clauses are reasonably treated as CPs with a null functional head, ‘intraposed’ zu-infinitives that exhibit no clause-union properties (which, although highly marked, do exist) cannot be taken to be null-headed CPs: they project no further than VP. Their argument derives from the fact that an intraposed, ‘non-coherent’ infinitive resists non-verbal material (such as rightward shifted PPs) at its right edge – a property which, interestingly, is shared with preverbal finite complement clauses in Bangla (which likewise must be verb-final). Bayer, Schmid and Bader exploit the fact that no restructuring is manifested in the Bangla cases to reject an account of the German adjacency facts that would have the syntax force a ‘coherent’ structure for intraposed infinitives; as such, the parser’s strong preference for a ‘coherent’ analysis is not handed down to it by the syntax. Rather, they derive the adjacency effect from a key insight of Bech’s (1955/1957) that fits in well with current thinking in minimalism (in terms of Agree), namely, that the matrix verb and the head of its zu-infinitival complement are engaged in a ‘status’-checking relationship.
Marcel den Dikken and Christina M. Tortora
Specifically, zu is a functional element prefixed to the infinitival verb that endows the verb with a particular ‘status’, and the matrix verb that selects the zu-infinitive must check this status. Status is signalled at the right edge of the infinitival clause (not at its left edge by a null C: Bangla shows clearly that complementisers in intraposed clauses cannot be initial), and since status checking would be interfered with by any non-verbal material to the right of the infinitive, such material is disallowed. This derives the adjacency effect in intraposed zu-infinitives, and, as Bayer, Schmid and Bader show, an approach along these lines can be extended beyond the verbal domain to the well-known Head-Final Filter effect in prenominal attributive modifiers. The C-head once again plays an important role in the paper by Marc Richards and Theresa Biberauer, which concerns itself with the question of how best to explain the distribution of expletives in the Germanic languages. Their paper also allows us to descend further down the tree, to Chomsky’s (1995) ‘light verb’ v, which Richards and Biberauer bring sharply into focus as well. Taking expletives to be ‘dummy arguments’, they approach the distribution of there and its ilk in the other Germanic languages by crucially exploiting the phase as a regulator of the base-generation sites of expletives. Their central hypothesis is that expletives may only be merged in the specifier positions of phase heads – C and v. The former, not surprisingly, introduces ‘high’ expletives such as German es and Icelandic það; the latter provides the merge-site of English-type there-expletives: there originates in SpecvP and raises locally to SpecTP to satisfy the EPP. Thus, there in an English expletive construction checks T’s ‘EPP feature’ in exactly the same way as do other constituents in SpecTP (i.e., by raising to that position). But why can’t an expletive be merged in SpecvP and raise to SpecTP in German and Icelandic as well? Richards and Biberauer’s answer to this question elaborates in an interesting way on Alexiadou and Anagnostopoulou’s (1998) hypothesis that the EPP can be satisfied by means other than NP-movement to SpecTP. The crux of their answer to why, say, German’s satisfaction of the EPP does not involve merging an expletive in vP with subsequent raising to SpecTP is that the EPP property of T, in German-type languages, is not in fact satisfied via raising of a nominal category (like an expletive) to SpecvP. Rather, such languages are parametrically predisposed to check T’s EPP property against the φ-features of the (rich, nominal) inflectional morphology on the verb. This forces these languages to raise all of vP into SpecTP, with vP then satisfying T’s EPP property. As a result, there is no EPP-role to play for a vP-merged expletive in such languages; they may nonetheless possess a vP-expletive (which will be redundant from the point of EPP-satisfaction) – and in fact, Richards and Biberauer argue that
The function of function words and functional categories
German da, the locative expletive, is precisely such an element. With the loss of rich inflection over time, languages that used to be able to rely on their rich, nominal inflection to have T’s EPP property satisfied via vP-raising cease to be able to meet the EPP’s demands via this route, and will come to demand that there be some nominal element in SpecvP to value T’s φ-features – a nominal expletive whenever no argument occupies SpecvP at any point in the derivation. Richards and Biberauer show that this perspective on the distribution of expletives, which ties it in directly with the distribution of ‘rich’ inflection (contra Zwart’s contribution to the volume, where the syntactic significance of morphological richness is called into question in the context of Verb Second phenomena), sheds a new and revealing light on the diachronic development of expletives in the Germanic languages. Whether an ‘expletive’ such as English there is an argument, a predicate, or a pleonastic element is a question that has given rise to a substantial amount of debate in the literature (e.g., Williams 1984; Moro 1997; Hoekstra & Mulder 1990; Den Dikken 1995; Belvin & Den Dikken 1997; Tortora 1997; Cresti & Tortora 2000; Hazout 2004; and Chomsky’s work over the years). Richards and Biberauer place themselves firmly in the expletive camp, analysing there as a ‘dummy argument’ – essentially a function word, therefore. The question of an element’s status as an argument or a ‘dummy’ is one that has also permeated the literature on reflexive markers such as German sich and Dutch zich. Marika Lekakou enters into this particular debate by arguing in detail that while German sich can be either an argument or what she calls a marker of valency reduction, Dutch zich is systematically an argument. She makes the interesting claim that this difference between the two cognate elements is rooted in the different organisations of the reflexive paradigms of the two languages. While the Dutch non-inherent complex reflexive zichzelf exists alongside the inherent reflexive zich, German sich selbst (in contrast with Dutch zichzelf ) is not a member of the reflexive paradigm in its own right; rather, it is merely an emphatic version of sich. She argues on this basis that the Dutch weak reflexive zich is specified for the feature [+inherent reflexive], while its German counterpart sich is not. German’s lack of a complex reflexive and the concomitant featural difference between zich and sich in turn has immediate consequences for the distribution of these elements in middles. Lekakou argues that a reflexive specified as [+inherent] cannot be used in anticausatives and middles – which accurately takes care of one of the more conspicuous differences between Dutch and German, namely, the fact that German middles systematically (and anticausatives predominantly) feature sich, while their (standard) Dutch counterparts do not (although Heerlen Dutch does have zich in middles; Lekakou
Marcel den Dikken and Christina M. Tortora
briefly discusses the repercussions of this and other properties that set Heerlen Dutch aside from standard Dutch). Middle constructions constitute one of the few contexts in which English allows the simple present tense to be used with verbs that are inherently eventive (e.g., this book reads easily). This difference is due by and large to the fact that middles are typically used as world-structure statements, not episodically. But there are a number of contexts in which the simple present in English can be used episodically, with reference to an event that is actually unfolding at the present time: sports commentaries (Beckham shoots and scores!) and performatives (I hereby pronounce you man and wife) are well-known cases in point. In his contribution, Guido Vanden Wyngaerd studies these contexts in detail. His central observation is that whenever an eventive verb in the English simple present is used to refer to the hic et nunc, the event denoted must be of ‘Very Short Duration’ (in contrast with languages like Dutch). He derives this from a novel approach to the Reichenbachian ‘speech time’ (S): in particular, while S is commonly understood in the tense literature as a point, Vanden Wyngaerd proposes that S is actually an interval, albeit a very short one. On the assumption that the event denoted must fit into this very short interval that represents the speech time, it follows that English present-tense eventive constructions can only make reference to an ongoing event at the present time if the event in question is of ‘Very Short Duration’. Vanden Wyngaerd goes on to show that his approach to the English simple present also accommodates its compatibility with stative verbs and generic sentences, proposing with regard to the former that they have point duration (and therefore fit into the very short interval representing the speech time as a matter of course). With regard to the latter, he proposes that they are of a kind with Individual Level predicates – i.e., predicates that ascribe a property to an entity without concern for its internal temporal make-up, and that, therefore, have a stative interpretation. Vanden Wyngaerd argues that the peculiar property of the English simple present which demands that events fit into the short interval of the speech time is not a property of tense itself but instead can be derived from the aspectual properties that distinguish English from other languages. While English aspectual distinctions between the perfect, the progressive, the perfect progressive and the rest are encoded with the aid of two binary features, [±extension] and [±completion], other languages employ just the single binary feature [±completion] to make the much simpler distinction between the perfect and the imperfect. Thus, English is more restricted in its use of the simple present because it has a richer feature inventory for aspect – much like, on Lekakou’s analysis of middles and reflexives, the idea that Dutch is more restrictive in
The function of function words and functional categories
its use of the simple reflexive because it has a richer feature inventory for reflexivity. At its core, Vanden Wyngaerd’s paper is a study that concerns itself with properties of the functional markers of tense and aspect in the structure of the clause. Like the other papers reviewed in the preceding paragraphs, therefore, it concentrates, in one way or another, on functional categories and/or function words in the sentential domain. The two final contributions to the volume, by contrast, address the structure of the noun phrase, once again with key reference to its functional elements and functional structure. Marit Julien’s paper is a detailed study of possessive noun phrases throughout Scandinavian, bringing together in a comprehensive way the empirical facts, and discussing them against the background of a uniform base configuration, with surface variation resulting from movement operations in the course of the overt-syntactic derivation. Base-generating possessors in SpecNP and postulating a NumP, an nP and an additional functional projection provisionally labelled ‘αP’ between the lexical NP and the D-head, Julien derives postnominal possessors via raising of the head of the possessed noun phrase to Num and further up to n, which licenses the (P-less) postnominal possessor (via agreement in the case of pronominal possessors, and via genitive Case in the case of full-nominal possessors). And by having possessors that are not licensed inside nP move to SpecDP, she accounts for surface orders in which possessors precede the rest of the noun phrase, with the nP-external functional head ‘Poss’ (realised as -s or a pronoun-like possessive element) then taking care of the licensing of the possessor. To get to SpecDP, the raised possessor must land in SpecnP on its way up, nP being a (strong) phase. Julien exploits this touch-down in SpecnP to account for the fact that prenominal possessors preclude the realisation of a suffixed definiteness marker on the possessed head noun. That is, assuming that this definiteness suffix is a spell-out of n, she recasts its absence in the presence of a prenominal possessor as a kind of ‘doubly-filled Comp’ effect: with SpecnP filled by the raised possessor (at the relevant stage of the derivation), n will remain empty. Julien goes on to meticulously demonstrate and derive the fact that prenominal possessors systematically make the possessed noun phrase outwardly definite (even if the possessors themselves are indefinite, despite persistent claims to the contrary in the literature on the ‘Saxon genitive’). She further derives the fact that postnominal pronominal possessors contribute their definiteness to the possessed noun phrase as well, as do postnominal fullnominal possessors (except those, found in older varieties of Scandinavian and in some varieties of Icelandic, that have a lexical (i.e., non-structural) genitive
Marcel den Dikken and Christina M. Tortora
Case). She establishes a connection between the definiteness of the possessed noun phrase, the presence of a [poss/def] feature on the possessor, and an agreement relationship between the possessor and n – but she argues that this Agree relationship between n and the possessor, which results in definiteness agreement, is not linked to Case; pronominal possessors agree with n but do not get their Case checked by it (instead, their Case comes from outside the possessed noun phrase altogether). She thus disconfirms the inextricable link between agreement and structural Case advocated in Chomsky’s recent work. While Julien is mostly concerned with the area between D and the lexical NP, Dorian Roehrs zooms in specifically on the left periphery of the extended noun phrase, looking at fillers of the D-head. He brings a battery of novel arguments (mostly from German) to bear on the question of whether pronouns are in D or not. Taking as its primary object of study the syntax of pronoun-noun constructions of the type us linguists, Roehrs’s central claim is that the Postalian approach to these phrases is correct: the pronoun sits in D. However, Roehrs argues that it is not actually born there; rather, all determiners, including pronouns, raise to D from a lower position in the DP, namely, the head of D’s ‘artP’ complement. The common or proper noun that may follow the pronoun (as in German du Idiot ‘you idiot’ or du Willi (du) ‘you “Willy”’, respectively) heads the complement of ‘art’, and serves as a predicate. This explains the fact that in du Willi (du), the proper noun is interpreted the way it is in predicate nominal constructions such as Du bist vielleicht *(ein) Willi ‘you are really a Willy’, parallel (both syntactically and interpretively) to Du bist vielleicht *(ein) Idiot ‘you are really an idiot’. The fact that an indefinite article is obligatory on the predicate nominal in these copular sentences, combined with the fact that no article of any kind is possible on the (common or proper) noun following the pronoun in the pronoun-noun construction, then gives Roehrs a first argument against an apposition approach to noun phrases such as du Idiot and du Willi (du) (which would entail that the projections of the pronoun and the common noun are in an appositive relationship), and in favour of his own complementation analysis. A detailed analysis of the difficult tangle of facts involving the distribution of strong and weak agreement inflection on prenominal adjectival modifiers of the common noun phrase following the pronoun gives him a second cogent argument in favour of a syntactic structure of pronoun-noun constructions that has the pronoun occupying D. Roehrs concludes his discussion by arguing that the tripartite typology of pronominal types laid out by Cardinaletti and Starke (1999) for free-standing pronouns is reproduced in its entirety in the realm of transitive pronouns (i.e., pronouns that take a common or proper noun phrase complement). That is, they, too, come in strong,
The function of function words and functional categories
weak and clitic forms. In sum, pronouns, regardless of whether they are on their own or accompanied by a complement, are a subspecies of determiners – hence quintessentially function words. Before closing, one final word is in order concerning this volume, which we believe has naturally emerged as a coherent collection of works with a common thread, reflected in the book’s title (and in this introduction). As an outgrowth of the Comparative Germanic Syntax Workshop, an additional feature shared by all of these contributions is their particular approach to problems in syntactic theory. Specifically, by focusing on the Germanic languages, each paper is concerned with the study of micro-parametric variation, whereby a number of overarching syntactic features shared by closely related languages are held constant, while minimal differences between the varieties are isolated (thus allowing the researcher to minimise the potential for confounding factors). This approach to the study of syntax has proven to be quite successful in recent years, having informed the theory in ways which the comparison of unrelated languages does not so readily afford. We are thus pleased that this selection of papers from the workshop allows us to present a collection which, in addition to having organically produced the common thread of functional syntax, also coheres with respect to an approach that is at the foundation of much exciting and successful work in the field today.
References Alexiadou, A. & E. Anagnostopoulou (1998). Parametrizing Agr: Word order, V-movement and EPP-checking. Natural Language and Linguistic Theory, 16, 491–539. Bech, G. (1955/1957). Studien über das deutsche Verbum Infinitum. (Det Kongeliege Danske Vi-denskabers Selskab; Dan. Hist. Filol. Medd. Bind 35, no. 2 (1955) & Bind 36, no. 6 (1957)). New edition 1983. Tübingen: Niemeyer. Belvin, R. & M. den Dikken (1997). There, happens, to, be, have. Lingua, 101, 151–183. Borer, H. (1984). Parametric Syntax: Case studies in semitic and romance languages. Dordrecht: Foris. Cardinaletti, A. & M. Starke (1999). The typology of structural deficiency: A case study of the three classes of pronouns. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe. Berlin: Mouton de Gruyter. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Cresti, D. & C. Tortora (2000). Aspects of locative doubling and resultative predication. In S. Chang, L. Liaw, & J. Ruppenhofer (Eds.), Proceedings of the 25th Annual Meeting of the Berkeley Linguistics Society (BLS25) (pp. 62–73). Dikken, M. den (1995). Binding, expletives, and levels, Linguistic Inquiry, 26, 347–354. Hazout, I. (2004). The syntax of existential constructions. Linguistic Inquiry, 35(3), 393–430.
Marcel den Dikken and Christina M. Tortora
Hoekstra, T. & R. Mulder (1990). Unergatives as copular verbs: Locational and existential predication. The Linguistic Review, 7, 1–79. Moro, A. (1997). The Raising of Predicates. Cambridge: Cambridge University Press. Tortora, C. (1997). The Syntax and Semantics of the Weak Locative. PhD dissertation, University of Delaware. Williams, E. (1984). There-insertion. Linguistic Inquiry, 15, 131–153.
Verb second as a function of Merge* Jan-Wouter Zwart
This article proposes a new approach to the Germanic verb-second phenomenon (V2). The background assumption is that the structure building operation Merge, which joins two elements, automatically creates a dependency relation between the elements merged, and that the dependency may be marked either by morphology or by position. In the latter case, a term of the dependent element is realized on the left edge of the dependent element, as a linker. The proposal is that verb second is positional marking of a dependency relation between the first constituent (subject, topic, wh-phrase, etc.) and its sister, via placement of the verb at that sister constituent’s left edge. The proposal makes it possible to incorporate a range of recalcitrant V2 phenomena within a unified theory of V2.
.
Introduction
A verb second (V2) construction is one in which the verb (by rule) appears directly after the first constituent. In this paper I propose to describe V2 as the positional marking of a dependency relation. The approach assumes that there is only one structure generating procedure in syntax, which is applied iteratively to the output of a previous application, Merge: (1) Merge Add x to y yielding <x,y>
I hypothesize that Merge as defined in (1) automatically creates a dependency relation S (for sisterhood) where x is invariably the antecedent (or nondependent) and y the dependent.1 I suggest furthermore that S can be (and perhaps universally is) marked on y, and spelled out on one of the terms of y. The proposal of this paper is that this dependency marking may be realized in two ways: by inflectional mor-
Jan-Wouter Zwart
phology (tense, agreement marking) or by position. In particular, the proposal is that V2 is positional marking of the relation between a fronted element and its sister, to the effect that the term of y spelling out S is realized as the leftmost element in y. It can be seen that V2, on this proposal, is really a verb-first requirement applied to the domain of the dependent in a dependency relation. This eliminates a property that has always been commented on as strange, namely reference to an arbitrary number (two) in the description of the pattern. It raises the question, though, why no V-last counterpart to V2 seems to exist (i.e. no pattern that requires fronting of x, creating <x,y>, to be accompanied by realization of the verb on the right edge of the dependent y – an asymmetry first noted by Kayne 1992, to my knowledge). I suggest that the verb in V2constructions is a member of a larger class of elements described as ‘linkers’, appearing more generally in constructions of predication or modification. We define linker as in (2): (2) Linker A linker is an element marking the left edge of y only when y is a dependent.
Linkers, then, are positional markers appearing as a function of Merge. There is no general requirement that a linker be a shifted term of the dependent y, it may also be a dummy element. While the V2-position is realized by a shifted verb in languages like Dutch, English uses either auxiliary verbs or dummies like did: (3) a. John kissed Mary b. Why did John kiss Mary ?
In (3b), x = why, y = John kiss Mary, and did is the linker between x and y appearing at the left edge of y. It will be seen that the linker, if present, is the designated element spelling out morphological dependency as well.Thus, the linker may be a dummy expressing tense, which is then (after fronting of x yielding <x,y>) realized in the V2-position to mark the dependency between x and y, as in Warlpiri-type ‘Aux second’ phenomena (Hale 1973), and, more generally, the ‘tense-second’ phenomena discussed by Koster (2003).2 The approach suggests that verb-first (V1) clauses (in languages showing V2 otherwise) are themselves dependents, so that the V1-effect is identical to the V2-effect, with the antecedent x unexpressed.3
Verb second as a function of Merge
The proposal made here will remain silent on the phrase structural realization of the linker (the left-edge element). There appears to be no objection to viewing the linker’s position as a head position, staying close to the analysis of V2 shaped by X´-theory (e.g. Chomsky 1986: 6). However, the analysis does not allow us to predicate any properties of that head position, in particular, to ascribe any agency to that head position or to any morphosyntactic features residing in it. In other words, our proposal entails that V2 is not triggered by the need to acquire, check, assign or eliminate formal features, and that there is at best an indirect connection with the presence of tense or agreement features within the clause. I submit that connections between morphosyntactic features and verb placement, if there are any, are to be explained by a consistency requirement of the type in (4): (4) Consistency If α, a term of y, spells out a dependency of y positionally, it also does so morphologically.
(4) follows on the conjecture that the linker in a dependent y (i.e. the verb in a V2 construction) has no other function than to spell out the dependency of y towards some x. Importantly, (4) works only in one direction, since not all languages employ the device of positional marking, and few languages (if any) employ it in all constructions. . General V2 properties The general aspects of the V2-phenomenon that the proposed analysis covers are: (5) General aspects of V2 a. V2 is the side-effect of a fronting operation b. Modulo parametric variation, V2 is insensitive to the type of element fronted
Traditional approaches to V2 concentrate on a general requirement forcing the verb (in independent clauses) to move to a position (C) occupied by the complementizer in embedded clauses.4 A second operation then moves an arbitrary category to a position to the left of C (later identified as the specifier position of CP), triggered by the V2 constraint: (6) The V2 constraint The verb must be second
Jan-Wouter Zwart
The V2 constraint (6) is unsatisfactory in that it predicates some requirement of the verb and triggers movement of some other category. Moreover, the movements satisfying the V2 constraint (subject placement, topicalization, expletive insertion, wh-movement) exist in non-V2 languages as well, suggesting that other triggers, bearing no relation to V2, are in force. If we take these triggers seriously, we may have to formulate the V2-phenomenon as in (7), with (7b) replacing the V2-constraint (6) when YP (= <x,y>) is the root: (7) a.
Merge x (= XP), a term of y (= Y’) with y (i.e. Move XP to its designated position Spec,YP) b. Move the verb to Y
We thus see a shift from a particular verb-movement trigger accompanied by generic XP-movements to particular (triggered) XP-movements accompanied by a generic verb movement. This shift entails that the target for the XPmovement (and hence the target for verb movement) may be variable, leading to a more dynamic analysis of the V2 pattern where the verb does not always occupy the position C (see Travis 1984, 1991; Zwart 1993) and a more dynamic analysis of clause structure more generally (Zwart 2003–2004).5 The general aspects of the V2 phenomenon in (5), captured more or less successfully by traditional approaches to V2, are covered by the analysis proposed here as well.
. Problems associated with V2 More interesting are particular problematic phenomena associated with the V2-pattern, some of which are listed in (8): (8) Difficult facts associated with V2 a. V2 asymmetries (between main and embedded clauses; construction specific ones; having to do with finiteness); b. nonstandard V2 phenomena (quotative inversion, conjunction-triggered inversion, apokoinou constructions); c. V2 deviations (V1, V3).
. V2 asymmetries Whether or not a language uses positional marking must be stipulated for each dependency. In Germanic, and perhaps universally, positional marking appears
Verb second as a function of Merge
to be limited to dependencies marking the end of a derivation, or the end of a well-defined subpart of the derivation. We call such a finite sequence of operations Merge a cycle, and state: (9) Positional dependency marking is limited to operations constituting a cycle.
A cycle is constituted as specified in (10): (10) Cycle In the unmarked case, a cycle is constituted iff: (a) no further operation Merge takes place, or (b) the nondependent is a lexical term (i.e. a noun, verb, or adjective).
This means that a root clause will constitute a cycle (a case of (10a)) and that the combination of a verb and an embedded clause will also constitute a cycle (a case of (10b)).6 Subject-initial root clauses, then, are the result of a sequence of operations Merge constituting a cycle. The final dependency relation, where x = the subject and y = the subject’s sister, is positionally marked in Continental WestGermanic and North Germanic, with the finite verb appearing as a linker at the left edge of y. The situation is different with embedded clauses, where a cycle is ended only where x = V and y = the embedded clause. In that case, the complementizer appears to function as the linker marking the dependency positionally. But the dependency between the subject of the embedded clause and its sister is not positionally marked, as this dependency does not mark the end of a cycle. It will be seen that this captures the traditional observation that the verb and the complementizer in V2-languages vie for a single position (Paardekooper 1955: 97; Den Besten 1977). As Den Besten showed, the fronted verb and the complementizer share a common distributional pattern, exemplified in (11) from Dutch, where both the verb and the complementizer appear to the immediate left of the weak pronoun subject ze: (11)
Jan-Wouter Zwart
However, not all fronted verbs share this distributional characteristic: in subject-initial main clauses, the fronted verb follows the subject: (11)
It is not straightforward, therefore, that the fronted verb always occupies the complementizer position, even if the competition between the finite verb and the complementizer seems real. In our proposal, the competition is not about an absolute position, but about the relative position constituted by the left edge of the dependent. If the size of the dependent is the same (i.e. a proposition including a subject), as in (11a) and (11b), so are the positions of the dependency markers (the complementizer and the verb). But if the size of the dependent varies, as in (11b) and (11c), so does the position of the dependency marker (the finite verb). The proposal that the complementizer is a positional dependency marker explains a curious and hitherto unexplained fact, namely that the specifier position of a declarative complementizer (dat in Dutch, dass in German, etc.) may not be occupied. Thus, fronting of an adverb in a root clause yields V2, but fronting across C in embedded clauses is impossible. Instead, the fronted adverb appears to the right of the complementizer (examples from Dutch): (12) a.
Gisteren heeft Jan Marie gekust yesterday has John Mary kiss-part ‘Yesterday John kissed Mary.’ b. *Ik heb gezegd [ gisteren dat Jan I have say-part yesterday that John c. Ik heb gezegd [ dat gisteren Jan I have say-part that yesterday John ‘I said that yesterday John kissed Mary.’
Marie Mary Marie Mary
gekust kiss-part gekust kiss-part
heeft ] has heeft ] has
This pattern is explained if the complementizer is a linker marking the dependency between the embedded clause and the matrix verb by appearing as the embedded clause’s leftmost element. Languages using positional dependency marking may differ as to which dependency they choose to mark positionally. Nothing excludes that a language views the combination of a subject and its sister in an embedded clause as the end of a cycle in need of positional marking (perhaps one of the marked cases of Note 6). This yields the embedded V2 phenomenon of Icelandic and Yiddish.7
Verb second as a function of Merge
Construction specific asymmetries are in evidence in residual V2 languages like English, where only the fronting of particular operator-like elements sets up a dependency which is positionally marked (as in (3b)). Here, little more needs to be said. As before, the positional marking requirement disappears in embedded clauses, suggesting that the relevant cycle is established only after merger with the matrix clause verb: (13) I wonder why (*did) John kiss *(ed) Mary
It is, however, remarkable that Germanic embedded interrogatives are rarely positionally marked when the embedded interrogative does not correspond to a yes/no-question: (14) a. I wonder if John kissed Mary b. I wonder (*if) why John kissed Mary
But cases like (14b) do exist, as noted by Hoekstra (1994) for the Dutch dialect spoken in the city of Amsterdam:8 (15) We moeten eens vragen of waar die heen gaat we must once ask-inf if where dem dir.prt goes ‘We should ask where he’s going.’
The logic of our analysis suggests that cases like (15) are in a sense unmarked, with the complementizer functioning as a linker between the matrix verb and the embedded interrogative. More common, however, is the pattern of some Germanic dialects allowing complementizers to appear after the wh-phrase (example from Dutch): (16) Ik wou weten waarom of dat Jan dat gedaan had I wanted know-inf why if that John that do-part had ‘I wanted to know why John did that.’
This suggests that in this particular construction (embedded wh-interrogatives) there is a tendency to mark the dependency between the wh-phrase and the proposition in its scope rather than the entire embedded clause. Parallel to (16) is the use of the verb as a linker in embedded wh-questions in Spoken Afrikaans (example from Biberauer 2002: 37): (17) Ek wonder wat het hy vandag weer aangevang I wonder what has he today again started ‘I wonder what he started today again.’
Jan-Wouter Zwart
Similarly, dialects of English spoken in Northern Ireland use a dummy verb as the linker between the verb and its complement clause and between the whphrase and its sister in embedded wh-questions (Henry 1995: 105ff.; data from Adger 2003: 343): (18) a. I asked did Medea poison Jason b. I asked who did Medea poison
One possibility explaining the choice of the linker in (16) could be that Dutch uses the complementizer as a dummy linker in these particular cases, on analogy with embedded yes/no-questions. Another asymmetry connected with V2 is that between finite and nonfinite verbs. Infinitives are not called upon as positional dependent markers in WestGermanic (i.e. they do not undergo V2).9 Nonfinite clauses in Dutch appear in two types of constructions, extraposed (19a) and interlaced with the matrix clause (19b): (19) a.
. . . dat Jan probeerde (om) het boek te lezen that John tried for the book to read-inf b. . . . dat Jan het boek probeerde (*om) te lezen that John the book tried for to read-inf Both: ‘. . . that John tried to read the book.’
We may take the complementizer om in extraposed infinitive clauses as a dummy positional dependent marker (a linker), blocking verb movement as in finite embedded clauses. In the type of (19b) (traditionally referred to as ‘verb raising’), material belonging to the embedded clause (such as het boek ‘the book’ in (19b)) is remerged to a constituent containing the matrix verb (probeerde ‘tried’ in (19b)), and the verbs appear to form a cluster. The embedded clauses in this type of construction are generally taken to be defective or transparent, suggesting that in our terms they will not constitute a cycle. If so, no positional dependent marking is called for.10 . Nonstandard V2 phenomena Nonstandard V2 phenomena include various types, some of which have received little or no treatment in the theoretical literature. .. Quotative inversion Most familiar will be the type of quotative inversion (Collins & Branigan 1997):
Verb second as a function of Merge
(20) I am so sick said John
√
( John said)
In English, quotative inversion is optional, apparently a residu of earlier English where V2 was much more pervasive. In strict V2 languages like Dutch and German it is obligatory: (21) Ik voel me zo ziek zei Jan (*Jan zei) I feel me so sick said John
Let us call the part exemplified by said John/zei Jan the quotative, and the part preceding the quotative the quote.11 The prosodic properties of the quotative, then, suggest that it be treated as backgrounded material: the intonation is low and flat, shown by Zwart (2002) to be characteristic of backgrounding in Germanic (cf. also Collins & Branigan 1997: 12). Backgrounding can be illustrated in various constructions, the most familiar of which will be right dislocation (example from Dutch, with small print indicating low pitch): (22) Ik ken hem niet die jongen I know him not that boy
Zwart (2002) argues that backgrounded material is generated in a high specifier position (i.e. merged last, in a bottom-up derivation), after which the remainder of the clause moves across it to the left (i.e. is remerged with the backgrounded material), inverting both the hierarchical and the linear ordering: (23)
The remainder can be a fully expanded clause, as in (24): (24) Waar komt hij vandaan die jongen ? where comes he hence that boy
The wh-phrase waar ‘where’ indicates that the remainder should be a CP, with V2 triggered by the fronting of the wh-phrase. It follows that the backgrounded material must occupy a position higher than CP, which is currently uncharted territory. If quotative inversion involves backgrounding, the quote = the remainder and the quotative = the background. Quotative inversion then takes the quote to a part of the structure that is beyond CP. On current understanding, then, the target for the V-movement is not C and is not associated with any formal features triggering the verb. That makes it a nonstandard V2-phenomenon.
Jan-Wouter Zwart
On the approach to V2 attempted here, quotative inversion is just another case of positional dependent marking. When the quote raises across the quotative, a dependency is created in which the quote = x (the antecedent) and the quotative = y (the dependent), and the verb appears at the left edge of the dependent.12 .. Conjunction-induced inversion Another nonstandard V2-phenomenon is conjunction-induced inversion, scorned by normative grammarians, but attested in many Germanic dialects at one stage or other: (25) Alles is nu reeds bepaald en kan ik hierin tot mijn spijt all is now already settled and can I herein to my regret moeilijk veranderingen maken hardly changes make ‘Everything is already settled and it is regretfully difficult for me to make any changes.’ (from a Dutch letter by Jan Toorop 1858–1929; in Van der Horst & Van der Horst 1999: 298)
It is attested in (at least) Old and Middle English (Kellner 1924: 289–290), Old, Middle, and Early Modern High German (Paul 1919: 78–81; Behaghel 1932: 31–36), Middle and Early Modern Dutch, surviving in written Dutch until around 1930 (Stoett 1923: 231; Van der Horst & Van der Horst 1999: 296– 299), Old and Early Modern Swedish (De Boor 1977: 195; Magnusson 2004), and Old French (Foulet 1963: 120, 287). It was originally certainly a feature of the spoken language, witness its appearance in isolated dialects such as Siberian Mennonite Low German (Jedig 1969: 145). This type of construction, called ‘Tante Betje’ in the Dutch tradition (after Charivarius 1940), is problematic for traditional approaches to V2, since the element inducing it is not a phrase but a head (the conjunction en ‘and’). But if we follow Kayne (1994) and Munn (1993) in taking coordination to involve regular X´-structure, with the conjunction taking the second member of the coordination as its complement, merge establishes a pair <x,y> with en = x (the antecedent) and y (the dependent) = the second member of the coordination: (25) en
kan ik ...
Verb second as a function of Merge
If we then take the combination of a conjunction and the second member to constitute a cycle, the inversion in the second member can again be described as positional dependent marking. More generally, traditional approaches to V2 are unable to account for inversion not triggered by fronting or merger of a phrase. The approach contemplated here is insensitive to the phrase structure status of the antecedent in the relevant dependency.13 .. Apokoinou constructions A third non-standard V2-phenomenon features in apokoinou constructions (Dutch ‘herhalingsconstructies’) of the type studied in De Vries (1910– 1911: Chapter 5) and Jansen (1981: Chapter 7), where the finite verb appears twice:14 (26) En dan was je tegenstander was neer and then was your opponent was down
(93)
While there appear to be various subtypes, the one illustrated in (26) can be analyzed as involving competition between the subject je tegenstander ‘your opponent’ and the topic dan ‘then’ for the first position of the clause (cf. Zwart 1998: 383). Abstracting away from verb placement, we get the following dependencies, where S1 is nested within S2 : (27) a. S1 < je tegenstander, neer was > b. S2 < dan, je tegenstander neer was (= S1 ) >
The apokoinou construction then results when both S1 and S2 show positional dependent marking (i.e. V2) with the antecedent of S1 ending up as the pivot in the final construction (called ‘overloopdeel’ in Sassen 1967).15 In this connection it is important to note that the apokoinou construction is a single utterance, constituting one prosodic domain (featuring a single nuclear pitch accent, on neer ‘down’ in (26)) and with various local dependencies holding between the parts preceding and following the pivot, including Negative Polarity Item (NPI) licensing (28a), selection (28b), and focus association (28c) (the pivot is in square brackets):16 (28) a.
en dan hoefde je [ vroeger ] hoefde je niet and then needed:npi you earlier needed:npi you not (93) naar de neutrale hoek to the neutral corner ‘In the old days you were not required to go to the neutral corner.’
Jan-Wouter Zwart
b. ik sta me [ op een morgen ] sta ik me te I stand me:refl on a morning stand I me:refl to (123) scheren shave ‘I’m shaving myself one morning.’ c. maar ik heb toen [ WEL ] heb ik toen [ drie keer but I have then foc-prt have I then three times kort na mekaar ] heb ik toen tegen Van Dam short after each other have I then against Van Dam (184) gebokst fought ‘But I did fight against Van Dam in those days, three times shortly after one another.’
In (28a), the NPI hoefde ‘needed’ preceding the pivot is licensed by the negation niet ‘not’ in the part following the pivot. In (28b) the reflexively used first person object pronoun me ‘me’ in the part preceding the pivot is selected by the verb scheren ‘shave’ in the part following the pivot. In (28c), a double apokoinou construction, the affirmative focus particle wel in the part preceding the second pivot (it is in fact the pivot of the first apokoinou construction) is associated with the focused object Van Dam in the part following the second pivot.17 In another type of apokoinou construction, the pivot is not a subject but a focused constituent (also (28c)): (29) ik heb [ nooit van mijn leven ] heb ik een wedstrijd gebokst I have never of my life have I a match fought (125) die gemaakt was rel fixed was ‘Never in my life have I fought a match that was fixed.’
In those cases, the subject also appears twice (ik in (28c) and (29)). Here the competition appears to be between a focus-initial and a subject-initial construction, yielding the pairs in (30) for (29): (30) a. S1 < nooit van mijn leven, ik een match gebokst heb. . . > b. S2 < ik, nooit van mijn leven heb ik een match gebokst. . . ( = S1 ) >
What is special about this type is the doubling of the subject in addition to the doubling of the verb. Accepting this as a special feature, the verb placement follows as described above, as positional marking of the dependent in each pair.
Verb second as a function of Merge
Another remarkable feature of the apokoinou construction is that the two verbs need not be identical. In those cases, the first (leftmost) verb is always less specific than the second: (31) Dat was [ in ’35 ] zal dat geweest zijn that was in 1935 mod:prob that be:part be ‘That must have been in 1935.’
(149)
Here a verbal complex consisting of a modal auxiliary zal, a perfective auxiliary zijn ‘be’, and a participle geweest ‘been’ is doubled by the simple copula was ‘was’. This might be taken as an indication that the doubled (leftmost) verb in the apokoinou construction is really a dummy, which may or may not be identical to the original verb. . V2 deviations18 Languages characterized by the V2 phenomenon regularly show deviations from the V2 pattern in which the verb shows up in first (V1) or third (V3) position (see Thráinsson 1986 for an early discussion). On our approach, these deviating patterns arise under two related circumstances: (32) a. V1: the cycle functions as a dependent b. V3: the dependent functions as a cycle
In other words, given a pair M = <x,y> with tree structure representation (33), the unmarked case is that where M constitutes a cycle, y is a dependent. The special cases in (32) then specify that either M is a dependent (in addition to being a cycle) or y is a cycle (in addition to being a dependent). (33)
M x
y
Positional dependency marking that spells out a verb at the dependent left edge then yields V1 if M is a dependent and V3 if y is a cycle. It remains to determine, then, under what circumstances these special situations may occur. .. V1 A striking fact about V1 constructions in V2 languages is that they are never independent declarative expressions. They can be classified as in (34), with examples from Dutch:
Jan-Wouter Zwart
(34) V1 constructions in Dutch a. yes/no-questions Kom je ook? come you too ‘Are you coming too?’ b. imperatives Kom nou eens op tijd! come now once on time ‘Be on time for a change!’ c. conditionals Kom je op tijd dan kun je mee eten come you on time than can you with eat ‘Be on time and you can join us for dinner.’ d. counterfactuals Was jij op tijd gekomen dan was er niks was you on time come-part then was there nothing gebeurd happen-part ‘If you had been on time, nothing would have happened.’ e. narrative inversions (connected discourse constructions) Kom ik daar binnen, zegt die vent . . . come I there inside says that guy ‘So I come in, and this guy says . . . ’ f. topic drop Ken ik niet know I not ‘Don’t know him/her/it.’
For most of these construction types, an analysis has been proposed in which the first constituent is an empty operator (see Katz & Postal 1964; Baker 1971; Huang 1984; Zwart 1997: 219 among others). If so, these reduce to ordinary V2 constructions. In the analysis contemplated here, the fronted verb would be a linker marking the dependency of the clause to an empty antecedent. A solid piece of argumentation in defense of empty operators in these constructions revolves around the fact that each V1 construction allows just a single interpretation, i.e. a topic drop construction cannot at the same time be interpreted as a yes/no-question or a conditional, etc. (Cardinaletti 1990). This suggests that each construction involves a designated empty element. It has been noted, however, that the empty operator itself has to be “sanctioned by preceding discourse or by pragmatics” (Cardinaletti 1990: 78). If this
Verb second as a function of Merge
is correct, the function of the fronted verb (on our assumptions) appears to be to signal the presence of an empty antecedent (the empty operator), which is itself a dependent of some discourse factor. This raises the question whether we could not interpret V1 directly as a linguistic sign indicating dependency of the construction on a discourse factor.19 In connection with this, it may be noted that the empty operator (Q) proposed for yes/no-questions is argued by Katz and Postal (1964: 97) and Baker (1970: 197) to be present in both yes/no-questions and wh-questions. If so, and if a fronted wh-phrase triggers inversion in wh-questions, it is not clear what triggers inversion in yes/no-questions. Also relevant is the observation that a declarative clause with the proper intonation is interpreted as a yes/no-question: (35) Je komt OOK? you come too ‘You’re coming too?’
This shows that the inversion does not correspond directly to a particular aspect of the semantics, but rather (as Katz & Postal 1964) argue, to pragmatics: whereas (35) expresses surprise about a state of affairs, (34a) expresses a request for information. We may hypothesize that the V1 character of yes/no-questions signals dependency of such an implicit request. Similarly, then, with imperatives, where the V1 character may signal dependency of an implicit statement conveying a directive (cf. Katz & Postal’s 1964: 76 I-morpheme). Again, neutral order V2-clauses may also function as imperatives, but these are not imperatives in the syntactic sense:20 (36) a.
Je trouwt met Govert! you marry with Govert ‘You are going to marry Govert!’ b. Je MOET met Govert trouwen! you must with Govert marry ‘You must marry Govert!’
That conditionals and counterfactuals are not independent utterances needs no argumentation.21 Finally, in narrative inversions and topic drop constructions, the element of discourse connectedness is obvious. I would like to suggest that, rather than stating that an empty operator is present which requires sanctioning by preceding discourse, the V1 character of the construction as a whole signals dependency of the relevant discourse factor.
Jan-Wouter Zwart
Summarizing, V1 constructions in V2-languages are all characterized by a perceived dependency of the construction as a whole to some factor of discourse organization or pragmatics. This raises the question of why discourse factors play a prominent role precisely in these V1 constructions. One answer, entirely consistent with the V2 analysis proposed here, is that discourse dependency is a function of the presence of an empty operator. But an interesting perspective opened up by the analysis proposed here would seem to be that it is the V1 character itself which suggests that the construction as a whole is a dependent. That V1 has this effect then follows from the hypothesis that a verb (in V2 languages) is fronted only to function as a positional dependency marking device. .. V3 V3-orders are commonplace in residual V2 or non-V2 languages (like Modern English), but we are interested here in V3-orders in strict V2-languages. Residual V2-languages with a history of strict V2 appear to have lost V2 in topicalizations first, then in subject-initial declarative constructions, and finally in wh-constructions. Thus we find reports of Dutch dialects with strict V3-orders in topicalizations (data from the dialect of Oostende, Winkler 1874: 364): (37) Zonder entwat te zeggen Wansje loat zen zwiins achter without something to say Wansje leaves his pigs behind ‘Without saying anything, Wansje leaves his pigs behind.’
The finite verb loat ‘lets’ in (37) is adjacent to the subject, as in ordinary subject-initial V2-constructions, suggesting that the combination of a subject and its sister is taken to constitute a cycle in this dialect. One way to align this type of V3 with standard V2 is to assign a special status to either the topic or the subject, such that they may be ignored in identifying the cycle relevant to positional dependency marking (cf. (9)–(10)). In principle, V3 (in a V2-language) may come about in two ways, given four constituents w, x, y, and z, merged in such a way that the structure in (38) results: (38) w
x y
z
First, the pair may be interpreted as constituting a cycle, while the pair <w,x> is not. This would result in positional marking of z, but not of x, yielding
Verb second as a function of Merge
V3. I propose to call w in this situation ‘extracyclic’ (falling outside the cycle relevant to positional marking of dependency). Second, the pair <w,x> may be interpreted as constituting a cycle, but y may somehow be interpreted as an interpolation, i.e. not as a genuine part of the dependent of w. This amounts to saying that not x but z is the dependent of w, and positional marking of the dependency would then affect z rather than x, again yielding V3. I propose to call y in this situation ‘extradependent’. This subsection has no ambition beyond suggesting that a number of cases of V3 in V2-languages may be identified as involving either extracyclicity or extradependence. In particular, no attempt is offered at an explanation of these V2-deviations. The case of (37), then, appears to be a typical case of extracyclicity: a cycle is constituted as soon as the subject Wansje is merged to the derivation (leading to verb fronting), and the further merger of the topic zonder entwat te zeggen apparently does not yield a cycle.22 The concept of extracyclicity is also helpful in describing structural V3orders of the type in (39) (from Dutch): (39) a.
Jan die ken ik niet John dem:nntr know I not ‘John, I don’t know.’ b. Dat het regent (dat) verbaast me niet that it rains that amazes me not ‘That it’s raining does not amaze me.’
In (39a), a fronted constituent is resumed by an agreeing demonstrative (nntr = nonneuter gender), which itself triggers inversion. We can say that the pair consisting of the demonstrative and its sister constitutes the cycle relevant for positional dependent marking (i.e. V2), while the fronted constituent itself is extracyclic. Accepting Koster’s (1978) analysis of subject clauses as in (39b), where fronted clauses are invariably resumed by a possibly silent demonstrative, we conclude that fronted clauses are always extracyclic. Similar to the cases in (39) appears to be (40a), where the consequent is introduced by an optional demonstrative time adverbial dan; this potentially implies that fronted conditional clauses are extracyclic as well: (40) a.
Als je komt (dan) bak ik een taart if you come then bake I a cake ‘If you come I’ll bake a cake.’
Jan-Wouter Zwart
b. Al schreeuwden ze ook nog zo hard, ze konden er even.if yelled they prt prt so loud they could there niet meer uit not anymore out ‘No matter how loud they cried, they could not get out anymore.’
Slightly different, but arguably still a case of extracyclicity, is the construction in (40b), where the first constituent, the clause introduced by al ‘even if ’, does not trigger V2.23 Finally, various speech act adverbials may appear in first position without triggering V2 (cf. Meinunger 2004: 73): (41) Echter / Eerlijk gezegd / Tussen haakjes, dit voorstel is however honest said between brackets this proposal is onacceptabel unacceptable ‘However/honestly put/incidentally, this proposal is unacceptable.’
In cases such as these the clause to the right of the speech act adverbial may be considered a complete proposition which the speech act adverbial modifies; the ‘complete’ character of the clause may play a role in determining cyclicity here.24 Other cases of V3 cannot be described as involving an extracyclic first constituent, which is what motivates the concept of extradependence. In Dutch, these involve sentence connecting adverbs like echter ‘however’, nu ‘(nontemporal) now’, dan ‘(nontemporal) then’, immers ‘as is known’, and daarentegen ‘in contrast’: (42) a.
Dit voorstel echter is onacceptabel this proposal however is unacceptable b. In diezelfde landstreek nu waren herders in that-same countryside now were shepherds ‘Now there were shepherds in that same countryside.’
Prosodically, these unstressed adverbs group with the first constituent, not with their sister:25 (43) a.
Dit voorstel echter dat is onacceptabel this proposal however dem:ntr is unacceptable ‘This proposal however is unacceptable.’ b. *Dit voorstel dat echter is onacceptabel this proposal dem:ntr however is unacceptable
Verb second as a function of Merge
We may take this to indicate that the relevant adverbs are not really part of the sister of the first constituent (i.e. of the dependent of the first constituent), hence ‘extradependent’.26 Another class of V3 constructions suggestive of extradependence appears in Mainland Scandinavian (Platzack 1986; Egerland 1998; Nilsen 2002). These involve focus-sensitive adverbs like Norwegian bare ‘(nontemporal) just’ and nesten ‘almost’ (Norwegian data from Nilsen 2002: 152): (44) a.
Jens John b. Jens John
bare gikk just left nesten gråt almost cried
In these constructions, the finite verb is focused, and the particular reading expressed here (‘John simply/virtually left/cried’) is lost when the adverb is realized to the right of the verb. A similar class of adverbs blocks V2 in German, but in contrast to Norwegian, the V3 order is not allowed, leading to a situation where these adverbs can be used only in embedded clauses or with analytic tenses (Meinunger 2001, 2004; German examples from Meinunger 2004: 56): (45) a.
. . . weil die Kommission nichts als meckerte because the committee nothing but grumbled ‘. . . because the committee did nothing but grumble’ b. Die Kommission hat nichts als gemeckert the committee has nothing but grumble-part ‘The committee did nothing but grumble.’ c. *Die Kommission meckerte nichts als the committee grumbled nothing but
Meinunger observes that the relevant class of adverbials needs to be proclitic to some host to its right, which is absent in (45c). What is not explained is why the V3-order of Norwegian is not available to German. In our terms, we might state that whereas focus sensitive adverbs may be extradependent in Norwegian, they may not in German.27, 28 Further research is needed to evaluate the viability of the concepts of extracyclicity and extradependence introduced here, and their interaction with the general mechanism of positional dependency marking in (9)–(10).
Jan-Wouter Zwart
. The relation of V2 to morphology The approach to V2 outlined here presupposes that syntactic and morphosyntactic dependencies are invariably sisterhood relations. It follows that subjectverb agreement is not a relation between the subject and the verb, but between the subject and its sister, with the subject the antecedent and the sister the dependent. The dependency may then be spelled out on a term of the sister, which, in the Germanic languages, is invariably the verb. In this approach, agreement is not mediated by a functional head, and features residing in functional heads are not taken to be responsible for verb movement in any way. It follows that V2 is not triggered by morphosyntactic features. In the literature, the only morphosyntactic feature taken to be directly involved with V2 is tense or finiteness, sometimes notated [+F] (e.g. Den Besten 1978). The approach outlined here takes tense morphology on the verb to be the spell-out of a dependency relation between a tense operator (the antecedent) and its sister (the dependent), which contains the verb. In other words, tense morphology on the verb is formally identical to agreement morphology, i.e. the spell-out of a dependency on a term of the dependent element. We take it to be uncontroversial that the feature tense is located outside the verb phrase (i.e. tense is a clausal property). It follows that tense is not an inherent feature of the verb, and indeed verbs may appear without tense, e.g. in nominalizations. However, the idea that tense is located in (or constitutes) a functional head assumes a theory of morphosyntactic dependency which we have abandoned here, namely a theory where morphology is the function of a direct dependency between a lexical and a functional head. Since such dependencies are not sufficiently local (they do not involve sisterhood), we propose that the grammar lacks them entirely. We submit that tense be viewed as an operator merged to the structure at some point in the derivation, creating a dependency which is spelled out on a term of the dependent element, leaving the question of whether tense is a functional head or a phrasal operator moot. The relevance of tense to V2 may then be stated in the following terms: the element designated to spell-out tense dependency is also the element designated for positional dependent marking (a subcase of the consistency principle (4)). We do not see any evidence beyond sheer conjecture that tense is more directly involved in the V2 phenomenon (a point shared with Anderson 2000; pace Evers 1982 and Koster 2003). Agreement morphology is often taken to be relevant to the question of whether V2 is generalized to embedded clauses or not, with ‘rich’ morphol-
Verb second as a function of Merge
ogy forcing generalized V2 (yielding what Vikner 1995: 131 calls ‘V-to-I languages’). On our view, generalized V2 results when merger of a subject invariably yields a cycle, i.e. a dependency requiring positional dependent marking. This raises the question what ‘richness’ of morphology contributes to the system of grammar we are contemplating. I take agreement morphology on the verb to be the morphological spell out of a relation between the subject and its sister, a larger constituent containing the verb. Richness of morphology addresses the structure of the morphological paradigm of the verb spelling out the agreement relation, hence is twice removed from the actual syntactic dependency relation triggering agreement and/or V2. It is not obvious, in the system we are considering, that richness of morphology of the verbal paradigm should be in any way related to a requirement of positional dependency marking of the dependent containing the verb. In this connection it may be useful to point out that the theory we are contemplating makes no use of a distinction between ‘strong’ and ‘weak’ morphosyntactic features. Previously, the association of strong features with rich paradigms may have had some intuitive appeal. But in a system where features do not trigger movements, there is no need for a distinction between strong and weak features, and the supposed correlation with rich morphology loses much of its appeal. This is quite apart from the circumstance that it has proved difficult to define the exact cut-off point between rich and poor morphology in connection with generalized V2.29
. A note on OT approaches to V2 The approach to V2 discussed here bears some resemblance to analyses of V2 proposed within the framework of Optimality Theory (OT) (Anderson 2000; Legendre 2001). I outline some of the differences here. In the OT-approaches to V2, the placement of the verb is described as the interaction of two constraints familiar from prosodic morphology, of which one requires the verb to be leftmost (EdgeMost (L)) and another bars the verb from initial position (NonInitial). If both constraints are ranked high with respect to other relevant constraints, and are ordered as in (46), the V2 pattern results.30 (46) NonInitial (Vfin , S) >> EdgeMost (Vfin , L, S)
Jan-Wouter Zwart
On our approach, the circumstance that the verb appears in the second position does not follow from an interaction of constraints, but from the following set of assumptions: (47) a. Dependency is a function of binary merge, yielding <x,y> b. In each pair <x,y>, the relation between x and y is marked on y (dependent marking) c. Dependent marking can be morphological or positional marking d. The element marking a dependency morphologically is the designated element for marking the dependency positionally e. Positional marking is done by lexicalizing the left edge of the dependent (i.e. the positional marker is a linker)
We take all of (47) to hold universally, with languages differing as to which dependencies they choose to mark, and how (i.e. morphologically and/or positionally). The V2 pattern then results when positional marking of the highest pair <x,y> in the clause is done by the verb. It will be seen that what the approaches have in common is the aspect of linearization (cf. also Chomsky 2001), but that the approach advanced here relates V2 to the derivational process of Merge, to the configurational properties of the output of Merge, and to a general theory of dependency marking. It follows from the approach advanced here that the verb in V2 constructions occupies the position to the left of the first constituent, not to the left of the first word (unlike second position phenomena involving clitics). This is because the first word of a complex first constituent is in a local dependency configuration with a sister that does not contain the verb. Unlike Anderson (2000) we need not invoke any conditions on movement or syntactic well-formedness to obtain this result. It also follows that Wackernagel (1892: 428) was right in questioning a general relation between V2 and second position clitic placement. Wackernagel conjectures that since V-final is common Indoeuropean, V2 is an innovation which may have started out with ‘light’ (one or two syllable) verbs as a subcase of his law fronting light elements. But he admits that there is little evidence for this scenario (and in fact counterevidence from Celtic and Greek), listing only second position copula placement in Latin and Lithuanian as relevant cases outside Germanic.31 From our perspective, second position copula placement is indeed related to V2, but not in terms of prosodic properties. Adopting a Small Clause analysis of copula constructions (Kayne 1984), copula constructions often involve raising of the Small Clause predicate to subject position, yielding a dependency
Verb second as a function of Merge
which may need positional marking. One possibility is that the copula is such a positional dependent marker, a clausal counterpart to linkers found more generally in the nominal domain (Den Dikken & Singhapreecha 2004).
. Conclusion In this paper I have argued for the following. The basic (in fact, only) structure building operation in the computational system of human language, Merge, creates a pair of sisters, one of which is the dependent of the other. In particular, when a constituent is fronted, the element it adjoins to (its sister) becomes its dependent. Dependency may be marked in two ways: morphologically or positionally. Morphological marking occurs when a term of the dependent (in the Germanic languages, this is the verb or an auxiliary) is marked for features which the nondependent carries inherently (agreement). The particular proposal of this paper is that positional marking occurs when a term of the dependent (or a dummy) is realized at the left edge of the dependent, functioning as a linker. In V2 languages, the designated element for positional dependent marking is the element which marks the same dependency morphologically as well, i.e. the finite verb. Verb second, then, is really verb first applied to the domain of the dependent of a fronted constituent. The proposal allows us to incorporate a range of recalcitrant V2 phenomena within a unified theory of V2. These phenomena include the well-known asymmetries associated with V2 (between main and embedded clauses, between finite and nonfinite verbs), nonstandard V2 phenomena (such as quotative inversion, conjunction-triggered inversion, and apokoinou constructions), as well as, possibly, deviations from the V2 pattern (V1 and V3). It is suggested that verb placement needs to be studied in the context of a theory of linkers expressing a relation between two elements joined by Merge, rather than in terms of features residing in functional heads triggering overt or covert head movement.
Notes * Thanks to Hans Bennis, Marcel den Dikken, and the anonymous reviewers for this volume. Also thanks to the participants of the 19th Comparative Germanic Syntax Workshop and of the Bloomington 2004 Syntax Fest. Finally, thanks to the participants of the CLCG Advanced Syntax Seminar of the University of Groningen for useful discussion and many
Jan-Wouter Zwart
helpful comments: Janneke ter Beek, Reineke Bok-Bennema, Olaf Koeneman, Jan Koster, Hidetoshi Shiraishi, and Mark de Vries. . I write the product of Merge as an ordered pair rather than as a set, on the proposal of Zwart (2004a) that Merge transfers one element at a time from the numeration to the current derivation (instead of selecting two elements and combining them in a symmetric fashion). The dependency discussed in the text is taken to be the automatic consequence of the process of asymmetric Merge, assuming that Merge turns the current stage of the derivation into a dependent (or the argument of a newly added functor). . I take morphological dependency marking (for tense, agreement, etc.) to be a function of Merge as well (Zwart 2004a, 2004b): dependency relations are sisterhood relations marked on (a term of) the dependent. In the case of subject-verb agreement, the dependent is the subject’s sister, which spells out the dependency morphologically via agreement on one of its terms (the finite verb, in the unmarked case). In the case of tense, we assume a dependency relation between a tense operator outside VP and its sister, which contains the verb; tense marking on the verb, then, is a subcase of dependency marking via agreement. The proposal is consistent with earlier analyses describing V2 as ‘Move Tense’ (Den Besten 1978; Evers 1982; Koster 2003), since realization of a tense-marked verb in the V2-position has the effect of realizing Tense in a relatively high position, but it differs from these analyses in identifying the trigger for V2 as the (parametrized) need to mark dependency by position. . The properties of consistent verb initial languages are arguably not comparable to those of V1 constructions in V2 languages. . See Vikner (1995: 51–64) for more extensive discussion of the possible triggers for V2. On the relation to tense, see Note 2. . This flexible analysis of the V2 pattern, where the V2-position is not identified as a designated functional head (C), but as the highest functional head is sometimes referred to as ‘asymmetric’ (cf. Schwartz & Vikner 1989, 1996). . Mergers not ending the derivation or not involving lexical nondependents may constitute a cycle in marked cases, possibly needed to describe language and construction specific variation. More generally, we can say that the positional marking property applies to each dependency, but is passed on to each next dependency (taking the derivation to proceed in a bottom up fashion) until the end of a cycle is reached. . As is well known (Vikner 1995: 129; Zwart 2003: 285–288), the embedded V2 phenomenon of Mainland Scandinavian, Frisian, and Colloquial Dutch is qualitatively different, requiring that these embedded clauses be viewed as root clauses; in other words, viewing the combination of the subject and its sister in these constructions as a cycle may fall under the unmarked case of (10a). . The order complementizer + wh-phrase also shows up in other dialects surveyed in recent years by the Meertens Institute in connection with the project Syntactische Atlas van de Nederlandse Dialecten (SAND), such as the dialect of Strijen discussed in Van Craenenbroeck (2004: 34). . To be sure, there is no principled reason within the system contemplated here, why nonfinite verbs should not be used as positional dependency markers. The Frisian infinitivus-proparticipio cases discussed in Hoekstra (1997: Chapter 2) and Zwart (2001) may be regarded
Verb second as a function of Merge
as instances of dependency marking by a nonfinite verb. Verb movement of nonfinite verbs is attested in a variety of languages, among which Icelandic (Thráinsson 1979: Chapter 5), French (Pollock 1989: 372f.), and Hungarian (É. Kiss 2002: 201–202). . A related construction is the nominal infinitive, where the lexical head is a nonfinite verb contained within a nominal constituent (DP): (i)
dat vervelende altijd maar stripboeken lezen that boring:adj always just comic-books read-inf ‘this boring (habit of) reading comic books all the time’
Here we see no argument for thinking that any subpart of the DP constitutes a cycle. The question of why no V2 takes place in nominalizations then reduces to the question of why no positional dependent marking takes place within DP. Here we have nothing new to contribute. Note that the determiner dat can be replaced by a possessive Jan z’n ‘John his’, where z’n arguably functions as a linker. . English also has the construction in (i), where inversion occurs in the quotative even if the quote does not precede: (i)
Said John, I am so sick
I take the inversion here to be unrelated to the quotative character of the construction (perhaps akin to narrative inversion, cf. (34e)). . The analysis in the text is similar to that of Collins and Branigan (1997) in assigning the quote a position outside the CP of the quotative, but Collins and Branigan take the subject-verb inversion to be a process internal to the quotative, rather than a phenomenon marking the dependency of the quotative to the quote. Discussion of the various properties of quotative inversion derived on Collins and Branigan’s analysis would take us too far afield at this point. . Icelandic Stylistic Inversion, illustrated in (i), presents another construction where V2 may be triggered by a head (Anderson 2000). (i)
Ég hélt að kysst hefðu hana margir stúdentar I thought that kiss:part had her many students ‘I thought that many students had kissed her.’
However, facts and references in Holmberg (2000) suggest that the fronted element in Stylistic Inversion may be a phrase as well. . The examples in this subsection are all Colloquial (Rotterdam) Dutch, taken from Jules Deelder, 2001, The Dutch Windmill (Amsterdam: De Bezige Bij), an apparently verbatim transcript of an interview with Dutch boxing legend Bep van Klaveren (1907–1992) on his life and career, which contains over one hundred examples of apokoinou constructions. Source page numbers are indicated with each example. (Thanks to Elzerieke Hilbrandie-Van Hooijdonk for assistance in compiling the corpus.) . The structure of (26), on this analysis, would be more or less as in (i) (for PredP, see Zwart 1994: 399): (i)
[CP dan was [TP [je tegenstander] was [PredP neer [VP ] ] ] ]
Jan-Wouter Zwart
. These facts argue against the suggestion that the apokoinou construction is a type of anacolouthon, if we define the latter as involving a contamination of two syntactic constructions (a mixing of two derivations), resulting in ‘a want of grammatical sequence or coherence’ (Webster’s dictionary, edition of 1913). Apokoinou and anacolouthon are also strictly separated in the seminal treatment of De Vries (1910–1911). . The examples show that repetition is not restricted to the finite verb: in addition to the subject, for which see below, the inherent reflexive pronoun me is repeated in (28b) and the temporal adverb toen ‘then’ is repeated in (28c). These facts raise a problem for the approach taken here, but it must be noted that the repeated material in these cases is prosodically weak, suggesting an analysis involving cliticization to the left edge marking element (the verb). We leave these facts for further study. . The discussion of the V2 phenomena in this section is necessarily cursory, indicating how they could be approached from the perspective taken here. . As pointed out to me by Erika Troseth, this raises a question as to the configurational nature of the dependency which the fronted verb is supposed to mark on the approach to V2/V1 contemplated here. The proposal in the text implies that a constituent y may be interpreted as a dependent as soon as it is marked by a dependency marking device, even in the absence of a structural antecedent with which y is merged. Discourse factors then come in to assist in the correct interpretation of the dependency. . In (36a), a future state of affairs is presented as a matter of fact; in (36b) the modal verb conveys the notion of obligatoriness or necessity. In this connection it is perhaps also relevant that a subject cannot be freely expressed in imperatives (cf. Portner 2004: Section 4.1). This would follow if the imperative denotes a property which can only be true of the addressee, as Portner argues, and if the addressee is already contained in the implicit directive statement of which the imperative construction is a dependent. . See Iatridou and Embick (1994) for an analysis of conditional inversion. . Since similar V3-orders do not occur (in the relevant dialects) where the first constituent is a wh-phrase, it must be the nature of the antecedent that determines whether or not a cycle is constituted. This is where reference to the marked cases hinted at in Note 6 becomes relevant. Loss of V2 can then be described as the gradual progression of extracyclicity across types of fronted constituents. . Alternatively, the construction in (40b) involves juxtaposition (asyndetic coordination), which is supported by the observation that the clause introduced by al may also follow the clause introduced by ze. . As noted by Meinunger (2004: 76–77) for German, some of these speech act adverbials may actually trigger V2, such as the type eerlijk gezegd ‘honestly put’, but others, such as echter ‘however’ and eerlijk ‘honest’, may not. . The relevant adverbs may als appear to the right of the finite verb (Dit voorstel is echter onacceptabel), suggesting they are not a term of the first constituent. . Alternatively, we might suppose that the positional spell-out rule (left edge dependent marking) which we argue yields V2 is sensitive to prosodic grouping, such that the left edge position be defined as the first position following enclitic materical.
Verb second as a function of Merge . Alternatively, we might attribute the Norwegian pattern to a spell-out mechanism which orders two elements competing for the left edge of a dependent (the verb and the focus sensitive adverb) in a particular way not available to German. Another analysis is proposed by Nilsen (2002), who takes the Norwegian facts to suggest that verb movement in Norwegian involves masked XP-movement, so that the finite verb occupies a specifier rather than a head position. . It is tempting to consider French Complex Inversion, illustrated in (i-b), as a case of extradependence as well: (i)
a.
b.
Jean il est venu John scl:3.masc.sg is come:part ‘John came.’ Pourqoui Jean est il venu? why John is scl:3.masc.sg come:part ‘Why did John come ?’
What seems entirely regular about this construction is the inversion of the subject clitic il with finite auxiliary est triggered by merger of the wh-phrase pourqoui; what is illunderstood is why the inversion yields V3 rather than V2, but this could follow if Jean in (i-b) is an extradependent. We leave this for further study. . A question not addressed in this article is what explains language and construction specific variation with respect to positional dependency marking. The system proposed here provides one clear option, namely between positional and morphological dependency marking. There appears to be a trade-off, since all Germanic languages show positional dependency marking in (main clause) wh-interrogatives, where no morphological marking appears to be available. But there are many languages and constructions to consider, and it seems too early to comment with any confidence on a basis for parametric variation. . The constraints specify a domain over which the linearization requirements hold (S, the clause) and an element to which they apply, the finite verb. . Wackernagel also lists V2 examples from Ancient Greek votive inscriptions, which seem to have special syntactic properties.
References Adger, D. (2003). Core Syntax. Oxford: Oxford University Press. Anderson, S. (2000). Towards an optimal account of second position phenomena. In J. Dekkers, F. van der Leeuw, & J. van de Weijer (Eds.), Optimality Theory: Syntax, phonology and acquisition (pp. 301–333). Oxford: Oxford University Press. Behaghel, O. (1932). Deutsche Syntax: Eine geschichtliche Darstellung IV. Heidelberg: Carl Winters. den Besten, H. (1977). On the interaction of root transformations and lexical deletive rules. Ms., MIT and University of Amsterdam. Published in H. den Besten, Studies in WestGermanic Syntax, PhD Dissertation, University of Tilburg (1989).
Jan-Wouter Zwart
den Besten, H. (1978). Auxiliary deletions and the interplay between local deletive rules and filters. Paper presented at the GLOW Colloquium, Amsterdam. Biberauer, T. (2002). Verb second in Afrikaans: Is this a unitary phenomenon? Stellenbosch Papers in Linguistics, 34, 19–70. de Boor, H. (1977). Studien zur altschwedischen Syntax. Hildesheim: Georg Olms. Cardinaletti, A. (1990). Subject/object asymmetries in German null topic constructions and the status of Spec,CP. In J. Mascaró & M. Nespor (Eds.), Grammar in Progress: GLOW essays for Henk van Riemsdijk (pp. 75–84). Dordrecht: Foris. Charivarius (1940). Is dat goed Nederlands? Amsterdam: Ploegsma. Chomsky, N. (1986). Barriers. Cambridge, MA: The MIT Press. Chomsky, N. (2001). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 1–52). Cambridge, MA: The MIT Press. Collins, C. & P. Branigan (1997). Quotative inversion. Natural Language and Linguistic Theory, 15, 1–41. van Craenenbroeck, J. (2004). Ellipsis in Dutch Dialects. PhD Dissertation, University of Leiden. den Dikken, M. & P. Singhapreecha (2004). Complex noun phrases and linkers. Syntax, 7, 1–54. É. Kiss, K. (2002). The Syntax of Hungarian. Cambridge: Cambridge University Press. Egerland, V. (1998). On verb-second violations in Swedish and the hierarchical ordering of adverbs. Working Papers in Scandinavian Syntax, 61, 1–22. Evers, A. (1982). Twee functionele principes voor de regel ‘verschuif het werkwoord’. Glot, 5, 11–30. Foulet, L. (1963). Petite syntaxe de l’ancien français. Paris: Librairie Honoré Champion. Hale, K. (1973). Person marking in Warlbiri. In S. R. Anderson & P. Kiparsky (Eds.), A Festschrift for Morris Halle (pp. 308–344). New York, NY: Holt, Rinehart and Winston. Henry, A. (1995). Belfast English and Standard English: Dialect variation and parameter setting. New York, NY: Oxford University Press. Hoekstra, E. (1994). Overtollige voegwoorden en de volgorde of + interrogativum/relativum. De Nieuwe Taalgids, 87, 314–321. Hoekstra, J. (1997). The Syntax of Infinitives in Frisian. PhD Dissertation, University of Groningen. Holmberg, A. (2000). Scandinavian stylistic fronting: How any category can become an expletive. Linguistic Inquiry, 31, 445–483. van der Horst, J. & K. van der Horst (1999). Geschiedenis van het Nederlands in de twintigste eeuw. The Hague/Antwerp: SDU/Standaard. Huang, C.-T. J. (1984). On the distribution and reference of empty pronouns. Linguistic Inquiry, 15, 531–574. Iatridou, S. & D. Embick (1994). Conditional inversion. Proceedings of NELS, 24, 189–203. Jansen, F. (1981). Syntaktische Konstrukties in Gesproken Taal. Amsterdam: Huis aan de drie grachten. Jedig, H. H. (1969). Ocherki po sintaksisu niznenemeckogo govora Altaiskogo kraja. Omsk. Katz, J. J. & P. M. Postal (1964). An Integrated Theory of Linguistic Descriptions. Cambridge, MA: The MIT Press. Kayne, R. S. (1984). Connectedness and Binary Branching. Dordrecht: Foris.
Verb second as a function of Merge
Kayne, R. S. (1992). Word order. Keynote address at the GLOW Colloquium, Lisbon, April 14. Kayne, R. S. (1994). The Antisymmetry of Syntax. Cambridge, MA: The MIT Press. Kellner, L. (1924). Historical Outlines of English Syntax. London: MacMillan and Co. Koster, J. (1978). Why subject sentences don’t exist. In S. J. Keyser (Ed.), Recent Transformational Studies in European Languages (pp. 53–64). Cambridge, MA: The MIT Press. Koster, J. (2003). All languages are tense-second. In J. Koster & H. van Riemsdijk (Eds.), Germania et alia: A linguistic Webschrift for Hans den Besten. http://odur.let.rug.nl/ ∼koster/DenBesten/contents.htm Legendre, G. (2001). Masked second-position effects and the linearization of functional features. In G. Legendre, J. Grimshaw, & S. Vikner (Eds.), Optimality-theoretic Syntax (pp. 241–277). Cambridge, MA: The MIT Press. Magnusson, E. (2004). The loss of referential pro in Swedish. Paper presented at the 19th Comparative Germanic Syntax Workshop, New York, June 5. Meinunger, A. (2001). Restrictions on verb raising. Linguistic Inquiry, 32, 732–740. Meinunger, A. (2004). Interface restrictions on verb second. Linguistics in Potsdam, 22, 51– 81. Munn, A. (1993). Topics in the Syntax and Semantics of Coordinate Structures. PhD Dissertation, University of Maryland. Nilsen, Ø. (2002). V2 and Holmberg’s Generalization. In J.-W. Zwart & W. Abraham (Eds.), Studies in Comparative Germanic Syntax: Proceedings from the 15th workshop on Comparative Germanic Syntax (pp. 151–173). Amsterdam: John Benjamins. Paardekooper, P. (1955). Syntaxis, Spraakkunst en Taalkunde. Den Bosch: Malmberg. Paul, H. (1919). Deutsche Syntax IV: Syntax (Erste Hälfte). Halle: Niemeyer. Platzack, C. (1986). COMP, INFL and Germanic word order. In L. Hellan & K. Koch Christensen (Eds.), Topics in Scandinavian Syntax (pp. 185–234). Dordrecht: Reidel. Pollock, J.-Y. (1989). Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry, 20, 365–424. Portner, P. (2004). The semantics of imperatives within a theory of clause types. To appear in K. Watanabe & R. B. Young (Eds.), Proceedings of SALT 14. Ithaca, NY: CLC Publications. Sassen, A. (1967). Syntactische implicaties van de zgn. herhalingsconstructie (‘dat is een gek geval is dat’). Handelingen Vlaams filologencongres, 26, 30–47. Schwartz, B. & S. Vikner (1989). All verb second clauses are CPs. Working Papers in Scandinavian Syntax, 43, 27–49. Schwartz, B. & S. Vikner (1996). The verb always leaves IP in V2 clauses. In L. Rizzi & A. Belletti (Eds.), Parameters and functional heads (pp. 11–62). New York, NY: Oxford University Press. Stoett, F. A. (1923). Middelnederlandse Spraakkunst: Syntaxis. The Hague: Martinus Nijhoff. Thráinsson, H. (1979). Complementation in Icelandic. New York, NY: Garland. Thráinsson, H. (1986). V1, V2, V3 in Icelandic. In H. Haider & M. Prinzhorn (Eds.), Verb Second Phenomena in Germanic Languages (pp. 169–194). Dordrecht: Foris. Travis, L. (1984). Parameters and Effects of Word Order Variation. PhD Dissertation, MIT.
Jan-Wouter Zwart
Travis, L. (1991). Parameters of phrase structure and V2 phenomena. In R. Freidin (Ed.), Principles and Parameters in Comparative Grammar (pp. 339–364). Cambridge, MA: The MIT Press. Vikner, S. (1995). Verb Movement and Expletive Subjects in the Germanic Languages. New York, NY: Oxford University Press. de Vries, W. (1910–1911). Dysmelie: Opmerkingen over syntaxis. Groningen: Municipal Gymnasium. Wackernagel, J. (1892). Über ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen, 1, 333–436. Winkler, J. (1874). Algemeen Nederduitsch en Friesch Dialecticon. The Hague: Martinus Nijhoff. Zwart, J.-W. (1993). Dutch Syntax: A minimalist approach. PhD Dissertation, University of Groningen. Zwart, J.-W. (1994). Dutch is head-initial. The Linguistic Review, 11, 377–406. Zwart, J.-W. (1997). Morphosyntax of Verb Movement: A minimalist approach to the syntax of Dutch. Dordrecht: Kluwer Academic. Zwart, J.-W. (1998). Where is syntax? Syntactic aspects of left dislocation in Dutch and English. In P. Culicover & L. McNally (Eds.), The limits of Syntax (pp. 365–393). San Diego, CA: Academic Press. Zwart, J.-W. (2001). Syntactic and phonological verb movement. Syntax, 4, 34–62. Zwart, J.-W. (2002). Backgrounding in Dutch. Paper presented at CGSW 17. Zwart, J.-W. (2003–2004). Een dynamische structuur van de Nederlandse zin; deel 1: dynamische syntaxis. Tabu, 33, 55–71. Zwart, J.-W. (2004a). The format of dependency relations. Lecture series at the 2004 Syntax Fest, Indiana University, Bloomington, June18 – July 1. Zwart, J.-W. (2004b). Local agreement. Ms., University of Groningen. To appear in C. Boeckx (Ed.), Agreement systems. New York: Oxford University Press.
Nonnative acquisition of verb second On the empirical underpinnings of universal L2 claims* Ute Bohnacker Lund University
Acquiring Germanic verb second is typically described as difficult for second-language learners. Even speakers of a V2-language (Swedish) learning another V2-language (German) are said not to transfer V2 but to start with a non-V2 grammar, following a universal developmental path of verb placement. The present study contests this claim, documenting early targetlike V2 production for 6 Swedish ab-initio (and 23 intermediate) learners of German, at a time when their interlanguage syntax elsewhere is nontargetlike (head-initial VPs). Learners whose only nonnative language is German never violate V2, indicating transfer of V2-L1 syntax. Informants with previous knowledge of English are less targetlike in their L3-German productions, indicating interference from non-V2 English. V2 per se is thus not universally difficult for nonnative learners.
.
Introduction
This paper investigates the acquisition of verb placement, especially verb second (V2), by native Swedish adults and teenagers learning German. Several recent publications (e.g. Platzack 1996, 2001; Pienemann 1998; Pienemann & Håkansson 1999; Håkansson, Pienemann, & Sayehli 2002) have claimed that learners, irrespective of their first language (L1), take the same developmental route in the acquisition of syntax of a foreign or second language (L2). Targetlike finite verb placement in a V2 language like German, Dutch or Swedish is seen as an exceptionally difficult phenomenon, a difficult syntactic constraint or parameter, to acquire. Even if both the L1 and L2 are Germanic V2 languages, learners are said to acquire V2 late, only partially, or never. Data from Swedes learning German, violating V2 in their L2 German, have been cited
Ute Bohnacker
as empirical support for these claims and for the resurrection of a hypothesised universal developmental path in L2 German verb placement – essentially the developmental stages proposed in the 1980s on the basis of Romance learners of German (e.g. Clahsen, Muysken, & Pienemann 1983; Clahsen & Muysken 1986). Explanations propounded in the literature are (i) the so-called “vulnerability” of the C-domain (Platzack 2001), (ii) SVO being a more basic word order (e.g. Clahsen & Muysken 1986; Platzack 1996), and (iii) “processability”, according to which the “canonical” word order of SVX and XSV (i.e. V3, with a fronted element before the subject) is easier to process – and to produce – than XVS (i.e. V2) (e.g. Pienemann 1998; Håkansson, Pienemann, & Sayehli 2002). The vulnerable C-domain, processability, and the idea of a universal developmental path have also found their way into language teacher training and linguistics courses and are sometimes uncritically presented as axioms and as the basis for classroom teaching methods. The present paper aims to show (a) that the above studies suffer from some methodological flaws, (b) that new studies of learners of a V2 language (German) with a V2 L1 (Swedish) yield strikingly different results, and thus (c) that the previous empirical underpinnings for proposed universals in the development of L2 verb placement are weak. I proceed as follows. Comparing finite verb placement in German and Swedish, I firstly draw attention to the existence of well-formed V3 main clauses in native Swedish which tend to be overlooked in generative syntax and are ignored in L2 research. If Swedish allows certain violations of the V2 requirement (but German does not), Swedish learners of German may produce certain V3 orders in their interlanguage German not because V2 is universally hard to acquire, but because these particular V3 orders, grammatical in their L1 Swedish, have transferred. I then discuss earlier research on the acquisition of L2 German verb placement by Swedes (Håkansson 2001; Sayehli 2001; Håkansson, Pienemann, & Sayehli 2002). A confounding variable in these studies is that their informants all knew English before learning German, making it impossible to rule out transfer from L2 English to L3 German. I therefore present quantitative and qualitative data from new research: (a) oral and written cross-sectional data from 23 Swedish teenage school pupils at intermediate proficiency level, and (b) micro-comparative quasi-longitudinal oral data from 6 Swedish adults learning ab initio German (3 with no English, 3 with L2 English). The results indicate that learners in both groups productively use V2 early on, pointing to transfer of the V2 property from their L1 Swedish to Ger-
Nonnative acquisition of verb second
man. This happens at a time when other aspects of their interlanguage syntax are clearly nontargetlike, e.g. early VPs are head-initial (VO), giving way to German-style head-final VPs (OV) over time. My data also show a categorical difference between those ab initio learners who do not know English (0% V3 in their interlanguage German) and those who do know English (45% V3). Knowledge of a non-V2 language (English) can thus make it initially “harder” to acquire another V2 language, which points to partial transfer of the nonV2 L2-English syntax to L3 German. My findings cast doubt on current claims concerning the role of a canonical (X)SVO word order and concerning the idea of Germanic V2 being difficult to acquire per se. I therefore argue that theories of acquisition based on these notions are misguided and should be abandoned in favour of transfer approaches (e.g. Schwartz & Sprouse 1994, 1996), and I advise caution when propounding absolute “universals” for the L2 acquisition of syntax.
. The V2 requirement in German and Swedish As is well known, German non-finite utterances require the verb or verbal element(s) to be in final position. Because of this fact, German is regarded as an OV language (at the relevant level of abstraction), and generative grammar standardly holds German to have a head-final VP (cf. e.g. Koster 1975; den Besten 1977/1983; Thiersch 1978; Haider 1986, 1993; Grewendorf 1988; Stechow & Sternefeld 1988; Schwartz & Vikner 1989, 1996; pace Kayne 1994). German finite clauses too have a requirement on verbs to be in final position, but in main clauses this only holds for the non-finite verbal elements of complex verbs, such as infinitives, participles and particles of separable prefix verbs (e.g. Drach 1937; Reis 1980; Grewendorf 1988: 19–25; Zifonun et al. 1997: 1498–1504). Generative grammar typically regards the positioning of these verbal elements as further evidence of German VP being head-final. In Swedish, the nonfinite verbal element(s) do not appear in final position, but to the left of the complement(s). Swedish is therefore regarded as a VO language with a head-initial VP. German main clauses require the second constituent to be the finite verb (V2), an observation going back to Erdmann (1886) and Paul (1919). For nonsubject-initial clauses, so-called “inversion” of the subject and the verb (XVS) is required (e.g. Zifonun et al. 1997: 1500). The same holds for Swedish (e.g. Teleman et al. 1999b: 10–13). The V2 phenomenon may obscure the basic verbfinal pattern of German if there only is a simplex verb (lese in (1)), but when
Ute Bohnacker
the clause contains a complex verb, both V2 and head-final VP are visible (habe . . . ausgemacht in (2)). (1) a.
Ich lese nicht so einen Schrott. I read not so a scrap ‘I don’t/won’t read such rubbish.’ b. *So einen Schrott ich lese nicht. c. So einen Schrott lese ich nicht.
(2) a.
Ich habe gerade das Licht ausgemacht. I have just the light out-switched ‘I’ve just switched off the light.’ b. *Gerade ich habe das Licht ausgemacht. c. Gerade habe ich das Licht ausgemacht.
(SVO)
(*OSV) (OVS) (SVO)
(*AdvSVO) (AdvVSO)
SVX is often said to be the most frequent word order in German, but inversion is very common too. The first position, the Vorfeld (Drach 1937; Reis 1980), can be occupied by virtually any constituent, phrasal or clausal, argumental or nonargumental, phonologically light or heavy, and with any semantic function (some modal particles excluded).1, 2 The same holds for Swedish. However, as corpus studies show, in natural discourse the two languages vary slightly in the frequencies of the types of items that occupy the Vorfeld. Hoberg (1981) finds, in a corpus study of written German, slightly more than 50% of the main clauses to be subject-initial, second come those starting with an adverbial adjunct. Similarly, Fabricius-Hansen and Solfjeld (1994) find 54% subject-initial clauses in written German, and 37% adverbial-initial ones. Westman (1974), on the other hand, in a corpus study of written Swedish, finds 64% of the main clauses to be subject-initial in regular prose texts, vs. 72% in formal legal texts. And in a large corpus study of spoken Swedish, Jörgensen (1976: 101– 105) finds 60% of the main clauses to be subject-initial in informal discussions, 62% in interviews, vs. 80% in read-aloud formal radio news. Rosén (2005) also finds 65% subject-initial clauses in native adult Swedish, vs. only 49% in native adult German, in elicitations of the same informal written genre. A comparison of these corpora suggests, then, that subject-initial clauses are more frequent in Swedish than in German. The second largest group of main clauses in both languages are those with an adverbial in first position (Westman 1974; Jörgensen 1976), but the range of these adverbials is more restricted in Swedish than in German (Rosén 2005). In the spoken Swedish corpora, temporal adjuncts predominate in the Vorfeld (Jörgensen 1976, for adult-to-adult speech; Josefsson 2004a: 110–111, for adult-to-child speech), wheras in German, the Vorfeld frequently hosts locational, temporal and other adjuncts (e.g.
Nonnative acquisition of verb second
Zifonun et al. 1997: 1607; Carroll & Lambert 2003: 282; Rosén 2005), and also arguments such as objects, which are rare in the Swedish Vorfeld. Generative grammars typically describe V2 as a two-step process, a syntactic double-movement transformation: leftward movement of the finite verb to a functional head position on the left sentence periphery, creating a V1 clause, plus movement of a constituent into the specifier position of that functional projection. In GB models, this functional projection is commonly identified as CP (e.g. Grewendorf 1988: 64–67). The element in SpecC is not always considered to have moved; expletives and non-subcategorised-for elements especially may be seen as base-generated.3 In symmetric V2 analyses, clause-initial subjects and non-subjects – which occupy the same linear position – occupy the same hierarchical position (e.g. Grewendorf et al. 1987; Schwartz & Vikner 1989, 1996), though they do not on an asymmetric V2 analysis, where subjectinitial clauses are smaller than non-subject-initial ones (e.g. Travis 1984, 1991; Zwart 1993; Barbier 1995; Branigan 1996). With the breaking-up of the CP domain into several functional projections in GB and Minimalist models, suggestions of where to locate the preverbal constituent and the verb have multiplied.4, 5 In (syntactic) acquisition research it is generally advisable to keep the – often ephemeral – formal syntactic apparatus to a minimum. I will therefore concentrate on the linear order of constituents (SVX, V1, V2, V3 etc.), and abstract away from analytical questions concerning the structural account of this linear order, as none of the basic and robust findings to be reported on hinges on specific syntactic analyses. I will occasionally refer to the old Principles and Parameters model of CP, IP and head-final VP for German, head-initial VP for Swedish, which readers can easily translate into the syntactic models of their choice.
. Exceptions to V2 Swedish is a robust V2 language, but it does allow certain exceptions to V2 which are not that widely known and typically ignored in the acquisition literature. If pockets of V3 are grammatical in Swedish, these might play a role in the acquisition of L2 German. I will discuss four such types: (i) clause-initial connective så ‘so’, (ii) clause-initial sen ‘then’, (iii) left-dislocated adverbials with resumptive så ‘so’, and (iv) V3 with certain adverbs.
Ute Bohnacker
. Clause-initial connective elements When a main clause is introduced by a coordinating conjunction (e.g. German und ‘and’, aber ‘but’, denn/weil ‘for’, Swedish och ‘and’, men ‘but’, för ‘for’), this conjunction is typically not regarded as a constituent of the clause, but as a linking word with no influence on word-order, as cj + [XVX] (e.g. Zifonun et al. 1997: 1578; Eisenberg et al. 1998: 400, 819). Thus, there is no exception to the V2 requirement. However, there is a gradient that relates coordinating conjunctions to conjunct/connective adverbials: Such adverbials also have a linking function, but are regarded as constituents of the clause. Hence, for the classification of L2 data a non-trivial problem arises. When learners produce a main clause introduced by a connective, is this a coordinating conjunction which does not “count” for V2 word order (cj + [XVS]; cj + [SVX]) – or is it an adverb that is a constituent in the Vorfeld of the clause ([AdvVS])? In Swedish, connective så ‘so’ is particularly problematic in this regard. One type of så, commonly used in colloquial spoken Swedish to indicate temporal succession or consequence (‘so/so then’), induces inversion, as illustrated in (3). V2-så is always unstressed, there is no prosodic boundary after så, and this så can be preceded by a coordinating conjunction (och ‘and’), suggesting that it is a proper constituent of the clause. (3) Swe. V2 (och) så köpte dom upp EM Airways. (and) so bought they up EM Airways ‘(And) then they took over EM Airways.’
However, there is also another type of connective så (‘so/so that/hence/and as a result’) which is not temporal, but conclusive/consequential. As illustrated in (4), conclusive så does not induce inversion in native Swedish, and V3 utterances are the result.6 (4) a.
Swe. V3
b. Swe. V3
så dom köpte upp EM Airways. so they bought up EM Airways ‘So they took over EM Airways.’ så redan i januari köpte dem upp EM Airways. so already in January bought they up EM Airways ‘So in January already they took over EM Airways. / So already by January they had taken over EM Airways.’
Conclusive/consequential så is always unstressed, and there is never a prosodic boundary after så, suggesting that it is a constituent of the main clause; but on
Nonnative acquisition of verb second
the other hand, this så cannot be preceded by a coordinating conjunction, suggesting that it might be a coordinating conjunction itself (cj + [XVX]). V3-så is a common feature of informal spoken Swedish and is found in written genres too, though there seem to be individual differences amongst speakers with regard to its frequency. Reference grammars (Holmes & Hinchliffe 1994: 465– 469, 476, 529; Teleman et al. 1999a: 730) are unsure about the status of V3-så (adverbial or conjunction). I am not aware of any generative analyses, but I assume that the V3-så construction would be formalised as adjunction to CP: [CP [så] CP ], i.e. as a potential exception to the V2 requirement. There is no V3 equivalent in German, compare: (5) a.
Ger. *V3 *so sie kauften EM Airways auf. so they bought EM Airways up b. Ger. V2 {so/daher/deshalb} kauften sie EM Airways auf. so/therefore/hence bought they EM Airways up ‘So they took over EM Airways.’
When L2 learners produce a V3 main clause introduced by what looks like a connective adverbial and thus apparently violate V2 (AdvSVX or AdvXVS), the “adverbial” – in their interlanguage grammar – might be a coordinating conjunction, plus a V2 clause or at least some element which in their V2-L1 allows V3. As we will see in Section 6, such cases with initial V3-so are produced by Swedes learning German.7 . Clause-initial sen (‘then’) with optional V3 Swedish linguists have noted the occurrence of V3 with clause-initial (temporal/sequential) sen ‘then’ as a nontarget feature of aphasic, SLI (i.e. specific language impairment) and L2 Swedish speech (e.g. Håkansson & Nettelbladt 1993; Platzack 1996: 382–382). However, such sen with optional V3 (sen-SVX, sen-XVS) – notably without a prosodic boundary after sen – is also attested for informal native Swedish (cf. Jörgensen 1976); compare the authentic examples in (6a), (7a), and (8), the latter from an adult addressing a child (CHILDES database, cf. Strömqvist, Richthoff, & Andersson 1993)). Thus there may be a mismatch regarding the actual and perceived use of sen, and acceptability seems to vary amongst speakers. V3-sen can be formalised as adjunction to CP, constituting a true exception to the V2 requirement, and has no V3 correspondence in German (cf. (6c–d)).
Ute Bohnacker
(6) a.
Swe. sen han gick. then he went b. sen gick han. then went he c. Ger. *dann er ging. then he went d. dann ging er. then went he ‘Then he left.’
(7) a.
man gör ju allt för dom, och sen man undrar om allt one does well all for them and then one wonders if all det där var så smart egentligen. this there was so good really b. . . . (och) sen undrar man om allt det där . . . ‘You do everything for them, don’t you, and then you start wondering whether that really is such a good thing.’
(8) ja, ja(g) ska mata dej. sen ja(g) ska äta lite också. yeah I shall feed you. then I shall eat little too ‘Yeah, I’ll feed you, and then I’ll have something to eat as well.’ (Markus’ mother, Mar26_10.cha)
. Pauseless left dislocations In both German and Swedish, an argument or an adjunct followed by a resumptive may occur to the left of the finite verb in a main clause, making the verb appear in third position. Syntactic analyses of left dislocations often assume a boundary: The initial element is extraposed and not part of the clause proper (e.g. Altmann 1981; Jörgensen & Svensson 1986: 139; Ekerot 1988, 1995: 85; Heringer 1989: 247), while the resumptive element is inside the V2 clause. There are no reliable prosodic boundary diagnostics such as a pause or lengthening; the dislocated constituent in German and Swedish is linked to the remainder of the sentence by progredient intonation, though it can also constitute a domain of its own with a separate accent (Zifonun et al. 1997: 518 for German). For this reason, then, left dislocations are plausibly treated not as violations of V2, but as adjunctions to a V2 clause, formalised as (base-generated) adjunctions to CP.8 Left dislocations are more frequent in informal speech than in writing, and unless the initial element is particularly complex/heavy, e.g. a clause, left-dislocations with a resumptive in written registers are frowned upon
Nonnative acquisition of verb second
(Jörgensen & Svensson 1986: 136; Ekerot 1988: 72–74, 1995: 86; Zifonun et al. 1997: 519; Teleman et al. 1999b: 99, 695). However, in spoken Swedish, one type of left dislocation with resumptive is extremely common: left-dislocated adverbials with a resumptive, XXresumptive VS, henceforth XXr VS (Ekerot 1988, 1995: 85-86; Teleman et al. 1999a: 670, 1999b: 10–11, 99, 694–695). The dislocated adverbial can have almost any function, but is often a frame-setting adverb, a PP, or a (temporal/conditional) clause. Resumptive så is always unstressed. German equivalents of this Swedish construction are rare, only occurring when the left-dislocated adverbial is a clause (cf. Ekerot 1988: 241–242), whilst for many other adverbials, e.g. temporal adverbs, left dislocation with a resumptive (so/da ‘so/then’) is ungrammatical in German, as shown below. (9) a.
Ger. [als ich bei meiner Oma war] ([da]r ) hab ich die when I at my granma was there have I the Katze gefüttert. cat fed b. Swe. [när jag var hos mormor] [så]r matade jag katten. I cat-the when I was at granma so fed ‘When I was at granma’s I fed the cat.’
(10) a.
Ger. [dann] [*so/*da]r fütterte ich die Katze. then so/then fed I the cat b. Swe. [sen] [så]r matade jag katten. then so fed I cat-the ‘Then I fed the cat.’
(11) a.
Ger. [Jetzt] [*so/*da]r habe ich die Katze gefüttert. now so/then have I the cat fed b. Swe. [Nu] [så]r har jag matat katten. now so have I fed cat-the ‘Now I’ve fed the cat.’
(12) a.
Ger. [Endlich] [*so/*da]r hab at-last so/then have b. Swe. [Äntligen] [så]r har jag at-last so have I ‘At last I’ve fed the cat.’
ich die Katze gefüttert. I the cat fed matat katten. fed cat-the
This fact should be borne in mind with regard to L2 acquisition. Swedes might – and in Section 6 we will see that they do – produce certain nontargetlike V3 clauses (XXVS) in their interlanguage German not because they have a problem with V2, but because V3 orders with subject-verb inversion and a
Ute Bohnacker
resumptive are permitted in their L1 syntax, a possibility that has not been investigated so far. . V3 with certain adverbs Swedish allows V3 word order in main clauses for a group of adverbs of modality and degree, also known as focalising adverbs (e.g. Egerland 1998; cf. Nilsen 2002 for Norwegian), e.g. bara ‘only/just’, liksom ‘like’, nästan ‘nearly’, rentav ‘fairly/it is no exaggeration to say’, kanske ‘maybe’ (cf. Teleman 1967: 164; Jörgensen 1976: 105–110; Jörgensen & Svensson 1986: 138; Platzack 1986a, 1998: 89; Ekerot 1988: 69–70; Wijk-Andersson 1991; Teleman et al. 1999b: 14, 100). These Swedish adverbs commonly occur between the subject and the finite verb (SAdvV), as illustrated in (13a)–(14a), but are ungrammatical in that position in German, cf. (13b)–(14b).9 (13) a. hon {bara/faktiskt/egentligen} sover. b. *sie {nur/eigentlich} schläft. she only/actually sleeps c. hon sover {bara/faktiskt/egentligen}. d. sie schläft {nur/eigentlich}. (14) a. hon kanske sover. b. *sie vielleicht schläft. she maybe sleeps c. hon sover kanske. d. sie schläft vielleicht. (15) a. kanske hon sover. b. *vielleicht sie schläft. maybe she sleeps c. kanske sover hon. d. vielleicht schläft sie.
Some of these V3-adverbs, especially kanske, can also occur in first position without inversion (AdvSV), cf. (15a) kanske hon sover ‘maybe she sleeps’ alongside (15c) kanske sover hon (maybe sleeps she). Of these adverbs, kanske, which historically is a modal verb taking a clausal complement (‘may be (that) she sleeps’) is the most well known, yet the other V3-adverbs, which cannot be traced back to modal-plus-clause origin, are equally common in spoken Swedish. V3 orders with these adverbs are not obligatory but relatively frequent in spoken Swedish (cf. e.g. Jörgensen 1976: 105–110; Wijk-Andersson
Nonnative acquisition of verb second
1991: 130–150; Josefsson 2004b: 165–167). Note that the corresponding German V3 clauses are always ill-formed (see the b and d examples in (13)–(15)).
. L2 acquisition of German(ic) verb placement: Common notions The following notions about the (adult) L2 acquisition of German(ic) verb placement have long been widespread in certain linguistic circles, and still are today: a. Irrespective of L1, it is hard or impossible to fully acquire V2. b. Learners start out with (and stick to) the canonical word order SVO. c. There is a universal developmental path in L2 German verb placement. Commonly suggested explanations of these notions today are: i. Vulnerability of the C-domain (Platzack 2001) ii. SVX as the underlying more basic word order (Clahsen & Muysken 1986; Platzack 1996) iii. Processability Theory (Pienemann 1998; Pienemann & Håkansson 1999; Håkansson 2001; Håkansson, Pienemann, & Sayehli 2002) As readers will recognise, these ideas are not entirely new. Vulnerability of the C-domain harks back to the idea that the topmost syntactic structure in a phrase is always acquired last in L1 and L2 (cf. e.g. Radford 1988; Vainikka & Young-Scholten 1994). Building on the assumption that language acquisition proceeds from the bottom up, from lexical to functional projections and on to higher functional projections, Platzack (2001: 371–372) hypothesises that the left periphery of the clause, or CP-level (C-domain), is “vulnerable”. Syntactic processes involving these projections, including V2, are said to be acquired last and only imperfectly by L2 learners and SLI children, and to be the first to be adversely affected in aphasia and attrition.10 Alternatively, Platzack (1996) claims that V2 is hard to acquire because SVO is the underlying basic and more economical word order (at least in his 1996 Minimalist framework, building on Kayne 1994). And according to Pienemann’s (1998) Processability Theory, V2 is hard to acquire because the “canonical” SVO word order and XSV (i.e. V3, with a fronted element before the subject) are “easier to process” than inverted XVS (i.e. V2). In essence, these ideas are intended to reflect the notion that certain word orders map function more straightforwardly than others, and that certain word orders (N-V-N, NP-V-NP) therefore are easier to process, recall the alleged semantic universal of “agent–action–patient” and the alleged
Ute Bohnacker
pragmatic universal of “topic first–focus last” (e.g. Bever 1970; Slobin 1973; Givón 1979a, b; Klein & Perdue 1997). And what is easier to process must also be easier to produce. Pienemann explicitly claims that L2ers employ a non-language-specific procedure to map semantic roles and surface forms both for reception and production. Additionally, they may utilise general cognitive principles (or strategies) of initialisation and finalisation (Pienemann 1998: 47, 83–85). For a lucid critique of this notion of processing, see e.g. White (1991) and Eubank (1993). V2, then, is seen as hard or impossible to acquire (fully), irrespective of L1. The hypothesised universal L2 developmental path towards V2 is essentially the same as proposed in the 1980s on the basis of cross-sectional data from Romance L1 learners of German, guest workers in Germany who were studied in the Heidelberger Projekt Pidgindeutsch and ZISA (Zweitspracherwerb italienischer, spanischer und portugiesischer Arbeiter, L2 acquisition of Italian, Spanish and Portuguese workers), by e.g. Meisel, Clahsen and Pienemann (1981), Clahsen, Meisel and Pienemann (1983) and Clahsen and Muysken (1986). An implicational hierarchy of originally 6 developmental stages was established (16), later reduced to 4 (with stages 1 and 2 collapsed and stage 5 scrapped). Starting with SVO, learners did not acquire surface OV order until stage 3 and surface subject-verb inversion (V2) not until stage 4, if ever. (16) Clahsen and Muysken’s (1986) developmental stages. Stage 1: S (AUX/Modal) V O Stage 2: (ADV/PP) S (AUX/Modal) V O Stage 3: (ADV/PP) S Vfinite O Vnonfinite Stage 4: XP Vfinite S O (Vnonfinite ) Stage 5: S Vfinite (ADV) O (Vnonfinite ) Stage 6: main clauses as at stage 5 embedded clauses: daß SOVfinite
At first, the stages in (16) were interpreted as learners starting out with canonical (X)SVO, lacking a UG-constrained grammar, gradually adding “unnatural” phrase-structure rules to this underlying SVO order via a cognitive problemsolving approach to mimic German (Clahsen & Muysken 1986, 1989). But reanalyses of the ZISA data (e.g. duPlessis, Solin, Travis, & White 1987; Tomaselli & Schwartz 1990) showed that the implicational stages could also be captured by assuming initial L1 transfer and subsequent resetting of parameters, i.e. as a succession of interlanguage grammars constrained by UG, such as in (17).
Nonnative acquisition of verb second
(17) Stage 1&2: Romance L1 grammar (SIVO) transfers: head-initial VP, head-initial IP. XP adjunction possible (resulting in V3). Stage 3: Parameter for VP headedness switches from VO to OV. IP remains head-initial (SIOV); V-to-I raising derives postsubject V. Stage 4: V2 becomes operative. IP remains head-initial. Stage 6: Parameter for IP headedness switches to head-final (SOVI).
The litmus test came with Vainikka and Young-Scholten’s (1994) study of elementary learners of German with L1 Korean and Turkish, i.e. SOV languages. They did not follow the implicational hierarchy in (16), but initially produced OV. Apparently they had transferred OV from their L1 to their interlanguage German. It followed that the hypothesis of a universal developmental L2 path could not be upheld in its entirety (see Hoekstra & Schwartz 1994). This finding is acknowledged – though sometimes grudgingly and in footnotes – even by staunch supporters of a universal SVO base order (e.g. Clahsen & Muysken 1986: 104, fn. 5; Håkansson 2001: 69, fn. 1; Platzack 2001: 371, fn. 17). Nevertheless, a number of researchers did and do continue to argue for a universal developmental L2 path as far as the acquisition of V2 is concerned, on the basis of the pairing of Swedish and German, as the next section will show.
. Existing studies of V2 in L1 Swedish learners of L2 German There are few studies of L1 Swedish and nonnative German. Håkansson (2001) reports on a written cross-sectional elicitation test, where 143 Swedish teenagers filled in forms with 6 simplex verb translation sentences and 4 sentences with gaps in their foreign language German. Håkansson tested 48 pupils after 2 months of classroom German 2 hours per week (7th grade), 58 after 1 year of German (8th grade), and 37 after 2 years of German (9th grade). Prior to testing, the pupils had had 4, 5 or 6 years of training in English. Overall, 18% (186/1037) of the students’ sentences are V3 (XSV) (Table 1, Håkansson 2001: 75). 53 of the 143 pupils produce V3 or nontargetlike subject omissions, though unfortunately Håkansson does not separate the two. Håkansson claims that her informants do not transfer V2 to their interlanguage German but start with XSV. However, as she does not provide numbers for SVX or for targetlike V2, nor exact figures for the different age groups, the empirical basis for her claim is perhaps not particularly strong.
Ute Bohnacker
Table 1. Word order in declarative main clauses: Written L2 German (Håkansson 2001) 2 mts to 2 yrs German
SVX
V2
V3
143 informants, 7th, 8th & 9th grade
no information given
no information given
18% (186/1037)
Table 2. Word order in declarative main clauses: Oral cross-sectional L2 German, 20 L1 Swedish teenagers (Sayehli 2001) SVX
V2
V3
8 mts German 10 informants, 6th grade
92% (143/155)
0% (0/155)
8% (12/155)
17 mts German 10 informants, 7th grade
85% (265/312)
3% (10/312)
12% (37/312)
Sayehli (2001) carried out a cross-sectional study where she elicited oral productions via a cut-and-reassemble cartoon strip description task from 20 Swedish pupils, 10 in 6th grade (after 8 months of classroom German) and 10 in 7th grade (after 17 months of German). All had 4–5 years of prior exposure to English. Sayehli’s results were published in Håkansson, Pienemann and Sayehli (2002), and are summarised in Table 2. The high frequency of SVX and the incidence of V3 (Adv-SVX) were interpreted as evidence for canonical SVO word order and for Pienemann’s (1998), i.e. Clahsen, Meisel and Pienemann’s (1983), hierarchy of universal stages of L2 German development.11 The absence of V2 in Sayehli’s 6th graders and the low number of (non-subject-initial) V2 in the 7th graders were interpreted as clear evidence against L1 transfer (e.g. Håkansson, Pienemann, & Sayehli 2002: 256–258). Sayehli’s results are interesting and important, and they do seem to suggest a lack of L1 transfer of V2, but then her database is perhaps not particularly large. For, out of the 155 main clauses that the 10 6th graders produce, nearly all are SVX, and only 12 are non-subject-initial, i.e. there are only 12 contexts overall where V2 could occur. In fact, 4 of the 6th graders (and 2 of the 7th graders) do not produce any adverbial-initial clauses, i.e. there are no contexts in their data where V2 could occur (Sayehli 2001: 27–29). Out of the 312 main clauses that the 10 7th graders produce, only 47 are non-subject initial, and in 22% (10/47) of them, targetlike V2 does occur. It is therefore imperative to verify Sayehli’s results for a larger database from Swedish elementary learners of German, i.e. with more than 12 potential contexts for (non-subject-initial) V2. The high number of SVX clauses is not particularly informative, as this word
Nonnative acquisition of verb second
Table 3. Word order in declarative main clauses: 3 advanced L2ers (Bohnacker 2003) 6-7 yrs German
SVX
V1
V2
V3
3 informants
50% (850/1686)
3% (52/1686)
46% (783/1686)
0.06% (1/1686)
245 311 294
13 17 22
216 266 301
1 0 0
which break down into Eva Tina Henrik
order occurs in all the three languages involved (Swedish, English, German). It might also be worth checking whether there are lexical or pragmatic reasons for the high percentage of SVX: Some learners might still have to acquire the relevant lexical items (adverbs, etc.) for producing adverbial-initial clauses. Secondly, the predominance of SVX over XVS in Swedish in general, as found in the corpus studies discussed in Section 2, may favour SVX in the learners’ interlanguage productions. And thirdly, SVX could also simply be an artifact of Sayehli’s elicitation method, as non-subject-initial clauses are dependent on discourse/pragmatics/information structure. Her informants were asked to describe the events depicted in a cartoon in front of them, and the experimenter prodded and scaffolded if they were taciturn, as becomes evident from Sayehli (2001: 23–25, 37–38, 54). Most importantly however, all of Sayehli’s subjects had 4 to 5 years of exposure to English. Therefore we still need to find out whether the interlanguage of these informants may not have been influenced by English. Before further investigating these issues empirically, it is worth pointing out that advanced L2ers have been shown to behave in stark contrast to Sayehli’s subjects. As part of a larger study, Bohnacker (2003) collected oral and written data from 3 very advanced Swedish post-puberty learners of German who as teenagers had been exposed to 3–4 years of classroom German at secondary school in Sweden and had spent 3 years of their adult lives in Germany. These advanced L2ers did not at all rely on SVX nor violate V2, but frequently use non-subject-initial V2 in their oral and written productions, as illustrated in Table 3. Whilst these data already call into question strong claims about V2 being hard or impossible to acquire fully, we also need to look at what less advanced learners do.
Ute Bohnacker
. New empirical studies . The intermediate learners Cross-sectional oral and written data were collected from 23 16-year-olds at secondary school in Sweden. Like Sayehli’s (2001) subjects, their first foreign language at school had been English, their second German. Only pupils with native Swedish were included. At the end of their third year of classroom German, they were asked to record their thoughts in German on the topic “Was ich in meiner Freizeit tue oder tun möchte” (what I do or would like to do in my spare time) in the form of an oral monologue. It was made clear to them that this was not a test influencing their school grades and they were encouraged to speak as much as possible on the tape. Recording took place in individual language laboratory booths under the supervision of their German teacher; communication between the pupils and the use of written materials were discouraged. This recording method was chosen to forestall short/elliptic utterances typically produced by elementary learners in interaction with an experimenter, to limit non-linguistic deixis, and to favour the production of longer utterances anchored in discourse by explicit linguistic means. This I hoped would encourage learners to produce a high ratio of complete sentences with verb and subject, and non-subject-initial declaratives. The resulting 23 recordings were each 20-to-30 minutes in length, and were subsequently transcribed and analysed. On a separate occasion, these 16-year-olds were asked to write a text in German on the same topic under the supervision of their teacher, not using written materials. This voluntary exercise yielded 10 written texts, which I then also analysed with regard to verb placement. When determining finite verb placement in main clauses, I only considered the first verb (i.e. the simplex verb, or the first verb of a periphrastic construction) and classified this verb as finite, even though morphological tense and agreement marking was often not targetlike, an approach also taken and argued for in other acquisition work (e.g. Lardiere 1998; Bohnacker 1999). I will now present my results in brief; a more detailed discussion can be found in Bohnacker (in progress). Immediately striking is the fact that the 16-year-olds hardly ever produce V3 word orders (4% in the written corpus, 4.5% in the oral corpus), but do produce sizeable percentages of targetlike nonsubject-initial V2 (52% in the written data, 32% in the oral data), see Tables 4 and 5. There is no extreme preponderance of SVX: 43% subject-initial clauses in the written data, and 62% in the oral data.
Nonnative acquisition of verb second
Table 4. Word order in main clauses: Written intermediate L2 German, 10 16-year-olds 3 yrs German
SVX
V1
V2
V3
V4
10 informants
43% (59/138)
–
52% (72/138)
4% (5/138)
1% (2/138)
50 9
– –
63 9
– 5
– 2
which break down into 8 informants 2 informants
Table 5. Word order in main clauses: Oral intermediate L2 German, 23 16-year-olds 3 yrs German
SVX
V1
V2
V3
V4
23 informants
62% (754/1220)
1.5% (18/1220)
32% (386/1220)
4.5% (55/1220)
0.2% (2/1220)
(plus 0.4% (5/1220) wh-questions)
Table 6. Types of V3 (and V4) clauses: Written intermediate L2 German 3 yrs German
XSV
SXV
XXVS
V4
8 informants 2 informants
0 5
0 0
0 0
0 2
Table 7. Types of V3 (and V4) clauses: Oral intermediate L2 German 3 yrs German
XSV
SXV
XXVS
V4
11 informants 12 informants
0 22
0 2
7 24
0 2
Violations of V2 are even rarer than the V3 percentage of 4.5% may suggest, as it includes some instances of XXVS with subject-verb inversion, listed in Tables 6 and 7. These are instances of a left-dislocated adverbial and a resumptive (XXr VS, probably modelled on Swedish, recall Section 3.3), exemplified in (18), and so-XVS (probably modelled on the homophonous Swedish conclusive/consequential V3-så, recall Section 3.1), and exemplified in (19). (18) [wenn man mehr Freizeit haben] [da]r wusste ich richtig was if one more sparetime have then knew I really what ich tun wollten. I do wanted ‘If one had more spare time, then I’d know what I’d want to do.’ (Dagny, oral, 3 yrs)
Ute Bohnacker
(19) so nun haben ich nicht mehr zu erzählen. so now have I not more to tell ‘So now I haven’t got anything else to tell you.’
(Dagny, oral, 3 yrs)
Subtracting such XXVS cases, the 16-year-olds produce only 2% nontarget (uninverted) V3 (and V4) in their main clauses. In fact, 8 out of 10 informants (written data, Table 6), and 11 out of 23 informants (oral data, Table 7) never violate the V2 requirement. The remaining 12 informants do so very rarely. For these 16-year-olds with 3 years of classroom German then, V2 does not seem to be much of a problem at all. Bohnacker (in progress) investigates the 2% nontargetlike V3 and V4 cases in more detail. 25% of these are [so + SVX] with the clause-initial element so/så, which again can be argued to have transferred from Swedish (Section 3.1). Most of the other V2 violations are of the type [adverbial + SVX], a V3 construction not found in Swedish, cf. (20). (20) [wenn ich nach Schule kam] ich mache meine Schulaufgabe. when I after school came I make my homework ‘When I get back from school, I do my homework.’ (Åsa, oral, 3 yrs)
As these pupils have had substantial training in English, one may conjecture that their V2 violations are due to syntactic transfer from English, where they would be grammatical, though I will postpone such discussion until Section 6.2. Readers might wonder whether these pupils simply are too advanced and have already mastered all the stages of the implicational developmental hierarchy up to V2 (recall (16)). If this were the case they should not only have mastered V2, but also VP-headedness (a “lower” stage) and therefore consistently use a head-final VP in their interlanguage German. Generally, learners with a head-initial L1 VP (e.g. Italian, Spanish, Portuguese, English, Swedish) have been found to produce nontarget head-initial VPs in their L2 German (e.g. duPlessis et al. 1987: 67–70; Vainikka & YoungScholten 1996b: 157; Pienemann 1998: 118–121; Håkansson 2001: 79; Sayehli 2001: 27–29, 36–37), whilst learners with a head-final L1 VP (Turkish, Korean) initially produce head-final VPs in their L2 German (e.g. Schwartz & Sprouse 1994: 335; Vainikka & Young-Scholten 1994: 276–277, 293, 1996a). However, both groups of learners have been reported to produce German-style OV long before V2. In some models of acquisition (e.g. minimal trees (Vainikka & Young-Scholten 1994, 1996a: 13–15, 24–25); vulnerable C-domain (Platzack 2001: 371–372); processability (Pienemann 1998: 99–111, 116); teachability
Nonnative acquisition of verb second
Table 8. Nonfinite verb placement in infinitival clauses, sentence fragments, and main clauses with a complex verb: Written intermediate L2 German 3 yrs German
Vnonfin X
XVnonfin
10 informants
12% (6/49)
88% (43/49)
Table 9. Nonfinite verb placement in infinitival clauses, sentence fragments, and main clauses with a complex verb: Oral intermediate L2 German 3 yrs German
Vnonfin X
XVnonfin
23 informants
26% (100/389)
74% (289/389)
–
100% (46/46) 84% (187/198) 52% (76/145)
which break down into 5 informants 11 informants 7 informants
16% (31/198) 48% (69/145)
(Pienemann 1984)), such findings have been taken to mean that L2ers must first have target headedness of the VP before they can go on to acquire V2. As I will show now, however, this is not the case: the 16-year-olds’ nonfinite verb placement (VP-headedness), especially in oral production, is a lot less targetlike than their finite verb placement (i.e. V2). Nonfinite verbal elements in native German occur in final position in finite main clauses with a complex verb (infinitive, participle, particle, etc.), and in nonfinite constructions (infinitival clauses, root infinitives/participles, sentence fragments), and in penultimate position in finite subordinate clauses with a complex verb (i.e. here the finite verb occurs in final position). I culled such nonfinite verbs from the learner data and determined their placement in relation to other constituents (Vnonfin X versus XVnonfin ). Many nonfinite verbs had to be excluded (e.g. utterances with a nonfinite verb but too few telltale constituents to determine headedness),12 but the remaining ones are informative enough; see Tables 8 and 9. The majority of nonfinite verbs in the learner data occur in final position and suggest an interlanguage grammar with a head-final VP. But 12% of the nonfinite verbs in the written data and 26% in the oral data show nontargetlike Vnonfin X placement. The breakdown of the informants into subgroups reveals that 7 of the 23 learners in fact place 48% of their nonfinite verbs in the Vnonfin X
Ute Bohnacker
pattern instead of the XVnonfin pattern; examples are given in (21)–(22). Compare these percentages to the near-perfect production of V2 (Tables 4 and 5 above). No implicational relation such as “if target V2, then target OV” could be detected – many informants who produce nontarget Vnonfin X nevertheless have perfect V2 (for discussion, see Bohnacker (in progress)). (21) ich wollte auch . . . schreiben Poesie und Gedichte. I wanted also write-inf poetry and poems ‘I’d also like to write poetry and poems.’ (Dagny, oral, 3 yrs) (22) ich habe nicht viel Zeit zu tun das. I have not much time to do-inf that ‘I don’t have much time to do that.’
(Yrsa, oral, 3 yrs)
For these learners of German then, the acquisition of V2 is not developmentally dependent on target headedness of the VP (here, OV) having been acquired first: acquiring V2 is much “easier” than acquiring VP headedness. This is an odd and problematic finding for acquisition models that assume a universal path of L2 development (e.g. Håkansson, Pienemann, & Sayehli 2002) or a C-domain that is vulnerable (and an invulnerable V-domain (e.g. Platzack 2001)), and also for models that invoke L1 transfer at lower, lexical projections only (e.g. VP) but not at higher, functional ones (e.g. CP) (e.g. Eubank 1993/1994; Vainikka & Young-Scholten 1994, 1996a, 1996b). Yet the same finding is unsurprising on an approach to second language acquisition that invokes the transfer of L1 syntax (e.g. Schwartz & Sprouse 1994, 1996): Swedish is V2 and its VP head-initial, whilst German is V2 but its VP is head-final. Transferring the Swedish L1 grammar to German will make finite verb placement in main clauses (V2) easy for the learner, but at the same time cause nontargetlike placement of nonfinite verbs – until the parameter setting for VP is changed to head-final. In order to add more empirical force to this reasoning, let’s look at the data from a different study, Swedish L1 ab initio learners of German. . The ab initio learners and the English puzzle In order to capture the initial stages of acquisition, I carried out another investigation, this time of adult ab initio learners of German. A micro-comparative setup was chosen to test the potential influence of knowledge of English, a factor that had not been controlled for in previous studies. I therefore had to locate native Swedes who did not know English (or any other foreign language),
Nonnative acquisition of verb second
a difficult task as English has been a compulsory subject in Swedish schools for decades. Eventually I did find them: 6 old age pensioners who were taking an ab initio German class and were willing to be recorded. All were native Swedes in their late sixties and had lived in Sweden all their lives. For 3 of the subjects (Märta, Signe, Algot), German was their first ever foreign language, their knowledge of English being limited to a few words and phrases (greetings, foodstuffs, swearwords). Their acquisition was to be compared with that of the other 3 informants (Rune, Gun, Ulf), who had learnt English in their youth and had been using it during their careers. Oral production data were collected twice, after 4 months of exposure (German classes, 3 hours per week) and after 9 months of exposure. The elicitation method was the same as with the 16year-old pupils: The ab initio learners were asked to record an oral monologue in German in the language lab on the topic “Was ich in meiner Freizeit tue oder tun möchte” (what I do or would like to do in my spare time) and to be as talkative as possible. This yielded four recordings at data point 1 (4 months of exposure) and six recordings at data point 3 (9 months of exposure). In addition, naturalistic oral production in dialogue form was elicited: At data point 2 (4 months of exposure), two of the learners were recorded in their home for 120 minutes each in one-to-one interaction with a monolingual speaker of German. More details on the methodology can be found in Bohnacker (2005). I then transcribed the recordings orthographically and analysed them. Oneword utterances and verb-less utterances were excluded. For investigating V2, sentence fragments and subordinate clauses, (subjectless) imperatives and subjectless main clauses were excluded. Main clauses with verb and subject were classified with respect to placement of the simplex verb, or of the first verb in a periphrastic construction. Just as for the intermediate learners, I treated these verbs as finite, even though morphological tense and agreement marking is often nontargetlike. Compared to earlier studies (e.g. Håkansson 2001; Sayehli 2001), my corpora contain a large number of utterances relevant to the acquisition of finite and nonfinite verb placement. The results are as follows. The 6 ab initio learners produce many subjectinitial main clauses, with an overall average of 67% (1055/1574), but they also produce 31% non-subject-initial clauses (Table 10), and 2% V1. SVX never is the exclusive word order, at neither 4 nor 9 months. This finding indicates that not all learners of German start out with “canonical” word order; it also suggests that empirical studies which only find SVX (e.g. Sayehli 2001) may have arrived at that result because of a small database or a method that mainly elicits SVX utterances.
Ute Bohnacker
Table 10. Word order in main clauses, percentages: Ab initio L2 German
The ab initio informants productively use non-subject-initial V2 already after only 4 months of exposure to German (14%–31%; see Table 10). Such early acquisition of V2 has not been documented for learners with a non-V2 L1, which suggests that L1 knowledge of a V2-language (Swedish) does make it easier to acquire V2 in a second or third language (German). Crucially however, there is a categorical difference between learners for whom German is their first nonnative language and those who have prior knowledge of English: The informants who do not know English (Märta, Algot, Signe, white rows in Table 10) do not produce V3 main clauses in their interlanguage German, but those informants who do know English (Rune, Gun, Ulf, rows shaded in grey in Table 10) do produce V3. Just as for the 16-year-old intermediate learners, V3 does not always indicate a violation of V2: 8% of the ab initio learners’ V3 clauses are instances of subject-verb inversion (XXVS), where the first element is an adverbial and the second a resumptive, as listed in Table 11 and exemplified in (23). If we exclude such instances of XXr VS from the V3 counts, the figures change very slightly,
Nonnative acquisition of verb second
Table 11. Types of V3 clauses, raw figures: Ab initio L2 German Data point
XSV
SXV
XXr VS
Rune1 Gun1 Rune2 Algot3 Rune3 Gun3 Ulf3 Total
13 9 22 0 17 23 12 91% (96/106)
0 0 1 0 0 0 0 1% (1/106)
0 0 1 1 2 5 0 8% (9/106)
but the difference between the two groups of learners is even clearer: Those who do not have knowledge of English never violate V2 (0%). (23) [Im Weihnacht] [dann] bjude ich ein Freund auch. in-the Christmas then inviteSWE I a friend also ‘At Christmas I also invite a friend.’ (Gun3, 9 months)
The difference between the two learner groups is even more prominent if we only consider non-subject-initial clauses and disregard the SVX clauses, which are uninformative with regard to subject-verb “inversion” (Table 12). Learners who know English produce on average 41% (97/235) nontarget V3 in their non-subject-initial main clauses, whereas those who do not know English never do (0%, 0/275). Examples of nontarget V3 are given in (24)–(28) below. Nearly all (96/106, cf. Table 11) are of the type XSV.13 (24) und dann ich lege detta kort und dann du nächst. and then I put [this card]SWE and then you next ‘And then, then I(’ll) play this card and then you(’ll play) the next.’ (Rune2, 4 mts) (25) wenn wir ist in Sommerhaus wir gehe promenad in Wald. when we are in cottage we go walkSWE in wood ‘When we’re at the cottage we go for a walk in the woods.’ (Rune2, 4 mts) (26) in Montags ich seh ein Programm um Essen # teve. in mondays I see a programme about food # tellySWE ‘On Mondays, I watch a programme on television about food.’ (Gun1, 4 mts) (27) aber manchmal sie helfe nicht, sie muß arbeiten, ja. but sometimes she help not she must work yes ‘But sometimes she doesn’t help, she’s got to work.’ (Gun3, 9 mts)
Ute Bohnacker
Table 12. Word order in non-subject-initial main clauses: Ab initio L2 German
(28) ich habe viel hören um Bodensee, so ich will gerne I have much hear about Lake-Constance, so I want gladly sehen das. see it ‘I’ve heard a lot about Lake Constance, so I’d like to see it.’ (Gun3, 9 mts)
As Table 13 shows, the first constituent in the XSV V3 clauses always is an adjunct, not an argument. 26% of the XSV V3 main clauses are introduced by the connective element so (‘so’). This use of so is strongly reminiscent of Swedish, where conclusive/consequential så does not require inversion (så + SVX), resulting in a V3 order, compare so ich will gerne sehen das ((28), interlanguage German) with Swedish så jag vill gärna se det (so I want gladly see it). så – and interlanguage so – allows adjunction to CP (or to IP, for readers who prefer to treat subjectinitial main clauses as IPs). We might therefore explain these so-V3 utterances as L1-induced from Swedish. The same V3 order occurs in English with conclu-
Nonnative acquisition of verb second
Table 13. Types of first constituent in XSV V3 clauses: Bulked data from Rune, Gun, Ulf Argument temporal 0
0% (0/96)
63% 60/96 (incl. 40 dann/denn/sen)
Adjunct locational connective so/så
other adverbial
0% 0/96
9% 9/96
26% 27/96
100% (96/96)
sive so (so I would like to see it), so that the utterances could also be interpreted as L2-induced. Indeed, two reviewers suggest that interference from L2 English is more likely than from L1 Swedish here, because informants with no knowledge of English do not produce so-V3 orders. The point is well taken, though on the basis of my naturalistic production data we do not know whether soV3 really is ruled out in the interlanguage of the informants without English; grammaticality judgment or elicited-production experiments would be necessary to decide this point. Two of the learners do start an utterance with så jag (so I) or so ich (so I . . . ) but then break off, leaving us uninformed about verb placement. The bulk of the ab initio learners’ V3 main clauses are introduced by other elements than so, often temporal adverbials (63%), and exhibit an AdvSVX word order not generally permitted in Swedish. Though many such V3 clauses involve initial dann ‘then’ (e.g. (24)), which might be construed as L1-induced, as V3 is marginally possible with the Swedish equivalent of dann, sen ‘then’ (Section 3.2), the learners also produce V3 with other clause-initial adverbials. I suggest that these are adjunctions to IP/CP in the learners’ interlanguage, transferred from English, where such adjunction is freely allowed, compare (24)–(28): und dann ich lege . . . ‘and then I put. . . ’; wenn wir ist in Sommerhaus wir gehe ... ‘when we’re at the cottage we go. . . ’; in Montags ich seh . . . ‘on Mondays I watch . . . ’; aber manchmal sie helfe . . . ‘but sometimes she helps . . . ’; so ich will . . . ‘so I’d like . . . ’ I therefore suggest that there is partial transfer of non-V2 English syntax to the learners’ German interlanguage grammar, and, in the particular case of so-adjunction (and potentially dann-adjunction), transfer from L1 Swedish and L2 English may work together. Only informants with previous knowledge of English produce XSV, but interestingly, transfer from L2 English seems to be most prevalent in connection with those phonologically light sentence-initial lexical elements that in Swedish also allow XSV, i.e. the V3-inducing connective adverbials: så (conclusive ‘so’) and sen (‘then’).
Ute Bohnacker
One might argue (with e.g. Håkansson, Pienemann, & Sayehli 2002) that the existence of nontarget V3 utterances in itself does not constitute evidence for transfer from English. This is true. However, we must not forget that there is a categorical difference between the two ab initio learner groups. It is the absence of nontarget V3 utterances in the learners who do not know English that remains a complete mystery under their approach – a mystery that is easily solved if we allow for L2-transfer from L2 English in the other group. Thus, L2 knowledge of a non-V2 language can make it more difficult to comply with the V2 requirement of the L3, even though the learner’s L1 is a V2 language. In this regard, Swedes who do not know English are at an advantage when learning German, but of course they need not be at an advantage as far as syntactic phenomena other than V2 are concerned – and as for other language constellations, knowledge of English may help or impede acquisition in quite different ways. It should be stressed that the interference of non-V2 English syntax on the learners’ interlanguage German is not total, but only intermittent. Recall that the ab initio learners who know English do produce ca. 50% targetlike V2 out of their non-subject-initial clauses after 4 months (Table 12), and ca. 60% after 9 months. And as readers will recall from Section 6.1, after 3 years of exposure, the 16-year-old Swedes are virtually perfect in observing the V2 requirement in their interlanguage German, which is earlier than what the literature generally reports of learners of German with an L1 that is not V2: Tutored learners of German with L1 English and untutored learners (with e.g. Romance, Turkish and Korean L1s) do not or do not fully adhere to the V2 constraint, despite years of exposure (cf. e.g. Clahsen & Muysken 1986; duPlessis et al. 1987; Ellis 1989; Schwartz & Sprouse 1994, 1996; Vainikka & Young-Scholten 1994, 1996a, b; Pienemann 1998: 118–130). The high frequencies of targetlike V2 at 4 and 9 months might make some readers wonder whether the ab initio learners are precocious and too advanced already. Yet this is unlikely to be the case considering their low MLUs, their small vocabulary and their nontarget morphology (Bohnacker 2005). And while they may be mastering V2, other aspects of their interlanguage syntax are strikingly nontargetlike, as will now be shown for VP headedness (VO/OV); in my corpora, head-final VP and V2 in interlanguage German are unrelated. Just as for the 16-year-olds (Section 6.1), I culled nonfinite verbs from the ab initio learner data and determined their placement; the results are given in Table 14. There is no substantial difference between the learners who know English (grey rows) and those who do not (white rows).
Nonnative acquisition of verb second
However, there is an important difference between the learners’ placement of nonfinite verbs after 4 months of exposure to German, when 87% (199/228) precede other material (nontargetlike Vnonfin X), and after 9 months of exposure, when only 29% (56/191) of their nonfinite verbs appear in Vnonfin X order. A plausible interpretation for this difference is that at 4 months, the learners have a head-initial VP in their interlanguage grammars (which they could have transferred from L1 Swedish), but in the 9-months recordings, they are changing to a head-final VP. Examples are given in (29) and (30). (29) nun haben ich spielt Boule vier Jahr. now have I played boules four year ‘I’ve now been playing boules for four years.’ (Target: nun habe ich vier Jahre Boule gespielt.)
(VO, Märta2, 4 mths)
(30) und dann solln ich Boule spielen. and then shall I boules play ‘And then I’ll play boules.’
(OV, Märta3, 9 mths)
This change from a VO to an OV interlanguage grammar is by no means complete at 9 months (recall also that many of the 16-year-old intermediate learners still have problems with nonfinite verb placement after 3 years of German), but it is a clear tendency. Moreover, it is entirely unrelated to the acquisition of V2: At 4 months, Märta and Algot produce 85% nontargetlike Vnonfin X (Table 14), at a time when their non-subject initial main clauses show perfect V2 (100%, 93/93, Table 12); and likewise at 4 months, Rune and Gun (who know English) produce 90% nontargetlike Vnonfin X (Table 14), at a time when, by contrast, ca. 50% of their non-subject initial main clauses are targetTable 14. Nonfinite verb placement in infinitival clauses, sentence fragments, and main clauses with a complex verb: Ab initio L2 German
Ute Bohnacker
like V2 (Table 12). For learners both with and without knowledge of English, then, acquiring V2 in German seems to be much easier and happens earlier than acquiring a head-final VP. Again, this empirical finding is sharply at odds with the claims and predictions of acquisition models that assume universal L2 developmental stages (e.g. processability theory (Pienemann 1998)), that assume that target lexical projections (VP) are developmentally prior to target functional ones (e.g. minimal trees (Vainikka & Young-Scholten 1994, 1996a, b), modulated structure building (Hawkins 2001: 73–75, 146)), or that assume that learners only have to grapple with the acquisition of the topmost levels of syntactic structure (e.g. vulnerable C-domain (Platzack 2001)).
. Conclusion In this paper I have explored the acquisition of verb placement by Swedish learners of German at elementary (ab initio) and intermediate levels, focusing on finite verb placement in main clauses (V2) and nonfinite verb placement (VP headedness). The empirical data suggest that contrary to claims in the literature, learners do not necessarily start out with “canonical” SVX. If there exists a canonical word order at all, all it means is a word order of high frequency, but it is certainly not exclusive. The findings also clearly indicate – contrary to what is often claimed – that Germanic V2 is not hard to acquire per se: With an appropriate elicitation method, it was shown that (non-subject-initial) V2 is productive and targetlike (100% contexts) already after just 4 months in Swedish ab initio learners of German as their first L2. I do not know of any similar results from learners with non-V2 L1s. Ab initio learners who know English and for whom German is the L3 also productively use (non-subject-initial) V2 after 4 months of exposure, but only in 50% of obligatory contexts: additionally, they produce nontargetlike V3, which indicates that knowledge of a non-V2 language (English) can make it harder to acquire V2. I have interpreted these results as clear evidence for L1-syntax transfer of the V2 property from Swedish to German (including modest evidence for L1-transfer of a small group of constructions that are exceptions to the V2 requirement), and as evidence for partial L2-syntax transfer from English to L3 interlanguage German. Hardly any traces of this hindering influence of English remain after 3 years of German exposure in the cross-sectional data of intermediate-level learners.
Nonnative acquisition of verb second
In contrast to these findings regarding V2, my informants do have problems acquiring the nonfinite verb placement of German. At first, after 4 months of exposure, the ab initio learners produce 87% nontargetlike VO orders. After 9 months, the percentage of such VO has dropped to 29%, presumably because VP headedness in the interlanguage grammars is being changed from head-initial to head-final. However, nonfinite verb placement is not completely targetlike yet, nor is it so for most of the intermediate-level learners after 3 years of exposure (26% VO). The acquisition of targetlike nonfinite verb placement (a syntactic phenomenon involving the VP domain) thus lags behind the acquisition of V2 (involving the C-domain) in Swedish learners of German. None of this is particularly surprising if we assume L1 syntax transfer in second language acquisition. On a transfer approach such as Schwartz and Sprouse’s (1994, 1996) Full Transfer/Full Access model, according to which learners initially produce and process L2 utterances entirely through the L1 grammar, and only later change their interlanguage syntax by acquiring new rules/constraints/parameter settings, we expect to find divergent L2 developmental routes with respect to the same target language for groups of learners with typologically distinct L1s. Thus we also expect to find groups of speakers of V2 languages who transfer the V2 property from their L1 to their interlanguage grammar (see e.g. Hulk 1991), and who therefore acquire V2 in a V2-L2 early and easily, even though this has not been documented empirically until now. The individuals acquiring V2 early and easily are our Swedish ab initio learners of German who do not know English, exhibiting L1 transfer of V2 in its purest form. On the other hand, L1 transfer of a head-initial VP is predicted to result in the production of nontargetlike head-initial VP utterances, and this is documented for the Swedish learners of German in my studies irrespective of whether they know English or not. Finally, there are the Swedish learners of German as an L3 who do know English, a group of learners that earlier research has focused on, unfortunately without paying any attention to the possibility of English influence. To capture the developmental path of these learners, existing models of syntactic transfer (such as Schwartz & Sprouse 1994) would need to be enriched to also allow for L2 syntactic transfer, yielding potentially divergent L3 developmental routes with respect to the same target language for groups of learners with the same L1 but with different, typologically distinct L2s. The assumption that there may be both L1-syntax and L2-syntax transfer in L3 acquisition is not particularly widespread today. However, much (non-generative) research on trilingualism has already documented L2 transfer alongside L1 transfer in L3 production: In the domain of the lexicon, L1
Ute Bohnacker
and L2 transfer to the L3 seems to be the rule rather than the exception (cf. e.g. Williams & Hammarberg 1998; Hammarberg 2001; Heine 2001; Ringbom 2001), and also Sayehli (2001) notes lexical transfers from L1 Swedish and L2 English in the productions of her L3 German learners. The same holds for my subjects here (for details, see Bohnacker 2005). Similarly for the domains of phonology and morphology, L1 and L2 transfer to the L3 has been documented for some learners (e.g. Hammarberg 2001: 32–35). There is no reason why syntax should be exempt from such transfer. As for predicting whether learners will transfer aspects of their L2 syntax, I think we should look to the factors that have been shown to condition L2 influence on the L3 in nonsyntactic domains, namely (i) L2 proficiency in the learner, (ii) perceived typological closeness of the languages involved, and (iii) recency of L2 use (cf. e.g. Cenoz, Hufeisen, & Jessner 2001). I have tried to show that the following notions about nonnative acquisition of syntax cannot be upheld empirically: a. Irrespective of L1, it is hard or impossible to fully acquire V2. b. Learners start out with (and stick to) the canonical word order SVO. c. There is a universal developmental path in German verb placement. As a consequence, I suggest that theories of nonnative acquisition that are based on these notions also lack an empirical underpinning and should therefore be abandoned, and this includes current generative models that postulate the existence of universally vulnerable (or universally invulnerable) syntactic domains. I believe it is time we took serious the complexity of different language constellations and the existence of transfer in nonnative acquisition. What needs to be done is to carry out methodologically sound comparative empirical studies of interlanguage for a larger variety of L1/L2 and L1/L2/L3 language combinations. Before results from these are in, it is advisable to tread cautiously when propounding “universal” developmental sequences for the acquisition of nonnative syntax.
Notes * I would like to thank the editors of this volume and two anonymous reviewers for their extremely helpful comments and suggestions, and the audiences at CUNY, Lund, Växjö, UMass Amherst and Boston University for valuable discussions. This work was supported by grant 142-2001-6286 from the Swedish Research Council (Vetenskapsrådet). An earlier
Nonnative acquisition of verb second
version of this paper has appeared in Working Papers in Scandinavian Syntax 74, December 2004. . Occasionally the Vorfeld does not contain one syntactic constituent, but only part of a constituent (discontinuous elements), or what seem to be two constituents. The latter are often analysed as two syntactic components merged into one information unit (e.g. Zifonun et al. 1997: 1626–1638) or as remnant movement (e.g. G. Müller 1998; S. Müller 2004). I will not discuss such cases here. . The Vorfeld can also be empty (yes/no questions, declaratives). V1 declaratives appear to be used more frequently and with a wider range of functions in spoken Swedish discourse than in German, cf. Önnerfors (1997) for German, Mörnsjö (2002) for Swedish. I will not discuss such cases here. . Certain analyses move the verb from the VP directly into C (e.g. Holmberg & Platzack 1988: 31; Haider 1993), others do so via one or more head positions in the IP domain (e.g. Grewendorf et al. 1987: 221; Grewendorf 1988: 219; Vikner 1991/1995). The existence and headedness of IP (TP, AgrP, etc.) in German is a matter of debate (cf. Haider 1993). . I am aware of other approaches that model V2 as single leftward movement of a larger constituent which includes the verb (e.g. G. Müller 2004). . A plethora of ideas exist about what might motivate and drive V2, e.g. a spec-head relationship (e.g. Zwart 1993; Grimshaw 1994), some (e.g. tense/finiteness) feature of the verb or on the position it moves to (e.g. Platzack 1986a, b; Platzack & Holmberg 1989), and/or some (e.g. topic/focus/contrast) feature of the XP constituent or the left-peripheral position it moves to (e.g. Grewendorf 2002; Frey 2004). The wide variety of elements that can occur in the Vorfeld – including nonreferential arguments (like the subjects of weather verbs), adverbials and V-projections – makes it difficult, I believe, to argue that they have an abstract grammatical feature in common (cf. also Haider 1993: 69–70). . There is also an adverb-of-manner type så ‘so/like this/this way’, which is not connective. It can occur inside the VP, it can be stressed, it can be preceded by a coordinating conjunction, and if used clause-initially, it induces inversion, resulting in a V2 clause, as shown in (i). (i)
a.
så köpte dom upp EM Airways. this-way bought they up EM Airways ‘This is how they took over EM Airways.’ b. *så dom köpte upp EM Airways.
. Incidentally – and in contrast to German – the Dutch equivalent of Swedish så, dus (‘so/thus/therefore/hence’), can also function as a clause-initial connective, adjoining to CP and giving a V3 order. Thus, in the Germanic V2 languages, the general ban on adjunction to CP seems to be stricter in certain languages than in others, and there seems to be cross-linguistic variation regarding the (small) set of elements that may adjoin to CP. . I will not discuss “V3” root clauses that involve a pause or intonation break between the initial constituent (left dislocated or otherwise) and the preverbal constituent. . Incidentally, Zwart (this volume) notes that in Dutch, a small set of “connecting adverbs” can intervene between the clause-initial constituent and the finite verb, e.g. echter ‘however’,
Ute Bohnacker
nu ‘now’, dan ‘then’, daarentgegen ‘in contrast’. Their semantics is somewhat different from the focalising adverbs of Swedish and Norwegian. For a discussion of related connectives in German, see Zifonun et al. (1997: 1637–1638). . For counterevidence, see Bohnacker (2003) for L2 acquisition, and de Roo (2003) for aphasia. . Sayehli (2001: 30–33) claims that Adv-SVX precedes OV in accordance with the implicational hierarchy, as she finds zero instances of OV in her 6th grade data, and only 7 in her 7th grade material. This may however be an artifact of data selection: Sayehli only counts uninterrupted declarative clauses with subject and verb, disregarding all sentence fragments, infinitival clauses, root infinitives, root participials, sentences with subject omissions, subordinate clauses and interrogatives. In order to investigate VO vs. OV, one would certainly want to look at the nonfinite verbal elements in all utterances, and not only at complex verbs in declaratives. . Other such uninformative cases are nonfinite verbs with a sentential complement, which must occur postverbally in German. Moreover, as the following authentic native German examples show, nonfinite verbs bearing focal stress can be optionally followed by a defocused adverb (i), and especially in informal speech, certain types of adjuncts may also occur postverbally, cf. (ii). (i)
ich hab sie gerade noch erWISCHT gestern. I have her just still caught yesterday ‘I reached her just in time yesterday.’
(ii) das Geld wurde alles aufgeteilt zwischen ihnen. the money was all divided between them ‘All the money was shared between them.’ . There is also 1 instance of SAdvV, with the Swedish focalising adverb bara ‘only’ (i). bara allows V3 word order in Swedish (cf. Section 3.4), and the learner may have transferred this to German, the same V3 order also being allowed in English, his L2. (i)
sage so, aber es stimme nicht, nein. Märta bara Märta onlySWE say so but it is-true not no ‘Märta only says that, but it isn’t true, no.’
(Rune2, 4 mths)
References Altmann, H. (1981). Formen der “Herausstellung” im Deutschen. Tübingen: Niemeyer. Barbier, I. (1995). Configuration and Movement in Germanic: Studies of first language acquisition of Dutch word order. PhD Dissertation, Cornell University. Bever, T. G. (1970). The cognitive basis for linguistic structure. In J. R. Hayes (Ed.), Cognition and the Development of Language (pp. 279–352). New York, NY: John Wiley. Bohnacker, U. (1999). Icelandic plus English: Language differentiation and functional categories in a successively bilingual child. PhD Dissertation (2nd version), Durham University.
Nonnative acquisition of verb second
Bohnacker, U. (2003). Residual word order problems in advanced adult L2 Swedish and German. In L.-O. Delsing, C. Falk, G. Josefsson, & H. Sigurðsson (Eds.), Grammar in Focus: Festschrift for Christer Platzack 18 November 2003, Vol. 2 (pp. 37–45). Lund: Wallin and Dalholm. Bohnacker, U. (2005). When Swedes begin to learn German: From V2 to V2. Forthcoming in Second Language Research. Bohnacker, U. (in progress). Detailed analysis of V2 in 23 Swedish learners of German. Branigan, P. (1996). Verb second and the A-bar syntax of subjects. Studia Linguistica, 50, 50–79. Carroll, M. & M. Lambert (2003). Information structure in narratives and the role of grammaticised knowledge. In C. Dimroth & M. Starren (Eds.), Information Structure and the Dynamics of Language Acquisition (pp. 268–287). Amsterdam: John Benjamins. Cenoz, J., B. Hufeisen, & U. Jessner (Eds.) (2001). Cross-linguistic Influence in Third Language Acquisition: Psycholinguistic perspectives. Clevedon: Multilingual Matters. Clahsen, H, J. M. Meisel, & M. Pienemann (1983). Deutsch als Zweitsprache: Der Spracherwerb als ausländischer Arbeiter. Tübingen: Narr. Clahsen, H. & P. Muysken (1986). The availability of universal grammar to adult and child learners: A study of the acquisition of German word order. Second Language Research, 2, 93–119. Clahsen, H. & P. Muysken (1989). The UG paradox in L2 acquisition. Second Language Research, 5, 1–29. den Besten, H. (1977/1983). On the interaction of root transformations and lexical deletive rules. 1977 ms. University of Amsterdam. In W. Abraham (Ed.), On the Formal Syntax of the Westgermania (pp. 47–121). Amsterdam: John Benjamins. de Roo, E. (2003). Split CP in Broca’s aphasia: Dutch topicalization data. Paper presented at GLOW Workshop II Generative Approaches to Language Development, Lund University, 12 April 2003. Drach, E. (1937/1963). Grundgedanken der deutschen Satzlehre (4th edition). Frankfurt/ Main. Reprinted 1963. Darmstadt: Wissenschaftliche Buchgesellschaft. du Plessis, J., D. Solin, L. Travis, & L. White (1987). UG or not UG, that is the question: A reply to Clahsen and Muysken. Second Language Research, 3, 56–75. Egerland, V. (1998). On verb-second violations in Swedish and the hierarchical ordering of adverbs. Working Papers in Scandinavian Syntax, 61, 1–22. Eisenberg, P., H. Gelhaus, H. Henne, H. Sitta, & H. Wellmann (1998). Duden Grammatik der deutschen Gegenwartssprache (6th edition). Mannheim: Dudenverlag. Ekerot, L.-J. (1988). Så-konstruktionen i svenskan: Konstruktionstypen “Om vädret tillåter, så genomföres övningen” i funktionellt grammatiskt perspektiv. Lund: Lund University Press. Ekerot, L.-J. (1995). Ordföljd, tempus, bestämdhet: Föreläsningar om svenska som andraspråk. Malmö: Gleerups. Ellis, R. (1989). Are classroom and naturalistic acquisition the same? A study of the classroom acquisition of German word order rules. Studies in Second Language Acquisition, 11, 305–328. Erdmann, O. (1886/1985). Grundzüge der deutschen Syntax und ihre geschichtliche Entwicklung. Vol. 1. Stuttgart: J. G. Cotta. Reprinted 1985. Hildesheim: Georg Olms.
Ute Bohnacker
Eubank, L. (1993). Sentence matching and processing in L2 development. Second Language Research, 9, 253–280. Eubank, L. (1993/1994). On the transfer of parametric values in L2 development. Language Acquisition, 3, 183–208. Fabricius-Hansen, C. & K. Solfjeld (1994). Deutsche und norwegische Sachprosa im Vergleich: Ein Arbeitsbericht. Arbeitsberichte des Germanistischen Instituts der Universität Oslo 6. Frey, W. (2004). The grammar-pragmatics interface and the German prefield. Sprache und Pragmatik Arbeitsberichte, 52, 1–39. Givón, T. (1979a). From discourse to syntax: Grammar as a processing strategy. Syntax and Semantics, 12, 81–112. Givón, T. (1979b). On Understanding Grammar. New York, NY: Academic Press. Grewendorf, G. (1988). Aspekte der deutschen Syntax. Tübingen: Narr. Grewendorf, G. (2002). Minimalistische Syntax. Tübingen: UTB. Grewendorf, G., F. Hamm, & W. Sternefeld (1987). Sprachliches Wissen: Eine Einführung in moderne Theorien der grammatischen Beschreibung. Frankfurt: Suhrkamp. Grimshaw, J. (1994). Minimal projection and clause structure. In B. Lust, M. Suñer, & J. Whitman (Eds.), Heads, Projections and Learnability (pp. 75–83). Hillsdale, NJ: Lawrence Erlbaum. Haider, H. (1986). V-second in German. In H. Haider & M. Prinzhorn (Eds.), Verb Second Phenomena in Germanic Languages (pp. 49–75). Dordrecht: Foris. Haider, H. (1993). Deutsche Syntax – generativ: Vorstudien zur Theorie einer projektiven Grammatik. Tübingen: Narr. Håkansson, G. (2001). Against Full Transfer: Evidence from Swedish learners of German. Lund University Department of Linguistics Working Papers, 48, 67–86. Håkansson, G. & U. Nettelbladt (1993). Developmental sequences in L1 (normal and impaired) and L2 acquisition of Swedish syntax. International Journal of Applied Linguistics, 3, 131–157. Håkansson, G., M. Pienemann, & S. Sayehli (2002). Transfer and typological proximity in the context of second language processing. Second Language Research, 18, 250–273. Hammarberg, B. (2001). Roles of L1 and L2 in L3 production and acquisition. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross-linguistic Influence in Third Language Acquisition: Psycholinguistic perspectives (pp. 21–41). Clevedon: Multilingual Matters. Hawkins, R. (2001). Second Language Syntax. Oxford: Blackwell. Heine, L. (2001). Der Einfluss vorher gelernter Sprachen auf das aktuelle Sprachenlernen. MA Dissertation, University of Hamburg. Heringer, H. J. (1989). Lesen lehren lernen: Eine rezeptive Grammatik des Deutschen. Tübingen: Niemeyer. Hoberg, U. (1981). Die Wortstellung in der geschriebenen deutschen Gegenwartssprache. München: Hueber. Hoekstra, T. & B. D. Schwartz (1994). On the initial states of language acquisition. In T. Hoekstra & B. D. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 1–20). Amsterdam: John Benjamins. Holmberg, A. & C. Platzack (1988). On the role of inflection in Scandinavian syntax. Working Papers in Scandinavian Syntax, 42, 23–42.
Nonnative acquisition of verb second
Holmes, P. & I. Hinchliffe (1994). Swedish: A comprehensive grammar. London: Routledge. Hulk, A. (1991). Parameter setting and the acquisition of word order in L2 French. Second Language Research, 7, 1–34. Jörgensen, N. (1976). Meningsbyggnaden i talad svenska. Lund: Studentlitteratur. Jörgensen, N. & J. Svensson (1986). Nusvensk grammatik. Malmö: Gleerups. Josefsson, G. (2004a). Input and output. In G. Josefsson, C. Platzack, & G. Håkansson (Eds.), The Acquisition of Swedish Grammar (pp. 95–133). Amsterdam: John Benjamins. Josefsson, G. (2004b). Non-target structures and non-target uses in child and adult Swedish. In G. Josefsson, C. Platzack, & G. Håkansson (Eds.), The Acquisition of Swedish Grammar (pp. 155–193). Amsterdam: John Benjamins. Kayne, R. (1994). The Antisymmetry of Syntax. Cambridge, MA: The MIT Press. Klein, W. & C. Perdue (1997). The Basic Variety (Or: couldn’t natural languages be much simpler?). Second Language Research, 13, 301–347. Koster, J. (1975). Dutch as a SOV language. In A. Kraak (Ed.), Linguistics in the Netherlands 1972–1973 (pp. 165–177). Assen: Van Gorcum. Lardiere, D. (1998). Dissociating syntax from morphology in a divergent L2 end-state grammar. Second Language Research, 14, 359–375. Meisel, J., H. Clahsen, & M. Pienemann (1981). On determining developmental stages in natural second language acquisition. Studies in Second Language Acquisition, 3, 109– 135. Mörnsjö, M. (2002). V1 Declaratives in Spoken Swedish [Lundastudier i nordisk språkvetenskap A59]. PhD Dissertation, Lund University. Müller, G. (1998). Incomplete Category Fronting. A derivational approach to remnant movement in German. Dordrecht: Kluwer. Müller, G. (2004). Verb second as vP-first. Journal of Comparative Germanic Linguistics, 7, 179–234. Müller, S. (2004). Zur Analyse der scheinbar mehrfachen Vorfeldbesetzung. Paper presented at Symposium Deutsche Syntax: Theorie und Empirie, Gothenburg University, 15 May 2004. Nilsen, Ø. (2002). V2 and Holmberg’s generalization. In J.-W. Zwart & W. Abraham (Eds.), Studies in Comparative Germanic Syntax (pp. 151–173). Amsterdam: John Benjamins. Önnerfors, O. (1997). Verb-erst-Deklarativsätze: Grammatik und Pragmatik. PhD Dissertation. Stockholm: Almqvist and Wiksell International. Paul, H. (1919). Deutsche Grammatik. Teil IV: Syntax. Vol. 3. Halle/Saale: Max Niemeyer. Pienemann, M. (1984). Psychological constraints on the teachability of languages. Studies in Second Language Acquisition, 6, 186–214. Pienemann, M. (1998). Language Processing and Second Language Development: Processability Theory. Amsterdam: John Benjamins. Pienemann, M. & G. Håkansson (1999). A unified approach toward the development of Swedish as L2. Studies in Second Language Acquisition, 21, 393–420. Platzack, C. (1986a). COMP, INFL, and Germanic word order. In L. Hellan & K. Koch Christensen (Eds.), Topics in Scandinavian Syntax (pp. 185–234). Dordrecht: Reidel. Platzack, C. (1986b). The position of the finite verb in Swedish. In H. Haider & M. Prinzhorn (Eds.), Verb Second Phenomena in Germanic Languages (pp. 27–47). Foris: Dordrecht.
Ute Bohnacker
Platzack, C. (1996). The Initial Hypothesis of Syntax: A minimalist perspective on language acquisition and attrition. In H. Clahsen (Ed.), Generative Perspectives on Language Acquisition (pp. 369–414). Amsterdam: John Benjamins. Platzack, C. (1998). Svenskans inre grammatik – det minimalistiska programmet: En introduktion till modern generativ grammatik. Lund: Studentlitteratur. Platzack, C. (2001). The vulnerable C-domain. Brain and Language, 77, 364–377. Platzack, C. & A. Holmberg (1989). The role of agreement and finiteness. Working Papers in Scandinavian Syntax, 43, 51–76. Radford, A. (1988). Small children’s small clauses. Transactions of the Philological Society, 86, 1–43. (= 1986 Bangor Research Papers in Linguistics, 1, 1–38). Reis, M. (1980). On justifying topological frames: ‘Positional field’ and the order of nonverbal constituents in German. Documentation et Recherche en Linguistique Allemande Contemporaine, 22/23, 59–85. Ringbom, H. (2001). Lexical transfer in L3 production. In J. Cenoz, B. Hufeisen, & U. Jessner (Eds.), Cross-linguistic Influence in Third Language Acquisition: Psycholinguistic perspectives (pp. 59–68). Clevedon: Multilingual Matters. Rosén, C. (2005). Deutsch als Fremdsprache: Zur Informationsstruktur schwedischer Deutschstudierender. Forthcoming in A. Bareis & I. Karhiaho (Eds.), Text im Kontext 6: Beiträge zur Sechsten Arbeitstagung schwedischer Germanisten. Göteborg: Gothenburg University. Sayehli, S. (2001). Transfer and syntax? A study on the acquisition of German word order by Swedish native speakers. MA Dissertation, Lund University. Schwartz, B. D. & R. Sprouse (1994). Word order and nominative case in nonnative language acquisition: A longitudinal study of (L1 Turkish) German interlanguage. In T. Hoekstra & B. D. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 317– 368). Amsterdam: John Benjamins. Schwartz, B. D. & R. Sprouse (1996). L2 cognitive states and the Full Transfer/Full Access model. Second Language Research, 12, 40–72. Schwartz, B. D. & S. Vikner (1989). All verb second clauses are CPs. Working Papers in Scandinavian Syntax, 43, 27–49. Schwartz, B. D. & S. Vikner (1996). The verb always leaves IP in V2 clauses. In A. Belletti & L. Rizzi (Eds.), Parameters and Functional Heads: Essays in Comparative Syntax (pp. 11–62). Oxford: Oxford University Press. Slobin, D. I. (1973). Cognitive prerequisites for the development of grammar. In C. A. Ferguson & D. I. Slobin (Eds.), Studies of Child Language Development (pp. 175–208). New York, NY: Holt, Rinehart and Winston. von Stechow, A. & W. Sternefeld (1988). Bausteine syntaktischen Wissens: Ein Lehrbuch der generativen Grammatik. Opladen: Westdeutscher Verlag. Strömqvist, S., U. Richthoff, & A.-B. Andersson (1993). Strömqvist’s and Richthoff ’s corpora: A guide to longitudinal data from four Swedish children. Gothenburg Papers in Theoretical Linguistics, 66. Teleman, U. (1967). Bisatser i talad svenska. In G. Holm (Ed.), Svenskt talspråk (pp. 160– 203). Stockholm: Almqvist and Wiksell. Teleman, U., S. Hellberg, & E. Andersson (Eds.) (1999a). Svenska Akademiens grammatik. Vol. 2: Ord. Stockholm: NorstedtsOrdbok.
Nonnative acquisition of verb second
Teleman, U., S. Hellberg, & E. Andersson (Eds.) (1999b). Svenska Akademiens grammatik. Vol. 4: Satser och meningar. Stockholm: NorstedtsOrdbok. Thiersch, C. (1978). Topics in German syntax. PhD Dissertation, MIT. Tomaselli, A. & B. D. Schwartz (1990). Analysing the acquisition stages of negation in L2 German: Support for UG in adult German. Second Language Research, 6, 1–38. Travis, L. de Mena (1984). Parameters and effects of word order variation. PhD Dissertation, MIT. Travis, L. de Mena (1991). Parameters of phrase structure and verb-second phenomena. In R. Freidin (Ed.), Principles and Parameters in Comparative Grammar (pp. 339–364). Cambridge, MA: The MIT Press. Vainikka, A. & M. Young-Scholten (1994). Direct access to X´ theory: Evidence from Korean and Turkish adults learning German. In T. Hoekstra & B. D. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 265–316). Amsterdam: John Benjamins. Vainikka, A. & M. Young-Scholten (1996a). Gradual development of L2 phrase structure. Second Language Research, 12, 7–39. Vainikka, A. & M. Young-Scholten (1996b). The early stages in adult L2 syntax: Additional evidence from Romance speakers. Second Language Research, 12, 140–176. Vikner, S. (1991/1995). Verb movement and the licensing of NP-positions in the Germanic languages. PhD Dissertation, University of Geneva. 2nd edition: Institut für Linguistik, Universität Stuttgart. Published 1995, Oxford: Oxford University Press. Westman, M. (1974). Bruksprosa. En funktionell stilanalys med kvantitativ metod. Lund: Liber Läromedel/Gleerup. White, L. (1991). Second language competence versus second language performanc: UG or processing strategies? In L. Eubank (Ed.), Point – Counterpoint: Universal Grammar in the Second Language (pp. 167–189). Amsterdam: John Benjamins. Wijk-Andersson, E. (1991). Bara i fokus: En semantisk-syntaktisk studie av bara och dess ekvivalenter i nysvenskt skriftspråk. PhD Dissertation, Uppsala University. Williams, S. & B. Hammarberg (1998). Language switches in L3 production: Implications for a polyglot speaking model. Applied Linguistics, 19, 295–333. Zifonun, G., L. Hoffmann, & B. Strecker (1997). Grammatik der deutschen Sprache [Schriften des Instituts für deutsche Sprache Vol. 7 (2)]. Berlin: Walter de Gruyter. Zwart, J.-W. (1993). Dutch syntax: A minimalist approach. PhD Dissertation, University of Groningen.
Clause union and clausal position Josef Bayer, Tanja Schmid and Markus Bader University of Konstanz, Germany
It is argued that German infinitival complements with zu are less homogeneous than usually assumed. They can be CPs only in extraposed or otherwise displaced position. In immediately pre-verbal position they are construed as coherent whenever possible. Non-coherent construal is subject to essentially the same licensing conditions and as such incompatible with a CP-analysis. Pre-verbal non-coherent as well as coherent infinitives are in a checking relation which relies on the PF-visibility of the infinitive’s head. Interveners which break the adjacency between zu+V and the matrix verb destroy this checking relation. Methodologically, the problem of clause union and clausal position is approached with the help of three empirical studies as well as a theoretical study that compares German pre-verbal infinitives with pre-verbal finite clauses in Bangla (Indo-Aryan).
.
Introduction
Inspired by Evers (1975), the classical generative view of German control infinitives with zu has been since the eighties that (i) the infinitive is a CP which, in the context of a certain class of coherence-inducing verbs, can be pruned so that reanalysis can take place, and that (ii) the infinitival CP can stay in situ, i.e. to the immediate left of the matrix verb, or undergo extraposition. Thus, the string in (1) should in principle be ambiguous between (2a) and (2b). (1) daß Max mir das Lexikon zu kaufen empfohlen hat. that Max me the lexicon to buy recommended has ‘that Max has recommanded to me to buy the lexicon.’ (2) a. [. . . [CP . . . V2 ] V1 ] b. [. . . V2 V1 ]
bi-clausal mono-clausal
(2a) would essentially be equivalent to the extraposition option in (3).
Josef Bayer, Tanja Schmid and Markus Bader
(3) daß Max mir empfohlen hat [das Lexikon zu kaufen]. that Max me recommended has the lexicon to buy ‘that Max has recommanded to me to buy the lexicon.’
However, for Bech (1955/1957) (1) would have been only a coherent, i.e. mono-clausal construction. We shall see below that with respect to language use, (2b) is preferred over (2a), and (3) is preferred over (2b). In other words, why should (2a, b) exist, if (3) exists? And in particular, why should (2a) as an intraposed version of the preferred extraposed version in (3) exist, if it is the least preferred construction anyway? In this article we will provide evidence that in German the structure in (2a) does not arise as such. There are two reasons for this: (i) If restructuring is an option, this option must be taken. In this case, (1) will always be structured as in (2b). (ii) If restructuring is impossible, the clausal infinitive is either not positioned as in (2a), or it is not a CP. If [. . . zu+V] is a CP with a null-complementizer, it may be scrambled or topicalized to SpecCP. If it is a bare clause without a complementizer, it can arise to the immediate left of the verb. In that case it is subject to the same licensing conditions as V2 in the mono-clausal structure seen in (2b). We will approach the question of clause union and clausal position from both an empirical and a comparative-theoretical perspective. The article is structured as follows: In Section 2, Infinitival Complementation in Language Processing, we will present the results of an empirical investigation of intra- and extraposed coherent and non-coherent constructions including frequency in a corpus of written language, graded grammaticality judgments, and processing load during reading. Section 3, Syntactic and Comparative Investigations, describes the syntactic structure of intraposed coherent and non-coherent infinitives, and the licensing of pre-verbal infinitives in immediately pre-verbal (‘intraposed’) position. This part contains among other things information about complementation in Bangla, a South-Asian SOV-language which resembles German in allowing both pre- and post-verbal complements, and seems to shed new light on the apparently free variation between (2a) and (3). Section 4 contains Conclusions in which we will try to integrate the findings presented. An appendix on Re-positioned ‘zu’ contains material that gives additional support to the account presented in Section 3.
Clause union and clausal position
. Infinitival complementation in language processing In this section, we will summarize several results from an ongoing series of experimental investigations of infinitival complementation in German. In particular, we will cite results from a corpus study, a questionnaire study, and a self-paced reading study that have investigated intraposition and other clause union properties. For reasons of space, we will discuss these experiments only insofar as they address the question as to which structure is preferably assigned to intraposed infinitival complements in German. With regard to the Human Sentence Processing Mechanism (HSPM), we will make three assumptions.1 First, the HSPM is a serial device, which means that it always computes a single, fully specified structure. Second, the HSPM works in an incremental manner, as captured in the Left-to-Right Constraint of Frazier and Rayner (1988) (cf. (4)). Third, when faced with a syntactic ambiguity, the HSPM is guided by economy principles like the Simplicity principle of Gorrell (1995) (cf. (5). (4) Left-to-Right Constraint (Frazier & Rayner 1988) Each item is incorporated into a constituent structure representation of a sentence (essentially) as the item is encountered. (5) Simplicity (Gorrell 1995) No vacuous structure building.
What do these assumptions predict when applied to a sentence with an intraposed infinitival complement? First, when parsing such a sentence, each word will be integrated immediately into an ongoing syntactic structure. For sentences with an intraposed infinitive, there will always be a local or global ambiguity as to whether the sentence ultimately receives a mono-clausal or bi-clausal analysis. Given Simplicity, the HSPM should impose the simplest structural analysis that is possible, which, for the constructions at hand, clearly is the coherent, mono-clausal structure. Thus, when considering data from language comprehension, we should find evidence for a preference of coherent infinitivals over non-coherent infinitivals. . Extraposition and intraposition: A corpus study A corpus study was conducted in order to obtain quantitative information about the 56 control verbs used in a following questionnaire study (cf. 2.2 below). The corpus search was conducted on the Mannheim COSMAS corpus
Josef Bayer, Tanja Schmid and Markus Bader
consisting of more than 1,500 million word-forms of newspaper texts and literature (http://www.ids-mannheim.de). A main objective of the corpus search was to extract the frequency of preverbal, i.e. intraposed, as opposed to postverbal, i.e. extraposed, infinitival complements. For each verb, sixty randomly selected occurrences were analyzed. With respect to the position of infinitival complements, we found a strikingly low rate of intraposed constructions in the corpus. The main results are the following: 1. Less than 4% of infinitival complements occurred in intraposed position. 2. Less than 20% of all verbs in our sample occurred with an intraposed verbal element. 3. These were the following nine verbs: beabsichtigen (‘to intend’), versuchen (‘to try’), drohen (‘to threaten’), aufhören (‘to stop’), angeben (‘to indicate’), hoffen (‘to hope’), raten (‘to recommend’), anfangen (‘to start’), beginnen (‘to begin’). With respect to the structure of the intraposed infinitival complements, all occurrences in our sample were such that matrix verb and embedded verb were adjacent. That is, we did not find sentences like the one in (6) in which the embedded infinitive is separated from the matrix verb by an intervening element, which automatically forces a non-coherent construction. (6) . . . , daß sie das Buch zu lesen mehrfach versucht hat. that she the book to read repeatedly tried has ‘that she repeatedly has tried to read the book.’
To sum up, according to the corpus study, intraposition is clearly disfavored for infinitival complements: Intraposed infinitival complements occurred at a low rate and only with a few verbs. This shows that they – although not ungrammatical – represent a marked option. That embedded verb and matrix verb are never separated in the intraposed constructions of our sample hints at a coherent, i.e., mono-clausal structure in cases where an intraposed infinitive is used after all. . Coherent and non-coherent constructions: A questionnaire study A questionnaire study was carried out in order to investigate native speakers’ intuitions with respect to various (non-)coherence-sensitive tests as found in the literature (cf. Bech 1955/1957; Grewendorf 1988; Stechow & Sternefeld 1988, among others). Only control matrix verbs were used in the bulk of this
Clause union and clausal position
study. The main objectives were to investigate (i) for which verbs restructuring is an option at all, and (ii) whether the preverbal, i.e., intraposed construction patterns with coherent or non-coherent structures. The study and the results are documented in full detail in Schmid, Bader and Bayer (forthcoming). Here, we will focus on the structural status of intraposed constructions: Are they preferably analyzed as coherent, non-coherent, or are both structures equally available? Altogether, seven different constructions involving infinitival complements were tested. Four constructions required a clause union structure (coherent constructions), two constructions disallowed clause union (non-coherent constructions), and one construction was ambiguous with respect to clause union according to standard generative references on this topic (e.g., Grewendorf 1988; Stechow & Sternefeld 1988). One sample sentence from the questionnaire study is shown exemplarily for two coherent constructions, the two noncoherent constructions, and the ambiguous construction in (7) through (9). (7) Coherence Tests a. Long scrambling of a pronoun Was den Aufsatz betrifft, so ist klar, warum ihn Karla zu what the article concerns so is clear why it Karla to schreiben beschlossen hat. write decided has ‘As for the article, it is clear why Karla decided to write it.’ b. Wide scope of negation Karla hat keinen Aufsatz zu schreiben beschlossen und Karla has no article to write decided and Friederike auch nicht. Friederike also not ‘Karla hasn’t decided to write an article, and neither has Friederike.’ (8) Non-Coherence Tests a. Extraposition of the infinitival complement Karla hat beschlossen, den Aufsatz zu schreiben. Karla has decided the article to write ‘Karla decided to write the article.’ b. Narrow scope of negation Karla hat keinen Aufsatz zu schreiben beschlossen und Karla has no article to write decided and
Josef Bayer, Tanja Schmid and Markus Bader
Friederike auch. Friederike too ‘Karla decided not to write an article, and so did Friederike.’ (9) Ambiguity with respect to Coherence? – Intraposition Karla hat den Aufsatz zu schreiben beschlossen. Karla has the article to write decided ‘Karla has decided to write the article.’
Coherent constructions have in common that they are transparent for otherwise clause-bounded processes. In the questionnaire study four constructions were tested for this property in total, two of which are given in (7): (7a) illustrates ‘long’ scrambling of a pronoun. When scrambling of a pronoun from a position in the embedded infinitive over the matrix subject leads to a grammatical result, this suggests a coherent construction. No clause boundary is present between the embedded verb and the matrix verb. The same argumentation holds when a negated element in the embedded infinitive takes scope over the matrix verb, as must be the case in sentence (7b). The wide scope reading of the negation that inheres the negative indefinite keinen Aufsatz (‘no article’) is forced by a negated conjunct (and . . . also not) following the main predicate. The two constructions that only allow a non-coherent, bi-clausal structure, are shown in (8). (8a) illustrates the extraposition of the infinitival complement. When extraposition leads to a grammatical sentence, the extraposed phrase must have been a constituent of its own. When the infinitival construction is opaque for trans-clause processes, the construction in question is non-coherent. This is the case in (8b). In this sentence, the negated element inside the infinitival complement may not take scope over the matrix verb but only over the embedded verb. In the questionnaire, the narrow scope negation was forced by a positive conjunct (and . . . too).2 The last construction tested in the questionnaire study is intraposition of the embedded infinitive with no additional material intervening between the embedded and the matrix verb. This is illustrated by the example in (9). Sentences with intraposed infinitives were included in the questionnaire study to see how acceptable they are, and especially to see whether they are perceived as ambiguous between a coherent and a non-coherent construction as suggested by, e.g. Grewendorf (1988) and von Stechow and Sternefeld (1988). In total, 56 control verbs were tested as matrix verbs. We concentrated on control verbs as matrix verbs in this study because they appear both in coherent and non-coherent constructions, and because there is still a debate in the lit-
Clause union and clausal position
Table 1. Materials – subclasses of control verbs Control
Infinitive in Additional Examples function of objects
Subject Subject
Accusative Accusative
Ø Dative
Subject Accusative object Dative object
PP PP Accusative
Ø Accusative Dative
versuchen (‘to try’), beschließen (‘to decide’) drohen (‘to threaten’), versprechen (‘to promise’) aufhören (‘to stop’), klagen (‘to complain’) auffordern (‘to ask’), ermahnen (‘to urge’) erlauben (‘to allow’), verbieten (‘to forbid’)
erature about the range of control verbs allowing clause union. The 56 control verbs were classified according to three criteria: 1. The control property of the verb: Subject control versus object control. 2. The syntactic function of the embedded infinitive: The embedded infinitive replaced either a direct object or a prepositional object. 3. Additional objects: In addition to the infinitival complement, the verbs either had no further object, or an additional dative object, or an additional accusative object. The resulting five subclasses according to these criteria are shown in Table 1. For the questionnaire study, 70 test sentences (14 sentences per verb class), and 60 filler sentences were randomly ordered. 35 participants were asked to rate the acceptability of each sentence on a scale from 1 (“very plausible, easy to understand”) to 5 (“very complicated, difficult to understand”). The main results of this study are shown in Figure 1 (for more details cf. Schmid et al. forthcoming). The y-axis shows the rating-scale from 1 to 5 4 3 2 1 mean coherence
intraposed
subj. control, Inf=acc subj. control, Inf=PP acc. obj. control
extraposed
narrow scope
dat. obj. control subj. control, Inf=PP, +dat
Figure 1. Main results of questionnaire study on coherent and non-coherent infinitival constructions
Josef Bayer, Tanja Schmid and Markus Bader
5, and the x-axis shows the different constructions. Whereas the results for the ambiguously coherent intraposed construction and the two non-coherent constructions (extraposed and narrow scope) are shown separately, the results for the four coherent constructions have been collapsed and are shown under the label “mean coherence”. The mean rating that has been assigned to each of the four constructions is given separately for the five verb classes. The main results of the questionnaire study concerning the status of intraposed infinitival complements can be summarized as follows: 1. The coherent constructions show a rather different picture when compared to the two non-coherent constructions (extraposed and narrow scope): The rating of coherent constructions showed a clear verb-class effect which was absent for non-coherent constructions. 2. Sentences with intraposed infinitival complements – which are assumed to be ambiguous between a coherent and a non-coherent structure – patterned with the coherence and not with the non-coherence tests in showing clear verb-class effects. This is support for the hypothesis that intraposed infinitivals are treated by the HSPM as coherent constructions. 3. The verb class that was judged best for the coherent constructions was the subclass of subject-control verbs in which the infinitival complement replaces the direct object. Next came the subclass of object control verbs in which the controller is a dative object, and the infinitival also replaces the direct object. That these two verb classes fared best on the coherence tests fits well with the hypothesis that coherence is only allowed when the infinitival complement is a direct object (cf. Bayer & Kornfilt 1990; Haider 1993: 251). 4. For each single verb type, extraposition was rated better than intraposition (mean 1.3 vs. 2.6). 5. Intraposed sentences with narrow scope of negation were rated as being rather deviant (mean 3.9). Because these sentences require a non-coherent structure, this deviance would at least partially follow if the HSPM indeed has a preference for interpreting intraposed infinitives as coherent. While the sentence-final conjunct und . . . auch (‘and . . . also’) which enforces the narrow scope reading might have contributed to the enhanced complexity of these sentences, conjuncts of this sort pose no problems as such. The questionnaire also contained different types of control sentences, among which sentences like (10). These received rather good acceptability ratings.
Clause union and clausal position
(10) Karla hat einen Aufsatz schreiben müssen, und Werner auch. Karla had an article write must and Werner too ‘Karla had to write an article, and Werner had too.’
. Scope of negation: A reading time experiment To test whether the linguistic intuitions that were obtained in the rating study are reflected in online language processing of (non-)coherent infinitives, we conducted a self-paced reading experiment. The main objective of this experiment was to test the hypothesis that intraposed infinitives are preferentially assigned a coherent, mono-clausal structure during on-line language comprehension. As described above, the results on intraposed infinitivals obtained in the prior questionnaire study provide some initial evidence for this hypothesis, insofar as intraposed infinitivals clearly patterned with the unambiguous coherent constructions in being dependent on verb-class. To test this hypothesis of a coherence-preference for sentences with intraposition, subjects had to read sentences in which an intraposed infinitive contained a negative quantifier such as keines von den Büchern (‘none of the books’). The test sentences always ended with the negated conjunct und . . . auch nicht (‘and . . . also not’): (11) Der Opa hat keines von den Büchern zu lesen versucht und the grandpa has none of the books to read tried and der Onkel auch nicht. the uncle also not ‘Grandpa didn’t try to read any of the books, and the uncle didn’t either.’
A negated conjunct is only compatible with the wide scope reading of negation, i.e., with a reading like it is not the case that grandpa tried to read any of the books. This reading in turn is only possible when the embedded infinitival complement is constructed coherently. 30 sentences of the type shown in (11) were constructed, with each sentence appearing in 6 conditions according to six verb classes. Five of these verb classes were the verb classes of control verbs already used in the questionnaire study. The sixth verb class were modal verbs. These are always coherent. The experimental method was the self-paced reading method with word-by-word non-cumulative presentation. This method works as follows. Participants are seated in front of a computer screen. They read sentences by pressing a response
Josef Bayer, Tanja Schmid and Markus Bader
button for each word. Upon each button press, the next word appears, and the last word disappears. Thus, always only a single word is visible. What do our hypotheses about the HSPM predict for reading such sentences? As already explained above, due to the Left-to-Right Constraint and Simplicity (cf. (4) and (5)), the HSPM will compute a coherent structure for a sentence with an intraposed infinitive. This means that the clause-final matrix verb (versucht in (11)) will be attached into a mono-clausal structure, irrespective of its lexical specification. As soon as it has been attached, the HSPM will engage in processes of checking and linking (cf. Mitchell 1994) with regard to the verb’s lexical properties. As a by-product, the HSPM will determine whether the verb is compatible with a coherent structure or not. In case the verb cannot engage in the formation of a coherent construction, some kind of reanalysis is called for, which should be reflected in enhanced reading times on the matrix verb. No such reanalysis is necessary for verbs which allow for the coherent construction. Thus, when taking the mean coherence rating that was obtained in the questionnaire study described in the last section as a measure of how well a verb is compatible with a mono-clausal structure, we expect that the better a verb was judged as coherent in the questionnaire study, the faster it should be read in the reading experiment. In sum, there should be a positive correlation between mean-coherence score in the questionnaire and reading times on the matrix verb. What should we expect if the HSPM does not prefer to compute a coherent structure, but instead has a preference for the non-coherent structure? Such a preference might result, for example, if the HSPM works according to a head-driven schema, that is, if the parser can insert a phrase into the ongoing syntactic representation only after its head has been encountered in the input. Such a parser, which would not work strictly incrementally, might first come up with a bi-clausal structure. When processing the matrix verb, there are then two possibilities. Either the parser simply sticks to the bi-clausal representation, or it reanalyzes it in case the verb has a strong tendency for coherence. Thus, we would expect either no correlation between reading times on the matrix verb and mean-coherence scores, or even a negative correlation because reading times should be slower for more coherent verbs which, under the head-driven schema, would induce reanalysis as an extra step. Turning to the clause-final negation, recall that this negation is only compatible with the wide scope construal of the negative quantifier in the embedded infinitival (keines von den Büchern (‘none of the books’ in (11)). This wide scope construal in turn is only compatible with a coherent construction. Whenever a matrix verb that is not allowed in a coherent construction induces
Clause union and clausal position
Table 2. Correlations between reading times and questionnaire results (starred correlations are significant)
clause union compatibility
infinite verb
matrix verb
spill-over region
auch ‘also’
nicht ‘not’
–0,035
0,479*
0,410*
0,292
0,231
reanalysis toward a non-coherent structure, reading times should increase on the clause-final negation. No such effect is expected when the construction stays coherent throughout the parsing process. Reading times of special interest were the reading times on the matrix verb, the “spill over” region (und . . . ), and the last two words (auch nicht). Since our main interest was in the relationship between reading times and coherence properties of verbs, we show in Table 2 the correlations between reading times and mean coherence scores as obtained in the questionnaire study presented in the preceding section. Reading times on the matrix verb as well as on the immediately following spill-over region showed a significant positive correlation with the mean coherence score: The better a verb scored on the coherence tests the faster it was read. This is exactly what was predicted under the hypothesis that a coherent structure is computed for these sentences. If a coherent structure is computed while reading the sentences, the matrix verb will fit into the ongoing structure according to its coherence properties. A coherence-inducing verb fits in well, allowing the sentence processor to proceed with the next words quickly. A verb disallowing coherence, in contrast, does not fit in well, and thus reading times will be slowed down. That this effect was not only visible on the matrix verb but also in the spill-over region is not unusual for self-paced reading experiments and probably reflects just some kind of inertia in button-pressing. If readers switched to a non-coherent structure when encountering a verb not allowing coherence, we would expect a correlation with mean coherence also on the final negation nicht because the negative conjunct acts as a trigger of coherence. Contrary to expectation, reading times on the two final words (auch nicht) requiring wide scope did not correlate significantly with the mean coherence scores obtained in the questionnaire study. This suggest that readers do not reanalyze the initial coherent structure toward a non-coherent structure on encountering a verb which resists clause union. That is, the mismatch between structural analysis and lexical requirements that occurs if a verb is not a clause union verb does not seem to be strong enough to trigger any kind of structural revision.
Josef Bayer, Tanja Schmid and Markus Bader
One possible alternative to assuming a general parsing preference for coherence might be that there is a semantic preference for the wide-scope reading (corresponding to the mono-clausal structure) as compared to the narrowscope reading (bi-clausal structure). However, in a further selfpaced-reading study which investigated sentences like (11) but without any negative element (e.g. Der Opa hat ein Buch zu lesen versucht, und der Onkel auch), a similar result was found, that is, a positive correlation between reading times and mean-coherence scores. Thus, while it is certainly true that in many instances a narrow-scope reading is less expected than a wide-scope reading (cf. Note 2), these further results show that a semantic bias alone cannot explain the whole pattern of results described above. On the other hand, this bias might be a further reason why we did not find a correlation at the clause-final negation. It might simply be that the syntactic parsing preference for coherence and the semantic bias towards wide scope were strong enough for subjects to hold on to the structure computed from the outset, even if this structure was not fully compatible with certain requirements of the verb. . Summary of empirical investigations All three of the empirical studies summarized above lend support to the hypothesis that intraposed infinitives are preferentially assigned a coherent, mono-clausal structure. Whereas extraposed infinitival complements are always clause-like and involve therefore a bi-clausal analysis, this option is strongly marked for intraposed infinitival complements being preferably analyzed as non-clause-like by the parser. In the next section we will turn to syntactic observations which pertain to the question of bi- versus mono-clausality.
. Syntactic and comparative investigations Quantitative and behavioral data as shown in the previous section are important contributions to linguistic theorizing. The strong preference in parsing pre-verbal infinitival complements as coherent constructions may invite the conclusion that Bech was right after all in not considering any clausal status for them. Syntactic structures such as (2a) would then simply not exist, and the quantitative and behavioral data would map onto the admissible structural assignments of competence grammar more or less directly. In the present section we will take a closer look at infinitival complementation in order to avoid premature conclusions. In doing so, we will also draw on data from Bangla (also
Clause union and clausal position
known as Bengali), an Indo-Aryan South Asian language in which finite complements may appear both in pre- and in post-verbal position. Our focus will be on the formal properties of complements as expressed by complementizers or by the infinitival marker zu in German. . The distribution of daß-CPs in German It is a well-known though still badly understood fact that German (as well as Dutch) resists the appearance of finite CPs in direct object (DO) position. (12) a.
Hans hätte vermutlich bezweifelt [daß Maria kommt]. Hans had-subj presumably doubted that Maria comes ‘Hans would have presumably doubted that Maria will show up.’ b. ?*Hans hätte vermutlich [daß Maria kommt] bezweifelt. c. Hans hätte [daß Maria kommt]1 vermutlich t1 bezweifelt. d. [Daß Maria kommt]1 hätte Hans vermutlich t1 bezweifelt.
Within GB-theory, Stowell (1981: 146) proposed a principle of Case Resistance (CRP) according to which “Case may not be assigned to a category bearing a Case-assigning feature”. Finite clauses such as those in (12) (and also toinfinitives) have a +Tense feature which is a Case-assigning feature. So CPs would be banned from the DO-position (as well as from the subject-position, P-object position etc.).3 (12a) would then be the result of extraposition as a strategy to evade this principle, similarly the topicalization seen in (12d). The scrambling operation in (12c) would be seen as adjunction outside the site of Case assignment. In more recent theorizing the question is more whether the CP would ever be merged as in (12b) or raised into the pre-verbal position. If no merging in or movement into this position is possible, structures like (12b) would never arise. . Distribution of infinitives in German and adjacency Infinitives are even more challenging because they do appear in pre-verbal position. If Stowell (1981) was right in subsuming to-/zu-infinitives under the categories blocked by Case Resistance they should not be allowed there. German zu-infinitives would be subject to the same fate as finite CPs. Given that there is no string-wise discrimination between coherence and non-coherence for verbs which allow both construals, it seems difficult to approach the question. Nevertheless, there is an interesting test that can shed light on the issue. This test rests on the fact that German (like many other SOV-languages) shows
Josef Bayer, Tanja Schmid and Markus Bader
the phenomenon of rightward movement (or stranding, depending on one’s analysis) of relative clauses or PPs of different phonological sizes. Instead of Ich habe mich [dafür] entschieden (‘I have decided on it’) one can also say Ich habe mich entschieden [dafür].4 As Haider (1993; 2003), van Riemsdijk (1994) and Bayer (1995) observe, such movement / stranding is not allowed when the infinitive occupies the left-adjacent position of the matrix verb.5 (13) Coherent construction: versuchen pre-V a. Ich habe [mich dafür zu entscheiden] versucht. I have ref there-for to decide tried ‘I have tried to decide on it.’ b. *Ich habe [[mich dafür zu entscheiden] dafür] versucht. pre-V+ex c. Ich habe versucht [[mich dafür zu entscheiden] dafür]. extraposed+ex d. Ich habe [[mich dafür zu entscheiden] dafür]1 schon I have ref there-for to decide there-for already scramb+ex mehrmals erfolglos t1 versucht. tried more-than-once successlessly ‘I have already more than once without success tried to decide on it.’
Extraposition leads to ungrammaticality only if the infinitive is in immediately pre-verbal position. Is this so because coherence is mandatory for versuchen as suggested by Haider (2003)? The intervention of dafür would block the process of head-to-head movement or base-generation of the V-cluster. Indeed, material can extrapose in such cases but only to the very end of the clause, i.e. to the right of versucht: Ich habe mich dafür zu entscheiden versucht dafür. The fact that this is impossible with verbs that disallow coherence is one of the classical tests for clause union in cases such as versuchen. Notice that the non-coherent verb auffordern (‘to ask’) does not allow this: *Ich habe ihn [sich dafür zu entscheiden] aufgefordert dafür (‘I asked him to decide on it’). The less widely known surprise is that even in the context of a non-coherence verb extraposition from the infinitival complement is blocked as long as the clause is in V-adjacent position: (14) Non-coherent construction: auffordern a. Ich habe ihn [sich dafür zu entscheiden] aufgefordert. pre-V I have him ref there-for to decide asked ‘I have asked him to decide on it’ b. *Ich habe ihn [[sich dafür zu entscheiden] dafür] aufgefordert. pre-V+ex c. Ich habe ihn aufgefordert [[sich dafür zu entscheiden] dafür]. extraposed+ex
Clause union and clausal position
d. Ich habe ihn [[sich dafür zu entscheiden] dafür]1 schon mehrmals t1 aufgefordert. scramb+ex ‘I have already more than once asked him to decide on it.’
What is so special about the left V-adjacent position that any sort of manipulation of the right edge of the infinitival complement will lead to a strong effect of ungrammaticality? The parser’s strong preference for coherent constructions which has been discussed in Section 2 could indicate that non-coherent intraposed infinitives such as (14a) are simply not admitted by the competence grammar. Given an input like (14a), the parser would then try to cope with it ‘as if it were’ a coherent construction. According to this reasoning, the deviance of (14b) could still be explained along the lines of obligatory V-cluster formation as predicted by head movement or base-generation of a coherent construction. Thus, the adjacency requirement either holds independently of coherence, or this particular positioning of the infinitive triggers coherence no matter what, i.e. even in the presence of a verb like auffordern which, according to general wisdom, does not trigger coherence.6 In order to approach this alternative we will next take a look at an SOV-language which shows no overt signs of clause union. . An answer from Bangla Bangla is an Eastern Indo-Aryan SOV-language with a mixed system of preand post-verbal finite complements. Like various other head-final languages, Bangla also shows rightward movement of backgrounded light elements that we have used as a test of adjacency or adjacency disruption in 3.2. Unlike German, Bangla allows also personal pronouns to the right of the verb.7 Why this is so is so far unclear, but does not matter here because we are simply using the construction as a test. (15b) shows that rightward pronoun movement is fine in post-verbal complements. (16b) shows that, just as in German, such movement leads to ungrammaticality in immediately pre-verbal position.8 The adjacency requirement is also relaxed when the complement has undergone leftward movement as seen in (16c). Between the topicalized complement and the main clause there is an intonation break. (15) Post-verbal complement a. ami wune -ch ilam [paw-er baqi-r kukur-za toma-ke I heard-have next-poss house-poss dog-cl you-obj
Josef Bayer, Tanja Schmid and Markus Bader
kamre-ch e]. bitten-has ‘I heard the next door’s dog has bitten you.’ b. ami wune-ch ilam [[paw-er baqi-r kukur-za toma-ke kamre-ch e] toma-ke]. (16) Pre-verbal complement a. ami [paw-er baqi-r kukur-za toma-ke kamre-ch e] I next-poss house-poss dog-cl you-obj bitten-has wune-ch ilam. heard-have ‘I heard the next door’s dog has bitten you.’ b. *ami [[paw-er baqi-r kukur-za toma-ke kamre-ch e] toma-ke]] wunech ilam. c. [[paw-er baqi-r kukur-za toma-ke kamre-ch e] toma-ke]]1 ami t1 wunech ilam.
How should (16b) be ruled out in Bangla? Coherence cannot be responsible for the following reasons: The verb of the dependent clause is finite and would have to move out of a finite clause in order to trigger clause union by virtue of verb raising. But even independently of finiteness, Bangla shows no signs of verb movement (head movement) or rearrangement of verb order as in West-Germanic. Although the strictly head-final nature of the language creates strings of adjacent verbs, there are no independent signs of a verbal complex. For example, none of the coherence tests that play a role in the syntax of German would yield a result in favor of coherence. There is also no indication that the two finite verbs in (16a), kamrech e and wunech ilam would form a prosodic constituent. In fact, there appears to be sustained a high tone on [ch e]σ of the first and a low tone on [ne]σ of the second. If the similarity between Bangla and German with respect to clausal licensing is not accidental, we may conclude from this that the restriction seen in (14b) of Section 3.2 is not the result of forced clause union but rather the result of a more general constraint that monitors the shape of a clausal complement in immediately pre-verbal position.9 This leads us to a question about the nature of the pre- and post-verbal complements which appear in Bangla finite complementation as well as in German infinitival complementation. Notice that the complements considered so far offer little to distinguish them. It looks as if the same type of complement can be placed either to the right or to the left of the matrix verb. From a theoretical point of view, this would be a disturbing situation because, given the same numeration, it would be unclear why the complement can in one case stay in post-verbal position whereas it
Clause union and clausal position
has to raise into pre-verbal position on other occasions. The issue can be addressed empirically because next to the bare finite clauses shown in (15) and (16), Bangla also has clauses with overt complementizers. Post-verbal clauses show the clause-initial complementizer je which is homophonous with the relative clause particle je (/3e/ ‘which’). Je-clauses can be linked to a pronominal or NP-like correlate in the matrix clause as seen in (17a). They are strictly impossible in pre-verbal position as shown in (17b) and (17c).10 (17) je-clauses a. ch ele-za (e kfzh a) wune-ch e [3e or baba aw-be]. boy-cl this news heard-has that his father come-will ‘The boy heard (it) that his father will come’. b. *ch ele-za [3e or baba aw-be] wune-ch e. boy-cl that his father come-will heard-has c. *[3e or baba aw-be] ch ele-za wune-ch e that his father come-will boy-cl heard-has Bayer (1996: 255)
Pre-verbal complements do exist, but they never have an initial complementizer. They rather show the post-verbal complementizer-like element bole; bole is a verbal form which consists of the stem bol (‘to say’) and the perfective participle morpheme -e. Bole is a quotative suffix as found in the Dravidian languages but also elsewhere in genetically and geographically unrelated languages. (18a) shows a bole-clause in immediately left-peripheral position of the matrix verb. (18b) shows that such a clause can move to the left, but cannot be related to a correlate. The question marks in (18c) should indicate that a bole-clause in post-verbal position is not fully normal but may occur. (18) bole-clauses a. ch ele-za [[or baba aw-be] bfl-e] wune-ch e boy-cl his father come-fut say-prt heard-has ‘The boy heard that his father will come’ b. [[or baba aw-be] bfl-e] ch ele-za (*e kfzh a) wune-ch e his father come-fut say-prt boy-cl this news heard-has ‘[That his father will come] the boy has heard’ c. ??ch ele-za wune-ch e [[or baba aw-be] bfl-e] boy-cl heard-has his father come-will say-prt Bayer (1996: 255)
Abstracting away from the somewhat unclear status of (18c), these clauses are in complementary distribution as summarized in Table 3.11
Josef Bayer, Tanja Schmid and Markus Bader
Table 3. Distribution of je- and bole-clauses
[ . . . V bole] [ je . . . V]
in pre-verbal position
in post-verbal position
– *
*/??/marked –
Although je and bole seem to both function as complementizers or subordinating morphemes, they have very different properties which go beyond lexical and etymological differences and also beyond the positional differences summarized in Table 3. Je-clauses allow extraposition in the same way as the C-less example in (15b), whereas bole-clauses insist on strict adjacency between the finite verb, the complementizer bole and the matrix verb according to which material from . . . cannot intervene in the positions indicated by in the configuration . . .Vfin ] * bole] * Vmatrix ]. The distribution of complements seen in Bangla conforms to a widely attested pattern of word order typology.12 Assuming that complements without an overt complementizer do not deviate from this pattern, Bayer (1996: 266f.) suggested that post-verbal complements may have a clause-initial zero complementizer, while pre-verbal complements may have a clause-final zero complementizer or simply the finite verb functioning as the complement’s head in place of bole. The adjacency effect observed in (16b) but not in (15b) can now be tied to the formal nature of the complement: Post-verbal complements are CPs which are licensed by an initial complementizer. The complement of C must be a finite clause, but there is no constraint on its right edge. Pre-verbal complements have no initial complementizer. They rely either on the verbal element bole or on the finite verb itself. Both appear invariably in final position. Material that moves to the right of this position seems to obliterate the complement’s formal features and turn it into an unusable structure. As a result of this digression to Bangla we can maintain that the adjacency effect cannot universally be reduced to restructuring by head-to-head movement (clause union). First of all, Bangla does not show the typical signs of restructuring; secondly, even if it would, the two relevant verbs in (16b) – kamrech e and wunech ilam – could not engage in such a process because they are both finite and clearly serve as heads of distinct syntactic domains. If we are allowed to use this result as a suggestion for the analysis of German, the no-intervention effect observed in non-coherent constructions such as (14b) should not prematurely be reduced to the parser’s preference for coherent constructions and resulting ignorance of non-coherent construal. With this result
Clause union and clausal position
Table 4. Status (Bech 1955/1957: Ch. 1) supine 1. status 2. status 3. status
-en zu -en ge- -t
lieb-en zu lieb-en ge-lieb-t
in mind, we will next explore the structural nature of pre- and post-verbal infinitival complements in German. . Status checking Bech (1955/1957) has suggested a formal theory of German infinitive constructions in which he draws a parallel between verb-form licensing and Case licensing. Infinitival complements are governed in analogy to nominal complements. Bech calls the relevant verb form Status. Table 4 shows the three Statuses for the verb lieben (‘to love’) that Bech distinguished for the verb-governed form, in his system the ‘supine’. A verb which merges with an infinitive must check its Status just like a verb that merges with an NP must check its Case. Modals, the future auxiliary werden and certain light verbs check 1st Status (-en), control and raising verbs check 2nd Status (zu . . . -en), the auxiliaries haben, sein and the passive auxiliary werden check 3rd Status (ge-. . . -t). Since verbs in 1st and 3rd Status are always coherent, we do not worry about them. The interesting cases are those which appear in 2nd Status. Here we find both coherent and noncoherent constructions. Thus, although Bech considered the string . . . V2 V1 ] as generally indicative of a coherent construction, a bi-clausal analysis cannot be excluded. The fact that Bangla shows precisely the same linear constellation and the same adjacency constraint but lacks clause union should be seen as support for the availability of a bi-clausal analysis. But then the important question arises how the Status of the complement clause can be checked. According to a widespread view, zu-infinitival clauses are CPs with a zero complementizer.13 While this makes sense for post-verbal infinitives and conforms to the typological pattern mentioned in Section 3.3, it is less clear that it does for pre-verbal (non-coherent) infinitives. In German one cannot find overt complementizers on argument CPs which would provide the right kind of evidence. However, Dutch has an optional complementizer om after the predicates proberen (‘to try’) and weigeren (‘to refuse’). And there, an om-complement is only possi-
Josef Bayer, Tanja Schmid and Markus Bader
ble in post-verbal (or perhaps in scrambled) position, but not in immediately pre-verbal position:14 (19) a.
Jan heeft geprobeerd [om weg te lopen]. Jan has tried C away to run ‘Jan tried to run away.’ b. *Jan heeft [om weg te lopen] geprobeerd.
Let us therefore consider the idea that in pre-verbal zu-infinitives with clausal status there is no complementizer. What would be the head of such a clause? There seems to be only one reasonable possibility: zu. Notice, however, that there have been convincing arguments against zu as a VP-final functional head that triggers rightward movement of the verb (cf. Haider 1993 and Vikner 2001). Following essentially Haider’s theory, we assume here that functional heads have distinct properties as a function of their positioning: Initial functional heads such as v, T, C etc. attract the verb and normally project a specifier with an EPP-feature. It is far from obvious that final functional heads behave alike. If they are treated on a par with initial functional heads, one has to accept the full consequences of a Kayne-style analysis. Alternatively, they could enter the derivation by virtue of morphological composition alone. In that case all there is is morphological merger between a functional element F and a lexical element X. When X projects in syntax, F will be carried along as a feature on X.15 F is not responsible for displacement; it does not give rise to a functionally motivated specifier etc. It is available for checking anywhere in the domain of X.16 Let us adopt such a system for the case at hand and assume that zu is a prefix which endows the verb with 2nd Status, and that the verb thus enriched projects to maximality. The Status of the clause will be straightforwardly determined by the head which is lexically V and functionally zu, i.e. <2nd status>.17 Instead of attributing the determination of Status to an empty C-head we can now say that Status is signaled directly at the right edge of the complement. Thus, the right edge of a clausal zu-infinitive may be all there is of functional information. Assume now that extraposed or stranded material somehow obliterates the right edge of the clause.18 If the right edge is the only formal signature of the clause this will lead to a licensing problem: the Status feature of the clause ceases to be visible for the selecting verb. This enables us to explain why pre-verbal clauses must be head-final in the strictest sense. Scrambled and extraposed/post-verbal clauses lack this constraint. In German they could be headed by a zero complementizer which is in an agreement relation with the feature <2nd status> provided by the final zu+V. There is much evidence that displaced material (even if it undergoes reconstruction) does not
Clause union and clausal position
need to echo the exact formal structure of its in-situ counterpart.19 The distribution of finite daß-CPs which has been considered in Section 3.4 shows exactly this. Thus, although daß-clauses are generally ungrammatical in immediately pre-verbal position they are grammatical in scrambled or post-verbal position. If we are on the right track, there are (i) special licensing requirements for complements which appear in left-adjacent position of the selecting verb, and (ii) these requirements hold independently of clause union, i.e. zu can be simply a prefix to a verb (in which case we have a mono-clausal construction) or prefix to a verb that heads a clause (in which case we have a bi-clausal construction). (iii) The adjacency requirement indicates that an agreement relation is at stake which is checked at PF, i.e. essentially a relation of concord as Juan Uriagereka (p.c.) suggested. This is compatible with a movement analysis in the sense of Kayne (1994) according to which the zu-infinitive raises to a checking position associated with the matrix verb. It is, however, equally compatible with a base-generation analysis according to which the zu-infinitive is directly merged into the domain in which the Status feature can be checked.20 The bi-clausal and the mono-clausal structures are shown in (20) and (21) respectively. The Status feature should indicate that the control verb (Vcontr ) searches for a complement that checks off this feature: (20) (=14a) bi-clausal / non-coherent Ü
Ü Ü
Josef Bayer, Tanja Schmid and Markus Bader
(21) (=13a) mono-clausal / coherent Ü Ü Ü
(20) roughly corresponds to a CP-analysis, and (21) corresponds to a VPanalysis with remnant movement in Wurmbrand (2003). While we will abstain from both a CP-analysis and remnant movement, the basic distinction remains the same.21 (21) presupposes the formation of verbal complexes without running through phrasal complementation and subsequent pruning operations. If computational steps in the grammar map onto steps of the parser, it would be wrong to assume that the unionized structure in (21) is derived from the bi-clausal structure in (20). The corpus, judgment and behavioral data we have seen in Section 2 convey consistently that the mono-clausal analysis is simpler and more readily available. It does not follow from this, however, that the less available parse in terms of a clausal analysis would be prevented by the competence grammar. To sum up, three things have emerged from the preceding considerations: (i) Pre-verbal clausal infinitives are in all likelihood complementizer-less. (ii) They are licensed via a checking process which relies on their Status feature, in our case , i.e. Bech’s 2nd status. (iii) The Status feature must be visible to the head with which the infinitive is merged; extraposed material destroys the required visibility. In the next section, we will try to make this last point more precise. . Status checking and agreement In the previous section it was assumed that material which has been extraposed from the infinitival clause or has otherwise been stranded in post-verbal position obliterates the Status feature and therefore prevents feature checking. The situation is familiar from the syntax of pre-nominal modification and has been
Clause union and clausal position
accounted for in earlier work with a filter called the Head Final Filter (HFF). The formulation of the HFF in (22) follows Escribano (2004: 5) (22) HFF: Base-generated pre-modifiers must be head-final
In Dutch and German, the pre-nominal inflected adjective must not be separated from the nominal by any non-agreeing interveners. The Dutch examples in (22) are from van Riemsdijk (1994). Witness that null-inflection is only possible in indefinite neuter singulars. Others show the schwa-ending -e. (23) a.
een snel genoeg vliegtuig a fast enough plane b. *een snel genoeg auto a fast enough car c. *een snell-e genoeg auto d. % een snel genoeg-e auto
(neuter) (non-neuter)
(23d) shows a frequently chosen repair strategy by which the left-adjacent inflection is retained at the cost of inflecting the wrong constituent, a deviance which is indicated by %. Genoeg is an adverb and would normally not be inflected. If it follows the actual adjective (snel) it seems to be treated like an adjective. German retains more morphology than Dutch. The following case of a masculine indefinite shows the same effects: (24) a.
ein genug groß-er Teller a enough big-agr plate b. *ein groß-er genug Teller c. % ein groß genug-er Teller
Although what van Riemsdijk refers to as a ‘frequent mistake’ violates a norm, the examples (23d) and (24c) are clearly distinct from the seriously ungrammatical examples in which the inflected adjective is separated from the agreeing noun by some non-agreeing intervener. The repair seen in (23d) and (24c) is an interesting proof of the obligatoriness of the adjacency relation between A+agr and N. In English, the adjectival inflection has been lost. As a result, N may be pre-modified by non-adjectival categories, and A and N may in principle be separated by non-adjectival interveners. (25) a. b. c. d.
the only book (that John has ever read) a big enough plate a tongue-in-cheek remark22 a wrong in my opinion view23
Josef Bayer, Tanja Schmid and Markus Bader
Agreement has to be checked in (23) and (24) by virtue of the categories D, A and N of which D and A show overt agreement morphology in modern German. We assume that uninterpretable (and interpretable but redundant) features have to be removed in core syntax, and that this can be done by the agree relation suggested in Chomsky (2001). In (24a) the AP agrees with N by virtue of the head A which bears the features agreeing with N, and D agrees with [AP N] by virtue of the features it encodes. Assume now that due to a PF-operation the head A is affected. How this can happen is not totally clear. Let us for concreteness follow the proposal of Truckenbrodt (1995) and assume that an extraposable element (in German usually an adverb or a PP) attaches to the right of its minimal containing XP for a prosodic reason, e.g.24 (26) a.
(Die Leute)φ (waren drauf stolz)φ the people were there-of proud b. (Die Leute)φ (waren drauf (stolz)ω (drauf )ω )φ
⇒ extraposition ⇒
A similar process applying to (24a) would lead to the ill-formed output in (24b). The PF-relevant agreement morphology of A (-er) would be inaccessible because [. . .großer] has now ceased to exist in favor of [. . . großer genug]. At this point we can return to the adjacency problem of intraposed infinitival clauses. Considering the most challenging case, namely the transition from (14a) to (14b), one can take extraposition as a prosodic realignment of the phonological word dafür. The relevant transition is as in (27). (27) a. . . . dafür zu entscheiden)φ b. . . . dafür (zu entscheiden)ω (dafür)ω )φ
⇒ extraposition ⇒
Although the verb form marked with the prefix zu remains the head of the infinitival clause, this head has now become inaccessible for PF. The string (zu entscheiden dafür) is not a proper exponent of the formal features of the PF-representation that is required to check off the feature of the verb auffordern. The uninterpretable feature will stay in the derivation and lead to a crash at the LF-interface. Why is it that extrapositions of this sort do not harm displaced (i.e. extraposed or scrambled) clauses? Our conjecture was that extraposed infinitives are headed by a null complementizer. This complementizer must be in a selection relation with the infinitival V-projection, but selection being distinct from agreement, we do not expect adjacency/head-final effects. Thus, one might want to correlate the absence of adjacency/head-final effects with selection of a CP.
Clause union and clausal position
Involvement of a CP-shell would also be predicted for infinitives in scrambled or topicalized positions. While this might be appropriate for German, we have to acknowledge that it is not appropriate for Bangla. Recall that in this language C-initial complements are strictly confined to the post-verbal position. Pre-verbal complements are either bare clauses or head-final bole-clauses. So at least for Bangla one may ask why scrambled infinitives should be exempted from PF-checking? One hypothesis could be that the infinitival complement is merged in or raised into the checking domain of the matrix verb, at which point it would be vulnerable by the process referred to in (26). Under this hypothesis the question is how these restrictions could be suspended after scrambling. One solution to this problem provides additional support for the PF-nature of rightward movement that has been argued for above. This solution says that PF-based processes of rightward movement are confined to the concrete occurrence in the flow of speech and are not relevant for the computation of filler-gap dependencies. Thus, the fact that some prosodic realignment in constituent X has taken place at a certain point in the derivation and has been spelled out at this point, cannot be held responsible for the representation that counts for the interpretation of X in its underlying (trace) position.25 We have now arrived at a sub-theory of infinitival licensing which can account for the non-intervention effect without abandoning the possibility that intraposed infinitives may retain their clausal status. The trees in (20) and (21) suggest that Status-checking is independent of clause union. Infinitives which are merged at the immediate left side of the verb (or have been raised into this position) and are spelled out there are subject to PF-checking at this place. As we could show, this holds for coherent infinitives (cf. 21) in the same way as for non-coherent ones (cf. 20). If we are right, the licensing of pre-verbal finite complements in Bangla should follow the same mechanism. Extraposed material which is prosodically aligned at the right edge of the Status checker will invariably destroy the constituent’s Status checking ability. We assume that phonological phrasing respects the clause boundary of non-coherent constructions. Thus, extraposition cannot skip over a potential prosodic attachment site.26 To sum up at this point, we have made a proposal as to why extraposition from an infinitival clause may render it useless as a Status-checker, and why such a process would only hold for clauses in immediately pre-verbal position.
Josef Bayer, Tanja Schmid and Markus Bader
. Conclusions We have started the present investigation with a question about intraposed (pre-verbal) infinitives in German: Does the grammar license infinitives with CP-status (cf. (2a)) in this position simultaneously with the mono-clausal construction by which matrix verb and dependent infinitive form a V-cluster (cf. (2b))? (2) a. [. . . [CP . . . V2 ] V1 ] b. [. . . V2 V1 ]
bi-clausal mono-clausal
This question was approached from two sides, 1. an empirical investigation which involved a corpus study, a questionnaire study and a processing experiment; 2. a syntactic investigation that concentrated on the positioning of infinitival complements and their formal properties. This part involved a comparative part in which we considered complementation in a more distant non-Germanic V-final language. The empirical studies referred to in 1. have consistently shown that post-verbal complements are the preferred option, and that sentences with intraposed infinitives are preferentially analyzed as coherent, mono-clausal constructions. Why should this be so? If UG allows clausal complementations as in (2a), why should they be dispreferred by the parser? The answer follows from properties of the HSPM. The study of head-final languages and constructions has revealed that the HSPM builds structure also in the absence of the verbal head, and that it does so under minimal assumptions about the input (cf. (4) and (5)).27 Considering an incomplete parse of (1) (daß Max mir nur das Lexikon zu kaufen empfohlen hat) as in (28), (28) daß Max mir nur das Lexikon . . . that Max me only the lexicon . . .
the HSPM has assigned a structure according to which Max is [DP-nom], mir is [DP-dat], and nur das Lexikon is [DP-acc]. Under the most parsimonious analysis, these three constituents are part of the same clause whose lexical head is a di-transitive verb.28 There is no function word which would signal to the HSPM a new clause boundary before the on-going clausal analysis is completed. Given that the grammar allows coherent constructions, there is no reason for the parser to first build a CP which is pruned down to a coherent construction once the matrix verb has been received and has been identified
Clause union and clausal position
as a coherence-inducing verb. The expectation is rather that there is only one parse, and that this parse is consistent with a mono-clausal, coherent construction. Since back-tracking and reanalysis are highly unfavorable options, it can be explained why the listener or reader would try to retain a coherent construction as long as possible. The fact that such an analysis tends to be retained even in the face of verbs which resist a coherent analysis is consistent with this explanation. One conclusion could then be that intraposed infinitives as in (2a) are not available for purely syntactic reasons, and that this is the reason why they would be forced into a mono-clausal analysis during parsing. Our syntactic investigation referred to under 2. has shown that this conclusion would be premature. To demonstrate this we concentrated on an adjacency (non-intervention) effect that has been observed in German intraposed infinitives, and can be described in the style of a filter as in (29). (29) *[. . . [[ . . .t1 . . .V] XP1 ] V]
The existence of such a constraint could mean that (2b) is in fact the only structural option available. XP would prevent this by blocking restructuring. We could show, however, that (29) has a wider coverage than expected in such a scenario. Bangla, the language we used for comparison, has intraposed finite complements, i.e. complements which cannot be affected by clause union for principled reasons. Nevertheless, (29) holds in Bangla as well. Since (29) does not hold for clauses which are arguably CPs, this result led to the conclusion that (2a) is available but that the complement may be inappropriately characterized by the label CP. Evidence was adduced that intraposed complements are formally typed via their right edge, that this typing has a function in checking features of the selecting verb, and that this typing is affected by an intervener as in (29). In this case it would be wrong to ascribe the typing of the intraposed infinitives to a C-head. In fact, typological research reveals a strong correlation of post-verbal clauses with initial C and pre-verbal (intraposed) clauses with final C.29 German does not have final C, but the fact that intraposed clauses and sub-clausal units insist on the peripherality of the verb (here zu+V) can be seen as a reflex of this typological generalization. Various questions remain which we cannot yet answer in a satisfactory way. Why would a language afford the luxury of intraposed clausal complements, which are known to cause processing problems, if extraposition or base-generation in post-verbal position is a much preferred option? How do languages with intraposed finite complements resolve the tension between structure and processing? Notice that these languages have a final C like Bangla
Josef Bayer, Tanja Schmid and Markus Bader
bole, if they use a special functional element at all. Given that new clauses are normally signaled where they begin, the clause boundary would in these languages be signaled ‘far too late’.30 Under the assumption that the HSPM is universal, the same dispreference for intraposed clauses should be expected as the one found in German. The examples from Bangla which have been considered here involved finite complements with or without overt signs of subordination. Here the overtly expressed subject of the dependent clause could be an advantage for the parser. Thus, even if no functionally marked left clause boundary is present, the appearance of a (nominative) subject could be used as a clue for inserting a clause boundary. This clue cannot be considered very reliable though. In Bangla, subject and direct object NPs can be morphologically distinguished only if the object NP is animate and then suffixed with -ke. But even then Case disambiguation can be temporally postponed by pre-nominal material like stacked possessive NPs which are frequently used, and which do not give a clue to the ultimately correct analysis. Bangla has, in addition, null subjects as well as non-nominative (genitive) subjects. In combination with the fact that there is no V-clustering or any other form of coherence, these possibilities could turn the processing of pre-verbal clauses into a risky enterprise. We would therefore predict that intraposed clausal complements are approximately as dispreferred in Bangla as they are in German. We do not have quantitative data, but intuitive judgments by native speakers point in the expected direction. Since clause union is not an option, intraposed bare clauses and bole-clauses are for many speakers best when they are topicalized, i.e. moved over the matrix subject, and prosodically disconnected.31 This is in full agreement with Hawkins’ (1990: 248ff.) generalizations about the placement of head-final CPs in SOV-languages. Even if various questions cannot be answered conclusively, we hope to have shown here that quantitative and behavioral data about syntactic structures and processes as well as cross-linguistic comparison can shed new light on issues of Germanic syntax that have been studied quite extensively over the past 30 years, and that theoretical and empirical work can fertilize each other.
Appendix: Re-positioned zu (23d) and (24c) in the text above show a repair strategy by which it is guaranteed that the pre-nominal AP’s inflection appears at the right edge and thus adjacent to the following noun. Given the adjacency requirement that has been observed in intraposed zu-infinitives, a similar repair strategy could be ex-
Clause union and clausal position
pected. There is indeed the phenomenon of “re-positioned (or misplaced) zu”. Bech (1963) was one of the first to provide relevant data and give an account of this phenomenon. The head-final verb order in German leads to the expectation that the placement of zu in a cluster of three verbs is as in (ia). However, German has a rule by which in a cluster V+V+haben the auxiliary haben is preposed, the so-called “haben inversion”, as seen in als er ihn hat kommen sehen instead of als er ihn kommen (ge)sehen hat (‘when he saw him coming’).32 If (i-a) is the input to this rule we expect (i-b), but (i-b) is entirely unacceptable. What happens instead is a re-positioning of zu as seen in (i-c). (i)
a.
Ich glaubte es tun können zu haben. I thought it do can to have ‘I thought I would be able to do it.’ b. *Ich glaubte es zu haben tun können. c. Ich glaubte es haben tun zu können.
⇒ haben inversion ⇒
⇒ zu re-positioning ⇒ Bech (1963: 291f.)
This re-positioning always targets the last verb.33 In the context of our considerations, this phenomenon can be seen as reflecting the strength of the requirement to mark the right edge of the infinitive construction with the relevant Status feature, similarly to the requirement of marking the right edge of an AP with the relevant features for gender, number and Case agreement. The infinitival complement with re-positioned zu in (i-c) appears in postverbal position. Under the assumption that it is headed by a zero complementizer (or is at least not in a PF-relevant checking relation with the matrix verb) its right edge may not be an issue. However given that in German zu-infinitives have the option of appearing as complementizer-less clauses in immediately pre-verbal position, the right-peripheral visibility of their Status is an issue. A comparison with Dutch may be interesting at this point. With the exception of scrambled clauses headed by om, the grammar of Dutch precludes te-infinitives from pre-verbal position. Within our account, this could be related to the fact that te does not undergo re-positioning. Consider the following run-of-the-mill examples from the press. (ii) a.
Aan het wetsvoorstel wordt een nieuwe regeling toegevoegd to the law-proposal was a new regulation added om sneller te kunnen bouwen. in-order faster to can build ‘A new regulation was added to the bill in order to speed up permissions for construction.’
Josef Bayer, Tanja Schmid and Markus Bader
b. Het Amerikaanse leger zegt in het zuiden van de opstandige the American army says in the south of the rebellious stad Fallujah een huis te hebben ontdekt dat . . . town Fallujah a house to have found which ‘The American army claims to have found a house in the south of the rebellious town of Fallujah which . . .’ c. pil zonder naar de dokter te moeten gaan pill without to the doctor to must go ‘contraceptive for which one does not need to see a doctor’
Acknowledgements This work was supported by the Deutsche-Forschungsgemeinschaft-sponsored Sonderforschungsbereich 471 “Variation and Evolution in the Lexicon”. For discussion and help with the data we would like to thank the audience of the 19th Comparative Germanic Syntax Workshop as well as Jogamaya Bayer, Ellen Brandner, Probal Dasgupta, Lyn Frazier, Hubert Haider, Jana Häußler, Baris Kabak, Andrew McIntyre, Amanda Pounder, Uli Lutz, Henk van Riemsdijk, Susanne Trissler, Juan Uriagereka and Susanne Wurmbrand. Written comments by Marcel den Dikken and three anonymous reviewers have substantially improved an earlier version of this article. The remaining inadequacies are, of course, exclusively our responsibility.
Notes . We do not have the space here to justify these assumptions in detail. For general reviews of human parsing, cf. Mitchell (1994); for a defense of these assumption for parsing verb-final languages, cf. e.g. Bader & Lasser (1994); Inoue & Fodor (1995). . One can, of course, not a priori exclude the possibility of a positive conjunct attaching to a negative sentence, but the result is often semantically awkward as seen in John ate nothing, and Bill didn’t either/ (??) and so did Bill or paraphrases like (??) What John did was eat nothing. . A generalization of the CRP followed with Hoekstra’s (1984) Unlike Category Condition (UCC) and various developments stemming from later work such as van Riemsdijk’s (1988) Unlike Feature Constraint (UFC). . For PPs of this kind it is important that they are backgrounded / phonologically light and can integrate into the prosodic phrase of the preceding focused verb, here the phonological word entSCHIEden. Without intending to suggest a particular analysis, we will refer to this phenomenon as ex(traposition).
Clause union and clausal position . When we speak of ‘adjacency’ we refer only to the case in which the verbal complex of the infinitive is linearly adjacent to the matrix verb by virtue of the absence of rightward movement / stranding related to the infinitival complement. We will not speak of the independent case in which the linear adjacency of the verbs is broken by scrambling of the clause or by verb-projection raising. . One reviewer suggested that there may be prosodic reasons for the non-intervention effect. As shown by Truckenbrodt (1995), PP-extraposition and relative clause extraposition must obey a prosodic constraint which requires that a prosodic constituent π attaches to the immediate right of another prosodic constituent of the same type. If the light PP dafür in (13) and (14) is a phonological word ω, it could attach to zu+V which is the next ω with the effect of blocking clause union. Truckenbrodt’s constraint is observed if the PP can attach at the end as in Ich habe mich dafür zu entscheiden versucht dafür. Assuming that non-coherence must be expressed prosodically by two phonological phrases (φ), this would perhaps explain why *Ich habe ihn [sich dafür zu entscheiden] aufgefordert dafür (‘I asked him to decide on it’) or *Ich habe ihn [sich dafür zu entscheiden] gezwungen dafür (‘I forced him to decide on it’) are out; dafür would illegally have skipped a potential attachment site. This would not explain, however, why the non-intervention effect still holds in non-coherent constructions such as (14b). . The transcription of the examples follows IPA. The glosses should be read as follows: -cl (classifier-style determiner), -poss (possessor/‘genitive’ Case), -obj (objective/‘accusative’ Case) . The original observation of this is due to Probal Dasgupta (p.c.). . In fact, we are quite sure that the observed similarity between Bangla and German is not accidental. Informal testing with speakers of the SOV-languages Japanese and Turkish, both of which allow light pronoun extraposition, revealed the same adjacency constraints as those observed in Bangla and German. Thanks to Baris Kabak for his help in clarifying this issue. . The same holds for ki-clauses in Hindi and all other Indo-Aryan languages we could obtain information about. It also holds for the SOV-languages Persian and Turkish both of which have post-verbal clauses with an initial complementizer ke or ki, cf. Dryer (1980). . For detailed discussion see Bayer (1995; 1996; 1999; 2001); Dasgupta (1980); Singh (1980). . According to Grosu & Thompson (1977), initial complementizers in pre-verbal complements are almost universally unavailable. . Cf. Wurmbrand (2001) to mention a more recent analysis. . Cf. Koster (1986: 123). . Arguments in favor of co-projection of this kind can be found in Reuland and Kosmeijer (1988), Haider (1993), Bayer & Kornfilt (1990; 1994) and elsewhere. . One immediate advantage of such a system is that Case checking is independent of positions. Given the possible base order non-nominative < nominative of German and other SOV-languages, there is a clear advantage over the standard position-based accounts of Case checking. Bayer (2001) points out other advantages in accounting for negation in South Asian languages.
Josef Bayer, Tanja Schmid and Markus Bader . For a caveat cf. the appendix. . We will try to make this more precise in Section 3.5. . Widely known examples are so-called ‘quantifier-floating’ and other split-NP-constructions. Cf. Note 25. . In this situation, complement and specifier are collapsed, a technical state of affairs which seems to characterize head-final constructions and their apparent non-distinctness of specifiers and complements quite accurately. . For arguments against clause union in terms of VP-remnants cf. Haider (2003). . From Escribano (2004: 26). . This is an observed example which may be considered ungrammatical. The fact that such constructions are produced by speakers of English shows, however, that the HFF cannot hold in full generality in this language. For a recent account of the HFF in purely phrase structural terms and without any reference to agreement cf. Escribano (2004). One of the problems of Escribano’s theory is that it cannot account for Dutch and German. Another problem is that it considers only modifiers but no complements or specifiers. It would thus be impossible to apply it to pre-verbal infinitival complements. . φ and ω are symbols for the phonological phrase and for the phonological word respectively. . For supporting demonstrations of this cf. Truckenbrodt (1995). According to the idea that the trace is a copy of the displaced element it should be clear that this copy is a syntactic and not a phonological entity. NP-split in German provides empirical support of this claim. Thus, although there is reliable evidence for movement (cf. van Riemsdijk 1989) in (i), the verbatim reconstruction of ein Auto leads to a seriously ill-formed construction as seen in (ii). The only permissible form would be the morpho-syntactically different form seen in (iii). (i)
[keines t1 ] leisten [Ein Auto]1 kann er sich leider car can he ref unfortunately none afford a ‘As for a car, he can unfortunately not afford any’
(ii) *(daß) er sich leider [keines [ein Auto]] leisten kann (iii) (daß) er sich leider [kein Auto] leisten kann . Notice that this amounts to the upward boundedness which Ross (1967) captured with the Right Roof Constraint. . For the processing of German, cf. Bader & Lasser (1994) and Bader (2000). . Sentence completion tasks from our experimental work in progress never yield constructions in which the provided constituents would distribute over more than a single clause. As realistic completions of (28) one can, thus, expect di-transitive verbs like zeigte (‘showed’), gab (‘gave’), verkaufte (‘sold’) etc. but not zu kaufen empfohlen hat (‘has recommended to buy’) or zu schenken dem Onkel empfohlen hat (‘has recommended to the uncle to buy’) etc.
Clause union and clausal position . Cf. Table 3, which describes Bangla. For reasons of space we cannot support this here with data from other languages. . Cf. Kimball (1973). . Hindi is closely related to Bangla but lacks pre-verbal finite complements altogether. Experimental results about the processing of center embedded non-finite clauses in Hindi (cf. Vasishth 2003) reflect the expected complexities. . This rule also extends to werden and sometimes even to modals, but less frequently so. . The purely linear nature of zu re-positioning should throw doubts on the representation in (20). In (20) the Status feature is projected with the verbal head of the clause. Repositioning of zu shows, however, that the appearance of zu as a prefix to the verbal head may only be an artifact of non-inversion. We have to leave this important issue for future research.
References Bader, M. (2000). On reanalysis: Evidence from German. In B. Hemforth & L. Konieczny (Eds.), German Sentence Processing (pp. 187–246). Dordrecht: Kluwer. Bader, M. & I. Lasser (1994). German verb-final clauses and sentence processing: Evidence for immediate attachment. In C. Clifton, Jr., L. Frazier, & K. Rayner (Eds.), Perspectives on Sentence Processing (pp. 225–242). Hillsdale, NJ: Lawrence Erlbaum. Bayer, J. (1995). On the origin of sentential arguments in German and Bengali. In H. Haider, S. Olsen, & S. Vikner (Eds.), Studies in Comparative Germanic Syntax (pp. 47–75). Dordrecht: Kluwer. Bayer, J. (1996). Directionality and Logical Form: On the scope of focusing particles and wh-insitu. Dordrecht: Kluwer. Bayer, J. (1999). Final complementizers in hybrid languages. Journal of Linguistics, 35, 233– 271. Bayer, J. (2001). Two grammars in one: Sentential complements and complementizers in Bengali and other South Asian languages. In P. Bhaskararao & K. V. Subbarao (Eds.), The Yearbook of South Asian Languages and Linguistics 2001 (pp. 11–36). New Delhi: Sage Publications. Bayer, J, & J. Kornfilt (1990). Restructuring effects in German. In E. Engdahl, M. Reape, M. Mellor, & R. P. Cooper (Eds.), Parametric Variation in Germanic and Romance [Edinburgh Working Papers in Cognitive Science 6] (pp. 21–42). University of Edinburgh. Bayer, J. & J. Kornfilt (1994). Against scrambling as an instance of Move-α. In N. Corver & H. van Riemsdijk (Eds.), Studies on Scrambling (pp. 1–15). Berlin: Mouton de Gruyter. Bech, G. (1955/1957). Studien über das deutsche verbum infinitum. (Det Kongeliege Danske Videnskabers Selskab; Dan. Hist. Filol. Medd. Bind 35, no. 2 (1955) & Bind 36, no. 6 (1957)). New edition 1983. Tübingen: Niemeyer. Bech, G. (1963). Grammatische Gesetze im Widerspruch. Lingua, 12, 291–299.
Josef Bayer, Tanja Schmid and Markus Bader
Chomsky, N. (2001). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 1–52). Cambridge, MA: The MIT Press. Dasgupta, P. (1980). Questions and Relative and Complement Clauses in a Bangla Grammar. PhD Dissertation, New York University. Dryer, M. S. (1980). The positional tendencies of sentential noun phrases in universal grammar. The Canadian Journal of Linguistics, 25, 123–195. Escribano, J. L. G. (2004). Head-final effects and the nature of modification. Journal of Linguistics, 40, 1–43. Evers, A. (1975). The Transformational Cycle in Dutch and German. Doctoral Dissertation, Rijksuniversiteit Utrecht. Frazier, L. & K. Rayner (1982). Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14, 178–210. Gorrell, P. (1995). Syntax and Parsing. Cambridge: Cambridge University Press. Grewendorf, G. (1988). Aspekte der deutschen Syntax. Tübingen: Narr. Grosu, A. & S. Thompson (1977). Constraints on the distribution of NP clauses. Language, 53, 104–151. Haider, H. (1993). Deutsche Syntax – generativ. Vorstudien zu einer projektiven Theorie der Grammatik. Tübingen: Narr. Haider, H. (2003). V-clustering and clause union. Causes and effects. In P. Seuren & G. Kempen (Eds.), Verb Constructions in German and Dutch (pp. 91–126). Amsterdam: John Benjamins. Hawkins, J. A. (1990). A parsing theory of word order universals. Linguistic Inquiry, 21, 223–261. Hoekstra, T. (1984). Transitivity. Grammatical relations in government-binding theory. Dordrecht: Foris. Inoue, A. & J. D. Fodor (1995). Information-paced parsing of Japanese. In R. Mazuka & N. Nagai (Eds.), Japanese Sentence Processing (pp. 9–63). Hillsdale, NJ: Lawrence Erlbaum. Kayne, R. S. (1994). The Antisymmetry of Syntax. Cambridge, MA: The MIT Press. Kimball, J. (1973). Seven principles of surface structure parsing in natural language. Cognition, 2, 15–47. Koster, J. (1986). Domains and Dynasties. Dordrecht: Foris. Mitchell, D. C. (1994). Sentence parsing. In M. A. Gernsbacher (Ed.), Handbook of Psycholinguistics (pp. 375–409). San Diego, CA: Academic Press. Reuland, E. & W. Kosmeijer (1988). Projecting inflected verbs. Groninger Arbeiten zur Germanistischen Linguistik, 29, 88–113. van Riemsdijk, H. (1988). The representation of syntactic categories. In Proceedings of the Conference on the Basque Language, Basque World Congress. Vol. 1 (pp. 104–116). van Riemsdijk, H. (1989). Movement and regeneration. In P. Benincá (Ed.), Dialect Variation and the Theory of Grammar (pp. 105–136). Dordrecht: Foris. van Riemsdijk, H. (1994). Adjunction and adjaceny. Ms. Tilburg University. Ross, J. R. (1967). Constraints on Variables in Syntax. PhD Dissertation, MIT. Schmid, T., M. Bader, & J. Bayer (in press). Coherence – an experimental approach. In M. Reis & S. Kesper (Eds.), Linguistic Evidence. Berlin: Mouton de Gruyter.
Clause union and clausal position
Singh, U. N. (1980). Bole an unresolved problem in Bengali syntax. Indian Linguistics, 41, 188–195. von Stechow, A. & W. Sternefeld (1988). Bausteine syntaktischen Wissens. Opladen: Westdeutscher Verlag. Stowell, T. (1981). The Origins of Phrase Structure. PhD Dissertation, MIT. Truckenbrodt, H. (1995). Extraposition from NP and prosodic structure. NELS, 25, 503– 517. Vasishth, S. (2003). Working Memory in Sentence Comprehension. Processing Hindi Center Embeddings. London: Routledge. Vikner, S. (2001). Verb Movement in Germanic and Optimality Theory. Habilitation thesis, University of Tübingen. Wurmbrand, S. (2001). Infinitives. Restructuring and clause structure. Berlin: Mouton de Gruyter.
Explaining Expl* Marc Richards and Theresa Biberauer University of Cambridge
We address a problematic asymmetry in the Probe-Goal-Agree system of Chomsky’s Minimalist Inquiries. Of the three Core Functional Categories, only C and T provide merge-sites for expletives (Expl); v lacks the Merge-Expl property. We claim that Merge-Expl is, instead, unique to phase heads (C/v); Spec-vP is consequently the only possible merge-site for there-type Expl. This eliminates a number of technical and conceptual problems associated with standard ‘Merge-TP’ approaches to Expl’s underlying distribution; moreover, it offers considerable empirical advantages. Building on a novel typology of EPP-satisfaction proposed in Biberauer & Richards (2004), we show how ‘Merge-vP’ allows a complete and perspicuous account of the surface distribution – diachronic and synchronic – of expletives across Germanic, including optional-expletive constructions in modern Dutch and Afrikaans.
. Introduction Chomsky (1998/2000) (Minimalist Inquiries, henceforth MI)1 marks a major conceptual shift in the analysis of feature-checking/agreement relations. Where earlier instantiations of checking theory relied on movement operations to feed the creation of local checking relations such as specifier-head, the Move component is now divorced from agreement proper, with agreement feeding movement (i.e. the converse of what had gone before). On this view, the ‘long-distance’ agreement effects familiar from expletive-associate constructions become primary, the T-associate agreement relation providing the template for a generalized operation of the form Probe-Goal, known as Agree. In dispensing with the need for “complex and unnatural” notions such as checking domain (MI: 125), the Probe-Goal incarnation of checking theory allows a considerable simplification and streamlining of this central component of the computational system, with the additional advantage of allowing a separate LF-cycle and its associated problems and redundancies to be eliminated.
Marc Richards and Theresa Biberauer
Most notably, since the associate now agrees with T in situ, LF-raising to the expletive position is finally dispensed with, so that the new binding possibilities and wide-scope interpretations erroneously predicted on the earlier expletivereplacement analyses (as pointed out in den Dikken 1995) are trivially absent – the associate simply does not move (MI: 123). Nevertheless, for all the conceptual and empirical advances that ProbeGoal system brings, there remain a number of problems and inconsistencies involving the very type of structures that the theory is designed around, namely expletive-associate structures. In particular, in its emphasis on the T-associate part of the picture, the nature and role of the expletive element itself in these relations has been largely neglected and, if anything, is now even more mysterious than it was on previous versions of the theory. Thus, Chomsky characterizes this long-distance agreement as a “T-associate relation that involves features only and is independent of the expletive” (MI: 126; our emphasis), but this emancipation of the expletive from the expletive-associate relation brings with it its own set of problems. As we discuss in Section 2, the expletive element (henceforth Expl)2 simply refuses to fit into the Probe-Goal system, exhibiting such anomalous behaviour as being a probe that does not project and satisfying EPP-features via (first, external) Merge rather than Move. Clearly, if we are to fully accept the Probe-Goal system of feature checking and all its attendant benefits, then these problems, central to the definitive exemplar of long-distance agreement, must be overcome. These problems, we believe, can only be overcome once we have a proper understanding of exactly where (and why) Expl enters the syntactic derivation; that is, once we have a fully explanatory account of Expl’s distribution. As Chomsky himself recognizes: (1) The problem of accounting for the distribution of EXPL in some principled way [. . . ] remains open. (BEA: 25 (fn. 45))
We take up this challenge in the remainder of this paper by investigating the facts and patterns of Expl-distribution across the Germanic languages (both the modern and historical varieties). The Germanic data, we argue, offers sufficient evidence for a complete account of Expl-distribution to emerge, one that captures not only Expl’s surface distribution (i.e. where it moves) but also its underlying distribution (i.e. where it merges). Only an analysis that can account for both of these dimensions of Expl-distribution can claim to be truly explanatory. Thus, in Part I of this paper, we give a principled account of the basic (‘underlying’) distribution of Expl; Part II then develops a principled account of the derived (‘surface’) distribution of Expl on the basis of the
Explaining Expl
conclusions of Part I and a novel theory of parametrized EPP-satisfaction (cf. Biberauer & Richards 2004).
Part I. Merge-Expl .
An unexpected asymmetry
What, then, are the possible Merge-sites for Expl? The best way to answer this question is by process of elimination. We can immediately exclude one class of positions: as a nonreferential, nonargumental category, Expl cannot merge into θ-positions (“Pure Merge in θ-position is required of (and restricted to) arguments”, MI: 103 (6)), that is, those positions licensed by the selector features of lexical heads. This restriction, which equally bars arguments from merging into non-θ (EPP-licensed) positions, places the lexical domain out of bounds for Merge-Expl.3 It therefore leaves us with the functional domain as the domain of Merge-Expl. Assuming the minimal clause-level functional structure given by the three Core Functional Categories (CFCs), there remain three possible Merge-Expl sites: the specifiers of C, T, and v. At this point, however, we encounter an unexpected asymmetry. Of the three CFCs, only C and T seem to actually provide Merge sites for Expl; v lacks the Merge-Expl property. Consider (2)–(4). (2) Merge to Spec-CP a. ‘V2-expletives’ / ‘expletive topics’ [vP [CP það klaruðu [TP margar mýs [vP ostinn there finished many mice cheese-the alveg [vP tsubj [VP tV tobj ]]]]]] completely ‘Many mice finished the cheese completely.’ [Icelandic, from Alexiadou & Anagnostopoulou 2001: (41)] b. ‘Wh-expletives’ / ‘partial movement’ Was glaubst du, welchen Mann sie liebt? what believe you which man she loves ‘Which man do you believe she loves?’ [German, from Felser 2003: (17)] (3) Merge to Spec-TP ‘Expletive subjects’ [TP There T [VP arrived a man]]
Marc Richards and Theresa Biberauer
(4) Merge to [outer] Spec-vP = n/a ?
There is a range of evidence to suggest that the Icelandic expletive það, as in (2a), is merged directly into spec-CP as a kind of ‘expletive topic’ when no other XP raises to the preverbal topic position (cf. Thráinsson 1979; Sigurðsson 1989; Falk 1993a, b; Holmberg & Platzack 1995; Bobaljik 2002; Bowers 2002), not least the fact that this element can never appear in the immediately postverbal, spec-TP position (see Section 4 below). Its role seems simply to be to ensure that the V2 requirement is met; in so far as the latter is a PF/ordering condition, það is perhaps to be inserted in the phonological component (Holmberg 2000; Bobaljik 2002; Wurmbrand 2004). The wh-expletives involved in partial movement structures such as (2b) would also seem good candidates for direct Merge into spec-CP, as they fail to exhibit the movement-related properties of the superficially similar ‘wh-copying’ construction (see Felser 2003: 551ff. for relevant discussion). The expletive subject in (3), of course, represents the most familiar kind of expletive, and its merger directly into the ‘subject position’ of the clause (i.e. spec-TP/IP) has long been a standard assumption in generative syntax. Under Government and Binding (GB) accounts, Expl is introduced into spec-IP to satisfy the Extended Projection Principle (EPP), as originally conceived, which required that every clause have a (structural) subject in spec-IP at S-structure (cf. Chomsky 1981, 1982). This analysis was carried over into minimalism and checking theory with only minor modifications – Expl is introduced into specTP for reasons of checking uninterpretable features, whether these be the Case feature on T (cf. Groat 1995, 1999), the Case and φ-features on T (Lasnik 1995), or purely the strong D-feature on T (the ‘purest’ implementation of the GB EPP; Chomsky 1995). It is when we come to vP that the asymmetry emerges, as empirical evidence for a class of vP-expletives would seem to be entirely lacking. For instance, whilst conceivable in principle, we do not find structures of the form in (5), with a theoretical vP-Expl alternating with Object Shift in (e.g.) Icelandic (cf. (2a)) in the same way that the TP-Expl in (3) alternates with raising of the associate subject. (5) *[CP Það klaruðu [TP margar mýs [vP Expl [vP alveg [vP tsubj [VP tV ostinn]]]]]]
Let us assume for the time being that this apparent empirical hole is a real one (we will return to refute it in our concluding section – see Section 7 below).
Explaining Expl
This restriction on Expl-distribution appears in various forms in Chomsky’s recent work. In both MP and MI, Merge-Expl is argued to be a property of T alone.4 Thus, in MP (p. 362 (196)), he writes: “Exp can only be in [Spec, T]”; in MI (p. 102 (5a)), taking the structure [XP [(EA) H YP]], he states: “If H is v/C, XP is not introduced by pure Merge” – i.e. pure Merge, of Expl, is restricted to H = T. The question then becomes, can we give a principled explanation for spec-vP’s lack of the Merge-Expl property? In MP (pp. 364–365), Chomsky attempts to derive the lack of Merge-Expl for v from the impossibility of associate raising to spec-vP at LF. Whether or not his reasoning goes through (and, tellingly, he concedes in his footnote 138 that this would still only bar associate raising of the external argument, i.e. raising from the inner to the outer specifier of vP, and not the raising of the internal argument to the latter position, so that our (5) should still in fact be possible), it fails to carry over to the Probe-Goal system of MI/DbP/BEA, where, as discussed above, associate raising (indeed, LF-movement generally) is replaced by the operation of (‘long-distance’) Agree. In MI (p. 109), on the other hand, the lack of Merge-Expl for C and v derives from their status as phase heads; Merge-Expl is argued to follow as a property of non-phase heads.5 This latter line of argumentation is promising, tying as it does the Merge-Expl property to an independent and indispensable corollary of the Probe-Goal derivational architecture – the phase. However, it clearly runs into empirical problems (besides the conceptual ones sketched in Note 5) given that C does not lack the Merge-Expl property (cf. the data in (2)). In light of the evidence that external Merge can in fact apply to spec-CP, Chomsky (BEA: 25) concedes that T’s Merge-Expl property (and v’s lack of it) cannot be derived from the phasal status of C and v after all and, essentially, admits defeat (viz. the quote in (1)). The lack of Merge-Expl for v thus remains a mystery. Nevertheless, we contend that the phase-theoretic approach to explaining Expl’s Merge-distribution is the right one and should not be abandoned, for the reasons just stated. Instead, we simply need a shift in perspective, such that it is not C’s Merge-Expl property that undermines the phase-based approach to Merge-Expl but rather T’s. Once we give up the long-held, deeply entrenched view that Merge-Expl is a property of T and instead attribute this property to v, the problem for a principled phase-theoretic account vanishes, as Merge-Expl then emerges as a unifying property of phase-heads rather than of non-phase-heads. Our claim, then, is that the data of the kind in (3) have been misanalysed. What have traditionally been analysed as TP-expletives are really vP-expletives. We will show how the derivation of a sentence such as (3) proceeds under a vP-
Marc Richards and Theresa Biberauer
Expl analysis in Section 3 (see (10)), where it will also be argued that such an approach immediately resolves the technical problems outlined in the Introduction and elaborated in the next section. For now, though, let us just point out one conceptual advantage of this approach. The phase is defined in MI as a lexical subarray containing a single instance of one of the phase-determining functional heads, C or v. As such, each phase constitutes a (separate) domain of external Merge (hence ‘Merge-over Move’ effects, however these are to be explained, are restricted to the phase). We would therefore expect the Merge-Expl property to be characteristic of phase heads – anything else is a departure from this null assumption. In other words, it would be much easier to derive the lack of Merge-Expl from T than from v – the latter would have to be stipulated (cf. Note 5), a distinctly suboptimal state of affairs from a minimalist point of view. In short, our approach can be summarized as in (6). That Merge-Expl should be a property of v rather than T is a priori to be preferred as the optimal outcome given that phases are the domains of external Merge. (6)
Merge-Expl Non-Merge-Expl Chomsky 1995, MI T C, v Present paper C, v T
In the next section, we strengthen our claim from the opposite angle – that is, not only can (and should) Merge-Expl be a property of v, as outlined above, but it cannot be a property of T given the workings of the Probe-Goal system.
. Why TP-Expl is inadequate Before examining in more detail the problems faced by the standard, prevailing accounts of Expl as merging directly into spec-TP, let us briefly review the core assumptions of the Probe-Goal-Agree system (MI/DbP/BEA) in which these problems arise. Central to this framework is the operation Agree, which establishes a feature-checking relation between heads and effects deletion of uninterpretable features under matching. Agree is initiated by the uninterpretable features of a Probe; these search for a corresponding set of interpretable (valued) features on a Goal. Agree is said to obtain when this matching of feature sets results in valuation of unvalued features. Specifically, the conditions for valuation/Agree are as follows:
Explaining Expl
(7) Definition: Agree(P[robe],G[oal]) if a. P c-commands G b. P and G are active c. P matches G for feature F d. G is interpretable for F (cf. DbP: 6) If all of (7a–d) obtain, then valuation occurs as in (7e): e. P values and deletes G if P is φ-complete; G values and deletes P if G is φ-complete
Agree thus replaces the arbitrary spec-head relation, previously central to checking theory, with basic c-command/sisterhood: the probe’s search space is simply the existing structure to which it is merged. Displacement to the specifier of the probe is no longer a necessary component of the agreement/featurechecking relation; rather, any such spec-head configuration requires a further trigger and is therefore secondary to (and parasitic upon) Agree proper. Specifically, Move (internal Merge) of G to (spec-)P obtains only if P is associated with an EPP-feature (“OCC”). Further, as spec-head relations are no longer primitive checking configurations, the system implies that the EPP-feature can only be checked/satisfied by Move (internal Merge). As Chomsky puts it: (8) If there is no SPEC-head relation, then the EPP-feature OCC cannot be satisfied by [external – MDR/MTB] Merge alone. (BEA: 11) It [. . . ] follows that external Merge does not suffice to check OCC. (BEA: 12)
That is, direct Merge of an element into the specifier of an EPP-bearing probe will fail to satisfy that EPP[-feature] as the specifier is not in any relevant relation with the probe/head. Hence Move (internal Merge) is required to satisfy a probe’s EPP, which in turn implies Agree (since Move is a composite operation comprising the components Agree + Piedpiping + Merge). Finally, Expl in this framework is taken to be a simple head with a defective and uninterpretable φ-set, perhaps simply [uPerson] (MI: 125, BEA: 12), hence it is active for probing and raising. This leads us directly to the first problem with this system: namely, if EPP requires internal Merge of G, how can T’s EPP be satisfied by Expl externally merged into Spec-TP?6 Moreover, how does T value Expl’s uninterpretable Person feature in this position? Chomsky’s solution (BEA: 12) is to allow Expl to probe T from its Merge site, spec-TP, a possibility that follows since Expl, as a simple head, c-commands T (and can therefore probe it, at least by (7a)). Agree(Expl, T) thus values (and deletes) EPP on T and [uPerson] on Expl.
Marc Richards and Theresa Biberauer
This solution only creates further problems, however. Even accepting that the EPP can be valued as a goal of Expl in the same way as uninterpretable Case/agreement features (a departure from MI/DbP), the valuing of Expl’s [uPerson] by T is not straightforward. Specifically, Agree(Expl, T) imposes a paradoxical requirement on T’s matching φ-set. For the structure [TP There T [VP arrived a man]] (cf. (3)), there are two logically possible derivations: (9) a. Agree(T, a man) precedes Agree(Expl, T) or b. Agree(Expl, T) precedes Agree(T, a man)
In either case, the Agree(Expl, T) operation will fail to obtain. (9a) is the expected derivation given a strictly step-by-step, cyclic application of Merge/Agree (cf. the Locus Principle of Collins 2002). However, Agree(Expl, T) here fails to meet condition (7b), since T will be rendered inactive by the preceding operation Agree(T, a man). The alternative derivation in (9b) arises as a possibility if we adopt Chomsky’s suggestion (BEA: 21–22) that operations are evaluated only at the phase-level, allowing a measure of phase-internal countercyclicity such as that implied by (9b). But Agree(Expl, T) now fails to meet condition (7d), since T’s φ-set is unvalued (uninterpretable) until Agree(T, a man) takes place. Agree(T, a man), therefore, both bleeds ((9a)) and feeds ((9b)) Agree(T, Expl), leaving us no way out, whichever order they apply in. Even if we assume the weaker version of phase-relativized cyclicity of DbP (p. 18), such that only the valuation/deletion part of Agree (and not the Match component) takes place at the phase level (i.e. at the point of SpellOut/Transfer), Agree(Expl, T) is still excluded. Whilst T in the (9a) derivation would then be able to remain active (“visible”) for (7b) even after Agree(T, a man), (9a) would now face the same problem as (9b), i.e. Agree(Expl, T) would fail to meet condition (7d), since T remains unvalued until the end of the phase. Unless we are willing to admit the existence of feature-chains, as in Frampton and Gutmann (2000), or otherwise to weaken condition (7d), this problem would seem irresoluble in a strictly derivational system that assumes Expl to merge to T. There is one further problem that cuts across all of the above, and that is Expl’s status as a probe. We have already touched on one problematic aspect of this (cf. (9) above): probing by Expl violates (featural/locus) cyclicity. If Expl and T are both probes, then Expl cannot be introduced to drive further operations until all the probe/selector requirements of T are satisfied (cf. MI: 132 (53)); in terms of MP (p. 234), “D[erivation] is canceled if α is in a category not headed by α.” The effect of this is that operations must be triggered by the
Explaining Expl
root node (i.e. the Extension Condition). However, Expl is not technically the root node here: it is contained in a category (TP) of which it is not the head (TP is not a projection of Expl, therefore T, not Expl, is the root node at the relevant point in the derivation). That Expl is not the root node becomes clear if we consider what would happen if Expl, as a probe, were itself associated with its own EPP-feature. Then, movement of Expl’s goal would have to target ‘spec-Expl’, an operation that would not extend the tree (but would instead make it ‘bushier’, as a result of adjunction to a specifier). Indeed, given that Merge-Expl is viewed as a free alternative to Move-Goal on the standard, TP-Expl approach, Expl’s hypothetical EPP could itself be satisfied via Merge-Expl, introducing another Expl probe that could then itself be associated with an EPP-feature, which could in turn be satisfied by Merge-Expl, and so on ad infinitum. This computational absurdity (an infinite regress of expletives inside each other’s specifiers) would be a considerable imperfection from a minimalist point of view, and strengthens the case for treating left branches as inactive/‘opaque’ (cf. Uriagereka 1999). Thus Expl may well be a head in Chomsky’s system, but it is also a nonprojecting ([+maximal]) category, unlike the root node. Since it is inactive for projection, it must also be inactive as a probe. In light of such problems, we believe that the standard position – viz. Merge-Expl as a property of T – is untenable under a serious and consistent application of the Probe-Goal logic.7 Given the desideratum that Merge-Expl should be a property of phase heads (Section 1), there would be little reason to maintain the TP-Expl approach if a Merge-vP analysis of Expl’s underlying distribution can overcome all of the aforesaid problems whilst accounting for the same full range of empirical facts. This is indeed the case, as we demonstrate in the next section.
. Raising a solution We can now make clear our proposal and show how it allows Expl to be fully integrated into the derivational architecture of the Probe-Goal system. Expl, we propose, merges in spec-vP, not spec-TP, and subsequently raises to T on the back of an Agree relation like any other formal subject.8 The assumption that Expl merges in spec-vP allows us to cut the Gordian knot underlying all three of the technical issues discussed in the previous section, as it is now no longer necessary to assume that Expl is a probe. Since spec-vP is in T’s Agree (ccommand/sister) domain, Expl can simply be probed and valued by T like any
Marc Richards and Theresa Biberauer
other nominal. And, like any other nominal, it is rendered active for goalhood by a Case feature (cf. Hazout 2004: 412). There is therefore no need to assume that Expl’s defective φ-set is unvalued: its [Person] has default, third-person value, hence it is not a probe.9 Since it is not a probe, we avert the cyclicity problems associated with Expl’s nonprojecting/[+maximal] status. If Expl merges inside T’s Agree domain (i.e. is contained inside T’s sister), then the EPP/external-Merge problem is also naturally overcome, since Expl now satisfies T’s EPP via Move (internal Merge), as required, in the same way that nominal arguments do.10 Indeed, given the core functional structure CT-v and the requirement that Expl satisfy EPP(T) via Move, the only possible Merge site for Expl is (the nonthematic) spec-vP once θ-positions are excluded (cf. Section 1). A similarly trivial solution is now available for the remaining problem of Section 2, namely the conflicting requirements on T’s φ-set implied by the Agree(Expl, T) relation in (9). Since Expl, firstly, is not a probe and, secondly, merges in spec-vP and not spec-TP, no Agree(Expl, T) relation need ever take place. Instead, the derivation of a structure such as [TP There T [VP arrived a man]] proceeds as in (10), with Expl valued by T via a canonical, fully legitimate Agree relation in step (10e). At that point, T values Case on Expl, but T’s φ-set remains unvalued as Expl is φ-incomplete (cf. (7e)). T therefore remains active for Agree(T, a man)) in step (10f). (10) Sample derivation: [TP There T [VP arrived a man]] (cf. (3)) a. Merge(V, DP) → [VP arrived [DP a man]] b. Merge v (defective/nontransitive) → [v [VP arrived [DP a man]]] c. Merge Expl11 → [vP there [v [VP arrived [DP a man]]]] d. Merge T → [T [vP there [v [VP arrived [DP a man]]]]] e. Agree(T, Expl) → [T[φ, EPP] . . . Expl[φ, Case] ] f. Agree(T, a man) → [T[φ, EPP] . . . DP[φ, Case] ] g. Merge(Expl, T) [i.e. Move] → [TP there [T[φ, EPP] [vP (there) [v [VP arrived [DP a man]]]]]
Expletive-raising is, of course, already standardly assumed for such structures as (11a), cf. DbP: 16–17 ((15a), (18a)). Our approach simply generalizes such a
Explaining Expl
raising derivation to all cases of Expl, including presentational clauses such as (3)/(10) and existentials such as (11b). (11) a. [TP There [VP seems [TP (there) to be a man in the garden]]] b. [TP There is [vP (there) [SC a man in the garden]]]
Our proposal can thus be viewed as a Probe-Goal implementation of earlier movement/raising analyses of Expl, such as Hoekstra and Mulder (1990), Zwart (1992), den Dikken (1995), Moro (1997), Groat (1999), and Sabel (2000). However, our account crucially differs from all of these (except Sabel 2000) in that we do not assume that Expl raises as a predicate (see Hazout 2004: 396–397 for arguments against such approaches). Expl, for us, is a dummy argument, not a (dummy) predicate, and so agrees with and raises to T like any other nominal/argument. As such, our Merge-vP account of Expl within the Probe-Goal framework comes very close to that of Bowers (2002), who independently proposes a low Merge site for Expl (spec-Pr, in his terms). Whilst a full comparison of our analyses is beyond the scope of this paper (for some comments, see Section 7), it is worth noting that our Merge-vP account, like that of Bowers, allows many further previously unexplained properties of Expl’s surface distribution to be straightforwardly derived (such as its restriction to unaccusative/passive predicates, i.e. those that lack an external argument, in so far as Expl and the external argument compete for the same specifier – we will refine this stance somewhat in Section 7). In the remainder of this paper, we focus on just one such problematic aspect of Expl’s surface (‘Move-Expl’) distribution, perhaps the major Expl-related issue from the perspective of comparative Germanic syntax, and show how our Merge-vP approach paves the way for a new and truly explanatory account of the phenomenon in question.
Part II. Move-Expl . Empirical support: EPP-satisfaction in Germanic . Introduction: Problematic data and previous accounts The distribution of Expl in Germanic has received a significant amount of attention in the literature to date (see Vikner 1995: Chapters 6 and 7 for an overview, and Rosengren 2002 for references to some subsequent discussions). What these discussions highlight is the fact that Expl-distribution in languages
Marc Richards and Theresa Biberauer
like English and also the Mainland Scandinavian languages (Danish, Norwegian and Swedish; henceforth MSc) seems, in general,12 to conform to what standard (Merge-TP) theories of Expl would lead us to expect: Expl must surface in spec-TP wherever an appropriate (i.e. ‘thematic’) subject is absent, thus in presentational contexts in English and additionally in impersonal passives in the MSc languages. Two languages whose behaviour is rather less in keeping with what standard theories would predict are German and Icelandic. The relevant problematic data are presented in (12) and (13) below: (12) a.
(German)
(13) a.
(Icelandic)
Es kam gestern ein Junge. Expl came yesterday a boy ‘There came a boy yesterday.’ b. Gestern kam (*es) ein Junge. yesterday came (Expl) a boy ‘Yesterday there came a boy.’ c. Es wurde getanzt. Expl became danced ‘There was dancing.’ d. Gestern wurde (*es) getanzt. yesterday became (Expl) danced ‘Yesterday there was dancing.’
Það hefur komið strákur. Expl has come boy ‘There came a boy.’ b. Í gær hefur (*það) komið strákur. yesterday has (Expl) come boy ‘Yesterday there came a boy.’ c. Það hefur verið dansað. Expl has been danced ‘There was dancing.’ d. Í gær hefur (*það) verið dansað. yesterday has (Expl) been danced ‘Yesterday there was dancing.’
As the examples show, both German and Icelandic exhibit the peculiar requirement that Expl be barred from overt realization in spec-TP, i.e. it may not surface in precisely the position in which it is standardly assumed to merge. Additionally, these languages also permit problematic structures such as those illustrated in (14) and (15) where it seems clear that raising of the nominative argument to spec-TP has failed to occur, but Expl-insertion is nevertheless impossible:13
Explaining Expl
(14) . . . daß (*es) dem Mann ein Buch geschenkt wurde. that (Expl) the.dat man a.nom book presented became ‘. . . that the man was given a book as a present.’ (German) (15) Í gær voru (*það) konunginum gefnir hestar. yesterday were.3pl (Expl) king-the.dat given horses.nom ‘Yesterday the king was given horses.’ (Icelandic)
Faced by data of this kind, adherents of the Merge-TP analysis of Expl have in general opted for one of three types of explanation: a. a pro-based one (cf. i.a. Safir 1985; Sternefeld 1985; Rizzi 1986; Koster 1987; Grewendorf 1989; Cardinaletti 1990); b. one that appeals to the idea that TP is simply not projected (cf. i.a. Abraham 1993; van Gelderen 1993; Haider 1993); and c. one that postulates an optional EPP-feature on T (cf. Wurmbrand 2004; and also i.a. Rosengren 2002 and Mohr 2004). Let us briefly consider each of these in turn. The central idea underlying the pro-based analysis is that languages that fail to license referential pro (so-called semi pro-drop languages; cf. Rizzi 1982, 1986) may license expletive pro in a very specific set of contexts, namely those where expletive pro is governed (essentially, c-commanded) by an appropriate (functional) head. Since (filled) C is identified as an appropriate governor for the licensing of expletive pro in German and Icelandic, the absence of overt expletives in spec-TP contexts such as those illustrated in (12)–(15) above appears to be explicable: expletive pro is licensed in these contexts as a suitable governor is present, while overt expletives are barred as the consequence of economy-based considerations such as, for example, the Avoid Pronoun Principle (cf. Chomsky 1981: 65) which stipulates that, wherever a covert pronominal is possible, its overt counterpart is barred unless an emphatic reading is required (cf. Travis 1984: 229). Viewed from a minimalist perspective, this type of account clearly faces a number of problems, not least the question of how the notion of government might be recast and also the more fundamental question of how the existence of an entity like expletive pro, a category lacking both phonological and interpretable grammatical features, can be justified (cf. MP: 349ff. in connection with the status of Agr-categories; and Manzini & Savoia 1997; Alexiadou & Anagnostopoulou 1998; Svenonius 2003 specifically in connection with pro). Quite independently of these theoretical considerations, however, there are also at least two empirical issues that raise serious doubts about the viability of
Marc Richards and Theresa Biberauer
a pro-based analysis. The first relates to structures such as those illustrated in (14) and (15). As indicated, it is impossible for Expl to be realized in structures of this kind, despite the fact that subject-raising has clearly failed to take place (though raising of the dative DP (quirky subject) clearly does take place in (15), an issue which we discuss in Note 18). Adherents of pro-based accounts therefore postulate a null expletive in contexts of this sort, arguing that pro is always licensed in the relevant cases. The problem with this proposal is, however, that the definiteness restrictions that usually apply to subjects in expletive structures do not appear to apply in structures of this type (cf. (14) above; see also Manzini & Savoia 1997 and Alexiadou & Anagnostopoulou 1998). The second empirical problem involves alternations of the kind illustrated in (16) and (17) below: (16) a.
. . . dat (er) gisteren een jongen kwam. that (Expl) yesterday a boy came ‘. . . that a boy came yesterday.’ b. . . . dat (er) gedanst wordt. that (Expl) danced becomes ‘. . . that there is dancing.’
(17) . . . dat (daar) gedans word. that (Expl) danced becomes ‘. . . that there is dancing.’
(Dutch)
(Afrikaans)
As (16) and (17) clearly show, the occurrence of an overt Expl is optional in both Dutch and Afrikaans (cf. Bennis 1986 and Biberauer 2003), a state of affairs which also previously obtained at different stages in the history of the other members of the Germanic family (see Section 5). For adherents of procentred theories, this is particularly problematic since theories of this type have no way of accounting for how overt and covert Expl can be licensed in identical contexts.14 Note that Travis’s get-out clause in respect of the Avoid Pronoun Principle does not rescue this type of theory in cases such as (16) and (17), as the overt realization of Expl can, per definition, not be related to considerations of emphasis. Evidently, therefore, pro-based accounts involve various serious shortcomings on both the theoretical and the empirical front (cf. Wurmbrand 2004 for further discussion). Another major approach to the problematic data in (12)–(15) assumes that TP is not projected in languages where Expl fails to surface in spec-TP. Van Gelderen (1993) and Haider (1993) put forward various specific arguments, both synchronic and diachronic, in favour of the plausibility of this proposal for Dutch and German respectively, but, to the best of our knowl-
Explaining Expl
edge, this type of proposal has not been made for Icelandic, a language which is frequently said to feature (independent) V-to-T movement (cf. i.a. Roberts 1993; Rohrbacher 1994; Vikner 1995: 138ff., 1997, 2001; Bobaljik & Thráinsson 1998). Even leaving aside the applicability of this analysis to Icelandic, however, it is not clear how a TP-less account can accommodate the Dutch data in (16) or, as we shall see in Section 5 below, the developments that took place during the history of English, a language which, in its Old and Middle stages, shared many of the characteristic properties of German and Dutch (cf. van Kemenade 1987). Furthermore, this type of analysis is also highly problematic for theoretical reasons. Consider, for example, the core minimalist assumption that T has universal semantics, being the locus of what Chomsky (BEA: 13; cf. also DbP: 9) designates “true tense”, and also the central role that T, as the locus of an uninterpretable φ-set, is assumed to play in argument licensing. Viewed against this theoretical background, postulating the absence of TP for German and Dutch would seem to be a rather drastic departure from ‘universal design’, one that seems rather under-motivated by the relatively small TP-related divergences that these languages appear to exhibit from those for which the existence of TP has never been questioned. As in the case of pro-based accounts, therefore, TPless accounts appear to face serious problems not only on the empirical front, but also on the theoretical one (once again, cf. Wurmbrand 2004 for further discussion). The third type of proposal attempts to account for phenomena of the kind illustrated in (12)–(15) and additionally for those in (16)–(17) by postulating what is essentially an optional EPP-feature.15 In terms of this proposal, T is therefore assumed, contra Chomsky (MI: 109), to be no different from other functional heads in being optionally associated with an EPP-feature: when this feature is present, it triggers the usual EPP-type phenomena (i.e. subjectraising and, where relevant, expletive-insertion), and when it is absent, these phenomena fail to surface, in the manner exemplified in (12)–(17) above. An immediate problem raised by this type of analysis is the question of predictive power: how can one employ an optional EPP-feature to account for an (allegedly) optional phenomenon without encountering the same sort of circularity that, for example, bedevilled attempts to account for verb-raising in terms of feature strength? Compare the logic in (18) and (19) in this connection (cf. Thráinsson 2003, who makes a similar point regarding verb-raising). (18) Q: When does verb-raising take place? A: When a strong V-feature is present. Q: How do we know that such a feature is present?
Marc Richards and Theresa Biberauer
A: Because verb-raising takes place. (19) Q: When do subject-raising and (overt) expletive insertion fail to take place? A: When an EPP-feature is absent. Q: How do we know that an EPP-feature is absent? A: Because subject-raising and (overt) expletive insertion fail to take place.
Proponents of this approach do, in fact, take this consideration into account and propose that the presence of an EPP-feature on T correlates with interpretive/semantic phenomena such as scope (cf. Wurmbrand 2004) and Diesingtype effects (cf. Diesing 1992). Thus, for example, T will bear an EPP-feature wherever a definite (specific) DP is present in the numeration; and T lacks an EPP-feature wherever the subject is not definite/specific, with the result that the well-known ‘Diesing effects’ emerge. This proposal is entirely in keeping with the DbP/BEA conception of EPP-features in terms of which optional EPPfeatures, in stark contrast to obligatory ones, must necessarily result in “new interpretations” (cf. DbP: 34 and, relatedly, the MI: 99/142, discussion of Full Interpretation (FI) and interface economy; Biberauer & Richards 2004 offers more detailed consideration of the implications of this distinction). Where it encounters particular difficulties, however, is in Expl contexts such as those illustrated in (16)–(17): as noted in connection with pro-based theories above, the overt versus covert realization of Expl does not result in any interpretive differences and, furthermore, the possibility of Diesing effects clearly does not arise in the case of impersonal passives. For structures such as these, then, the advocates of ‘optional EPP’ proposals presumably have to postulate EPPfeature optionality that has no interpretive reflex, i.e. the kind that suffers from the circularity demonstrated in (19) above and that also runs counter to the strict requirement on optional EPP-features (deriving ultimately from FI) that would otherwise appear to provide a plausible motivation for this type of analysis. Clearly, therefore, this proposal also cannot account for the distribution of Expl in an entirely nonstipulative manner. Having surveyed the major proposals that have been put forward to date to account for Expl-distribution in those Germanic languages which appear to exhibit ‘inconsistent’ EPP-behaviour, we conclude that none of these can provide us with a principled account of why Expl would, either consistently or sporadically, fail to surface in spec-TP. In what follows, we propose a new and, we argue, more explanatory analysis.
Explaining Expl
. A new proposal Recall that in terms of our conclusions in Sections 2 and 3 above, Merge-Expl in spec-TP is universally excluded on principled grounds. As such, the question for us is therefore clearly not (20), but, instead, (21): (20) Why does Expl fail to surface in its Merge site in the languages in question? (21) Why do the languages in question lack the Move-Expl property which results in Expl appearing in spec-TP in languages like English?
What we propose is that the way in which T’s universally present (i.e. obligatory) EPP-feature is satisfied holds the key to the answer. More specifically, we propose that certain languages achieve the satisfaction of T’s EPP-feature (henceforth: EPP-satisfaction) by means of vP-raising, and that it is precisely these languages which present the cluster of Expl/subject-raising properties illustrated in (12)–(15). Let us establish how this works. Since Alexiadou and Anagnostopoulou (1998), there has been widespread acceptance of the idea that EPP-satisfaction may be achieved in one of two ways: a. via DP-raising, i.e. raising the contents of spec-vP to spec-TP (cf. English); b. via V-raising, i.e. V(v)-to-T movement (cf. Greek, Italian – the ‘pro-drop’ option). Biberauer (2003) and Biberauer and Richards (2004) propose a third means of EPP-satisfaction, namely: c. via vP-raising, i.e. raising of the entire complement of T to Spec-TP, a mechanism employed by SOV Germanic. Consider the derivation of the structure in (14) above, which is presented in (simplified) tree-form below:16 (22) . . . daß dem Mann ein Buch geschenkt wurde. . . . that the.dat man a.nom book presented became
Marc Richards and Theresa Biberauer
Ü
Ü
The derivation that we assume proceeds as follows: (23) Sample derivation: a. Merge lexical V (geschenkt) and its direct object/theme (ein Buch) b. Merge indirect object/recipient (dem Mann) within lexical domain (see Note 16) c. Merge the passive auxiliary as head of defective vP d. Merge T and raise wurde to T e. Raise T’s vP-complement to spec-TP f. Merge the complementizer (daß) as head of C
For us, therefore, the absence of Expl in structures like (14)/(22) emerges as entirely expected: since EPP-satisfaction is achieved by means of vP-raising to spec-TP, there is no possibility for an expletive to be merged in this position (cf. the expectation set up by the standard, Merge-TP analysis, above). Furthermore, this analysis also enables us to account for the fact that subject-raising is apparently ‘optional’ in German (cf. the discussion of Wurmbrand’s proposal in Section 4.1): in cases like (14), subject-raising fails to take place since (defective/passive) v in this case lacks the optional EPP-feature that would trigger interpretively motivated DP-raising to its (outer) specifier. EPP-satisfaction therefore results in the raising of a vP in which the hierarchical order of the arguments reflects their Merge order. In the case of the raised-subject counterpart of (14)/(22) (. . . daß ein Buch dem Mann geschenkt wurde), we assume that the crucial difference is the presence of an optional EPP-feature on v, giving rise to the difference in interpretation. This optional feature triggers DP-movement to v’s specifier, whereafter merger of T and vP-raising take place as usual. The same analysis carries over to Dutch, which exhibits similar structures (cf. van Gelderen 1993: 149ff.), and also to Icelandic structures like (15) and others that
Explaining Expl
fail to exhibit the expected patterns of raising of the nominative argument (see Note 18 for more on the Icelandic situation). However, we have yet to fully account for the absence of Expl in (14)/(22), as a further possibility remains on our account. Specifically, we must now answer the following question: why, apparently (see below), can Expl not merge into spec-vP and thus raise to spec-TP as part of the moved vP? As a prelude to addressing this question, we must first elaborate on the theoretical and empirical basis for our proposal that vP-raising is a plausible and well motivated EPP-satisfaction mechanism alongside the DP- and V(v)-raising modes that are already well established in the literature. Firstly, let us consider how vP-raising fits into the overall system of EPPsatisfaction proposed by Biberauer & Richards (2004) and alluded to above. In terms of this system, there is parametric variation as to the manner in which T’s EPP-feature is satisfied in different languages. Specifically, languages are claimed to vary as to the source (location) and size of the nominal (‘D’-)category that values T’s φ-set (i.e. the goal in the T-initiated Agree relation that feeds the movement operation into spec-TP). The relevant source and size ‘choices’ made by the modern Germanic languages are given in (24): (24) The crosslinguistic parametrization of EPP-satisfaction Language Source of φ-features (goal) Size of EPP-moved category English, MSc D(P) in outer spec-vP DP (to spec-TP) Italian (pro-drop) φ-features on V-morphology v (to T) German, Icelandic φ-features on V-morphology vP (to spec-TP) Dutch, Afrikaans D(P) in outer spec-vP vP (to spec-TP)
In terms of this system, then, we expect to find not only the two standard cases of EPP-satisfaction via spec-raising and head-raising (instantiated by English and Italian respectively), but additionally also EPP-satisfaction via vP-raising which, in parallel to the standard cases, may be either spec-driven or headdriven (cf. Afrikaans and German respectively). Given current minimalist technology – in particular, Probe-Goal-Agree, internal Merge and piedpiping – the availability of the two species of vP-raising listed in (24) must be viewed as the null assumption: the vP-raising modes could only be excluded by stipulation as they simply amount to ‘Italian + piedpiping’ (head-piedpiping) and ‘English + piedpiping’ (spec-piedpiping).17 The typology of EPP-satisfaction that therefore emerges from (24) is given in (25):
Marc Richards and Theresa Biberauer
(25) Typology of EPP-satisfaction18 Probe [D]-on-Vf – piedpipe vP Head-raising (Italian) + piedpipe vP Head-piedpiping (German, Icelandic)
Probe [D] in outer spec-vP Spec-raising (English, MSc) Spec-piedpiping (Afrikaans, Dutch)
As noted in Biberauer & Richards (2004), this parametrization is further supported empirically since it allows a number of problematic subject-related phenomena to be accounted for. In addition to the ‘unraised subject’ cases that we have already discussed above, it also permits a principled explanation of a case of ‘true’ optionality that surfaces in Modern Spoken Afrikaans (MSA). The relevant phenomenon is illustrated in (26) below: (26) a.
Ek weet dat sy dikwels Chopin gespeel het. I know that she often Chopin played has ‘I know that she has often played Chopin.’ b. Ek weet dat sy het dikwels Chopin gespeel. I know that she has often Chopin played ‘I know that she has often played Chopin.’
As (26) shows, MSA permits apparent ‘embedded V2’ structures alongside the prescriptively correct V-final structures that represent the only permissible embedded-clause structure in standard Afrikaans. Biberauer (2003) shows that the prescriptively barred structures cannot be analysed as genuine V-in-C V2 structures for a number of reasons, including the fact that they obligatorily feature a subject clause-initially and some form of auxiliary (temporal, passive or modal) in second position, and also the fact that they are interpretively equivalent to (26a)-type structures, receiving an embedded-clause interpretation rather than the matrix-clause interpretation that is generally ascribed to genuine embedded V2 clauses (i.e. CP-recursion structures; cf. Iatridou & Kroch 1992) in so-called “limited embedded V2 languages” (cf. Vikner 1995). The structure in (26b) is therefore analysed as a TP in which the subject DP has raised to spec-TP, i.e. English-style DP-raising has taken place. Biberauer & Richards (2004) argue that the optionality in (26a) and (26b) falls out simply as a by-product of the spec-piedpiping mode of EPP-satisfaction, which is able to alternate in a principled, restrictive, yet semantically vacuous manner between piedpiping (i.e. vP-raising) and non-piedpiping (i.e. spec-raising) in the case of T’s EPP-satisfaction due to the formal equivalence of these two operations from the perspective of the grammar (see Biberauer & Richards 2004 for full discussion).19 The importance of the data in (26), then, is that it highlights the
Explaining Expl
plausibility of postulating the kind of EPP-satisfaction system schematized in (24) and (25): it allows a principled and unified analysis of the stable, ‘true’ optionality exhibited in MSA as part of a single grammatical system. With the EPP-satisfaction typology in (25) in place, let us now return to the question with which we are centrally concerned here, namely how the absence of Expl from spec-TP can be accounted for on our Merge vP/Move-Expl analysis given the possibility of merging Expl into spec-vP and then raising it to specTP as part of the moved vP. Consideration of the typology in (25) reveals that the lack of vP-Expl would seem to correlate with the head-piedpiping mode of EPP-satisfaction: as shown in (12)–(15) above, German and Icelandic are the languages which bar the occurrence of Expl in spec-TP, and these are also the languages which meet T’s EPP-requirements by means of head-piedpiping. Why should this correlation hold? Consider (24) again. As indicated there, the goal probed by T in headpiedpiping languages is the same as that targeted by the languages traditionally classified as pro-drop languages, namely Vf ’s interpretable φ-set. Following Alexiadou and Anagnostopoulou (1998), we assume that the crucial property facilitating Vf ’s status as goal is the presence of rich (‘nominal’) inflection in the verbal paradigm: what is required is what Alexiadou & Anagnostopoulou describe as “verbal agreement morphology with the categorial status of a pronominal element”, i.e. agreement morphemes which we can assume to be stored as independent lexical items in the lexicon (cf. Bobaljik & Thráinsson’s 1998 “separate agreement morpheme”). This leads us to the following (diachronic) prediction: (27) Loss of verbal agreement morphology will be a sufficient condition for a language to shift from being a ‘Probe D-on-Vf ’ language to a ‘Probe D in spec-vP’ language, i.e. to shift from being a head-piedpiper/-raiser to being a spec-piedpiper/-raiser. This is because the loss of agreement morphemes from the verbal paradigm will force the language in question to look elsewhere to find an appropriate goal category to satisfy T’s morphological requirements.
In other words, loss of suitably rich verbal inflection will lead to the requirement that spec-vP be filled in order to supply the necessary nominal goal feature(s). In the absence of an argument (raised or otherwise) in spec-vP, this requirement must be fulfilled by means of Expl. In the case of headpiedpiping/-raising languages, the use of Expl is rendered unnecessary (but not impossible – see below and Sections 6–7) by virtue of the verb’s nominal agreement morphology which guarantees a φ-bearing goal for T. The loss of
Marc Richards and Theresa Biberauer
unambiguous verbal agreement morphology is therefore predicted to coincide with the emergence (and resulting obligatoriness) of Expl. We can therefore characterize Expl as follows: (28) Expl is a (last-resort) strategy for supplying vP with the nominal feature (φ-set) necessary for feeding spec-piedpiping/-raising into spec-TP and, thereby, the satisfaction of T’s EPP-feature.
With (28) in place, we now have an answer to the question that has been the main focus of this section: Expl’s unavailability as an element which can be merged in spec-vP in head-piedpiping (and head-raising) languages is a diachronic contingency, a consequence of the fact that these languages have, throughout their history, retained the kind of D-bearing agreement morphology that T can successfully probe in order to establish an Agree (φ-valuing) relation on the back of which an internal Merge operation can then take place to satisfy T’s EPP-feature. Consideration of the modern Germanic languages indicates that the characterization of Expl given in (28) is synchronically borne out: all the inflectionally impoverished languages (English, the MSc languages and, with the exception of structures like (17) to which we return in Section 6, Afrikaans) consistently require Expl to surface in spec-TP, and Richards & Biberauer (2004) furthermore show that Faroese tað also fits the predicted pattern. In other words, Expl appears in all those languages in which the diachronic rationale for the introduction of Expl obtains. It should be noted, however, that the implication set up by (27)/(28) is one-way only, i.e. from impoverished verbal inflection to the presence of Expl, and not from the presence of Expl to impoverished verbal inflection. That is, we predict only that DP-piedpipers/-raisers must have Expl, not that headpiedpipers/-raisers must not (rather, they simply may not – as stated in (27), loss of verbal inflection is simply sufficient, not necessary, for the emergence of Expl). The synchronic possibility thus remains of a rich agreement system that nevertheless possesses a vP-Expl (redundantly for the purposes of (28)). In Sections 6 and 7 we will see that German in fact instantiates such a language (i.e. it does possess a vP-Expl after all, namely da). Before that, in the following section, we consider the diachronic side of our theory and show that the aforesaid implication yielded by the prediction in (27) appears to be well supported by the historical facts surrounding the emergence of Expl in Germanic.
Explaining Expl
. Diachronic support: The emergence of Expl in the history of Germanic We established in the preceding section that the loss of ‘rich’ verbal inflection will force T to ‘look elsewhere’ to satisfy its morphological requirements; this, in turn, will force spec-vP to be consistently filled. Two indicators of a change in the importance of the contents of spec-vP will therefore be: a. thematic subjects exhibiting raising-to-subject behaviour in unaccusative and passive contexts; and b. the appearance of Expl in spec-TP, or rather, in our terms, in the linear position immediately following Vf in V2 clauses and overt complementizers in embedded clauses. As this paper is primarily concerned with Expl, we will not consider (a) here, but see Biberauer 2004 for some discussion of (a) and also more detailed discussion of (b) than the brief overview that we offer here. Since Early Germanic was inflectionally rich, the attestation of structures featuring unraised subjects and lacking Expl is unsurprising, not only on our analysis, but also on more traditional analyses (cf. Section 4.1 above). What is significant in the present context, however, is what subsequently transpired in the history of the Germanic family and, particularly, the sequence in which the relevant developments unfolded. Let us briefly consider these, starting with the languages that we have classified as head-piedpipers and then moving on to those that we would classify as spec-piedpipers and spec-raisers, respectively. . The diachrony of Expl in German and Icelandic As noted in the detailed descriptions of the diachrony of expletives in German by Lenerz (1985) and Abraham (1993), expletives are barely attested in Old High German (ca. 750–1000). The same was true of Old Norse (cf. Sigurðsson 1993). Expl subsequently became a fixture in clause-initial position (i.e. specCP) in both German and Icelandic, a development that coincides historically with the regularization of V2 structures as the stylistically neutral form for declarative clauses. This is therefore clearly not a subject-related development, but rather one which can be viewed as a ‘last-resort’ strategy for supplying CP with an appropriate specifier in the absence of an appropriate topical XP (cf. the discussion of það as an ‘expletive topic’ in Section 1 above; Breckenridge 1975; Lenerz 1985; Bennis 1986; Koster 1987; Grewendorf 1989 and Abraham 1993 all propose a parallel characterization of Expl es). As both German and
Marc Richards and Theresa Biberauer
Icelandic retained their rich inflectional morphology throughout their history, it is unsurprising on our analysis that neither of these languages were forced to develop clause-internal Expl: they were able to remain head-piedpiping languages throughout and consequently T remained able to probe D-on-Vf, obviating the need for the development of an obligatory vP-Expl. . The diachrony of Expl in Dutch and Afrikaans To the best of our knowledge, very little systematic work has been done on the diachrony of Expl and how this specifically relates to morphological changes in either of these languages, but the little that is known would seem to indicate that our prediction is once again correct. Thus Burridge (1993: 163ff.) notes that the earliest extant records of Dutch reveal the same large-scale absence of Expl, with Expl only beginning to appear clause-initially in matrix clauses during the 14th century when V2 was regularized; however, it consistently remained absent from inversion structures and V-final embedded clauses. The rise of optional clause-internal Expl (cf. (16) above) was therefore clearly a separate development from the establishment of clause-initial Expl. Burridge’s (1993: 225ff.) observations in connection with “morphological levelling” during the course of Middle Dutch (ca. 1150–1500) and, specifically, to the fact that “Middle Dutch did not possess an entirely dependable subject agreement paradigm on the verb” (ibid.) points to the possibility that this development may have correlated with a loss of morphological richness, as we predict. In the Afrikaans case, morphological considerations once again appear to be key in the regularization of clause-internal Expl: the 17th-century Hollandic variety that formed the Dutch basis of Afrikaans would have been an asymmetrical V2 language with Expl fully established in spec-CP, but only surfacing sporadically in clause-internal position. Morphological impoverishment proceeded rapidly (cf. Ponelis 1993: 394ff.) and, by the time Afrikaans was standardized in 1925, clause-internal Expl had clearly become a systematic feature in presentational structures. In impersonal passives, however, it was then and has since remained merely optional, a point to which we return in Section 6 below. . The diachrony of Expl in MSc and English It well known that both the MSc languages and English underwent significant morphological impoverishment during the course of their history (cf. i.a. Falk 1993a, b in connection with Swedish and Fischer et al. 2000: 70ff. in connection
Explaining Expl
with English). What is particularly noteworthy in the present context, however, is how this loss of inflectional morphology affected the Expl system employed by these languages. Falk (ibid.) shows for Swedish that Expl only surfaced sporadically in clause-initial position during the first half of the Early Modern Swedish period (ca. 1500–1700), noting (1993b: 161–162) that she failed to find “any examples with expletive det [our Expl – MDR/MTB] in the inverted subject position of main clauses prior to the loss of agreement [our emphasis – MDR/MTB]”. Furthermore, she (1993b: 164) also observes that “expletive det prior to 1600 is almost exclusively found in existential constructions”, i.e. not in impersonal passives. At this stage, therefore, det was functioning in the same way as German and Icelandic Expl, namely as a dummy (spec-)CP-element. After 1600, however, two things changed: a. number agreement (i.e. the last vestiges of ‘rich’ inflectional morphology) was lost in the mid-17th century (cf. Falk 1993b: 159); and b. expletive det began to appear in impersonal passives and also clauseinternally. Falk (1993b: 166) summarises and accounts for these facts as follows: (29) . . . expletive det in a grammar with governing I [i.e. for Falk, I which can license pro by virtue of the presence of sufficiently rich agreement – MDR/MTB] does not have the function of a subject position marker, but that of a construction marker instead: expletive det is almost exclusively used in the first position of existential constructions. This use is fully compatible with the existence of agreement. After the loss of agreement, we find for the first time expletive det with a purely grammatical function, i.e. it fills the otherwise illicit, empty subject position. [our emphasis – MDR/MTB]
For us, these developments can straightforwardly be accounted for as the consequence of the loss of sufficiently rich verbal agreement which therefore entailed the loss of the D-morpheme on Vf that T previously targeted to satisfy its morphological requirements. In the absence of this D-morpheme, it became necessary to ensure that spec-vP was always filled and, consequently, det began to be merged in this position. Strikingly similar developments took place in English. Old English (ca. 800–1100) only featured Expl sporadically, and the introduction and regularization of Expl during Middle English (ca. 1100–1450) proceeded on the same positional basis as that outlined for Dutch, Afrikaans and Swedish above (cf. i.a. Breivik 1983; Allen 1995; Haeberli 1999; Williams 2000). Significantly, clause-internal there only began to appear with any kind of regularity once
Marc Richards and Theresa Biberauer
morphological impoverishment had become significant (i.e. during the middle Middle English period; cf. Williams 2000). . Summary of the diachronic facts As shown above, the rise of Expl in all the Germanic languages appears to have been positionally determined, with Expl consistently appearing in clause-initial position in the first instance. The fact that it may (a) never become established clause-internally (cf. Icelandic) or (b) only become established in this position once its occurrence in clause-initial position is well established (cf. Dutch, MSc and English) strongly suggests that the initial rise of (clause-initial) Expl is not conditioned by subject-related considerations. In our terms, this development can be interpreted as the establishment of a last-resort strategy to supply CP with an appropriate specifier in the absence of appropriate topical material, i.e. this is a strategy which arises quite independently of what happens in the vPphase. As far as vP-Expl is concerned, morphological considerations appear to be crucial, with languages only requiring vP-Expl once their agreement morphology has become sufficiently impoverished, making it impossible for T to value its φ-set on the basis of D-on-Vf. T therefore looks for the relevant Dfeatures elsewhere and it thus becomes crucial for spec-vP to be filled with an appropriate nominal. Where such a nominal is missing – either by virtue of its remaining in situ (cf. presentationals)20 or by virtue of its being entirely absent (cf. impersonal passives) – it therefore becomes essential for spec-vP to contain a last-resort, D-bearing Expl (cf. (28)); hence the rise of clause-internal Expl. Note that the fact that clause-internal Expl did not finally become established ‘overnight’ in any of the Germanic languages is also explicable under our analysis: inflectional systems typically disintegrate gradually and therefore we would initially only expect vP-Expl to surface in structures where Vf lacks appropriate agreement marking. Once this option has become established to some extent, vP-Expl can be expected to appear more generally, both where it is absolutely required and where this is not strictly the case. The possibility of omitting vP-Expl would, however, be expected to remain wherever sufficiently rich morphology is available until the language’s grammatical system has been reanalysed as a spec-piedpiping/spec-raising one. Our proposal therefore enables us to account not only for the initial and ultimate distribution of vP-Expl in Germanic, but also for the erratic distribution that it exhibited across a relatively long time span during intermediate stages in its history. This latter empirical fact has never been amenable to satisfactory explanation on other analyses.
Explaining Expl
In the following section, we consider a specific instance of (apparently) erratic Expl-distribution, namely the delayed and, in some cases, only partial regularization of Expl in impersonal passives. Whilst this staggered emergence remains unexplained on other accounts, we show that it is readily amenable to an analysis in our terms.
. Synchronic optionality: Evidence from Dutch and Afrikaans As noted in Section 4.1, both Dutch and Afrikaans failed to regularize vP-Expl in impersonal passives, with the result that structures of the kind illustrated in (16)–(17) are standardly available in these languages. In the Afrikaans case, this is particularly striking since Expl is certainly not optional in presentational contexts. Compare (30) in this connection: (30) . . . dat *(daar) gister ’n skip gesink het. that (there) yesterday a ship sunk has ‘. . . that a ship sank yesterday.’
The question that we are now faced with is why it should be that impersonal passives are compatible with the absence of an expletive when presentationals evidently require one. In other words, how can T satisfy its morphological requirements in impersonal passives in Afrikaans? And, similarly, how is this possible in Dutch and how was it possible at the relevant point in the history of Swedish? Our proposal is that impersonal passives can survive without an obligatory D-supplying expletive in spec-vP because these structures always contain a category that bears appropriate D-features: the passive participle. Following Baker et al. (1989), we assume that passive morphology expresses an absorbed argument and that, as such, the passive morpheme can be viewed as a D-bearing element (i.e. the kind of element that T can successfully probe and displace into its specifier via vP-piedpiping. This option is therefore available only to the [+piedpiping] grammars in (25), accounting for its absence from specraising languages, such as English; see Biberauer & Richards 2004 for further discussion). Our Expl-analysis therefore offers a principled explanation of both the synchronic and the diachronic distributional peculiarities of vP-Expl, an explanation which has proved elusive on previous approaches. Let us finally consider two further cases of Expl-related optionality that our account is able to explain. The first of these, which surfaces in Dutch
Marc Richards and Theresa Biberauer
impersonal passives, was brought to our attention by Hans Bennis and is illustrated in (31): (31) a.
. . . dat (er) gedanst wordt. that (Expl) danced becomes b. . . . dat *(er) wordt gedanst. that (Expl) becomes danced ‘. . . that there is dancing.’
As the example indicates, er is only optional in impersonal passives featuring a preverbal passive participle; where this participle surfaces postverbally, the expletive is obligatory. As Bennis himself notes, this phenomenon is inexplicable on existing accounts of Expl-distribution. On our account, however, it receives a straightforward explanation. We have already accounted for the optionality in (31a): er is omissible in this case thanks to the presence of D-features on the passive participle gedanst. In the case of (31b), however, gedanst has clearly failed to undergo vP-raising to Spec-TP: as consideration of the structure in (22) and of the discussion of the optionality exhibited by MSA in respect of structures of the kind illustrated in (26) will reveal, we assume that finite auxiliaries in Germanic are spelt out in T. If this is the case, gedanst in (31b) cannot have raised to Spec-TP, a state of affairs which parallels MSA (26b) where the lexical verb has also clearly failed to undergo raising. Our proposal is therefore that Dutch (31b) parallels MSA (26b) in being a spec-raising structure, i.e. one which employs the non-piedpiping mode of EPP-satisfaction that is optionally available to spec-piedpiping languages as an alternative to piedpiping of vP (cf. the discussion in Section 4.2 and Biberauer & Richards 2004). If this is indeed the case, it is clear why the er-less version of (31b) cannot converge: in this case, the numeration contains only a single lexical item bearing the D-features that T needs to value its φ-set, namely gedanst; the passive participle therefore cannot remain in situ, but has to undergo (vP-)raising to spec-TP in order to satisfy T’s morphological requirements. Another, though rather different optionality phenomenon that our account enables us to explain in a way that is precluded by other analyses is illustrated in (32)–(33): (32) a.
. . . dat *(daar) gister ’n skip that (there) yesterday a ship b. . . . dat (*daar) ’n skip gister that (there) a ship yesterday ‘. . . that a ship sank yesterday.’
gesink sunk gesink sunk
het. has het. has
Explaining Expl
(33) a.
. . . daß (da) gestern ein Schiff that (there) yesterday a.nom ship b. . . . daß (*da) ein Schiff gestern that (there) a.nom ship yesterday ‘. . . that a ship sank yesterday.’
versunken sunk versunken sunk
ist. is ist. is
As (32) shows, Expl is obligatory in Afrikaans wherever raising-to-subject does not occur (cf. Notes 18 and 20); in our terms, Expl is therefore obligatory wherever v’s EPP-added (nonthematic) specifier would otherwise remain unfilled. Significantly, Expl is, however, barred when this kind of raising has occurred (cf. (32b)). On our analysis, this follows from the fact that we view Expl and derived subjects as competitors for the same spec-vP slot (cf. Section 1 above): wherever subject-raising has taken place, Merge-Expl is impossible. These data, which Merge-TP accounts would be unable to explain (as the (b) examples are generable under such analyses), suggest that the real locus for ‘Merge-overMove’ effects is vP, not TP: Expl and raised subjects are in complementary distribution because they compete for the same Merge site, but this site is now spec-vP, not spec-TP.21 Our account, unlike Merge-TP ones, can simply reduce these Afrikaans (and German) facts to the class of Merge-over-Move effects, which is now a vP-internal phenomenon (one which may then simply fall out from the requirement that the vP-subarray be exhausted before T is merged (i.e. before the vP-phase is complete), without the need to appeal to spurious economy principles such as ‘Merge is cheaper than Move’). Furthermore, this account extends to cover the German data in (33): here, da is optionally present in structures with unraised subjects like (33a), but barred from structures in which the subject has raised. Others have characterized da as a “locative” with “modal particle” characteristics (Rosengren 2002: 175ff.) or as a “semi-locative [. . . ] which exists alongside the expletive es” (Koeneman & Neeleman 2001: 228). If we assume, contra these authors, that da is in fact an expletive and that it is, furthermore, a vP-expletive of the kind that may optionally surface in Dutch (cf. (16) above), this distribution becomes explicable along the same lines as that outlined for Afrikaans daar. The three-way typology predicted by our implication from poor verbal agreement to presence of Expl (cf. end of Section 4.2) is thus complete: Icelandic has rich agreement and categorically lacks vP-Expl, German has rich agreement but possesses an optional vP-Expl, and English/MSc lack rich agreement and exhibit an obligatory vP-Expl. The one remaining logical possibility, namely the language type that combines poor verbal agreement with the cate-
Marc Richards and Theresa Biberauer
gorical lack of vP-Expl, is predicted to be absent by this entailment, a prediction which the Germanic languages (past and present) unequivocally confirm.
. Concluding remarks We have shown in the foregoing that careful consideration of a wide range of issues, both technical and empirical, supported by robust evidence from the Germanic languages, both synchronic and diachronic, leads to the inevitable conclusion that Expl, previously analysed as a TP-merged element, must in fact be merged at the vP stage of the derivation. The Merge-Expl property thus falls into line as a property of the domains of external Merge (i.e. of phases, C and v). With this much in place, new light can be shed on the perennial problem of absent TP-expletives in German and Icelandic, rich-agreement languages which, we argue, satisfy T’s EPP via vP-raising to spec-TP, an option that is directly implied by the Alexiadou and Anagnostopoulou (1998) approach to EPP-satisfaction in combination with the operation Piedpipe. In this way, we arrive at a fully explanatory account of the formal distribution – both underlying (basic, Merge) and surface (derived, Move) – of Expl, one that simply deploys the independently available minimalist machinery in a logical and consistent manner. Once viewed in the manner we propose, the precise and central role Expl plays in EPP-satisfaction can truly emerge, which, in turn, provides the reasons for why such an element should exist in the first place (something which had become increasingly obscure under the Probe-Goal system of MI/DbP/BEA). Let us finally return to the asymmetry identified in Section 1, where we still have some unfinished business. With TP-Expl excluded on principled, universal grounds, expletives now fall into two types (vP and CP); for Germanic, the expletives discussed in this paper fit into this typology as follows: (34) Expl typology (Germanic) Language Merge-vP Afrikaans daar Dutch er English there, it Faroese tað German da Icelandic Swedish det
Merge-CP (er)
es, was það
Explaining Expl
Since vP-expletives do, after all, exist, we must now reconsider the apparent empirical hole that we previously schematized in (5), repeated here as (35). (35) *[CP Það klaruðu [TP margar mýs [vP Expl [vP alveg [vP tsubj [VP tV ostinn]]]]]]
If our central proposal is correct and TP-expletives are to be reanalysed as vPones, then the hole represented by (35) should not exist. Instead, we would predict that structures of the kind in (35) should indeed be attested. This is because we attribute the Merge-Expl property to v’s nonthematic, EPP-added specifier: as a nonargument, Expl cannot be merged into the θ-related, external argument specifier of (transitive, nondefective) v (cf. Section 1). As such, vP-Expl should still be able to freely cooccur with a true external argument, giving rise to transitive-expletive constructions (TECs), as in (35), since the external argument and vP-Expl occupy different specifiers. In other words, (35) is the (pre-raising) structure that our analysis assigns to TECs. As we saw in (32)–(33), it is only derived subjects (i.e. internal arguments of passive and unaccusative predicates) that are predicted to be in complementary distribution with Expl, as only these elements target the same specifier of v.22 It is here that our approach makes a different prediction from that of Bowers (2002), for whom Expl is in complementary distribution with all subjects, derived or otherwise (as all occupy his spec-Pr). Bowers’s Merge-vP(/PrP) analysis, unlike ours, would predict that (35) is impossible (and, indeed, that TECs are impossible unless the expletive is a CP-merged one, a view that he defends (2002: 199ff.)). To test these subtly divergent predictions, we simply need a language that possesses both CP- and vP-expletives, as only here could a double-expletive TEC like (35) potentially occur. As the typology in (34) shows, Icelandic is no good in this respect; however, German meets the necessary requirements, as here we find both the CP-Expl es and, as argued at the end of Section 6, the vPExpl da. We therefore predict that structures of the form in (35) should indeed be found in German. This prediction is borne out:23 (36) Es hat da jemand einen Apfel gegessen Expl has Expl someone an.acc apple eaten [from Koeneman & Neeleman 2001: 229]
The typological hole in (35), then, is indeed only apparent, in line with our analysis (and contra that of Bowers). We conclude that it is only by restricting ourselves to the minimal CFC architecture, with multiple specifiers of v, that a
Marc Richards and Theresa Biberauer
full account of Expl-associate alternations can be attained, as only then can we discern the true patterns of complementary distribution within the vP-domain.
Notes * We would like to thank the CGSW 19 audience for their helpful comments and feedback, especially Josef Bayer, Hans Bennis, Jonathan Bobaljik, Dorian Roehrs and Susi Wurmbrand, as well as our two anonymous reviewers. We are particularly grateful to Marcel den Dikken and Christina Tortora for their thorough and encouraging comments and suggestions. The research reported in this paper was supported by an Arts and Humanities Research Board postgraduate award (Richards) and by AHRB project grant AR14458 (Biberauer). . In the following, MP = Chomsky (1995), MI = Chomsky (1998/2000), DbP = Chomsky (1999/2001), BEA = Chomsky (2001/2004). . We are concerned in this paper only with the there-type, ‘pure’ expletive, that which canonically takes a DP associate, and not with the quasi-argumental or CP-associated it-type of expletive. . By “lexical domain”, we mean VP and the thematic [‘inner’] specifier of vP – that is, we assume v with multiple specifiers, replacing earlier Agr projections, following MP: 349ff. Note that the restriction under discussion allows a local implementation of (anti-)‘Mergeover-Move’ effects, cf. MI: 111; Groat (1999: 39). . The two statements that follow, then, set aside the CP-expletives in (2), which, as they do not enter into expletive-associate relations (at least not of the kind that we are interested in here), are arguably a different kind of entity from (there-type) Expl. Still, as Expl is separated from the T-associate relation under current theory (see Introduction), C’s Expl property can no longer be conveniently ignored, a point to which we immediately return. . Essentially, assuming that EPP-features are added to phase heads only once the relevant phase is completed (that is, once the lexical subarray is exhausted), internal Merge is left as the only option for satisfying any such EPP-feature (thus excluding external Merge of Expl). However, this logic fails to go through unless we can independently exclude selection of Expl into v’s subarray; it therefore applies more naturally to T, if anything, than to v, thus supporting our claim (below) that it is T that lacks the Merge-Expl property. . The problem is a familiar one, having arisen in earlier incarnations of minimalist checking theory. Thus Chomsky (MP: 311–312) is forced to make an exception for the checking of T’s strong D-feature by direct Merge of Expl (checking relations otherwise require nontrivial chains) – cf. Groat (1999: 36–37). However, it arguably no longer arises under the revised assumptions of Chomsky (2005a, b), in which the EPP property is reduced to a generalized “edge” (selectional) feature common to internal and external Merge alike. The other problems discussed below still remain, however. . A reviewer notes the possibility that Expl could be a projecting head selecting TP as its complement. Whilst this would serve to render Expl a legitimate probe (as root node), T
Explaining Expl
would remain an illegitimate goal (by (7b)), and the problematic possibility of Expl in specExpl would remain. . We therefore assume that all VP-types, and not just transitive (agentive) ones, are selected by a light verb, in line with (amongst others) Harley (1995), DbP, Bowers (2002), Legate (2003). . Note that our imputing interpretable φ-features to Expl is not in conflict with Expl’s nonreferential, empty semantics – after all, expletive it is standardly assumed to comprise a complete and interpretable φ-set, controlling third-person singular agreement, with no implications for its semantic dummyhood. . A reviewer points out that Nomura (2003) also (independently) argues for a Merge-vP approach to Expl-distribution on the basis of perceived problems with the BEA position, namely that the Agree relation between Expl (in spec-TP) and T involves residual spec-head agreement and is thus an undesirable exception to the rule in (7a). However, we would dispute that this is actually what the problem is. Chomsky makes Expl the probe precisely to avoid a spec-head relation (which is no longer a primitive notion and so would imply movement of Expl from a lower position (cf. (8)), something which Chomsky’s commitment to Merge-TP does not allow him to even consider). It is thus not the case that T probes Expl but rather that Expl probes T, under c-command, just as required by (7a). Indeed, it is precisely because Expl is not the goal for Chomsky (but rather the probe) that the real problems of the BEA account, as described in Section 2 above, emerge. . Ignoring for expository purposes the Agree(v, a man) operation that would take place between steps (10b) and (10d), valuing v’s defective (incomplete) φ-set and leaving DP’s Case unvalued (cf. (7e)). . As has frequently been observed, Locative Inversion structures would seem to constitute one exception to this generalisation (cf. i.a. Bresnan 1994 and Falk 1993a: 269ff.). . On the other hand, raising of the dative DP across the subject clearly has occurred in (15). See Note 18 below for more discussion of this point. . The postulation of ‘competing grammars’ (cf. Kroch 1989), one of the pro-licensing German type and the other of the pro-less English type, is of course a possibility, but this clearly complicates the account and, in cases like Dutch and Afrikaans and also in the case of all the older forms of Germanic which permitted alternations of the type illustrated in (16) and (17), it raises the question of how alternations of this kind can/could have coexisted in apparently stable variation for as long as they do/did. See Richards and Biberauer (2004) for more detailed discussion of the problems associated with postulating ‘competing grammars’ in cases of this kind. . It should be noted that the various researchers who have made proposals of this type assume rather different clausal architectures – Wurmbrand employs the three Core Functional Categories mentioned in Section 1 (C, T, v), while Rosengren (2002) adopts a Rizzi-style ‘split-CP’ (cf. Rizzi 1997) and Mohr (2004) an even more elaborated, ‘cartography’-inspired functional structure (cf. Cardinaletti 2004; Rizzi 2004). The optionality that Rosengren and Mohr postulate in respect of available subject positions therefore relates to the presence versus absence of ‘higher’ subject positions (spec-FinP and spec-RefP, respectively), rather than to the presence or absence of an EPP-feature specifically on T (cf. main text). Since
Marc Richards and Theresa Biberauer
the optionally present higher subject positions that they postulate fulfil the same role as the specifier of Wurmbrand’s EPP-endowed T (i.e. they provide a slot to which a subset of subjects can raise, namely those that clearly have not remained within their Merge domain for whatever reason – scope, Diesing-type considerations, etc.), we treat these proposals as a single type, referring to Wurmbrand’s sparser clause-structure for ease of exposition. . As noted in the main text, (22) represents a simplified illustration of the structure associated with the German sentence in question. Our concern here is simply ease of exposition; however, we follow a long tradition of researchers (including, more recently, McGinnis 1998 and Anagnostopoulou 2003) in assuming that the vP-internal structure of ditransitives is in fact richer than that of their transitive/intransitive counterparts. Topicalisation/focus structures such as (i) (pointed out to us by an anonymous reviewer) do not therefore commit us to an analysis in terms of which the V-bar node in (22) undergoes raising. In the case of (i), the indirect object has in any case undergone short scrambling (object shift to the vP-edge, across noch nie); whether the VP in (22) thus simply topicalizes as a remnant, or whether that VP does not even include the base position of the indirect object in the first place (as would be the case if the latter is generated in the specifier of an applicative head, in line with the aforesaid researchers), is irrelevant for our present purposes. (i)
Ein Buch geschenkt wurde dem Mann noch nie. A.nom book presented became the.dat man still not ‘A book, the man has never been given.’
. Note, therefore, that the widely discussed ban on comp-to-spec raising (cf. i.a. Pesetsky & Torrego 2001; Abels 2003; Boškovi´c 2003; Collins 2003 and Kayne 2003) does not, on current minimalist assumptions, rule out the possibility of vP-raising: violations of the general ban on comp-to-spec movement are expected to be possible wherever such movement piggybacks on a licit Agree relation. Given Agree(T, v[D]), all that is required is an EPP-feature on the probe (T) for its complement to be piedpiped along with the targeted goal (D-on-v) into the specifier of its selecting head. . The place of Icelandic in this scheme (given as a +piedpiping, head-seeking language in (25)) should be qualified. As a reviewer points out, the dative DP in (15) is a quirky subject (cf. Zaenen, Maling, & Thráinsson 1985) and thus standardly assumed to undergo raising to spec-TP (like any other raised subject). Thus even though the nominative is in situ in (15), raising of the dative subject across the verb does indeed occur here, hence its preverbal position. On our analysis, T’s EPP is satisfied by vP-raising. The domain of ‘traditional’ EPP effects such as raising to subject (i.e. A-movement with passive, unaccusative and raising predicates) is therefore shifted from TP to vP. Thus we assume that the dative DP in (15) raises within vP to the vP-edge position (spec-vP), satisfying EPP on v rather than T (note that v, as a phase head, is a more natural locus for optional EPP-features than is T, cf. MI: 109ff.). See (32) and (33) below for discussion of the equivalent structures in German (which we take to be a head-piedpiper like Icelandic, cf. (25)), where we also argue that the domain of Merge-over-Move effects is likewise shifted from TP to vP on our analysis. That vP-raising in Icelandic does not yield head-final orders of the kind in German (22) or Afrikaans (26a) is simply due to Icelandic being a symmetrical V2 language (“general embedded verb-second” in Vikner’s 1995 terms): the embedded clause in (22) would be a V2
Explaining Expl
clause in Icelandic, thus undoing the effects of vP-raising at the surface. The lack of verbfinal orders in truly non-V2 clauses in Icelandic (i.e. indirect questions and certain adverbial and relative clauses (cf. Thráinsson 2003) and nonfinite complements of control verbs, etc.) is, we suggest, the result of the Icelandic equivalent of the optionality that we outline below for Afrikaans. That is, the [+piedpiping] languages will, under certain conditions (see Biberauer & Richards 2004), exhibit systematic optionality in the ‘size’ dimension, allowing just the material in spec-vP to be moved in Afrikaans (cf. (26)) and just the head of vP to be moved (to T) in Icelandic. For reasons that are not yet fully clear to us (and which we therefore leave for future research), this [–piedpiping] option is conditioned by V2 (V-toC), obtaining only in non-V2 clauses (hence the resultant independent V-to-T movement is detectable only in non-V2 clauses). . It should be noted that questions of ‘closeness’ (locality, minimality) do not arise here since the principled alternation that we propose rests on differences in the amount of material that the goal (which remains constant) piedpipes, and not on the location of different goals. Thus vP-raising can alternate with DP-raising as the goal is identical in both cases (viz. D(P) in spec-vP), regardless of whether vP- or just DP-raising then takes place. . The ‘nominal in situ’ characterisation of presentationals/existentials offered in the main text is intended to capture the observation that structures of this kind (cf. (i)) differ from ordinary declaratives such as that illustrated in (ii) in that they contain a nominal which has remained in its Merge position rather then undergoing raising-to-subject: (i)
There arose an immense tumult in the foyer
(ii) An immense tumult arose in the foyer In terms of our analysis, an immense tumult may remain in situ in (i) because expletive there is merged in spec-vP, thereby precluding the raising-to-subject operation that usually takes place in ordinary declaratives (cf. (ii)). As stated in Note 18, our proposal implies that vP and not TP is the domain in which raising-to-subject operations take place. See the discussion of (32)–(33) in Section 6 for further consideration of this matter. . Note that the same is not true for languages like Icelandic because Expl in this case is a CP-element which is merged during the CP-phase (cf. Section 1). . Expl is also predicted to compete with Object Shift/Scrambling, assuming that these operations target this same specifier of vP, too. This prediction would also seem correct – see Richards (2004) for discussion, where a more detailed extension of the vP-Expl analysis to TECs is also undertaken. . A reviewer asks whether our analysis has anything to say on the matter of why ‘pure’ (vP-) expletives, like German da and English there, are often morphological locatives, whereas CP-expletives are not (and, instead, tend to be third-person pronominals). In fact, the vP-analysis is able to shed considerable new light on this question. Because Expl is now merged in spec-vP, the T-associate Agree relation would be blocked by a pronominal expletive (e.g. it, es, það) since these elements, as pronouns, have a full set of φ-features (person and number; cf. Note 9). In order for T-associate Agree to obtain across Expl (and thus for the associate DP’s Case feature to be valued), the intervening expletive must therefore be ‘φ-transparent’. As argued in Section 3, there-type Expl is indeed φ-defective (it has
Marc Richards and Theresa Biberauer only a [Person] feature, most plausibly default third person; cf. the φ-incomplete [uPerson] probe of MI/DbP/BEA). We may assume, then, that locative elements are preferred as vP-expletives since they lack (full) φ-sets and therefore, by (7e), do not intervene for Tassociate Agree. This purely formal explanation for the differences in morphological form of CP- and vP-expletives would not be possible on the standard approaches that treat the latter as TP-merged elements, since Expl in spec-TP would not intervene for Agree(T, associate).
References Abels, K. (2003). Successive Cyclicity, Anti-locality, and Adposition Stranding. PhD Dissertation, University of Connecticut. Abraham, W. (1993). Null subjects in the history of German: from IP to CP. Lingua, 89, 117–142. Alexiadou, A. & E. Anagnostopoulou (1998). Parametrizing Agr: Word order, V-movement and EPP-checking. Natural Language and Linguistic Theory, 16, 491–539. Allen, C. (1995). Case Marking and Reanalysis: Grammatical relations from old to early modern English. Oxford: Oxford University Press. Anagnostopoulou, E. (2003). The Syntax of Ditransitives: Evidence from clitics. Berlin: Mouton de Gruyter. Baker, M., K. Johnson, & I. Roberts (1989). Passive arguments raised. Linguistic Inquiry, 20, 219–251. Bennis, H. (1986). Gaps and Dummies. Dordrecht: Foris. Biberauer, T. (2003). Verb Second (V2) in Afrikaans: A minimalist investigation of word order variation. PhD Dissertation, University of Cambridge. Biberauer, T. (2004). Reconsidering the EPP and Spec-TP in Germanic. Cambridge Occasional Papers in Linguistics (COPiL) 1, 15–40. Biberauer, T. & M. Richards (2004). Keeping movement motivated: Optionality without the tears. To appear in Cambridge Occasional Papers in Linguistics (COPiL) 2. Bobaljik, J. (2002). A-chains at the PF-interface: Copies and ‘covert’ movement. Natural Language and Linguistic Theory, 20, 197–267. Bobaljik, J. & H. Thráinsson (1998). Two heads aren’t always better than one. Syntax, 1, 37–71. Boškovi´c, Z. (2003). On the locality of left branch extraction. Unpublished Ms., University of Connecticut. Bowers, J. (2002). Transitivity. Linguistic Inquiry, 33, 183–224. Breckenridge, J. (1975). The post-cyclicity of es-insertion in German. Papers from the 11th Regional Meeting of the Chicago Linguistic Society (CLS 11), 81–91. Breivik, L. (1990). Existential ‘there’: A synchronic and diachronic study. Oslo: Novus Press. Bresnan, J. (1994). Locative inversion and the architecture of Universal Grammar. Language, 70, 72–131. Burridge, K. (1983). Syntactic Change in Germanic: Aspects of language change in Germanic with particular reference to Middle Dutch. Amsterdam: John Benjamins.
Explaining Expl
Cardinaletti, A. (1990). Impersonal Constructions and Sentential Arguments in German. Padua: Unipress. Cardinaletti, A. (2004). Toward a cartography of subject positions. In L. Rizzi (Ed.), The Structure of CP and IP: The Cartography of Syntactic Structures (pp. 115–165). Oxford: Oxford University Press. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1982). Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, MA: The MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Chomsky, N. (1998/2000). Minimalist inquiries: The framework. In R. Martin, D. Michaels, & J. Uriagereka (Eds.), Step by Step (pp. 89–156). Cambridge, MA: The MIT Press. Chomsky, N. (1999/2001). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale: A Life in Language (pp. 1–52). Cambridge, MA: The MIT Press. Chomsky, N. (2001/2004). Beyond explanatory adequacy. In A. Belletti (Ed.), Structures and Beyond: The cartography of syntactic structures, Vol. 3 (pp. 104–131). Oxford: Oxford University Press. Chomsky, N. (2005a). Three factors in language design. Linguistic Inquiry, 36, 1–22. Chomsky, N. (2005b). On phases. Unpublished Ms., MIT. Collins, C. (2003). The Distribution of Particles in Quotative Inversion. Unpublished Ms., Cornell University. Diesing, M. (1992). Indefinites. Cambridge, MA: The MIT Press. den Dikken, M. (1995). Binding, expletives and levels. Linguistic Inquiry, 26, 347–354. Falk, C. (1993a). Non-referential Subjects in the History of Swedish. PhD Dissertation, University of Lund. Falk, C. (1993b). Non-referential subjects and agreement in the history of Swedish. Lingua, 89, 143–180. Felser, C. (2003). Wh-copying, phases, and successive cyclicity. Lingua, 114, 543–574. Fischer, O., A. van Kemenade, W. Koopman, & W. van der Wurff (Eds.). (2000). The Syntax of Early English. Cambridge: Cambridge University Press. Frampton, J. & S. Gutmann (2000). Agreement is feature sharing. Unpublished Ms., Northeastern University. van Gelderen, E. (1993). The Rise of Functional Categories. Amsterdam: John Benjamins. Grewendorf, G. (1989). Small pro in German. In G. Grewendorf & W. Sternefeld (Eds.), Scrambling and Barriers (pp. 294–315). Amsterdam: John Benjamins. Groat, E. (1995). English expletives: A minimalist approach. Linguistic Inquiry, 26, 354–365. Groat, E. (1999). Raising the case of expletives. In S. Epstein & N. Hornstein (Eds.), Working Minimalism (pp. 27–43). Cambridge, MA: The MIT Press. Haeberli, E. (1999). Features, Categories and the Syntax of A-positions. Synchronic and diachronic variation in the Germanic languages. PhD Dissertation, University of Geneva. Haider, H. (1993). Deutsche Syntax Generativ. Tübingen: Gunter Narr. Harley, H. (1995). Subjects, Events, and Licensing. PhD Dissertation, MIT. Hazout, I. (2004). The syntax of existential constructions. Linguistic Inquiry, 35, 393–430. Hoekstra, T. & R. Mulder (1990). Unergatives as copular verbs: Locational and existential predication. The Linguistic Review, 7, 1–79.
Marc Richards and Theresa Biberauer
Holmberg, A. (2000). Scandinavian stylistic fronting: How any category can become an expletive. Linguistic Inquiry, 31, 445–483. Holmberg, A & C. Platzack (1995). The Role of Inflection in Scandinavian Syntax. New York, NY: Oxford University Press. Iatridou, S. & A. Kroch (1992). The licensing of CP-recursion and its relevance to the Germanic verb-second phenomenon. Working Papers in Scandinavian Syntax, 50, 1–24. Kayne, R. (2003). Antisymmetry and Japanese. Unpublished Ms., New York University. van Kemenade, A. (1987). Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Koeneman, O. & A. Neeleman (2001). Predication, verb movement and the distribution of expletives. Lingua, 111, 189–233. Koster, J. (1987). Domains and dynasties: The radical autonomy of syntax. Dordrecht: Foris. Kroch, A. (1989). Reflexes of grammar in patterns of language change. Language Variation and Change, 1, 199–244. Lasnik, H. (1995). Case and expletives revisited: On greed and other human failings. Linguistic Inquiry, 26, 615–633. Republished in Lasnik, H. (1999). Minimalist Analysis (pp. 74–96). London: Blackwell. Legate, J. A. (2003). Some interface properties of the phase. Linguistic Inquiry, 34, 506–516. Lenerz, J. (1985). Zur Theorie syntaktischen Wandels: das expletive ‘es’ in der Geschichte des Deutschen. In W. Abraham (Ed.), Erklärende Syntax des Deutschen (pp. 99–136). Tübingen: Narr. Manzini, R. & L. Savoia (1997). Null subjects without pro. UCL Working Papers in Linguistics, 9, 301–313. McGinnis, M. (1998). Locality in A-movement. PhD Dissertation, MIT. Mohr, S. (2004). Clausal Architecture and Subject Positions: Impersonal constructions in the Germanic languages. PhD. Dissertation, University of Stuttgart. Moro, A. (1997). The Raising of Predicates. Cambridge: Cambridge University Press. Neeleman, A. (1994). Complex Predicates. PhD Dissertation, University of Utrecht. Nomura, M. (2003). Expletives Move! Proceedings of WECOL [available at http://homepage. mac.com/masashi.nomura/linguistics/work.html]. Pesetsky, D. & E. Torrego (2001). T-to-C movement: causes and consequences. In M. Kenstowicz (Ed.), Ken Hale: A Life in Language (pp. 355–426). Cambridge, MA: The MIT Press. Ponelis, F. (1993). The Development of Afrikaans. Frankfurt: Peter Lang. Richards, M. (2004). Object Shift and Scrambling in North and West Germanic: A case study in symmetrical syntax. PhD Dissertation, University of Cambridge. Richards, M. & T. Biberauer (2004). Optionality, loss of inflection and the rise of expletives: Why Faroese is a VO Afrikaans. Paper presented at the Workshop on Faroese, Reykjavík. Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris. Rizzi, L. (1986). Null objects in Italian and the theory of pro. Linguistic Inquiry, 17, 501–558. Rizzi, L. (1997). On the fine structure of the left periphery. In L. Haegeman (Ed.), Elements of Grammar (pp. 281–337). Dordrecht: Kluwer. Rizzi, L. (2004). Locality and the left periphery. In A. Belletti (Ed.), Structures and Beyond: The cartography of syntactic structures, Vol. 3 (pp. 223–251). Oxford: Oxford University Press.
Explaining Expl
Roberts, I. (1993). Verbs and Diachronic Syntax. Dordrecht: Kluwer. Rohrbacher, B. (1994). The Germanic VO Languages and the Full Paradigm: A theory of V to I raising. PhD Dissertation, University of Massachusetts, Amherst. Rosengren, I. (2002). EPP: A syntactic device in the service of semantics. Studia Linguistica, 56, 145–190. Sabel, J. (2000). Expletives as features. In R. Billerey & B. Lillehaugen (Eds.), WCCFL 19 Proceedings (pp. 411–424). Somerville, MA: Cascadilla Press. Safir, K. (1985). Missing subjects in German. In J. Toman (Ed.), Studies in German Grammar (pp. 193–229). Dordrecht: Foris. Sigurðsson, H. (1989). Verbal Syntax and Case in Icelandic: In a comparative GB approach. Lund: University of Lund Publications. Sigurðsson, H. (1993). Argument-drop in Old Icelandic. Lingua, 89, 247–280. Sternefeld, W. (1985). On case and binding theory. In J. Toman (Ed.), Studies in German Grammar (pp. 231–287). Dordrecht: Foris. Svenonius, P. (2003). Introduction. In P. Svenonius (Ed.), Subjects, Expletives and the EPP (pp. 3–28). Oxford: Oxford University Press. Thráinsson, H. (1979). On Complementation in Icelandic. New York: Garland. Thráinsson, H. (2003). Syntactic variation, historical development, and minimalism. In R. Hendrick (Ed.), Minimalist Syntax (pp. 152–191). Oxford: Blackwell. Travis, L. (1984). Parameters and Effects of Word Order Variation. PhD Dissertation, MIT. Uriagereka, J. (1999). Multiple spell-out. In S. Epstein & N. Hornstein (Eds.), Working Minimalism (pp. 251–282). Cambridge, MA: The MIT Press. Vikner, S. (1995). Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press. Vikner, S. (1997). V0 -to-I0 movement and inflection for person in all tenses. In L. Haegeman (Ed.), The New Comparative Syntax (pp. 189–213). London: Longman. Vikner, S. (2001). Verb Movement Variation in Germanic and Optimality Theory. Habilitationschrift, University of Tübingen. Williams, A. (2000). Null subjects in Middle English existentials. In S. Pintzuk, G. Tsoulas, & A. Warner (Eds.), Diachronic Syntax: Models and mechanisms (pp. 164–187). Oxford: Oxford University Press. Wurmbrand, S. (2004). Licensing case. Unpublished Ms., University of Connecticut, Storrs [available at www.wurmbrand.uconn.edu/research/files/Agree.pdf]. Zaenen, A., J. Maling, & H. Thráinsson (1985). Case and grammatical functions: The Icelandic passive. Natural Language and Linguistic Theory, 3, 441–483. Zwart, J.-W. (1992). Dutch expletives and small clause predicate raising. In K. Broderick (Ed.), NELS 22 Proceedings (pp. 477–491). Amherst: GLSA.
Reflexives in contexts of reduced valency: German vs. Dutch Marika Lekakou
This paper addresses the question of why Dutch middles do not admit the reflexive element zich, whereas German middles require its apparent counterpart sich. Following Steinbach (2002), German sich can be a canonical argument, but it can also function as a valency-reduction morpheme. I show that nonargument sich appears in unergative structures. By contrast, I argue that Dutch zich can only be an argument. What has been analyzed as nonargument zich is a syntactic object in contexts of inherent reflexivity. I relate the difference between zich and sich to the structure of the corresponding paradigms: German lacks a counterpart of zichzelf. The analysis makes correct predictions for other Germanic languages, and concerns, besides middles, anticausatives and inherent reflexives as well.
.
Introduction
The primary aim of this paper is to account for the contrast between Dutch and German middles concerning the presence of a reflexive element.1 The contrast is provided in (1). (1) a.
Dit boek leest (*zich) makkelijk. this book reads refl easily b. Dieses Buch liest (sich) leicht. this book reads refl easily ‘This book reads easily.’
Setting the presence of the reflexive in (1b) aside, one of the main claims defended here is that syntactically German middles behave on a par with their Dutch counterparts. Ackema & Schoorlemmer (1994, 1995) show that middles in Dutch and English are unergative and lack a syntactically active agent.2 On the other hand, in languages like French and Greek, middles are syntactically
Marika Lekakou
indistinguishable from (reflexive) passives, and the implicit agent is syntactically represented, to the effect that a by-phrase is licit, cf. Condoravdi (1989), Tsimpli (1989), Lekakou (2002, 2003), Ackema & Schoorlemmer (2002). There is thus no cross-linguistically consistent way to syntactically define ‘the middle construction’ (Condoravdi 1989). The middle is best thought of as a particular interpretation that independently existing structures receive. The question then is, what determines the choice of structure on which the middle interpretation will be parasitic in a given language?3 In Lekakou (2002, 2003, 2004, 2005) I have proposed that the factor responsible for the choice between unergative and unaccusative middles relates to the aspectual paradigm of the aforementioned languages. In this paper, I argue for the significance of the reflexive paradigm as a means of explaining the difference within the unergative middles in (1). Let me briefly discuss how the aspectual paradigm is implicated in the realization of the middle semantics, and why on this approach German is predicted to have unergative middles. I treat personal middles as disposition ascriptions to an internal argument. (For impersonal and adjunct middles, see Lekakou 2005.) Disposition ascriptions are subject-oriented generic sentences. The variation in middles reduces to the variation in the way different languages encode genericity, and in particular to the (un)availability of imperfective morphology which realizes the generic operator (Gen). More in particular, I propose that the implicit agent in middles is interpreted as an inherently generic indefinite, which I dub ONE*. Much like (overt) one, ONE* needs to be licensed by a modal operator. In the case of middles, ONE* is licensed by the generic operator, Gen. I propose a correlation between the status of ONE* and the status of Gen that licenses it: for ONE* to be represented in the syntax, Gen must also be present in the (morpho)syntax.4 The typology of middles on this view is as in (2). The criterion for availability of morphosyntactic Gen is given in (3). (2) A passive structure is employed to convey the middle interpretation iff Gen is encoded in imperfective morphology. Otherwise, the middle is parasitic on an unergative verb. (3) A language encodes Gen in imperfective morphology iff in at least one tense it has two distinct verb forms for generic and nongeneric uses, in other words, iff genericity ⇒ imperfectivity.
Like English and Dutch, German lacks morphosyntactic encoding of genericity. Habitual and episodic sentences employ the same verbal forms, as the following examples illustrate:
Reflexives in contexts of reduced valency
(4) a.
John ging gestern nachmittag zu Fuß zur Schule. John went yesterday afternoon on foot to-the.dat school ‘Yesterday afternoon, John went to school on foot.’ b. Als Jugendlicher ging John zu Fuß zur Schule. as youngster went John on foot to-the.dat school ‘In his youth, John used to go to school on foot.’
(5) a.
John ist gestern nachmittag zu Fuß zur Schule John is yesterday afternoon on foot to-the.dat school gegangen. gone ‘Yesterday afternoon, John went to school on foot.’ b. Als Jugendlicher ist John zu Fuß zur Schule gegangen. as youngster is John on foot to-the.dat school gone ‘In his youth, John used to go to school on foot.’
The prediction made for German is therefore that its middles are syntactically unergative. This is borne out. German middles systematically pattern with unergatives and not unaccusatives, as argued extensively by Cabredo-Hofherr (1997).5 Similarly to unergatives, they select haben ‘have’, cf. (6). Moreover, past participles of middles as prenominal modifiers are impossible, witness (7). Finally, topicalization of the surface subject with the participle, which is only possible with subjects of unaccusative verbs, fails in middles, as can be seen in (8). The syntactic inactivity of the implicit agent is supported by the impossibility of a by-phrase, cf. (9). (6) a.
John ist/*hat zur Schule gegangen. John is/has to-the.dat school gone b. John hat/*ist gesungen. John has/is sung c. Das Buch hat/*ist sich immer gut gelesen. the book has/is refl always well read-part ‘The book has always read well.’
(7) a.
der the b. *das the c. *das the
zerbrochene Stock broken stick gesungene Kind sang child sich gut gefahrene Auto refl well driven car
Marika Lekakou
(8) a.
Ein Stock zerbrochen ist schon einmal. a stick broken has already once b. *Ein Kind gesungen hat schon einmal. a child sang has already once c. *Eine Kurzgeschichte gelesen hat sich schon immer schnell. a short story read has refl always quickly
(9) *Dieses Buch liest sich von irgendwem leicht. this book reads refl by anyone easily
After this brief exposition of the approach to middles taken here, let us return to the contrast in (1). If German and Dutch belong to the same group of languages with respect to middle formation, as suggested by the above, the contrast in (1) is all the more mysterious.6 The presence of sich in structures like middles indicates that the reflexive in German can function as a marker of argument structure manipulation. As Steinbach (2002) discusses at length, sich can be a canonical argument in non-nominative position, and it can also be a ‘nonargument’. On the latter guise, the reflexive marks (the effects of) arity operations, in the sense of Reinhart (2000). Reinhart defines two arity operations, reduction and saturation. Reduction eliminates a theta role from the predicate’s theta grid, and its application to the external argument of a transitive verb results in the anticausative variant. Saturation effects binding of the variable corresponding to the external theta role by an existential or a universal operator. The former derives passives, the latter middles. The proposal I put forward with respect to (1) is that contrary to German sich, Dutch zich cannot serve as a valency-reduction morpheme, but is restricted to always appear in argument positions. A similar claim has been made by Sells et al. (1987); some of their evidence will be reviewed in what follows. This proposal concerns more structures besides middles, and it also generates a new question: why should sich and zich differ in this way? This question is addressed in Section 2, where I provide an explanation that capitalizes on the structure of the reflexive paradigm of German and of Dutch. The approach outlined explains the data in (1) and, as I show in Sections 3 and 4, also makes correct predictions for German and Dutch anticausatives and inherent reflexives. In Section 5, I turn to other Germanic languages, which also validate the proposal made here. Section 6 summarizes.
Reflexives in contexts of reduced valency
. Reflexive paradigms . Sich selbst is not zichzelf, and sich is not zich In this section, I argue that German sich and Dutch zich have different properties, and that, furthermore, this is due to the differing nature of the anaphoric systems to which they belong. In particular, Dutch has a complex anaphor, zichzelf, which German lacks. This has been noted by a number of authors, see Faltz (1985), Sells et al. (1987), Fagan (1992), Safir (1996) among many others. The following examples from the literature illustrate. (10) a.
Hans sah/ haßt sich (selbst). Hans saw/ hates refl SELF b. Jan zag/ haat *zich/zichzelf. Jan saw/ hates/ refl/refl-SELF ‘Jan saw/hates himself.’
(11) a.
Max spricht über sich. Max speaks over refl b. Max praat over *zich/ zichzelf. Max speaks over refl/ refl-SELF ‘Max talks about himself.’
German Dutch
German Dutch
Whereas Dutch has to use the complex anaphor zichzelf with non-inherently reflexive verbs, e.g. ‘see’, German can express the same coreference relation without the addition of SELF. Moreover, when German employs selbst in (10a) above, Faltz reports, the sentence “involves contrast on the object NP, whereas the Dutch sentence [in (10b)] is neutral” Faltz (1985: 130). For Dutch, it is the received wisdom that zich is used with inherent reflexives, and zichzelf for noninherent reflexives (cf. Reinhart & Reuland 1993, Reuland & Reinhart 1995), a view I will follow and elaborate on in this paper. I adopt Steinbach’s (2002) view of selbst as a focus particle and not as part of a complex anaphor sich selbst. German does not possess a (grammaticalized) complex anaphor, contrary to Dutch.7 I will not repeat here the arguments in favour of this treatment of selbst, but I would like to mention another indication that suggests the nonidentity of zichzelf and sich selbst. According to Everaert (1986), although most reflexive idioms in Dutch employ zich, there are some instances of idioms with zichzelf (Everaert 1986: 49), e.g. zichzelf niet meer zijn (‘to no longer be oneself ’) and buiten zichzelf van woede zijn (‘to be besides oneself from anger’). The German counterparts of these expressions feature anything but the alleged complex anaphor: sie ist nicht länger sie selbst/
Marika Lekakou
*sich selbst (‘she is no longer herself ’), sie ist außer sich / *sich selbst vor Wut (‘she is besides herself with anger’). More generally, there seem to exist no idioms with sich selbst. This follows, on the view that selbst is a focus particle, since there are no idioms that contain focus particles.8 It is quite telling that Reuland & Reinhart (1995) in effect grant sich the status of zichzelf : they propose that German sich can occupy two structural positions, either the one corresponding to zich (the determiner of NP), or the one corresponding to -zelf (the head of NP). The two structures correlate with the potential of bearing stress, because only the head of NP can bear stress. It has been observed (Everaert 1986) that zichzelf can, but zich cannot be stressed or topicalized. Reuland and Reinhart (1995) claim that the contrast found in Dutch between zich and zichzelf surfaces in German as the contrast between unstressed and stressed sich respectively, but this cannot be true. As Steinbach (2002) shows, nonargument sich (i.e. sich that appears with inherent reflexives, middles and anticausatives) cannot be stressed or topicalized, and so qualifies as an instance of unstressable sich. We therefore expect to find zich in the corresponding contexts, but this expectation is not met. Out of the aforementioned structures, zich adorns only inherent reflexives. . The organization of reflexive paradigms As suggested by the above, and as will be further demonstrated in subsequent sections, sich and zich are quite different. In this section, I examine the way the reflexive paradigm of German differs from the Dutch one. My aim is to motivate a correlation between the (non)existence of a complex anaphor in the reflexive system and the different status of ‘simplex’ anaphors like zich and sich. In other words, my aim is to show that the structure of the reflexive paradigms in which they belong determines the properties of zich and sich stated in (12):9 (12) a. Sich can be an argument or a marker of valency reduction. b. Zich can only be an argument.
According to (12), zich-sentences in Dutch are syntactically transitive. On the other hand, German sich in some cases contributes a syntactic object, and in others indicates argument structure manipulation. (A more precise characterization of nonargument sich will be given in Section 3.) Before examining the reflexive paradigms themselves, let me summarize the gist of the approach, which attributes the dual status of sich to the unavailability of a complex anaphor in German.
Reflexives in contexts of reduced valency
My claim is that the difference between Dutch zich and German sich reduces to the distinction between a pronominal paradigm that has and a pronominal system that lacks the dimension [+/–Inherently Reflexive], for which I use the shorthand [+/–Inh.Refl.]. The existence of zichzelf in Dutch means that the Dutch paradigm instantiates the dimension [+/–Inh.Refl.]. The lack of a complex anaphor in German means that the system lacks this dimension. The only dimension that exists in German is [+/–Reflexive], which I abbreviate as [+/–Refl.]. In order to relate this to the potential of weak anaphors to function as markers of valency reduction, I make the assumption in (13): (13) If an element has a specification for the feature [inh.refl.], it will be restricted to argument positions.
(13) says that [inh.refl.] can only characterize argument reflexives, and not nonargument reflexives, viz. markers of derived intransitivity. This means that a paradigm in which the dimension [+/–Inh.Refl.] is realized cannot contain any elements that are non-referential, i.e. that can be used as markers of valency reduction. This seems a plausible assumption to make, in view of the nature of inherent reflexivity, which I discuss more extensively in Section 4. I argue that inherent reflexivity, i.e. an obligatory coreference relation, amounts to the lack of alternatives to the value denoted by the anaphor. This may be enforced either lexically, as in the case of verbs like zich schamen, ‘be ashamed’, or contextually, as in the case of verbs like zich wassen, ‘wash oneself ’. Since the notion of inherent reflexivity implicates (the potential for alternatives to) coreference, it seems plausible to associate it with true anaphors (argument reflexives), and not with markers of arity operations, which arguably have no reference. I apply the proposal outlined in Pinker (1984) for the principles governing the acquisition of inflectional paradigms to the way in which reflexive paradigms are structured, relatively to the availability of certain distinctions. Let us start with the simplest pronominal paradigm, the one which lacks a weak reflexive. This is the case of Afrikaans and Frisian, depicted in Table 1. Table 1. Afrikaans pronominal system
1 2 3 1 2 3
Nominative
Accusative
ek jy hy/sy/dit ons julle hulle
my jou hom/haar/dit ons julle hulle
Marika Lekakou
Because there exists no weak reflexive, no elements in the system will be used as markers of derived intransitivity. Personal pronouns in such a paradigm are restricted to argument positions (see Section 5).10 If, however, there exists a (weak) reflexive in the language, it expands the prononimal paradigm by adding a dimension of reflexivity, ‘dragging along’ the personal pronouns. At this point, the weak reflexive can appear in argument positions, as do the elements in adjacent cells, but, moreover, it can function as a nonargument, i.e. as a marker of valency reduction.11 This is the situation in German, as depicted in Table 2. It follows that personal pronouns in this kind of paradigm will have a double life, just as sich does. This is true. As observed by Plank (1993), in middles with first or second person subject, the reflexive as a ‘marker’ agrees with the subject in person (and number):12 (14) Du ziehst dich/ *sich schwieriger an als dein you-nom dress-2sg you-acc/ refl harder part than your Bruder. brother ‘You are harder to dress than your brother.’
The Dutch paradigm is richer due to the existence of zichzelf. Complex reflexives are formed by attaching -zelf to the weak pronouns and to zich. The dimension that zichzelf adds to the paradigm of Dutch is [Inh.Refl.]. First and second person weak pronouns are canonical personal pronouns in nonreflexive contexts; they surface with inherently reflexive predicates, but with noninherently reflexive predicates they require the addition of zelf. Zich thus stands in a paradigmatic relation to zichzelf : the former occurs in contexts of inherent reflexivity (in cases where no alternatives to a coreference relation exist), the latter in contexts of noninherent reflexivity. The relevant parts of the Dutch system are given in Table 3. The crucial difference between German and Dutch, namely the fact that there is no complex anaphor in German, is reflected in the Table 2. German personal and reflexive pronouns
1 2 3 1 2 3
Nominative
Accusative –Refl/+Refl
ich du er wir ihr sie
mich dich ihn/sich uns euch sie/sich
Reflexives in contexts of reduced valency
Table 3. Dutch accusative personal and reflexive pronouns
1 2 3 1 2 3
–Reflexive
+Reflexive: inherent
+Reflexive: noninherent
me je ’m ons jullie ze
me je zich ons je zich
mezelf jezelf zichzelf onszelf jezelf zichzelf
availability in Dutch but not German of a grammaticalized contrast between inherent and noninherent reflexive anaphors. The generalizations of the system (in conjunction with (13)) are the following. If the paradigm lacks a [+/–Refl.] dimension, then the elements in the paradigm will always appear in argument positions. If the paradigm has a [+/–Refl.] dimension but lacks a further speficification for that dimension, namely [+/–Inh.Refl.], then the elements in the paradigm will be able to surface both as arguments and as markers of arity operations. Finally, if the paradigm encodes the distinction [+/–Inh.Refl.], then the distribution of the elements in it will be restricted to argument positions. The approach makes the following predictions: (15) If REFL occurs with middles, then no SELF-anaphor exists in the language. (16) If REFL occurs with middles, then REFL occurs also with anticausatives (more often than not). (17) Personal pronouns can only appear with middles or anticausatives in certain kinds of paradigms, namely the ones that lack the [Inh.Refl.] dimension. In the complement set of these paradigms, personal pronouns in middles and anticausatives are illicit.
Independent support in favour of (15) comes in the form of a seemingly mysterious correlation between the occurrence of reflexives in middles and their ability to support reciprocal readings. The correlation, which is due to Maaike Schoorlemmer (p.c.), is stated in (18). REFL stands for reflexive element. (18) Schoorlemmer’s Generalization If REFL occurs in middles, REFL can be interpreted as a reciprocal.
Dutch middles do not feature zich, and Dutch zich cannot be interpreted as a reciprocal (even though it can take plural antecedents). German sich is obligatory in middles, and can give rise to reciprocal readings. But why should such
Marika Lekakou
a correlation exist? It would seem that what lies behind (18) is something like the following: for an anaphor to be able to occur on the middle interpretation of the predicate, it has to be sufficiently bleached of semantic content, in which case it will also apparently be sufficiently underspecified so as to be compatible with a reciprocal reading. I will now show that the feature with respect to which a middle-/reciprocal-reflexive needs to be underspecified is [Inh.Refl.], just as the approach outlined above predicts. Consider the following observation, due to Safir (1996) (where SIG stands for the ‘weak’ reflexive in Germanic and Mainland Scandinavian): (19) Safir’s Generalization SIG-type atoms may be interpreted as reciprocals only if the SIG-type atom is bound locally by a coargument of a non-inherently reflexive predicate. (Safir 1996: 567)
I read (19) as a necessary and sufficient condition on reciprocal interpretations of simplex reflexives. According to (19), if a language has an anaphor designated for inherent reflexives, it will not be used as a reciprocal. The generalization can be made to follow from the logic of reciprocity and inherent reflexivity. Recall that inherent reflexives are predicates which impose an obligatory coreference relation on their arguments: they are obligatorily reflexive verbs. Reciprocity, on the other hand, by definition requires the existence of alternatives (in other words, the possibility of a non-coreferential relation). It follows that if an anaphor is restricted to contexts of inherent reflexivity, it will be unable to convey a reciprocal interpretation. This is the situation with zich: since it occurs in contexts where no alternatives to a coreference relation are available, zich is incompatible with a reciprocal interpretation.13 The combination of (18) and (19) yields the following: a reflexive will be employed in middles iff it is not restricted to contexts of inherent reflexivity. A [+Inh.Refl.] anaphor necessarily stands in a paradigmatic relation with a [–Inh.Refl.] anaphor. But specification for [+/–Inh.Refl.] can only occur on referential elements (by (13)), hence elements in the paradigm will not be able to serve as valency reduction markers. Therefore, their occurrence in middles will be illicit. This is what happens in Dutch, where the dimension of (non)inherent reflexivity distinguishes between the two anaphors, zich and zichzelf. Because German lacks a complex anaphor and the relevant dimension, sich is not restricted to argument positions and can thus function as a valency reduction marker. Since sich is not limited to contexts of inherent reflexivity, it is compatible with a reciprocal interpretation.
Reflexives in contexts of reduced valency
In the remainder of the paper, I turn to the evidence concerning the predictions in (15)–(17) that the paradigm approach generates. Before doing that, however, I wish to briefly point out what the proposed account may have to say about other languages. For example, Romance reflexive clitics also comply with (18), but there are significant differences between the latter and German sich, which prevent treating them on a par (cf. Reinhart 2003 and Reinhart & Siloni 2003). First, there is a difference in the status of the two reflexives: French se is a clitic, sich is not. Second, se-structures select ‘be’, sich-sentences select ‘have’. Third, as shown by Labelle (1992), se-anticausatives pattern with unaccusatives and ‘plain’ anticausatives behave as unergative. As we will see in Section 3, the converse is true of German: reflexive anticausatives are unergative, ‘plain’ anticausatives are unaccusative. Fourth, there is no passive sich; if German and French patterned together, why do reflexive passives exist only in French but not in German? Quite independently of these syntactic differences, the reflexive paradigm of e.g. French is quite similar to the German one, in that there does not exist a complex anaphor that relates to se. See Zribi-Hertz (1995) for discussion of lui-même (the apparent equivalent of a SELF-anaphor in French) and Safir (1996, 2004) for the differences between the atoms SAME and SELF. A thorough investigation of the reflexive paradigms of Romance and Slavic languages will have to remain a project for future research.
. German sich German sich can undoubtedly function as a canonical argument (recall the examples in (10) and (11)). I will have nothing particular to say about argument sich. My proposal concerns the nonargument reflexive, as it appears in examples like (20), namely with inherent reflexive, anticausative and middle verbs. My proposal for these cases appears in (21): (20) a.
Hans schämt sich. Hans shames refl ‘Hans is ashamed.’ b. Der Stock biegt sich. the stick bends refl ‘The stick bends.’ c. Das Buch liest sich leicht. the book reads refl easily ‘The book reads easily.’
Marika Lekakou
(21) Nonargument sich marks the externalization of an internal thematic role.14
Externalization of an internal role means that a thematic role that is normally realized internally, in the object position, is assigned externally, to the subject position, in violation of Baker’s (1988) UTAH (cf. the rule ‘Externalize Theme’ of Williams 1981). On this view, what nonargument sich does is indicate/instigate violations of the UTAH: the Patient/Theme argument projects directly to the subject position, and does not surface there via movement from the complement-to-V position. In what follows, I show that (21) can successfully generalize over all occurrences of nonargument sich. The evidence in favour of (21) is that inherent reflexives, middles and anticausatives alike feature a base-generated subject, as indicated by their behaviour with respect to the unaccusativity diagnostics. Recall that middles have already been shown to fail these diagnostics. . Anticausatives German has two kinds of anticausatives, those which feature the reflexive, and those which do not. In this respect, German is not unlike Greek (Alexiadou & Anagnostopoulou 2004), French (Labelle 1992) or Italian (Folli 2002) – although Schäfer (2003) shows that the aspectual differences between plain and marked anticausatives that have been observed for these other languages do not exist in German. (22) Der Stock biegt *(sich). the stick bends refl ‘The stick bends.’ (23) Die Vase zerbricht (*sich). the vase breaks refl ‘The vase breaks.’
Reflexive anticausatives are apparently the predominant way of reducing the basic verb’s external argument; ‘plain’ anticausatives are significantly fewer than reflexive anticausatives. Moreover, newly-coined anticausatives in German, i.e. reduced forms of transitive verbs such as digitalisieren, ‘digitalize’, and html-isieren, ‘htmlize’, employ sich, cf. (24). I thank Florian Schäfer (personal communication) for supplying these two observations.
Reflexives in contexts of reduced valency
(24) Die Medizin digitalisiert sich mehr und mehr. the medicine digitalizes refl more and more ‘Medicine is getting more and more digitalized.’
A crucial difference between plain and reflexive anticausatives is that only the former pass the unaccusativity diagnostics. Formally reflexive anticausatives in German are unergative. First, ‘plain’ anticausatives select sein, ‘be’; reflexive anticausatives select haben, ‘have’, see (25). Second, only past participles of ‘plain’ anticausatives can be used as prenominal modifiers; past participles of reflexive anticausatives as prenominal modifiers are illicit, cf. (26). Finally, topicalizing the surface subject with the past participle yields a grammatical sentence if the verb does not feature sich and ungrammaticality if it does, as seen in (27). (25) a.
Die the b. Der the
Vase ist/*hat zerbrochen. vase is/has broken Stock hat/*ist sich gebogen. stick has/is refl bent
(26) a.
zerbrochene Vase broken vase sich gebogene Stock refl bent stick
die the b. *der the
(27) a.
Eine Vase a vase b. *Ein Stock a stick
zerbrochen ist mir schon einmal (in der Küche). broken is me-dat already once in the kitchen gebogen hat sich während des Sturms. bent has refl during the-gen storm
. Inherent reflexives We now turn to inherent reflexives. Contrary to unaccusatives and similarly to unergatives, inherent reflexives select haben and not sein, cf. (28). Moreover, topicalization of the subject with the past participle is impossible, witness (29): (28) Hans hat/*ist sich geschämt. Hans has/is refl shamed ‘Hans was ashamed.’ (29) *Ein Kind geschämt hat sich noch nie. a child shamed has refl yet never
Marika Lekakou
A final piece of evidence that inherent reflexives are unergative is the fact that they can undergo impersonal passivization (Sells et al. 1987, Plank 1993). Since this process is only possible for unergative verbs, inherent reflexives cannot be unaccusative in German, nor can they be syntactically transitive (cf. Steinbach 2002). I will return to this point presently. (30) a.
Es wurde sich geschämt. it was refl shamed ‘People were ashamed.’ b. Hier wird sich täglich gewaschen. here is refl daily washed ‘One washes oneself daily here.’ c. Es wurde sich um die alten Leute gekümmert. it was refl of the old people care-taken ‘One took care of the old people.’ d. Es wurde sich auf den Fußboden gesetzt. it was refl on the floor sat ‘People sat on the floor.’
(Plank 1993)
(Schäfer 2004)
(Schäfer 2004)
The class of formally reflexive verbs that can undergo impersonal passivization is as yet not well defined. Schäfer (2004 and references therein) suggests that it is not only inherent reflexives that are elibigle for impersonal passivization, but medial verbs more generally in the sense of Kemmer (1993), as long as their subject can be interpreted as agentive. According to Kemmer (1993: 58), medial (or middle) verbs are “semantically intermediate between true reflexive events and prototypical one-participant events. [. . . ] Inherent in their meaning is the lack of expectation that the two semantic roles they make reference to will refer to distinct entities”. This is compatible with the notion of inherent reflexivity that will be formulated in Section 4.5. Note that, if the restriction on agentivity accurately describes what constrains German impersonal passivization, we can explain why middles and anticausatives cannot undergo this process, even though I have been arguing that they are unergative: their subjects are never agents (even animate ones).15 Impersonal passivization is an important test for our purposes, because it shows not just that sich-verbs have a base-generated subject, but also that sich cannot be a syntactic object, as Steinbach (2002) argues. If nonargument sich-sentences were transitive, impersonal passivization should be impossible, contrary to fact. As suggested above, whatever restricts the process in German is responsible for the fact that only inherent reflexives, but not middles
Reflexives in contexts of reduced valency
and anticausatives can undergo it.16 On the other hand, a characterization of nonargument sich as in (21) can capture all the data under consideration.
. Dutch zich Let us now turn to Dutch. It is undisputable that the Dutch weak reflexive can occupy a thematic position. The question I am concerned with is whether zich can only occupy a thematic position. In other words, can (the relevant occurrences of) zich be analyzed as a nonargument reflexive, similarly to German sich? The answer to this is negative, and the explanation involves the nature of the paradigm in which zich belongs. If zich can be a valency-reduction morpheme, it is entirely mysterious why it cannot mark the effects of middlesaturation and unaccusative-reduction and only surfaces with inherently reflexive verbs: (31) Hans schaamt zich. Hans shames refl ‘Hans is ashamed.’ (32) *Het metaal buigt zich onder grote druk. the metal bends refl under great pressure ‘The metal bends under great pressure.’ (33) *Dit boek leest zich makkelijk. the book reads refl easily ‘The book reads easily.’
In what follows, I provide evidence that there is no nonargument zich. In particular, I will argue that the purported cases of ‘valency reduction’ zich in fact involve transitive syntax. Both the unaccusative analysis of zich-sentences (Everaert 1986) and the view of zich as a marker of reduced unergative entries (Everaert 2002; Reinhart 2003) will be argued to be wrong. For one thing, zich-verbs fail the unaccusativity diagnostics. Moreover, no zich-verbs can be input to impersonal passivization, which I argue is because the sentences are transitive. . Everaert’s (1986) terminatives In Dutch, addition of the prefixes ver- or over- (or the particle in) to unergative verbs makes the presence of zich obligatory. When the original verb is
Marika Lekakou
transitive, its direct object becomes an optional prepositional phrase, cf. (35). Examples follow, all from Everaert (1986). (34) a.
Hij eet/werkt/schreeuwt. he eats/works/shouts. b. Hij overeet/overwerkt/overschreeuwt zich. he overeats/overworks/overshouts refl ‘He overeats/overworks (himself)/strains his voice.’
(35) a.
Eva at de appel. Eva ate the apple b. Eva overat zich aan de appels/*de appels. Eva overate refl on the apples/the apples ‘Eva gorged herself on the apples.’
As Everaert concedes, ver- and over- are transitivizing prefixes, and it is therefore hardly surprising that an additional argument becomes obligatory. Zich is, in this case, a regular object. But what led Everaert to argue that such cases are unaccusative? Note that zich-verbs select hebben, and not zijn, contrary to what an unaccusative analysis of such verbs predicts. Everaert claims that zich-verbs, for example ‘terminatives’ like (37a) and inherent reflexives like (37b), pattern with unaccusatives in disallowing -er affixation: (36) a.
de werker ‘the worker’ b. de eter ‘the eater’ c. *de ontganer ‘the escaper’
(37) a. *de (zich) overeter ‘the overeater b. *de (zich) vergisser ‘the forgeter’
However, it is possible to attribute the incompatibility of formally reflexive verbs with -er affixation to the i-within-i condition Chomsky (1981: 212) which states that the coindexation in (38) is illicit. Er-affixation based on zich-verbs fails, because it corresponds to an illicit (referential) coindexing, whereby the reflexive and the NP/DP within which it is embedded have the same reference. (39b) is ungrammatical for the same reason:17,18
Reflexives in contexts of reduced valency
(38) *[i ...αi ...] (39) a. *[NP [VP zichi wass] -er]i b. *[DP een [wasser van zichi ]]i a washer of refl
The second argument Everaert brings forward in favour of an unaccusative analysis of zich-verbs is their inability to undergo impersonal passivization. Given the aversion of unaccusatives vis-à-vis passivization, one explanation for this is that zich-verbs are unaccusative: (40) Er wordt gegeten. there was eaten (41) *Er wordt gevallen. there was fallen (42) *Er werd (zich) vergist. there was refl forgotten (43) Er wordt gerend. there was run (44) *Er wordt zich overeten. there was refl overeaten
There is, however, a different explanation for the data above. It is not only unaccusatives, but transitives as well that fail to undergo impersonal passivization. An analysis of zich as occurring in the object position can explain the impossibility of impersonal passives just as well.19 In all, there are no convincing arguments in support of zich as an unaccusativity-inducing (or signaling) morpheme. The data which were offered as evidence of zich-verbs being unaccusative are entirely compatible with the proposal that zich is an argument. On the other hand, if zich were a marker of unergative reduction, we would expect impersonal passivization to apply to at least some cases, as we saw happens in German. In the following subsections, I look at more environments which feature zich, and which have been analyzed as involving derived unaccusative verbs, with the aim of showing that a transitive analysis is preferable. . Everaert’s (1986) inchoatives Consider the following instances of zich, which occur in what looks like the anticausative counterpart of a causative verb:
Marika Lekakou
(45) a.
Hij verspreidde het gerucht. he spread the rumour b. Het gerucht verspreidde zich. The rumour spread refl ‘The rumour spread.’
(46) a.
Zij manifesteerde haar ongenoegen. she manifested her dissatisfaction b. Haar ongenoegen manifesteerde zich. her dissatisfaction manifested refl ‘Her dissatisfaction manifested itself.’
(47) a.
De chemicus verbond de zuurstof met stikstof. the chemist combined the oxygen with nitrogen b. De zuurstof heeft zich met stikstof verbonden. the oxygen has refl with nitrogen combined ‘The oxygen combined with nitrogen.’
The main reason to consider zich as a marker of unaccusative-type reduction in these cases is the apparent existence of a verbal alternation. But it is not clear that the (b) variants do not simply involve the transitive verb: to start with, they do not exhibit syntactic unaccusativity. If the verbs in (b) were unaccusative, they should select zijn, ‘be’, instead of hebben, ‘have’ – a prediction which is not borne out (witness (47b)). Moreover, their past participles should be able to act as prenominal modifiers, which is not the case: (48) *de zich met stikstof verbonden zuurstof the refl with nitrogen combined oxygen (49) *het zich verspreide gerucht the refl spread rumour
The reservations against an unaccusative analysis of the (b) sentences are strengthened, when we consider versions of these sentences with an animate subject: (50) De agenten verspreiden zich. the policemen disperse refl ‘The policemen disperse.’ (51) Zij manifesteerde zich als een diva. she manifestes refl as a diva ‘She presented herself as a diva.’
Reflexives in contexts of reduced valency
(52) Zij heeft zich met Karel verbonden. she has refl with Karel combined ‘She has commited herself to Karel.’
The interpretation that the subject receives in such cases is agentive, which is at odds with an unaccusative analysis of the verbs in question. For instance, consider the following scenario, which involves an unaccusative interpretation for (50): the policemen are fed into a special teleportation device, through which they then get distributed in an area, where a police investigation is conducted. (50) cannot be used in such a situation. Moreover, deliberately is licit, and the subject can control into a purpose clause: (53) Zij heeft zich opzettelijk met Karel verbonden (om het land she has refl deliberately with Karel combined for the country niet uitgezet te worden). not removed.from to become ‘She has deliberately commited herself to Karel (so as to not be expelled from the country).’
(53) arguably involves transitive syntax for the matrix verb, a fact that cannot be accounted for on the unaccusative analysis of zich. Since the structural analysis of these cases would presumably not vary depending on the animacy of the subject, we are forced to conclude that the sentences in (45b)–(47b) also involve transitive syntax. My suggestion for these latter cases, which feature inanimate subjects, is that we are presenting the inanimate subject as though it were animate. The effect of this is that the truth conditions of the sentence become virtually indistinguishable from the truth conditions of a sentence containing an unaccusative verb. Note that the strategy of presenting inanimate entities as though they were animate is quite common in human language, as witnessed by examples like (54):20 (54) The verb wants to move to the second position of the clause.
It is quite telling that researchers have often observed that Dutch zich does not normally surface in canonical causative-anticausative alternations (cf. Sells et al. 1987). According to Fagan (1992), the most productive way of forming anticausatives does not employ the weak reflexive (Fagan 1992: 174). She notes that the reduced forms of newly-coined verbs like finlandiseren ‘finlandize’ and resocialiseren ‘resocialize’ do not tolerate zich. In fact the presence of zich is licit, but only if we construe the sentences as involving an agent subject. Recall that the situation in German is the exact opposite of this: newly-coined
Marika Lekakou
anticausatives, and in fact a large part of anticausatives in general, require the reflexive. (55) a.
Gorbatsjov tracht Roemenië te finlandiseren. Gorbachev tries Rumania to finlandize ‘Gorbachev is trying to finlandize Rumania.’ b. Roemenië finlandiseert (*zich). Rumania finladizes refl ‘Rumania is finlandizing.’
(56) a.
De regering besloot de delinquenten te resocialiseren. teh government decided the delinquents to resocialize ‘The government decided to resocialize the delinquents.’ b. De delinquenten resocialiseren (*zich). the delinquents resocialize refl ‘The delinquents are resocializing.’
This is exactly what is expected on the view defended here: zich does not mark unaccusative-type reduction, and in fact never occurs with true unaccusatives. The cases which feature the weak reflexive are syntactically transitive. . Everaert’s (1986) psych-movement verbs Consider next the following examples, which exemplify another instance of what seems to be a causative–anticausative alternation: (57) a.
Die uitslag ergert haar. that result annoys her b. Zij ergert zich aan die uitslag. she annoys refl on that result ‘She gets annoyed at that result.
(58) a.
Deze gedachte interesseert this thought interests b. Zij interesseert zich voor she interests refl for
haar. her deze gedachte. this thought
(59) a. Dit argument verbaast haar. b. Zij verbaast zich over dit argument. she surprises refl over this argument
Again, there is no evidence for an unaccusative syntax for the (b) sentences. Auxiliary selection and the impossibility of prenominal past participles suggest that the (b) sentences involve a base-generated subject, like the (a) variants.
Reflexives in contexts of reduced valency
Moreover, the impossibility of impersonal passivization again is explained by the fact that zich occurs in the object position, and transitive sentences are not input to impersonal passivization. This class of verbs is different from the one discussed in the previous subsection, as the purported reduced variants already feature an animate subject (and in fact obligatorily so, since it is an experiencer argument). Although at this stage it is unclear to me what the right analysis for these data is, note that similar semantics as the one associated with the (b) examples is encoded in the following sentence, which involves what is normally regarded as a causative verb, maken, ‘make’, and the weak reflexive in object position. An unaccusative analysis of (60) does not seem plausible: (60) Marie maakt zich vrolijk over Jans gedrag. Marie makes refl merry about Jan’s behaviour ‘Mary gets merry about Jan’s behavour.’
Furthermore, it is interesting in this connection to look at languages that lack weak reflexives. In e.g. Afrikaans, bound pronouns are employed in more or less the same cases as the ones which feature zich. The Afrikaans equivalents of the sentences under consideration feature a bound pronoun: (61) Sy verheug haar in die uitslag. she rejoices her in the result ‘She rejoices at the result.’ (62) Sy ontstel haar oor die houding. she upsets her over the attitude ‘She gets upset about the attitude.’ (63) Sy verbaas haar oor die argument. she amazes her about the argument ‘She is amazed with the argument.’
In Section 2, I mentioned the prediction that in a system without simplex anaphors, personal pronouns cannot be markers of valency reduction and are restricted to argument positions. In Section 5, I will provide empirical support for this. Let me mention already that anticausatives do not employ a bound pronoun in Afrikaans. Therefore, there are good reasons to believe that the sentences in (61)–(63) above involve transitive syntax. And since in Afrikaans experiencer-subject sentences employ transitive syntax, it does not seem implausible that the same analysis applies to Dutch as well. The transitive analysis also applies to the class of verbs which Everaert (1986) deemed ‘idioms’, to which we now turn.
Marika Lekakou
. Inherent reflexives To exemplify this class of zich-verbs, consider cases like the following, where the sentence is ungrammatical in the absence of zich: (64) Jan schaamt zich/*Karel. Jan shames refl/Karel ‘Jan is ashamed.’ (65) Zij gedraagt zich/*Karel. She behaves refl/Karel ‘She behaves herself.’ (66) Hij vergist zich/*Karel. he forgets refl/Karel ‘He is mistaken/mistakes Karel.’
I take the semantics associated with inherent reflexivity to involve the obligatory co-indexation of two thematic roles, as schematically indicated below: (67) V(θi , θi )
As may have become obvious in the course of the discussion, I do not assume that cross-linguistically the semantics of inherent reflexivity in (67) is neccessarily mapped onto identical syntax. There exist three possibilities for the syntax of inherent reflexives: (a) both theta roles are assigned in the syntax, (b) only the external theta role is assigned in the syntax, (c) only the internal role projects to syntax. All these possibilities are attested. The first option is taken up in Dutch. Recall that Dutch inherent reflexives cannot undergo impersonal passivization, cf. (68), because zich–sentences are transitive. In Section 5 we will see more cases of transitive syntax for inherent reflexivity. On the other hand, German inherent reflexives are syntactically unergative, and hence can passivize, cf. (69):21 (68) *Er wordt zich geschaamd/vergist/slecht gedragen. there became refl shamed/forgotten/badly behaved. (69) Es wurde sich geschämt/gefürchtet. it was refl shamed/feared
To complete the picture, in French, inherent reflexives behave as syntactically unaccusative, cf. Dobrovie-Sorin (2004) and references therein. It seems to me that we would gain no true insight if we tried to account for the diversity at-
Reflexives in contexts of reduced valency
tested across languages on the assumption that the syntax of inherent reflexives is uniform. To summarize, I have argued that there are no good reasons to extend the view of the dual status of German sich to Dutch zich, which instead always occupies an argument position. I have pointed to the paucity of arguments in favour of an unaccusative analysis of verbs featuring zich, and to the alternative explanation of the data, according to which sentences featuring zich are syntactically transitive. An unergative analysis finds no support; it wrongly predicts that (at least some) zich-verbs can undergo impersonal passivization. . On a generalized notion of ‘inherent reflexivity’ My claim has been that the weak reflexive in Dutch is not a marker of valency reduction, but a canonical argument. If this is so, then the question that arises is, in which cases can we expect it to appear? The question concerns not only the relative distribution of zich and zichzelf, but also the contexts which allow what Everaert deemed ‘inchoative’ and ‘psych-movement’ zich. I cannot fully address this question here, but I believe it is possible to characterize all occurrences of zich that we have encountered as inherently reflexive, as is the traditional view of zich (cf. also Reinhart & Reuland 1993), in which case the remarks in the previous subsection are relevant for the other classes of zichverbs that we have been looking at. In what follows I offer some suggestions as to how this project may be pursued.22 There seems to be a common denominator in all the occurrences of the weak reflexive in Dutch that we have been considering: zich surfaces when no alternatives to the value denoted by it are available, either because the verb is lexically inherently reflexive (like zich schamen), or because the context forces or requires a ‘no-alternatives’ interpretation. On this, I follow König & Siemund (2000: 48), who suggest that “what zelf adds to the meaning of the reflexive is the evoking of alternatives to the value given”. Given the affinity of SELF morphs and focus marking (see König & Siemund 2000), the proposal that the complex reflexive is chosen whenever alternatives to its value are available is not novel, but has been implicit in many accounts in one form or the other (and not just in those accounts which capitalize on SELF as a focus marker.23 Focus presupposes a set of alternatives; the focused element belongs to this set of alternatives. While narrow focus is based on the existence of a set of alternatives, from which an element is selected to the exlusion of the rest, zich expresses the lack of alternatives. The presence of zich thus gives rise to an interpretation according to which there are no alternatives to the coreference
Marika Lekakou
relation established between zich and its antecedent; hence inherent reflexive verbs are obligatorily reflexive. And so it is entirely expected that zich cannot be focused and hence is not stressable, whereas zichzelf can. Consider the verb wassen, which is one of the verbs which can take either zich or zichzelf. In the actual world, since it so happens that adults are not washed by anyone other than themselves (under normal conditions), use of zich is appropriate. Zichzelf is not excluded in this context. It does not, however, correspond to the description of this ‘default’, but rather implies the existence of alternatives, namely that the set of ‘washables’ in this context was not a singleton set. Now imagine we live in a slightly different world, where it is not customary for people to actually wash themselves; instead, each of us is washed by someone else, and we each wash an individual distinct from our own selves every day. In this context, the weak anaphor is illicit and zichzelf is preferred. This recalls the distinction made in König & Siemund (2000) between conventionally other-directed and (conventionally) non-other-directed situations, and also the notion of ‘middle verb’ that Kemmer (1993) develops (see the quote in Section 3.2). It is possible to extend such an explanation to other instances of the weak reflexive, such as the one given in (70). (70) a.
Eva at de appel. Eva ate the apple b. Eva overat zich aan de appels/*de appels. Eva overate refl on the apples/the apples ‘Eva gorged herself on the apples.’
The predicate ‘overeat’ is inherently reflexive, in that it is impossible to ‘overeat someone else’. In other words, it is impossible to eat to such an extent that someone else becomes full. Overeating is something that one can only do to oneself. We predict that use of zichzelf, which encodes the existence of alternatives, instead of zich, which encodes the absence of alternatives, will be degraded. Finally, note that Ackema (1995) discusses causative verbs like oplossen, ‘dissolve’, whose anticausative variant seems to optionally take zich, and concludes that the alternants with the weak reflexive are “the inherently reflexive variants of the transitive causative” (p. 234). This is precisely the stance taken here, according to which there are in fact no zich-anticausatives, only inherently reflexive variants of transitive verbs.
Reflexives in contexts of reduced valency
. Afrikaans and Frisian Lastly, let us turn to languages which lack weak reflexives, namely Afrikaans and Frisian. I repeat the prediction for these cases from Section 2: (71) Personal pronouns can only appear with middles or anticausatives in certain kinds of paradigms, namely the ones that lack the [Inh.Refl.] dimension. In the complement set of these paradigms, personal pronouns in middles and anticausatives are illicit.
(71) corresponds to an assumption standardly (albeit tacitly) made, which follows from the approach presented in Section 2. There it was argued that, because of the structure of the relevant paradigm, personal pronouns always appear in argument positions. It follows that pronouns can never serve as ‘valency reduction markers’, hence they cannot appear with middles or anticausatives. In Afrikaans middles and anticausatives indeed appear ‘unadorned’, cf. (72b) and (72c): (72) a.
Hy het hom misgis. he has him mis-guessed ‘He was mistaken.’ b. Jy terg (*jou) so maklik. you tease you so easily ‘You tease so easily.’ c. Die metaal buig (*hom). the metal bends him ‘The metal bends.’
On the other hand, inherent reflexives as in (72a) feature a bound personal pronoun, hom. What explains the discrepancy in (72) is that inherent reflexives in Afrikaans, as in Dutch, are syntactically transitive. The prediction is that (72a), in virtue of involving a transitive verb, cannot be input to impersonal passivization. The prediction is borne out. (73) *Daar word hom (dikwels) misgis. there becomes him often mistaken
(71) applies similarly to Frisian, which also lacks a simplex reflexive anaphor and whose inherent reflexives also feature a bound personal pronoun (Reuland 2000: 15):
Marika Lekakou
(74) Pier skammet him. Pier shames him ‘Pier is ashamed of himself.’
(71) predicts that middles and anticausatives will not feature the element him, since the bound pronoun cannot be a marker of arity operations. It is clear from the exposition in Abraham (1997) that bound pronouns do not appear in the anticausative variant of causative verbs. We predict that middles too will not employ the bound pronoun. The prediction is borne out (examples provided by Jarich Hoekstra, p.c.): (75) Dit boek lêst maklik. this book reads easily ‘This book reads easily.’ (76) Dizze skuon dûnsje noflik. these shoes dance comfortably ‘Dancing is comfortable in these shoes.’
. Conclusion The starting point of this paper was the question of the (non)occurrence of the reflexive in Dutch and German middles. The difference between German and Dutch middles was related to differences between German sich and Dutch zich. Specifically, I argued that zich always occurs as an argument, but is restricted to contexts of inherent reflexivity. The sentences which feature it are therefore syntactically transitive. On the other hand, German sich can be a canonical anaphor in argument position but also a nonargument; in the latter case, sich does not contribute an accusative object (as in Steinbach 2002), but partakes in an unergative structure and, more precisely, indicates that the syntactic subject is the recipient of an internal thematic role. My proposal was to attribute the dual status of sich to the lack of a complex anaphor in German. More generally, the proposed account relates the potential of weak reflexives to the structure of the paradigm in which they belong; it makes predictions not only for middles but also for anticausatives and inherent reflexives; and it accounts for the fact that in languages like Afrikaans and Frisian, neither middles nor anticausatives can employ a bound pronoun, and that only inherent reflexives do. The data also suggest that a conception of inherent reflexivity as the universal output of unaccusative- or unergative-type reduction is empirically inadequate.
Reflexives in contexts of reduced valency
Notes . An extended version of this paper appears as the fourth chapter of Lekakou (2004a). I am indebted to Ad Neeleman for invaluable comments and suggestions. For their help with the data, I thank Dirk Bury, Nicole Dehé and especially Theresa Biberauer, and for helpful discussion I thank Florian Schäfer, the audience of 19th CGSW and the participants in the Argument Structure Reading Group in Tromsø. The paper has benefited significantly from comments by two reviewers and by the editors of this volume. The usual disclaimers apply. This research was supported by a scholarship from the Lilian Voudouris Foundation. . The behaviour of Germanic middles thus constitutes a challenge to the UTAH (Baker 1988: 46). In Section 3, we will see more instances of UTAH-violations in German. . On this view, the term ‘middle construction’ actually refers to the independently available stucture which conveys the middle interpretation. I will be using such abbreviations for the sake of convenience throughout. . The properties of dispositional generics, as well as the aspectual systems of the languages in question are examined in detail in Lekakou (2005). . Cabredo-Hofherr (1997) tests middles against all the available unaccusativity diagnostics, of which I have included three. She also addresses reservations about the validity of the diagnostics. . Reinhart (2000), Reinhart & Siloni (2003), Marelj (2004) classify German as a ‘Syntax’ language, which in effect means that it will pattern with Greek and Romance with respect to middles. According to Marelj, the presence of the reflexive in middles is a criterial property of a syntactic derivation. The evidence just reviewed, however, suggests that German middles pattern with their Dutch and English counterparts. . This view of German reflexives has been challenged by Reinhart & Siloni (2003), according to whom sich selbst is a complex anaphor. One of their arguments concerns dative sich. I refer the reader to Steinbach (2002), who shows that the predictions made by Reinhart & Reuland (1993), Reuland & Reinhart (1995) especially with respect to dative sich are not borne out. Their second argument is typological, and concerns the well-known correlation between locality of binding and morphological complexity of anaphors. Given the other available evidence from German, however, the correlation cannot be absolute. . One of the reviewers points out sich selbst der Nächste sein (‘charity begins at home’) and sich selbst im Wege stehen (‘to be one’s own worst enemy’) as potential counterexamples. However, the selbst that occurs in these cases can float, much like the Dutch focus zelf and unlike the -zelf that occurs in zichzelf, cf. daß ich selbst zuallererst mir der Nächste bin; er selbst sich im Wege steht. Moreover, selbst can be omitted, at least for some speakers: jeder ist sich der Nächste; sie steht sich immer im Wege. These facts suggest that what we have again is focus selbst, and not the alleged complex anaphor. . The claims advanced in this paper concern the so-called nonargument uses of reflexive elements. In other words, I will not be concerned with occurrences of the reflexives in indisputably argument positions, such as subjects of complements to ECM predicates, or complements to prepositions.
Marika Lekakou . In fact, the paradigm of Afrikaans and Frisian is not quite as described in the main text, because there exists a complex, SELF anaphor, which contrasts with personal pronouns in terms of the [+/–Inh.Refl.] distinction. In other words, the structure of the pronominal system of Afrikaans and Frisian resembles that of Dutch. Either way the result is the same: the elements in both Afrikaans/Frisian and Dutch cannot function as markers. See the discussion in the main text. . The term ‘reflexive’ is thus not entirely appropriate, as the uses it can be put to are not always (semantically) reflexive. . Inherent reflexives pattern the same way, as do anticausatives, although it is extremely difficult to make sure we are actually testing anticausatives and not their transitive, canonical argument reflexive variant. . This observation has already been made by Everaert (1986: 94). Everaert’s explanation also employs the conflicting semantic requirements of inherent reflexity and reciprocity, although his view of the syntax of inherent reflexives is different from the one developed here. . Lidz (2001a) proposes something somewhat similar to (21) for the verbal reflexive in Kannada, which according to him is not associated with a particular meaning, but occurs when a certain mismatch arises between the two tiers of argument structure (Grimshaw 1990). . By contrast, in Dutch, it seems that impersonal passivization applies more freely to any unergative verb, as long as its single argument is animate. The reasons to suspect this are the following. According to Ackema & Schoorlemmer (1995), middles with animate subjects can marginally undergo impersonal passivization in Dutch, as in the following examples (Ackema & Schoorlemmer 1995: 103): (i)
Er wordt hier over het algemeen niet gemakkelijk omgekocht. there is here over the general not easily bribed.
(ii) Er wordt lekker makkelijk gewogen vandaag; het zijn allemaal dikzakken. there is pretty easily weighed today it are all fatsos That Dutch impersonal passivization can target nonagentive animate arguments more generally is also suggested by the possibility of the nonagentive unergative verb rondhangen, ‘to hang around’, to passivize: (iii) Hier wordt te veel rondgehangen. here is too much around.hung The verb rondhangen is considered nonagentive by Everaert (1986) because it cannot be embedded under the causative verb laten; its unergativity is supported by the fact that it selects hebben and its past participle cannot appear as a prenominal modifier. . Note that Steinbach’s claim that nonargument sich is an object in the syntax is equivalent to positing expletive objects, for which there is no independent motivation. See CabredoHofherr (1997) for further evidence that at least middle-sich does not behave as a syntactic object. A different problem for Steinbach is that his analysis cannot account for the fact that Dutch middles do not feature zich.
Reflexives in contexts of reduced valency . Everaert proposes an account of auxiliary selection which captures the facts, albeit through an ad hoc characterization of ‘be’-taking verbs as non-distinct from [+N]. The fact remains: zich-taking verbs, like transitives and unergatives, select ‘have’, and this argues against their purported unaccusativity. . Note that, for principled reasons, omitting zich is not an option. Inherently reflexive verbs like vergissen and overeten are ungrammatical without zich, and so are their -er nominals. Wasser is grammatical, but does not mean ‘self-washer’. . Reinhart (2000), for whom zich marks unergative reduction, attributes the impossibility of impersonal passivization with zich-sentences to the impossibility of applying two arity operations on the same verbal entry. That this explanation cannot be correct is indicated by the data from impersonal passivization of reflexive verbs in German reviewed in Section 3. The evidence from German also runs counter to the slightly different explanation of the Dutch data offered in Reinhart & Siloni (2004). . When the subject is inanimate, deliberately is illicit. But the same holds of (54). Perhaps deliberately requires animate subjects, whereas what we are actually doing in the cases under consideration is impute to the inanimate subject control of the action denoted by the verb. The issue relates to a more general question, concerning the kinds of external arguments that exist. For evidence that we need more than just Agents and Causers (in Reinhart’s 2000 terms), see Doron (2003). . Impersonal passivization of some zich-taking verbs is possible in Heerlen Dutch. Moreover, this dialect employs zich in middles and anticausatives, cf. Cornips & Hulk (1996). Given that the dialect also lacks the standard Dutch reciprocal elkaar, ‘each other’, and uses zich instead, my prediction is that Heerlen Dutch lacks zichzelf. The situation with Heerlen Dutch, however, is extremely complicated due to contact with standard Dutch. There is another dialect, also spoken in the South, Obachsberg Dutch, which confirms my predictions. For dicussion of both these dialects, see Lekakou (2005). For their help with this issue, I thank Hans Bennis, Leonie Cornips and especially Sjef Barbiers. . See Lidz (2001a) for a view entirely compatible with the one defended here. Lidz argues that there are no syntactic differences between zich and zichzelf. The differences are semantic, and relate to the notion of inherent reflexivity. Safir (2004) also discusses the different interpretations conveyed by zich and zichzelf. . See for instance Rooryck & Vanden Wyngaerd (1997). Ter Meulen (1998) provides a formalization of the no-alternatives idea.
References Abraham, W. (1997). Kausativierung und Dekausativierung zwischen dem Friesischen und Deutschen: Sichtbarkeitskriterium als paradigmenbedingung. Us Wurk. Tydskrift foar Frisistyk, 46, 3–22. Ackema, P. (1995). Syntax Below Zero. Doctoral Dissertation, OTS Utrecht. Ackema, P. & M. Schoorlemmer (1994). The middle construction and the syntax-semantics interface. Lingua, 93, 59–90.
Marika Lekakou
Ackema, P. & M. Schoorlemmer (1995). Middles and nonmovement. Linguistic Inquiry, 26, 173–197. Ackema, P. & M. Schoorlemmer (2002). Middles. Ms. University of Nijmegen and University of Utrecht, SynCom Project. Alexiadou, A. & E. Anagnostopoulou (2004). Voice morphology in the causative-inchoative alternation: Evidence for a non-unifed structural analysis of unaccusatives. In A. Alexiadou, E. Anagnostopoulou, & M. Everaert (Eds.), The Unaccusativity Puzzle (pp. 114–136). Oxford: Oxford University Press. Baker, M. (1988). Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press. Cabredo-Hofherr, P. (1997). The German Middle Construction. DEA Dissertation, Université Paris 7. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Condoravdi, C. (1989). The middle: Where semantics and morphology meet. MIT Working Papers in Linguistics, 11, 18–30. Cornips, L. & A. Hulk (1996). Ergative refexives in Heerlen Dutch and French. Studia Linguistica, 50, 1–21. Dobrovie-Sorin, C. (2004). Se-Si type anaphors. SynCom Project. Available at http://www. llf.cnrs.fr/fr/Sorin/SE_SI_Carmen Sorin.pdf Doron, E. (2003). Agency and voice: The semantics of the Semitic templates. Natural Language Semantics, 11, 1–67. Everaert, M. (1986). The Syntax of Reflexivization. Doctoral Dissertation, Rijksuniversiteit Utrecht. Everaert, M. (2002). Zich-reflexives in Dutch and the theta system. Talk given at the Workshop on Argument Structure and Reflexivization, Utrecht 2002. Fagan, S. (1992). The Syntax and Semantics of Middle Constructions. Cambridge: Cambridge University Press. Faltz, L. M. (1985). Reflexivization. A study in Universal Syntax. New York: Garland. Folli, Raffaella (2002). Constructing Telicity in English and Italian. Doctoral Dissertation, University of Oxford. Grimshaw, J. (1990). Argument Structure. Cambridge, MA: The MIT Press. Kemmer, S. (1993). The Middle Voice. Amsterdam: John Benjamins. König, E. & P. Siemund (2000). Intensifers and reflexives: A typological perspective. In Z. Frajzyngier and T. S. Curl (Eds.), Reflexives: Forms and functions (pp. 41–74). Amsterdam: John Benjamins. Labelle, M. (1992). Change of state and valency. Journal of Linguistics, 28, 375–414. Lekakou, M. (2002). Middle semantics and its realization in English and Greek. UCL Working Papers in Linguistics, 14, 399–416. Lekakou, M. (2003). Greek passives on the middle interpretation. To appear in Proceedings of the 6th International Conference on Greek Linguistics. Lekakou, M. (2004). Middles as disposition ascriptions. In C. Meier & M. Weisgerber (Eds.), Proceedings of Sinn und Bedeutung 8 [Konstanzer Arbeitspapiere Linguistik] (pp. 181–196). Konstanz: Universität Konstanz. http://www.ub.uni-konstanz. de/kops/volltexte/2004/1383/.
Reflexives in contexts of reduced valency
Lekakou, M. (2005). In the Middle, Somewhat Elevated. The semantics of middles and its crosslinguistic realization. Doctoral Dissertation, University College London. Lidz, J. (2001a). Condition R. Linguistic Inquiry, 32, 123–140. Lidz, J. (2001b). The argument structure of verbal reflexives. Natural Language and Linguistic Theory, 19, 311–353. Marelj, M. (2004). Middles and Argument Structure across Languages. Doctoral Dissertation, OTS Utrecht. ter Meulen, A. (1998). On the economy of interpretation: Semantic constraints on SEreflexives in Dutch. UiL-OTS Working Paper, Utrecht. Pinker, S. (1984). Language Learnability and Language Development. Cambridge, MA: Harvard University Press. Plank, F. (1993). Peculiarities of passives of reflexives in German. Studies in Language, 17, 135–167. Reinhart, Tanya (2000). The Theta System: Syntactic realization of Verbal Concepts. OTS Working Papers in Linguistics. Reinhart, T. (2003). The Theta system – an overview. Theoretical Linguistics, 3, 229–290. Reinhart, T. & E. Reuland (1993). Reflexivity. Linguistic Inquiry, 24, 657–720. Reinhart, T. & T. Siloni (2003). Thematic arity operations and parametric variations. OTS Working Papers in Linguistics. Reinhart, T. & T. Siloni (2004). Against the unaccusative analysis of reflexives. In E. Anagnostopoulou & M. Everaert (Eds.), Studies on Unaccusativity. The syntax-lexicon interface (pp. 159–180). Oxford: Oxford University Press. Reuland, E. (2000). The fine structure of grammar: Anaphoric relations. In Z. Frajzyngier & T. S. Curl (Eds.), Reflexives: Forms and functions (pp. 1–40). Amsterdam: John Benjamins. Reuland, E. & T. Reinhart (1995). Pronouns, anaphors and case. In H. Haider, S. Olsen, & S. Vikner (Eds.), Studies in Comparative Germanic Syntax (pp. 241–268). Dordrecht: Kluwer. Rooryck, J. & G. Vanden Wyngaerd (1997). The self as other. A minimalist approach to zich and zichzelf in Dutch. Proceedings of NELS, 28. Safir, K. (1996). Semantic Atoms of Anaphora. Natural Language and Linguistic Theory, 14, 545–589. Safir, K. (2004). The Syntax of Anaphora. New York, NY: Oxford University Press. Schäfer, F. (2003). The morphological patterns of anticausatives and their interpretations. Talk given at Kleinwalsertal, July 2003. Schäfer, F. (2004). Passivierte Reflexivkonstruktionen. Ms. University of Stuttgart. Sells, P., A. Zaenen, & D. Zec (1987). Reflexivization variation: Relations between syntax, semantics, and lexical structure. In M. Iida, S. Wechsler, & D. Zec (Eds.), Working Papers in Grammatical Theory and Discourse Structure (pp. 169–238). Stanford: CSLI. Steinbach, M. (2002). Middle Voice. Amsterdam: John Benjamins. Tsimpli, I. (1989). On the properties of the passive affix in Modern Greek. UCL Working Papers in Linguistics, 1, 235–260. Williams, E. (1981). Argument structure and morphology. The Linguistic Review, 1, 81–114. Zribi-Hertz, A. (1995). Emphatic or reflexive? On the endophoric character of French luimême and similar complex pronouns. Journal of Linguistics, 31, 333–374.
Simple tense* Guido Vanden Wyngaerd K.U. Brussel
The Simple Present tense in English only refers to the present moment with eventive verbs if the event has Very Short Duration. I account for this restriction by assuming, first, that the speech time itself has Very Short Duration and, second, that simple aspect in English requires the event to be contained in the speech time interval. In most other languages, this requirement does not exist, so that longer events are also compatible with the Simple Present. States are compatible with the Simple Present because they have point duration, and therefore are invariably contained within the speech time interval. Generic sentences combine eventive verbs with a stative interpretation and should be treated like Individual Level Predicates.
.
Introduction
In this paper, I want to discuss some properties of the Simple Present tense in English, and compare it to its equivalent in a number of other European languages. It turns out that the English tense is rather special in that it can only refer to now with eventive verbs under a fairly limited set of conditions. These conditions can be summarised under the requirement that the event at issue have Very Short Duration. I shall propose to account for this restriction by requiring that only events that fit into the short duration of the speech time and do not extend beyond it are compatible with the Simple Present in English. In other languages the event does not need to be contained in the speech time interval. Next, I shall look at the use of the Simple Present with stative verbs, and argue that states are compatible with the Simple Present because they have point duration, and therefore are invariably contained within the speech time interval. Certain facts relating to durational adjuncts will reveal a contrast between English and other languages that is highly reminiscent of the contrast found with eventive verbs, and I shall show that they can be accounted for by the same assumptions. Finally, I shall devote some comments to the issue
Guido Vanden Wyngaerd
of generic sentences, which combine eventive verbs with a stative interpretation. I shall argue that they should be treated on a par with Individual Level Predicates. First, however, I shall start out by sketching a few background assumptions concerning the nature of tense and aspect in general (Section 2). Section 3 deals with the use of the Simple Present with eventive verbs. States will be dealt with in Section 4, and generic sentences in Section 5.
. Background I adopt a Reichenbach-inspired system for the representation of Tense, which assumes three time points and two relations. (1) time points: S, R, E relations: precedence (_) simultaneity (,)
I deviate from Reichenbach, however, in that I assume two conceptually different levels in the analysis, which I take to coincide with the tense-aspect distinction (see also Comrie 1985: 6).1 I take the relative ordering of S and R to be determined by tense, whereas the relationship between E and R is a matter of aspect. Aspect itself is a concept which involves two rather distinct notions, one being aspectual class (internal aspect), and another grammatical aspect (external or presentational aspect). Schematically, this gives the following picture: (2) Tense Aspect (external) Present: S,R Simple Past: R_S Progressive Future: S_R Perfect
Aspect (internal) state activity accomplishment achievement
The internal aspect (or Aktionsart) concerns the well-known four-way aspectual class distinction familiar from the work of Vendler (1967) and Dowty (1979). As far as external aspect is concerned, we can distinguish simple, progressive and perfect aspect; formally, the latter two are expressed by a combination of an auxiliary and a bound morpheme (BE+ING and HAVE+EN, respectively), whereas the former is characterized by the absence of any grammatical marking.2 Perfect and progressive aspect may combine to yield the perfect progressive. Tense in English is expressed by bound morphemes for Present and Past, and a free morpheme for the Future (WILL). The bound
Simple tense
morphemes may be applied to the free morpheme, yielding a present future and a past future tense.3 The full system is represented in the table below: (3) Pst
Fut
Simple
Perfect
Progressive
Pres
–
–
V she works
Past
+
–
V+ed she worked
(Pres) Future
–
+
WILL V she will work
(Past) Future
+
+
WOULD V she would work
HAVE V+en BE V+ing she has worked she is working HAVE BE+en V+ing she has been working HAVE+ed V+en BE+ed V+ing ashe had worked she was working HAVE+ed BE+en V+ing she had been working WILL HAVE V+en WILL BE V+ing she will have worked she will be working WILL HAVE BE+en V+ing she will have been working WOULD HAVE V+en WOULD BE V+ing she would have worked she would be working WOULD HAVE BE+en V+ing she would have been working
The mapping of these three levels onto syntactic representations is straightforward: Tense is represented at the highest level, i.e. in TP. External aspect occupies a lower projection, call it AspP. Tense and external aspect combine in a completely predictable way in English, as the above table shows. The internal aspect, finally, is a property of the VP (Verkuyl 1972). As far as the interpretation is concerned, I assume that the semantic contribution of Tense is to specify the relationship of the speech time S to a reference time R. The semantics of external aspect is to specify how E relates to R. Take R to be a point on the time line, represented as ------ in (4). I take E itself to be not so much a point on the time axis, but rather a representation of the aspectual class of the event. Thus an activity like read or run has no inherent endpoint and has dynamism. The latter property is represented by the dots °°°° in the representation in (4). The absence of boundaries to the event is represented by the absence of the symbols (which marks the start of an event) and (which marks its end). (4) Progressive: ------------R------------- John was/is/will be reading. °°°°°E°°°°°
Guido Vanden Wyngaerd
Progressive aspect presents an event as ongoing, i.e. it implies that E has a certain extension before and after R, i.e. that E surrounds R; the progressive is furthermore restricted to dynamic events; (4) graphically depicts these properties of extension and dynamism. Note that these two properties are properties of both the internal and the external aspect; put differently, in this case, the external aspect does not add anything that is not already present in the internal aspect of the event. On the other hand, the former does not conflict with the latter either. This is different with the simple (external) aspect, which I shall argue below conflicts with the extension property of the internal aspect, at least as far as most eventive verbs are concerned. Further note that the representation of the relationship between E and R will be identical in the present, past and future tenses: in each case the event E covers an interval extending before and after the reference time. The location of the reference time with respect to the speech time concerns the relationship of R to S, and depends on the particular tense chosen. The perfect aspect does add something to the internal aspect of the event: it presents the event as terminated at the reference time; this is shown by adding a right boundary to the event in (5): (5) Perfect: --------------R------------- John had/has/will have run. °°°°E°°° Perfect Progressive: -------------R------------- John has been running. °°°°--E--°°°°
This is true, whether the sentence be present, past or future; as before, the relationship of the reference time to the speech time depends on the tense chosen. The interpretation of the perfect progressive is somewhat more complex, and involves both the aspects of extension and termination, which gives an interpretation like that of a temporary standstill (the dotted line (- - -) under the timeline in (5) represents stativity). Since neither progressive nor perfect aspect is the topic of this paper, I shall not go into the treatment of them any further here.4
. The Simple Present with non-stative events Having outlined the general background that I assume, I now turn to a discussion of the Simple Present Tense. In this section, I focus on eventive verbs. Comparing English with other languages, one observes an intriguing difference, as illustrated below:
Simple tense
(6) *What do you do? I read a book. (7) Wat doe je? Ik lees een boek. (=(6); Dutch) (8) a.
Sten äter ett apple. (Swedish) (exx. from Giorgi & Pianesi 1997: 153) Sten eats an apple b. Han skriver et brev. (Norwegian) He writes a letter c. Ole kommer. (Danish) Ole comes
(9) -------------S,R-----------------°°°E°°°
These cases involve accomplishments, i.e. telic events with dynamism, as shown in (9), which, by the nature of their internal aspect, extend beyond R. The Simple Present cannot refer to the present time with eventive verbs in English (with a number of exceptions, to which I return), whereas in a language like Dutch cases like (6) are fine. Cross-linguistically, the behavior of English is exceptional, whereas that of Dutch appears to be exemplary for all the languages of the Germanic and Romance domain (Giorgi & Pianesi 1997: 153–154). Dowty (1979: 167ff.), following up on Taylor (1977), suggests that (6a) is ruled out because, on the one hand, activity and accomplishment/achievement verbs denote intervals larger than a moment, and, on the other hand, the time of utterance is always a moment.5 As Giorgi & Pianesi (1997) note, this explanation fails to account for the data in languages other than English, where combinations of the Simple Present Tense and activity/accomplishment/achievement verbs are possible.6 A further problem with both the Taylor/Dowty and Giorgi and Pianesi accounts is that there is a fairly large and fairly heterogeneous class of exceptions to the generalization that the English Simple Present does not combine with eventive verbs. Although generally mentioned, these are rarely accounted for in the literature on the English present tense. These cases include sports commentary, instructions and demonstrations, stories in the historical present, performatives, plot summaries of plays, novels, movies, etc. The following examples are mostly taken from Swan (1995), except where otherwise indicated: (10) commentaries Smith passes to Devaney, Devaney to Barnes, Barnes to Lucas – and Harris intercepts – Harris to Simms, nice ball – and Simms shoots!
Guido Vanden Wyngaerd
(11) stories a. So I open the door, and I look out into the garden, and I see this man. He’s wearing a pink shirt and a policeman’s helmet. ‘Good morning,’ he says. . . b. So I’m sitting there on that bench, reading my newspaper, when this guy walks up to me and starts beating me. Suddenly a man in uniform takes me by the arm and asks me what I’m doing there. (Declerck 1991: 89) (12) stage directions Mallinson enters. The girls immediately pretend to be working hard. William assumes a businesslike air, picks up two folders at random and makes for the door. (Quirk et. al. 1985: 183) (13) instructions, directions and demonstrations a. First I put a lump of butter into a frying pan and light the gas; then while the butter’s melting I break three eggs into a bowl, like this. . . b. OK, let’s go over it again. You wait outside the bank until the manager arrives. Then you radio Louie, who’s waiting around the corner, and he drives round to the front entrance. You and Louie grab the manager. . . c. ‘How do I get to the station?’ ‘You go straight on to the traffic lights, then you turn left. . . (14) performatives a. I promise never to smoke again. b. I agree. c. I pronounce you man and wife. d. We thank you for your recent inquiry.
(Quirk et. al. 1985: 180)
(15) here comes. . . etc. a. There goes our bus. b. Off you go! (16) plot summaries a. In Act I, Hamlet meets the ghost of his father. The ghost tells him . . . b. In The Scarlet Letter Hester meets her lover in the forest. c. In Chapter 3, Max builds a house.
Before going into a discussion of these data, I should like to discuss some previous analyses, both of the cross-linguistic difference mentioned earlier, and of the exceptional English facts in (10)–(16).
Simple tense
Zagona (1992) ties the availability of present-moment readings to the presence of verb movement, noting that the correlation with verb movement holds in Germanic and Romance (though not in Chinese and Korean). She develops an account of this contrast in terms of binding theoretic principles: in languages without V-movement, the governing category for the VP is IP, which does not contain an antecedent, a temporal subject, the latter residing in CP. If, however, the verb is raised to I, then its governing category is CP and it finds an antecedent within CP. The resulting interpretation is one where the event time is anaphoric to the speech time. The correctness of Zagona’s generalization seems to hinge on the status of Portuguese. Thus Zagona appears to assume that it patterns with Spanish and Dutch in allowing (7). Discussing the availability of an imperfective reading with achievement predicates in the present tense, Giorgi & Pianesi (1997: 153) likewise speak of “a difference between English, on the one hand, and all the other languages of the Germanic and Romance domain, on the other.” On the other hand, Schmitt (2001) argues that Portuguese behaves like English in this respect. Oliveira & Lopes (1995) also specifically discuss Portuguese data, mentioning a number of examples of eventive sentences where the Present Tense is used deictically, i.e. to refer to the speech time, in contexts which are similar to those used in English (e.g. sports commentary, performatives, instructions and demonstrations, etc.). They state that “[e]xcept for these two types of contexts [i.e. sports commentary and performatives-GVW], it is difficult to find any form of the Present with a strict deictic value in contemporary European Portuguese” (Oliveira & Lopes 1995: 105). They also note that “[w]ith accomplishment predicates, the Simple Present is rarely used” (1995: 109), and they assign a question mark to the examples illustrating this claim: (17) a.
?O João escreve um romance.
‘João writes a novel.’ b. ?A Patricia pinta o carro. ‘Patricia paints the car.’ ? c. O João viaja até Faro. ‘João travels to Faro.’
I will let these facts stand as they are, and not defend a particular position on the status of Portuguese at this point, leaving it as a topic for further research. The viability of Zagona’s account must likewise await further enquiry. Schmitt (2001) develops an account of the English/non-English contrast in terms of De Swart’s (1998) analysis. The account states that the English present tense selects for states only, whereas in non-English present T selects for ho-
Guido Vanden Wyngaerd
mogeneous predicates (i.e. states and activities). Moreover, the latter languages have a coercion operator which ‘coerces’ an accomplishment into an activity reading. All languages have a coercion operator converting accomplishments into habitual readings. Such an account in terms of selection restrictions imposed by T on its complement does not really explain why the English situation is so rare cross-linguistically, however. It also forces one to assume, rather counterintuitively, that the cases in (10) through (15) are stative. Moreover, the introduction of the powerful mechanism of the coercion operator raises the question why this mechanism is as restricted as it is, e.g. why activities and accomplishments in English cannot be converted into states to satisfy the selection requirement of English T.7 Let us now return to the issue at hand, which is that of the eventive readings of the sentences in (10) through (16). Despite giving an initial impression of great diversity, I will demonstrate that, with the exception of (16), they all share the property of describing events which have Very Short Duration (VSD), in a way that may preliminarily be represented as in (18):8 (18) -------------S,R-----------------VSD-E
This is most obvious in a case like (10), where very short events are being recounted as they happen. As Swan (1995: 446) observes, for longer actions and situations, the Present Progressive is used, even within the register of sports commentary: (19) Oxford are drawing slightly ahead of Cambridge now; they’re rowing with a beautiful rhythm; Cambridge are looking a little disorganized. . .
Whereas (10) involves short events that last no longer than the time it takes to describe them, the events in (19) have a longer duration and require the Present Progressive. This invalidates the claim made by Langacker (1982: 290) and Michaelis (1998: 27) to the effect that, since events like those in (10) are somehow ‘stereotyped’ or ‘scripted’, or represent ‘formulaic occurrences’, the usual requirements are relaxed.9 In the same vein, Schmitt (2001: 438) argues that the events in so-called play-by-play accounts (like (10)) lack internal structure, and that they therefore behave like states. Apart from the fact that the events in question certainly do not feel like states, such accounts are questionable because of a case like (19), which shows that events are not randomly compressible in this way. In the account to be presented below, VSD-events fully preserve their dynamic nature, and eventive present tenses in English are
Simple tense
principally restricted to VSD-events, ruling out longer ones such as those in (19). I shall come back to this issue below, where I discuss the difference between the actual duration of an event and the manner of presentation of events (see (26) and surrounding discussion). Stories like (11) likewise show an alternation between Simple Present and Present Progressive, with the former being used for the short events which follow one another in rapid succession and which therefore carry the story forward. The Present Progressive, on the other hand, is used for the events with longer duration, which do not necessarily succeed the other events, and which are easily considered to constitute the background. What is special about these cases, however, is that the events are presented as taking place at the speech time, but that this is a fictitious speech time, since the events being recounted are of course in reality past events: “the speaker, as it were, forgets all about time and imagines, or recalls, what he is recounting, as vividly as if it were now present before his eyes” (Jespersen 1909, IV, 19 – my emphasis, GVW). The speaker pretends to be in the past and views the events from that fictitious past perspective as taking place now. This use of the present as a so-called historic or narrative present is of course not restricted to English. As in the other cases discussed, the speech time and the event time coincide, but the speech time is an imaginary speech time, which is not the actual speech time. A largely similar story can be told about stage directions (12), which describe events being acted out in the fictional universe of the play. Within that universe, however, the events coincide with the time of utterance. The restriction to VSD events also appears to hold here. The same is true of instructions and demonstrations, as in (13). Demonstrations (as in (13a)) typically involve the description of events as they occur. In the case of instructions (i.e. (13b) and (13c)), we rather appear to be dealing with a fictitious speech time, but which, unlike the cases in (11), is situated in the future. Performatives (14) are another nice case illustrating the isomorphy between the utterance time and the event time: in a very literal way, the progression of the event of pronouncing someone man and wife, for instance, coincides with the utterance time, which itself of course also progresses in time. The examples in (15) likewise involve events which happen simultaneously with the time of utterance. Cases like those in (10), (11), (14), (15) and (13) seem moreover to invalidate the claim by Goldsmith & Woisetschlaeger (1982) that the Simple Present describes structural properties, not observable phenomena. Whereas I believe that Goldsmith & Woisetschlaeger are correct for a large subclass of the English Present Tense sentences (cf. Section 4 below), I disagree with them on the cases
Guido Vanden Wyngaerd
just discussed, which describe observable phenomena all right, but those with the special property of having VSD. How can the restriction to VSD events be accounted for? As far as Dutch/Scandinavian/etc. are concerned, it seems safe to assume that E may extend before and after R, as depicted in (9) above. English presents a more complicated case, in that it is not at all obvious how a distinction could be made between VSD events and others in terms of their relationship to S,R. This is particularly true if we conceive of S as a point with zero duration, or a divider, which divides the time-line into a past and a future portion (Giorgi & Pianesi 1997). (20)
S,R -----------|-----------E
However short an event is, it will always extend preceding and following S,R, just as in (9). We therefore need to adopt an alternative approach. This approach assumes that S, the speech time, is not to be thought of as a point, but rather as an interval with a certain extension. Needless to say, this temporal extension is very short, but it is nevertheless real. Put differently, S also has the property VSD, rather than being punctual in the mathematical sense, i.e. having zero duration (see also Langacker 1982: 286). This seems intuitively plausible: the act of speaking itself is one that takes up a certain amount of time; also, the speech act is dynamic, i.e. each stage of it is different from the preceding and the following stage. These factors imply a certain duration, as I shall argue more extensively below. At the same time, the speech time interval is not arbitrarily long: if a person holds a one hour monologue, the speech time does not have the same extension. Rather, the speech time is a continually shifting interval of minimal length.10 The restriction to VSD-events observed in the English Simple Present can now be formulated as follows: only events which can be conceived as ‘fitting into’ the length of the speech time, i.e. which do not extend beyond it, are compatible with the Simple Present in English: (21) ----------|°°S,R°°|---------------°°° VSD-E
Formulaically, this can be represented as follows: (22) English Simple Present tense S=R⊇E
Simple tense
Clearly, the requirement that E be contained in S,R is absent from languages like Italian and Dutch, as configurations of the type in (9) are acceptable. Having formulated the distinction, it would of course be nice to be able to present some deeper explanation for it. A first observation to make is that the difference probably does not reside in any of the properties of the present tense, but should be looked for in the different properties of (external) aspect. This is because I have defined the latter as a specification of how E relates to R, and the relationship of E to R is the domain in which I have just argued that English and non-English are different. In what respect, then, could English simple external aspect be taken to differ from the other languages at issue? Let us think of the English system of external aspect as involving two binary features, on the one hand extension (before and after R), and completion or termination at R on the other. This gives the following four logical possibilities: (23) English Simple Perfect Progressive Perfect Progressive
extension – – + +
completion – + – +
This table gives us a way of approaching the question why the configuration in (9), repeated here, is ruled out in English. (9) --------------S,R--------------------E---
The negative value for the feature ‘extension’ in the Simple aspect will imply that E cannot extend beyond the boundaries set by S itself, i.e. must fall within the confines of the speech time, as in (21). I shall furthermore assume that the other languages at issue have only one binary feature, yielding only two possible categories for external or presentational aspect: (24)
completion Perfect + Imperfect –
In this system, there is no feature ‘extension’ which could block the configuration in (9); as a result, cases like (7) and (8) are fine. Put differently, the twocategory system does not make a distinction between ‘simple’ and ‘progressive’ presentational aspect.
Guido Vanden Wyngaerd
A final comment is in order on the issue of the duration of an event. Duration is, first of all, determined by internal aspect. Thus atelic events (activities, e.g. run, walk, swim, push a cart, drive a car, tease Mary) typically lack an end-point and therefore will not be able to occur in any of the VSD contexts sketched above. Telic events may vary in duration, depending, for example, on the size of the internal argument, which provides a measure for the amount, and hence the duration, of verbal activity (e.g. run 10 meters vs. run 10 kilometers, read War and Peace vs read the headlines, etc.; see Vanden Wyngaerd 2001). This fact has its effect on which telic VPs can be construed as being VSD and which cannot, as is shown by the following contrast (e.g. as stage directions): (25) a.
Max sits down on the couch, picks up a paper, reads the headlines, stands up again, and exits. b. *Max sits down on the couch, picks up a novel, reads War and Peace, stands up again, and exits.
A second consequence is that my explanation of (6) ultimately relies on there being a conflict between the properties of the internal aspect (duration) and the requirements imposed by the external aspect (VSD). But even two identical VPs do not always need to have the same duration. Actual duration of identical VPs is to some extent variable: thus an event like run a mile will have a variable duration depending on the speed of the runner, which itself is a function of his or her fitness, the nature of the ground, etc. Such factors consequently allow for a certain degree of freedom in the manner of presentation of an event: the event may be presented as either very short, or as having a certain extension, i.e. as ongoing, as in the following pair, suggested by a reviewer: (26) a. Magda refuses to let the police in. She closes the door in their face. b. Magda is refusing to let the police in. She’s closing the door in their face.
The reviewer goes on to suggest that manner of presentation rather than actual event duration is the relevant property with the present tense. As the reviewer observes, in (26a) “there is more of a sense of immediacy” than in (26b), which tends to focus more on the internal structure of the event. I would suggest that the sense of immediacy felt in (26a) is due to the fact that the event has VSD, or at least is presented as having VSD. Because actual event duration is to some extent variable, different points of view on an event are possible, and the same event may also be presented as extending beyond the speech time interval, as in (26b). At the same time, not all events have VSD-potential, because the mini-
Simple tense
mal actual duration of some events will, under any circumstance, exceed VSD. For example, there is no way in which events like run 10 kilometers or read War and Peace could ever appear in the Simple Present, because their actual duration is not compressible in the way required by the present tense (as shown by (25b), for example). In sum, manner of presentation may lead to some amount of variation such as that illustrated by (26), but at the same time such variation is clearly subject to the requirements imposed by the present tense (see (22)). The above analysis extends to the Simple Past and the Simple Future, with a few differences, however. Tensologically, the Past Tense involves the formula R<S and the Future S
In these contexts, the event read a book is compatible with the Simple Past/Future, in contrast to (6a) above. It has been suggested that there is a sense in which the book-reading event in (27a) is viewed as a punctual one, e.g. in the description of a series of successive events. Dowty (1979: 189) suggests that, in cases like this, time may be viewed as ‘compressed’, ‘in that the distinction between moments and intervals larger than a moment is obscured’. I will argue that, rather than involving compression of the event time, a case like (27) involves stretching of the Reference time to denote an interval. In the Present, S=R, and since S is VSD, R is VSD as well. In the Past and the Future, there is no identity requirement, however, but merely a precedence requirement between S and R. This allows R to denote an interval of arbitrary length. In fact, it would seem that in a case like (27), the adverb (yesterday/tomorrow) specifies the length of the interval. As long as a past (or future) event can be taken to fit into that interval in view of its internal aspect, as in (27), the sentence is acceptable. But one can easily construct examples where the event does not fit into the interval R: (28) a. #Yesterday, I got up, had breakfast, built a house, and then had lunch. b. In 1996, I built a house. c. #On May 25 2004/Between 1986 and 1990, I wrote my PhD dissertation.
Although, as a reviewer points out, (28c) seems fine if one considers those fields of academic inquiry where dissertations could easily be written in a day, and
Guido Vanden Wyngaerd
(28a) is acceptable if building a house uses techniques that drastically shorten building time, such observations actually confirm our claim that the adverb specifies the interval within which the event must fit. The point is that, if the building time does exceed the one-day interval specified by yesterday, then, even if the subject did some house-building within that interval, the sentence is unacceptable. Schematically, the situation with respect to the Simple Past is as follows: (29) --------+-----R-----+------S------°°°E°°°
The Simple Future is identical, except for the relative positions of S and R. This is means that I assume that for all the simple tenses in English, the following formula holds: (30) English Simple tense R⊇E
. The Simple Present: States Let us next turn to the discussion of states, starting out with a case where a contrast can be observed between English and other languages, which is highly reminiscent of the contrast in (6) and (7) above. This contrast concerns the fact that adjuncts specifying the length of a state are incompatible with the English Simple Present, whereas they can co-occur with the Simple Present in Dutch, for example: (31) a. *Max lives in Manchester for years. b. *Betty knows French all her life. (32) a.
Max woont al jaren in Manchester. ‘Max has lived in Manchester for years.’ b. Betty kent al heel haar leven Frans. ‘Betty has known French all her life.’
Above I argued that the English Simple Present required a particular sort of association between E and R, i.e. one where E is not allowed to extend beyond R. In Dutch, by contrast, such temporal extension is allowed. Something similar appears to be going on in (31) and (32): the state is given a certain temporal extension by means of the adjunct, and such extension is not compatible with the Simple Present in English, but it is in Dutch.
Simple tense
Can this contrast be given the same explanation as before? In order to consider this question, we need to look at states. It is well-known that states in general are compatible with the Simple Present in English: (33) a. b. c. d.
Betty knows French. Susie loves Mary. Henry is ill. The door is open.
This seems at first sight to be in contradiction with our earlier claim that the English Simple Present involves a feature [–extension], as the states in these examples would appear to have a temporal extension that stretches beyond that of the speech time. Hoekstra (1992), however, has proposed that, contrary to appearances, one subclass of the stative predications, the Individual Level Predicates (ILPs), in fact have what he calls ‘point duration’. He further suggests that the impression of temporal extension derives from the fact that individual-level predicates describe essential properties of individuals. “[T]o say of John that he knows French would be saying that at a certain point in time, he has the property of knowing French. Given that it is an individuallevel predicate, its extension over other instances of John can be taken as an implication” (Hoekstra 1992: 159).11 This may be represented schematically as in (34), where the star (*) is intended to represent point duration. (34) -------------|°°S,R°°|---------------| * E [VP know French]
The lexical content of the VP determines that it ascribes some essential property to the subject at S. Since the property is an essential one, it is reasonable to assume that it holds at times other than S, but this is not strictly speaking something that (33a), and sentences with ILPs in general, assert. Let us now return to the contrast in (31) and (32), which involve a state accompanied by a for-adverbial specifying duration. A number of authors have observed that for-adverbials introduce boundedness, both with states and activities (Vet 1980; Moens 1987; Naumann 1995). The following sentences (from De Swart 1998) show that activities with a for-adverbial pattern with accomplishments: (35) a.
Andrew was swimming → Andrew swam.
Guido Vanden Wyngaerd
b. Andrew was swimming for three hours Andrew swam for three hours c. Eve was drawing a circle Eve drew a circle.
Let us assume that the durational adjuncts in these sentences are VP-modifiers, i.e. semantically, they modify the denotation of the VP, i.e. the temporal extension of the event (see also De Swart 1998: 357). In this respect, these modifiers differ semantically from adverbs like yesterday and tomorrow, which I take to be tense modifiers. The result will be that the denotation of the VP will be one where the state has temporal extension. (36) -----------------|°°S,R°°|-------------------------------------E for years
This situation is not allowed in English by (22), but it is in other languages, as we saw.12 Although Hoekstra’s proposal was only intended to apply to ILPs and not states in general, I think it can be fruitfully extended in that direction. In fact, Hoekstra (1992) defines the difference between states and dynamic events as follows: dynamic events show temporal progress: each stage or time-slice of the event is different from the preceding and the following ones. States do not show this dynamism: each stage of the state is identical to the other stages. A straightforward way of accounting for this distinction is to assume that states never have successive stages, but always have point or zero duration (see also Langacker 198213 ). This generalisation not only applies to ILPs like intelligent, but also to stage-level predicates like ill and drunk and the like. The following generalisation can now be made: anything that has temporal extension has dynamism, and vice versa: anything dynamic has temporal extension.14 The difference between ILPs like intelligent and SLPs like drunk in this analysis is ultimately a lexical difference, not an aspectual or tensological one, as seems plausible. It is therefore not surprising that we find the same contrast with respect to durational adjuncts with SLPs: (37) a. *They are drunk for two hours. b. Ze zijn nu twee uur dronken. ‘They have been drunk for two hours now.’
Sentence (37a) is analogous to those in (31), and ruled out for the same reason: simple aspect in English does not permit extension beyond the speech time. The fact that states have point duration does not imply that they are not linked to the reference time at all. This is obviously true for stage-level states,
Simple tense
such as (33c–d), where the state holds at the time specified by the tense, i.e. now in the case of the present, some time in the past for the past tense, and a future time for the future tense. But the same is true for ILP’s, which could be called timeless. One possible interpretation of the latter characteristic could be that the predication is not anchored to a reference time at all. However, I do not believe that this is the case. ILP’s can be past or future in a variety of circumstances. (38) a. Dinosaurs ate leaves. b. Jack had blue eyes. c. Before retiring, Ann worked for an insurance company.
Past tense ILPs are appropriate when the subject no longer exists, because it has become extinct or has died, or because there has been some change in the characteristics of the individual concerned (on the former of these effects, see Musan 1997). Under analogous circumstances, ILPs can also appear in the future tense (39) a. b. c. d.
Your children will have blue eyes. By the end of this month, everyone on this course will know French. Summers will be wetter, winters warmer. Future generations will work in the service industry.
In much the same way, I take ILPs in the present tense to refer to ‘now’ or the speech time (see also Carlson 1982).
. Generic sentences as ILP Having looked at nonstative events in Section 3 and at states in Section 4, I now turn to an important use of the Simple Present tense, which combines aspects of both of those discussed earlier. These are the so-called generic or characterizing sentences which are based on dynamic verbs, i.e. cases like those in (40): (40) a. b. c. d.
Rick smokes (cigars). Ann works for an insurance company. Max plays the piano. She runs the marathon in less than 3 hours.
The puzzle posed by such sentences is that their interpretation is stative like that of the ILPs discussed in the previous section, yet the category of the internal
Guido Vanden Wyngaerd
aspect of the verbs used in them is clearly dynamic. This is shown by the fact that the very same verbs can be used in dynamic predications: (41) a. b. c. d.
Rick is smoking a cigarette. Ann is working for an insurance company. Max is playing the piano. She is running the London marathon.
Although a full discussion of the issue of genericity would clearly lead me too far afield, I do want to briefly sketch the broad outlines of the account I adopt for these cases. Basically, the proposal I should like to make is that cases like these are to be assimilated to cases of ILP. Before discussing my proposal, I want to comment briefly on a popular alternative analysis, which assumes that sentences like (40) involve a generic operator which quantifies over episodes (events or situations) that are taken to be constitutive for the meaning of the generic sentence (e.g. Carlson 1977). For example, Giorgi & Pianesi’s analysis of John smokes goes like this: (42) Gen t([TP(t,I) ∧ contextually-relevant(I) ∧ TP(s,I)] ∧∃e (smoke(e) ∧ Cl(e) ∧ at(e,t))
This is to be read like this: for generically many times t which are a temporal part (TP) of the contextually relevant interval I, where I contains the speech event, there is an event of smoking occurring at t.15 This analysis is exemplary for an approach which assumes there to be a form of quantification over instances of smoking. There are numerous problems with such an approach. First, an inductive analysis of this kind does not really explain the stative nature of sentences such as those in (40) above: as already noted, despite involving dynamic verbs, these sentences in fact have all the properties of stative predications. Carlson (1995: 232) further points out that, “if the linguistic meaning of a generic is derived from some operation over value substitutions for variables in episodic sentences, then any generic sentence should retain its original aspectual class”, which clearly is not the case in examples like (40).16 He furthermore observes that quantification does not normally have this stativising effect: (43) a. John pushed carts to Cleveland. b. Every man pushed a cart to Cleveland. c. John pushed a cart to Cleveland on two occasions.
Simple tense
All of (43) are eventive; if indeed genericity were to involve some sort of quantification by a generic operator, it is unclear why this type of quantification alone would lead to stativisation. A further drawback of an analysis of characterising sentences in terms of generic quantification over episodes is that it does not easily generalise to cases of ILP involving lexically stative predicates, such as (44), or all of the adjectival and nominal ILPs, e.g. (45): (44) a. John knows French. b. Tap water contains chlorine. c. An element consists of atoms of only one kind. (45) a. John is a smoker. b. John is intelligent.
It seems desirable to give a unified analysis of (40a) and (45a), given that, to all intents and purposes, they have the same meaning. Two options are open at this point: one could take the proposal in terms of generic quantification over instances developed for cases like those in (40) to be fundamental, and try to extend it to cases of ILP as in (45). On the other hand, one could adopt an analysis for ILP of the be intelligent type (such as (45b)) which does not involve generic quantification over instances, and then try to extend that analysis to the cases like those in (40). I shall argue that the latter strategy is to be preferred over the former. Before doing that, however, we need to discuss one further important problem for the quantificational approach, which is that many generic sentences do not involve instances: (46) This car runs on kerosene.
As noted by Goldsmith & Woisetschlaeger (1982: 81), “neither understanding nor substantiating the claim of [(46)] requires any reference to particular events”. The car in question might in fact be standing in a museum, and never have actually run at all; a mere knowledge of the car’s design suffices for someone to be able to utter (46) truthfully. Carlson (1995: 225) makes a similar point, in stating that generic sentences are not in principle based on “any array of observed (or even unobserved) instances”; he gives such examples as the following in evidence: (47) a. Bishops move diagonally. b. This machine crushes up oranges and removes the seeds. (said of a new machine to be later destroyed accidentally in shipping)
Guido Vanden Wyngaerd
c.
Tab A fits in slot B. (cut-out toy instruction on a cereal box which is thrown out) d. The Speaker of the House succeeds the vice president. e. Sally handles the mail from Antarctica. (an unfulfilled office function)
Even without instances, these sentences are perfectly interpretable. This problem arises even more pressingly for the case of ILPs based on nominal and adjectival predicates, supposing one wanted to extend the quantificational analysis to those cases. Thus it hard to see what the situations quantified over could be with predicates like the following (Krifka et al. 1995: 38 and Carlson 1995: 232): (48) be married, be a bachelor, own, be thirty years old, be a student, be male, be a natural satellite of the Earth, be a mammal
This problem is even more acute in the case of predicates which apply only to kinds, such as be rare/widespread/common/extinct/etc. This circumstance has not stopped people from trying to extend the generic quantification analysis from cases involving lexically episodic verbs (such as the ones in (40)) to all cases of ILP. Thus Krifka et al. (1995) propose an analysis of (44a) in terms of generic quantification over hypothetical instances of ‘showing knowledge of French’. (49) λxGEN[x,s;](x in s; x shows knowledge of French in s)17
Similarly, Chierchia (1995) presents an analysis of John is intelligent as involving generic quantification over situations in which John is. Though Chierchia’s representation of ILP is formally quantificational, it is far from clear what the content of the quantification is, both with respect to the operator, the restrictor and the variable bound by the operator. Given the difficulties inherent in such an approach, I shall instead argue that the unification should go the other way round, i.e. take the cases of ILP involving lexically stative predicates as basic, and extend the analysis of those to cases involving lexically dynamic verbs. How can sentences involving dynamic verbs acquire an interpretation as an ILP, which is, by definition, a state? And how does the stativity come about? Above, I defined stativity in terms of point duration. Let us assume that dynamic VPs have the option of being interpreted statively, i.e. as having point duration:
Simple tense
(50) -------------|°°S,R°°|---------------| * E [VP smoke cigars]
While this may at first sight appear to be an ad hoc move, one can in fact observe other instances outside the domain of generic sentences where such a ‘stativity shift’ occurs. Consider a verb like ‘run’, which would be classified as dynamic on all accounts. Yet contexts abound where this verb receives a stative interpretation: (51) a. b. c. d. e. f. g.
The road runs through the village. The exhibition runs at the Norwegian Folk Museum in Oslo. Opening hours run from 11.30 until midnight. The text in the ad runs something like this: ‘Don’t mess with Texas!’ Conan Doyle’s stories ran in ‘The Strand’ magazine. The contract runs for a year. Diabetes appears to run in families.
Why is there stativisation in cases like (51)? Observe that all of the subjects in (51) are inanimate. In Rooryck & Vanden Wyngaerd (1998) it is argued that language encodes a difference between animate and inanimate NPs as regards the representation of time-slices: animate NPs can either be viewed as individuals, i.e. in abstraction from times and locations, or as consisting of successive spatio-temporal stages. Inanimate NPs lack this distinction. It so appears that the spatio-temporal make-up of the subject affects the interpretation of the predicate: if the subject is inanimate, the predicate is stative.18 The same is true for an animate subject which is viewed at the level of the individual: no internal temporal structure is present and the predicate is necessarily stative. Additional cases illustrating this phenomenon are given below (from Rooryck & Vanden Wyngaerd 1998 and Oehrle 1976, respectively): (52) a. John (just) stated that empty categories must be properly governed. b. The ECP (*just) stated that empty categories must be properly governed. (53) a. Nixon (just) gave Mailer a book. b. The book/Mary (*just) gave Nixon an ulcer.
In these cases an alternation between an agentive and a nonagentive interpretation can be observed. In (52), the distinction is obvious, whereas in (53) the agentive interpretation corresponds with the literal transfer reading (as
Guido Vanden Wyngaerd
illustrated by (53a)), and the nonagentive reading is one where the subject is interpreted as a cause that brought about a certain state of affairs without there having to be literal transfer involved. As these sentences show, the adverb just is awkward with the nonagentive reading. The agentive readings involve a verb which is interpreted dynamically, and consequently also a subject with internal temporal structure. The nonagentive readings involve a verb which is interpreted statively, and a subject without internal temporal structure, either because it is inanimate (as in (53) with the book as its subject), or because it is viewed as an individual ((28b) with Mary). As the name suggests, ILP involves a consideration of the subject as an individual. An individual can be defined as an entity which is considered abstracting away from its internal temporal structure. Under the conception I am defending here of stative predicates as having zero duration, the same is true of the predicate: the internal temporal make-up of the verb phrase is ignored. This necessarily results in a stative interpretation. In other words, verb phrases are subject to the same possibilities that exist for noun phrases; in particular, they have the possibility of being viewed as punctual, i.e. abstracting away from their internal temporal structure. Schematically, this can be represented as follows: (54) -----|°°S,R°°|--------| * TP * * I E [DP Rick] [VP smokes cigars]
Under the dynamic interpretation, the situation is the opposite: the VP is considered as consisting of successive temporal stages, corresponding to a similar series of successive temporal stages in the subject: (55) ------|°°S,R°°|--------°°°°°°°°° TP °°°°° °°°°°° [DP Rick] [VP is smoking a cigar]
Why should there be this connection between the interpretation of the VP and the subject? Let us look at (55) first. One could say that this is a graphical representation of Carlson’s observation that stage-level predicates predicate something of a spatio-temporal stage of the subject: the successive temporal episodes of the VP are predicated of successive temporal episodes of the sub-
Simple tense
ject. These stages cannot be predicated of the individual since the individual does not possess stages. In a similar vein, an individual-level predicate can only be predicated of an individual, as in (54), not of a spatio-temporal stage.19 Clearly, the remarks in this section can only serve as a first approximation of the issue of generic sentences, an issue that cannot possibly be treated satisfactorily within the space of this article. I have merely attempted here to sketch the general approach which I think should be taken, as well as added a few observations which I feel ought to be relevant for its solution.
. Conclusion In this paper, I have first set out a few basic assumptions concerning what I take to be the proper treatment of tense and aspect: tense is basically a deictic notion, whereas aspect concerns the temporal structure of an event. Next, I looked at the use of the Simple Present tense in English and other languages such as it is used to refer to events happening now. The English tense was shown to be restricted to events with Very Short Duration. Stative predications also occur happily in the Simple Present in English, which was attributed to their zero duration character. Finally, I discussed the issue of characterising or generic sentences based on dynamic verbs, arguing that they are to be analysed on a par with cases of ILP.
Notes * The material in this paper was presented at the 19th Comparative Germanic Syntax Workshop at CUNY, New York (June 2004), and at the University of Antwerp conference ‘On the expression of Time and Space’ (September 2004); I wish to thank the audiences at those conferences for their useful remarks and suggestions. The paper has also benefited a great deal from the exacting comments of two anonymous reviewers, as well as those by the editors of this volume. All remaining errors are my own. . Reichenbach’s system suffers from a number of shortcomings. Bouchard (1984) has observed that in the system proposed by Hornstein (1977), which is based on Reichenbach, there is no predictable way of associating a particular combination of S, R, and E with any given syntactic form: “we are not given any procedure to recognize the effective realization of these basic tenses in the syntax of the grammar” (1984: 92). Moreover, various authors have pointed out that not all natural language tenses imply an ordering of the three time-points S, R and E relative to one another, the Future Perfect being a case in point.
Guido Vanden Wyngaerd
(i)
J’aurai fini. I will have finished
Comrie (1981, 1985: 70), Bouchard (1984) and Vikner (1985) have argued that the meaning of the future perfect does not impose a relative ordering of S and E. This can easily be achieved by distinguishing two distinct levels, which allows S and E to be unordered relative to one another: (ii) Level I: S_R Level II: E_R In the system I adopt, the same effect is achieved by taking the relationship of R to S (Level I in Bouchard’s system) to be a matter of tense, and the relationship between E and R (Bouchard’s Level II) one of aspect. . A reviewer asks how the notion of simple aspect is to be understood, i.e. as a morphological or a semantic notion. Morphologically, simple aspect is the absence of any grammatical marking of aspect. Semantically, its interpretation differs from language to language, as will be argued below: in English, simple aspect involves a double negative requirement (lack of extension and lack of completion), whereas other languages have a mere perfect-imperfect opposition (see the diagrams in (23) and (24) below and the surrounding discussion). . One could object that the presence of two future tenses is somehow due to the periphrastic nature of the English future tense, which is expressed through the combination of a modal and a bare infinitive, rather than morphologically. A closer look at data from other languages, however, reveals that this approach cannot be correct. Even languages that have a morphological future tense, like French, allow the combination of past and future tense: (i)
Past: (il) arriv+ait ‘(he) arrived’ Future: (il) arriv+er+a ‘(he) will arrive’ Past Future: (il) arrive+er+ait ‘(he) would arrive’
Reichenbach calls the latter tense the posterior past, noting that it is a tense “which grammar does not officially recognize as a tense” (1947: 298). Its interpretation appears to be one involving two reference points, as in: (ii) a. b.
I did not expect that he would win the race. John left for the front. He would never return.
The first or matrix event is located in the past, establishing a past reference point, relative to which the second or embedded event is located in the future (Comrie 1985: 75). . The distinction between internal and external aspect is not original, nor do previous authors necessarily use the same terms to refer to the same distinction (see e.g. Smith 1991; Hoekstra 1992; Depraetere 1995; De Swart 1998; Demirdache & Uribe-Etxebarria 2000). Overall, the system adopted here is very much like the one proposed by De Swart (1998), who distinguishes the following syntactic and semantic hierarchy: (i)
[ Tense [ Aspect* [ eventuality description ]]]
Schmitt (2001) is also based on De Swart (1998).
Simple tense . A reviewer finds this comment ‘a bit unfair’ towards Dowty because in none of the Dutch cases ‘a description of the event and its result is possible’, suggesting that ‘Dowty’s intuition is not incorrect’. I think the cases in (10) to (15) do invalidate this intuition, which moreover I have been unable to find in anything approaching the formulation given by the reviewer in Dowty’s text. . Giorgi & Pianesi (1997) (henceforth G&P) present an account which is slightly more complicated, but which ultimately boils down to the same restriction as proposed by Dowty. G&P argue that the English eventive predicates are special in that they are inherently perfective. This is in turn a result of the absence of agreement features on English verbs; for the derivation to converge, the categorial features [+V,–N] need to be supplemented with the feature [+Perf], for reasons that we shall not go into here. As a result, “a verb in English is always perfective” (1997: 164). These assumptions account for (6), modulo a following step in the argument. In short, perfective events are closed and closed events cannot be simultaneous with punctual ones; since the utterance time is punctual, (6) is ruled out. The claim that perfective events cannot be simultaneous with the speech time follows from G&P’s punctuality constraint: (i)
A closed event cannot be simultaneous with a punctual event.
The process of closure involved in perfectivisation amounts to adding a boundary to a process, without, however, affecting the temporal properties of the process. In other words, a closed event can be decomposed into a processual part and a boundary. An essential characteristic of a process is that it has temporal structure. By definition, punctual events lack temporal structure. While these assumptions explain why (i) should hold, it seems that Giorgi and Pianesi’s account is in essence no different from the Taylor/Dowty account mentioned earlier: the speech time is punctual, and eventive predicates are nonpunctual. The nonpunctual character essentially derives from the processual part contained in a perfectivised eventive predicate. Crucially, G&P’s account also suffers from the same defect as Dowty’s: it fails to account for the fact that in languages other than English, eventive predicates do combine with the Simple Present Tense. In those languages, verbs are not inherently perfective, so that only the processual part of the event remains, and it is precisely this processual part which G&P take to be incompatible with the speech time. Put differently, if perfective events are considered to be incompatible with the speech time on account of the temporal structure, present events are incompatible a fortiori, since they only consist of the processual part. A further problem concerns the fact that the languages in which the equivalent of (6) is possible also include Mainland Scandinavian (see (8) above), which, from the point of view of verbal morphology are even more impoverished than English, raising the question why these verbs are apparently not inherently perfective. . In this connection, a reviewer raises the learnability issue: how does the learner of English figure out the relevant properties of the English present tense? Whilst I have no answer to offer to this question at this stage, it seems to me that this problem is not unique to the account I am proposing. . The sentences in (16) represent a different case: it will be obvious that (16c), for example, does not describe a VSD event. At the same time, it is also obvious that the events described by these sentences do not relate to the utterance time at all, i.e. they are not deictically an-
Guido Vanden Wyngaerd
chored to the speech time. This clearly sets them apart from the other uses of the Simple Present. Instead, it seems as if the events at issue are anchored to a location, which is usually overtly mentioned in a PP (in Act I, in The Scarlet Letter, in Chapter 3, etc.). There furthermore appears to exist a requirement, not dissimilar from the one to be discussed below, to the effect that the event at issue be contained in the location. This suggests interesting avenues for the analysis of temporal relationships in terms of spatial concepts, which we hope to pursue elsewhere (see also Demirdache & Uribe-Etxebarria 2000). . Michaelis (1998: 26) argues that the events in these reports fill slots in highly ritualised scripts, supporting her argument with the following example: (i)
Jones hits a high fly to left field. Ryan chases it. ??A streaker appears on the field.
This sequence does indeed have a flavour of slight deviance, which one could indeed account for in terms of requirements imposed by the ‘script’. But that fact in and of itself does not tell us why the present tense is usable in such cases. Observe that replacing the present by the progressive in the final sentence of (i) does not appear to improve the sequence. Michaelis goes on to argue that these events are construed as lacking internal structure. But as will be argued in the text, although there can indeed be some variability with respect to way the length of events is perceived or construed, such variability is not limitless (see (26) and surrounding discussion). . Dehaene (1997: 102) observes that memorising numbers and the like often involves using a verbal memory loop. This memory can hold data for only about two seconds. That is, short term memory is not limited in the number of elements it can contain, but is limited to a very short temporal interval. As a result, languages which monosyllabic names for the numbers, like Chinese, allow more digits to be remembered in short-term memory than languages with phonologically more complex number names, since less of those will fit into the same temporal interval. For concreteness, one might assume that the speech time, which I propose has the property VSD, is to be identified with the temporal length of this short term memory, which is obviously crucially involved in the act of speaking. . Comrie (1985: 39–40) proposes something analogous for habitual sentences. . A. Henry (p.c.) points out that in Belfast English, the restriction exemplified by (31) does not hold, i.e. it is possible to say things like: (i)
I’m here two weeks/since Wednesday.
Still, the equivalent of (6) continues to be ruled out in this dialect, suggesting that the two phenomena may not have a common explanation. . The present account differs from Langacker’s in that he considers stative predications containing verbs (e.g. hate, resemble, etc.) to be imperfective rather than stative. There is little, if any, empirical evidence to support such a move, however. . A clear exception to this are sentences where temporal extension is contributed by an adjunct, such as (32) above. The generalization only holds at the lowest VP-level, therefore. . The closure operator Cl represents the inherently perfective nature of all English eventive predicates. In an eventive sentence like John eats an apple, the logical representation
Simple tense
(i)
∃e∃t∃x[eat(e) ∧ Theme (e,x) ∧ apple(x) ∧ Agent (e,John) ∧ Cl(e) ∧ t=S ∧ at(t,e))
involves some sort of identity of e and S, which leads to ungrammaticality on account of G&P’s punctuality constraint (see note 6). In a generic sentence, however, their analysis does not establish a direct link between the event and the speech time. Instead, (42) requires that S be a part of the temporal interval which also includes e. . A reviewer states that it would be good if it could be shown explicitly that sentences of the type in (40) do not retain their original aspectual class under the relevant operation. The obvious answer to this is that there is no empirical evidence at all showing that they do retain their aspectual class. . The formula (49) instantiates the following general schema: Q[x1 ,. . .,xi ;y1 ,. . .,yi ] (Restrictor [x1 ,. . .,xi ]; Matrix [{x1 },. . .,{xi }, y1 ,. . .,yi ]) Where x1 ,. . .,xi are the variables to be bound by Q, and y1 ,. . .,yi are the variables to be bound existentially with scope just in the matrix. . A reviewer points out that it cannot be right to say that, if the subject is inanimate, the predicate is stative, given such cases as Sweat was running down his forehead. However, as discussed in Rooryck & Vanden Wyngaerd (1998), the distinction between animate and inanimate NPs is to a certain extent flexible, in that inanimate things may be viewed as animate, i.e. as evolving through time, particularly if they show the property of internally driven or autonomous movement, or if they possess technological complexity. The former appears to be the case in the example mentioned by the reviewer. . States can be predicated of spatio-temporal stages of individuals, but only if they are adjectival (e.g. be ill, happy, drunk, etc.). That is, all stative verbs are individual-level (e.g. hate, love, know French, etc.), including the states based on dynamic verbs discussed here. Put differently, there are no stative sentences of the type *John happies, meaning John is happy. See Langacker (1982) for a possible account of this restriction.
References Bouchard, D. (1984). Having a tense and time in grammar. Cahiers linguistiques d’Ottawa, 12, 89–113. Carlson, G. (1977). Reference to Kinds in English. PhD Dissertation, University of Massachusetts at Amherst. Carlson, G. (1982). Generic terms and generic sentences. Journal of Philosophical Logic, 11, 145–181. Carlson, G. (1995). Truth conditions of generic sentences: Two contrasting views. In G. Carlson & F. Pelletier (Eds.), The Generic Book (pp. 224–237). Chicago: University of Chicago Press. Chierchia, G. (1995). Individual-level predicates as inherent generics. In G. Carlson & F. Pelletier (Eds.), The Generic Book (pp. 176–223). Chicago: University of Chicago Press. Comrie, B. (1981). On Reichenbach’s approach to tense. CLS, 17, 24–30. Comrie, B. (1985). Tense. Cambridge: Cambridge University Press.
Guido Vanden Wyngaerd
Declerck, R. (1991). Tense in English. Its structure and use in discourse. London: Routledge. Dehaene, S. (1997). The number sense: How the mind creates mathematics. London: Lane. Demirdache, H. & M. Uribe-Etxebarria (2000). The primitives of temporal relations. In R. Martin, D. Michaels, & J. Uriagareka (Eds.), Step by Step. Essays in minimalist syntax in honor of Howard Lasnik (pp. 157–186). Cambridge, MA: The MIT Press. Depraetere, I. (1995). On the necessity of distinguishing between (un)boundedness and (a)telicity. Linguistics and Philosophy, 18, 1–19. Dowty, D. (1979). Word Meaning and Montague Grammar. Dordrecht: Reidel. Giorgi, A. & F. Pianesi (1997). Tense and Aspect. Oxford: Oxford University Press. Goldsmith, J. & E. Woisetschlaeger (1982). The logic of the English progressive. Linguistic Inquiry, 13, 79–89. Hoekstra, T. (1992). Aspect and theta theory. In I. M. Roca (Ed.), Thematic Structure. Its role in grammar (pp. 145–174). Dordrecht: Foris. Hornstein, N. (1977). Towards a theory of tense. Linguistic Inquiry, 8, 521–557. Jespersen, O. (1909). A Modern English Grammar on Historical Principles. London: George Allen & Unwin. Krifka, M., F. J. Pelletier, G. Carlson, A. ter Meulen, G. Link, & G. Chierchia (1995). Genericity: An introduction. In G. Carlson & F. J. Pelletier (Eds.), The Generic Book (pp. 1–124). Chicago, IL: The University of Chicago Press. Langacker, R. (1982). Remarks on English aspect. In Paul Hopper (Ed.), Tense-Aspect: Between syntax and semantics. Amsterdam: Benjamins. Michaelis, L. (1998). Aspectual Grammar and Past-Time Reference. London: Routledge. Moens, M. (1987). Tense, Aspect and Temporal Reference. PhD Dissertation, University of Edinburgh. Musan, R. (1997). Tense, predicates, and lifetime effects. Natural Language Semantics, 5, 271–301. Naumann, R. (1995). Aspectual Composition and Dynamic Logic. PhD Disseration, University of Düsseldorf. Oehrle, R. (1976). The Grammatical Status of the English Dative Alternation. PhD Dissertation, MIT. Oliveira, F. & A. Lopes (1995). Tense and aspect in Portuguese. In R. Thieroff (Ed.), Tense Systems in European Languages II (pp. 95–110). Tübingen: Max Niemeyer. Quirk, R., S. Greenbaum, G. Leech, & J. Svartvik (1985). A Comprehensive Grammar of the English Language. London: Longman. Reichenbach, H. (1947). Elements of Symbolic Logic. New York, NY: Free Press. Rooryck, J. & G. Vanden Wyngaerd (1998). The self as other: A minimalist approach to zich and zichzelf in Dutch. Proceedings of NELS 28, 359–373. GLSA, University of Massachusetts, Amherst. Schmitt, C. (2001). Cross-linguistic variation and the present perfect. Natural Language and Linguistic Theory, 19, 403–453. Smith, C. (1991). The Parameter of Aspect. Dordrecht: Kluwer. Swan, M. (1995). Practical English Usage. Oxford: Oxford University Press. Swart, H. de (1998). Aspect shift and coercion. Natural Language and Linguistic Theory, 16, 347–385. Taylor, B. (1977). Tense and continuity. Linguistics and Philosophy, 1, 199–220.
Simple tense
Vendler, Z. (1967). Linguistics in Philosophy. Ithaca, NY: Cornell University Press. Verkuyl, H. (1972). On the Compositional Nature of the Aspects. Dordrecht: Reidel. Vet, C. (1980). Temps, aspect et adverbes de temps en français contemporain. Genève: Droz. Vikner, S. (1985). Reichenbach revisited: One, two, or three temporal relations? Acta Linguistica Hafniensia, 19, 81–98. Wyngaerd, G. Vanden (2001). Measuring events. Language, 77, 61–90. Zagona, K. (1992). Tense-binding and the construal of the present tense. In C. Laeufer & T. Morgan (Eds.), Theoretical Analyses in Romance Linguistics (pp. 385–398). Amsterdam: John Benjamins.
Possessor licensing, definiteness and case in Scandinavian* Marit Julien
In Scandinavian, there is considerable variation in the surface syntax of constructions with DP-internal possessors. I propose in this paper that all the different possessive constructions are derived from the same basic configuration by movements related to licensing and to focus. Another main topic is the definiteness seen in most Scandinavian possessed DPs. I argue that this definiteness is a consequence of an agreement relation between the possessor and the functional domain of the possessed DP. In the case of a pronominal possessor we will also see that even though it agrees with the possessee, the possessee does not license its case. I conclude that despite claims to the contrary, agreement does not necessarily lead to structural case licensing.
.
DP-internal possessors in Scandinavian
There are many ways to realise a DP-internal possessor in Scandinavian. Several of the relevant constructions will be exemplified below, so I will not go into details here, but for introductory purposes the general picture can be sketched as follows. In some varieties, such as Standard Swedish and Standard Danish, all DP-internal possessors are prenominal. In other varieties, such as Faroese, Icelandic, and Norwegian, possessors can be either prenominal or postnominal. In the case of a pronominal possessor, prenominal position has to do with focus in these varieties. The same is true for nonpronominal possessors in Icelandic. In Faroese and Norwegian, on the other hand, the position of a nonpronominal possessor is a consequence of the way it is licensed. I should make it clear from the beginning that I use the term ‘possessor’ to refer to constituents that bear a possessor relation to the head noun in the syntax. From a semantic point of view, this relation can be interpreted in a number of different ways. Still, it appears that certain relations are more typical
Marit Julien
or canonical possessor relations than others. Vikner & Jensen (2002) mention the following four relations: (1) the inherent relation, with inherently relational nouns as possessees, such as teacher or sister, (2) the part-whole relation, where the possessee is a part of the possessor, as in Anna’s nose, (3) the agentive relation, where the possessor is an agent and the possessee a product or result, as in Anna’s poem, (4) the control relation, where the possessor is animate and the possessee is an item controlled by the possessor, as in Anna’s bike. Legal possession is one type of control relation, but by no means the only one. For any given possessive construction, one or more of these four canonical relations will normally be available as an interpretation, even with little or no contextual support. However, as Vikner & Jensen point out, a wide range of other interpretations are often possible as well, in particular if the context is rich enough and suitable. For example, Anna’s nose could be the nose that Anna modelled or performed surgery on. And in some cases, the most plausible interpretations are non-canonical ones. For example, my street would probably be the street that I live on, or alternatively, the street that was named after me. But the point is that the variation in interpretation is not necessarily reflected in the syntax. In English, the possessive construction my X is compatible with all kinds of canonical and non-canonical interpretations. Although there are exceptions – for example, there are languages where alienable possession is expressed differently from inalienable possession – a possessor construction in the syntax simply means that there is some semantic relation between the possessor and the possessee. The exact nature of the relation has to be determined outside of the syntax. It is, however, the syntax of possessive constructions I will be dealing with here, and I will set the interpretation issue aside. One of the goals of this paper is to demonstrate that all the different possessive constructions found in Scandinavian can be derived from the same basic configuration, such that surface variation results from movements that take place for licensing or for other reasons. Another main point is that possessors make their containing DPs definite in Scandinavian, except in conservative Icelandic. I will argue that the definiteness of possessed DPs is a consequence of an agreement relation that is established between the possessor and the functional domain of the possessed DP. In the case of a pronominal possessor we will also see that even though it agrees with the possessee, its case is not licensed by the possessee. Hence, it appears that despite claims to the contrary, structural case is not a reflex of agreement. I begin with a presentation of Scandinavian postnominal possessors, in Section 2. In Section 3 I sketch the syntactic structure of DPs with postnominal possessors, which I take to be relevant for all Scandinavian possessed DPs since
Possessor licensing, definiteness and case in Scandinavian
on my view prenominal possessors have moved from postnominal position. Then in Section 4 I give a more detailed analysis of the syntax of postnominal pronominal possessors. In Section 5 I deal with postnominal nonpronominal possessors and in Section 6 with prenominal possessors. In Section 7 I elaborate on the claim that possessed DPs in Scandinavian are definite, and my conclusions are summed up in Section 8.
. Postnominal possessors In several varieties of Scandinavian, DP-internal possessor expressions can appear in postnominal position. In Faroese these are most often realised as PPs regardless of whether the possessor is pronominal or nonpronominal – see (1).1 (1) a.
tann svart-i def.m.sg.nom black-def.m.sg.nom Jákup-i Jákup-dat ‘Jákup’s black horse’ b. tann svart-i def.m.sg.nom black-def.m.sg.nom mær me.dat ‘my black horse’
hestur-in hjá horse-def.m.sg.nom at (Far)
hestur-in hjá horse-def.m.sg.nom at
In Norwegian, nonpronominal postnominal possessors are realised as PPs, as in (2a), whereas pronominal postnominal possessors appear without a P layer, as (2b) shows. (2) a.
den gaml-e katt-a åt ein granne def.sg old-def cat-def.f.sg to indef.m.sg neighbourM ‘a neighbour’s old cat’ b. dei to svart-e katt-e-ne mine def.pl two black-def cat-pl-def.pl my.pl ‘my two black cats’
(Nor)
A striking property of these postnominal possessors is that they trigger definiteness marking on the possessed noun. This holds even when the possessor itself is indefinite, as in (2a). A noun that combines with a postnominal possessor cannot normally have an indefinite determiner:
Marit Julien
(3) *ei katte åt ein granne indef.f.sg catF to indef.m.sg neighbourM
(Nor)
Inherently relational nouns are exceptional in this respect, however – they can head indefinite nominal phrases. In Icelandic, the possessor is realised in the same way as elsewhere in such cases: (4) einn vinur minn one.m.sg.nom friend.sg.nom my.m.sg.nom ‘one of my friends’
(Ice)
In Mainland Scandinavian, certain inherently relational human-denoting nouns (those that do not involve a unique relation from the possessor’s point of view) can be indefinite if they appear with a partitive possessive PP. I give two examples in (5a). The construction is ungrammatical with other nouns – see (5b). (5) a.
{ei tante/ ein kollega } av meg an aunt/ a colleague of me ‘an aunt/a colleague of mine’ b. *{ein sjef/ ei katte} av meg a boss/ a cat of me
(Nor)
But note that the partitive possessive construction requires a preposition that is different from the preposition we find in other possessive PPs. Hence, I will follow Zamparelli (1998) and assume that the partitive possessive PP construction is syntactically different from ordinary possessive PP constructions, and consequently I will ignore it here. Unlike possessive PPs, other PPs inside the nominal phrase, even PPs that express a part-whole relation, allow the head noun to be definite or indefinite. In (6) we see that an indefinite noun can combine with a non-possessive PP with a definite noun and vice versa. (6) a.
eit hjørne av duk-en indef.n.sg cornerN of tablecloth-def.m.sg ‘a corner of the tablecloth’ b. hjørn-et av ein duk corner-def.n.sg of indef.m.sg tableclothM ‘the corner of a tablecloth’
(Nor)
Note that again a different preposition is used here compared to the possessor construction in (2a). Other part-whole relations may require yet other prepositions, as in the following examples:
Possessor licensing, definiteness and case in Scandinavian
(7) a.
bein-et på {ein stol/ stol-en } leg-def.n.sg on indef.m.sg chair/ chair-def.m.sg ‘the leg of a/the chair’ b. lomm-a-ne i {ei bukse/ buks-a } pocket-pl-def.pl in indef.f.sg trouser/ trouser-def.f.sg ‘the pockets of a/the pair of trousers’
(Nor)
There is also a semantic difference between these expressions of part-whole relations on the one hand and possessive PPs on the other. Whereas a possessive PP construction will in principle allow a number of interpretations, as pointed out above, each of the constructions in (6) and (7) can only be taken to denote a part-whole relation. For these reasons I conclude that they do not belong in the same class as real possessor constructions, and I will leave them aside as well.2 In Icelandic, postnominal pronominal possessors are syntactically similar to their counterparts in Norwegian. Postnominal nonpronominal possessors, on the other hand, trigger definiteness marking on the possessed noun for some speakers but not for others. I show this in (8). (8) a.
gaml-i köttur-inn old-def.m.sg.nom cat-def.m.sg.nom ‘my old cat’ b. gaml-i köttur(%-inn) old-def.m.sg.nom cat-def.m.sg.nom ‘the teacher’s old cat’
minn my.m.sg.nom
(Ice)
kennara-ns teacher.gen-def.m.sg.gen
In addition, a pronominal possessor will share the case that the possessed noun carries, whereas nonpronominal possessors, which do not appear in PPs in Icelandic, always have genitive case. We see this from the examples above. The pronominal possessor in (8a) has nominative case, just like its containing nominal phrase, whereas the nonpronominal possessor in (8b) has genitive case although its containing nominal phrase has nominative case. To make the point even clearer I add two examples in (9) where the larger nominal phrase has dative case (from Delsing 1998). As we see, the pronominal possessor in (9a) now appears in its dative form, but the nonpronominal possessor in (9b) retains its genitive case. (9) a.
af kött-num mínum of cat.dat-def.m.sg.dat my.m.sg.dat ‘of my cat’
(Ice)
Marit Julien
b. af kötti/ %kött-num kennara-ns of cat.dat/ cat.dat-def.m.sg.dat teacher.gen-def.m.sg.gen ‘of the teacher’s cat’
Thus, there appear to be two properties of nominal phrases with postnominal possessors that in particular need to be explained. Firstly, there is the fact that they are definite, with the exception of Icelandic nominal phrases containing nonpronominal possessors, where there is variation among speakers. Secondly, there is their behaviour with respect to case, best seen in Icelandic: why is it that pronominal possessors, whose paradigms include a genitive case form, nevertheless share the case of their containing nominal phrase, unlike nonpronominal possessors, which invariably appear in the genitive case? What I will argue in the following is that these two properties are connected, such that the case of the possessor as well as the definiteness of the possessee are consequences of the agreement relation that holds between the possessor and a functional head in the projection of the possessee.
. The syntactic structure of DPs with postnominal possessors Let us first consider the syntactic structure of a DP with a postnominal possessor. We have already seen from the examples in (1) and (2) that when the DP also contains an adjective, we have a free determiner in initial position, then follow the adjective, the head noun, and after that the possessor. This indicates that the possessor is located quite low down in the DP. On the other hand, the examples in (10) show that a postnominal possessor will precede an internal argument of the noun, and in addition, the possessor can bind into that argument. den ny-e forstå-ing-a hennesi av segi of 3refl def.sg new-def understand-ing-def.f.sg hers (Nor) sjølv self ‘her new understanding of herself ’ b. ?den overbevisande framstilling-a åt kandidat-eni av segi the convincing presentation-def to candidate-def of 3refl sjølv self ‘the candidate’s convincing presentation of himself ’
(10) a.
Possessor licensing, definiteness and case in Scandinavian
Also note that the head noun precedes not only its internal argument but also the possessor. On the assumption that the noun is first merged between its two arguments, this means that it must subsequently have moved out of its base position. Yet the noun follows the adjective and the prenominal determiner. This shows that it has not moved to D, the highest head of the DP. Instead, it is the prenominal determiner that occupies D. I conclude that below the adjectives there is a functional head that the head noun moves to. I will call this head n, since just like v, it defines a (strong) phase and is able to license arguments. Arguably, n is also where suffixed definite articles are generated (with the exception of Danish, where the suffixed definite article is a realisation of D – see Julien 2002). Hence, the suffixing of these markers is a consequence of the N movement (as proposed e.g. by Taraldsen 1990; Kester 1993; Santelmann 1993; Sigurðsson 1993; Giusti 1994; Sandström & Holmberg 1994). Similarly, I take the suffixing of plural markers, which are closer to the nominal stem than definiteness markers (see (2b)) to be the result of the noun having moved through Num, a head that encodes singular/plural and is located between N and n.3 On my analysis the DP in (10a) has the syntactic structure shown in (11). We see here the movement of N to Num and n. We also see that I take the possessor to sit in Spec-NP. In addition, I take adjectival phrases to be specifiers of functional heads, called α, in the projection line of the noun. I will not deal with the syntax of adjectives in the following, however, but concentrate on the relation between the possessed noun and the possessor. (11)
Ü
Ü
Marit Julien
As for the possessed DP in (10b), I would take its syntactic structure to be similar, but with a PP and not a pronoun in Spec-NP. The reason why (10b) is not quite as good as (10a) is that binding has to take place out of a PP in the former case. Binding out of PPs is well known from the verbal domain (as in Ernest talked to Louisei about herselfi ), and the phenomenon has been subject to some debate in the linguistic literature. The problem is that the classical binding theory as laid out in Chomsky (1981) predicts that binding should not be allowed out of a PP, since the binder then does not c-command the bindee. Several proposals have been put forward to account for binding out of PPs, but as it turns out, not all of them can be extended to the nominal domain. There is no room here for a detailed discussion of the binding relation in (10b), but Julien (to appear) argues that it can be accounted for by means of the theory presented in Reinhart and Reuland (1993). In any case, it is clear that (10b) is but one instantiation of a phenomenon that is also seen elsewhere, so that it will need an explanation that does not specifically refer to possessor relations.
. Pronominal postnominal possessors Concerning pronominal possessors, I have already shown with Icelandic examples (Icelandic being the Scandinavian variety where morphological case is best retained) that these possessors share the case of the possessed noun. That is, they get their case from the external element that licenses the case of the DP as a whole. This means that there is no case-licensing relation between the pronominal possessor and any other element inside the DP. There is however an agreement relation between the pronominal possessor and the suffixed definiteness marker of the possessee. In (8a) and (9a), for example, these two elements agree visibly in gender, number and case. Now consider the following examples of Icelandic definite nouns: (12) a.
arm-s-ins b. arm-sg.gen-def.sg.gen ‘the arm (gen)’
arm-ar-nir arm-pl.nom-def.m.pl.nom ‘the arms (nom)’
(Ice)
We see that not only the suffixed definiteness marker but also the number marker inside it bears case. I take this to mean that case is a feature of every functional head in the DP, from Num upwards. As for the gender of the noun, Ritter (1993) argues that it can be a feature of the Num head or of the nominal root itself. In Scandinavian, I think there is reason to believe that gender is a feature of the root or stem, since the gender of the noun does not co-vary
Possessor licensing, definiteness and case in Scandinavian
with the form of the plural marker. The distribution of phi-features inside the nP is then as shown in (13), where I take the DP köttnum mínum from (9a) as my example. (13)
Ü
Here the noun itself is specified for gender. The Num head specifies the number, and this is also where case first shows up in the derivation. The pronominal possessor is generated with a set of unvalued gender, number and case features, in addition to the phi-features that are determined by the reference of the possessor – in the present example, first person singular. When the possessor is merged in Spec-NP, its unvalued gender feature will be valued by N. When Num is merged, the number and case features of this head will value the corresponding unvalued features in the possessor. The n head is also generated with unvalued gender, number and case features, which get valued by the features present in NumP. The consequence is that n agrees with the possessor. Now in addition to its phi-features, the possessor has a possessive feature, which is reflected in its special possessive form. The possessive feature is not identical to genitive case, since a possessive pronoun can appear in any case form. Instead, the possessive feature is in reality a definiteness feature. Since n necessarily also has a definiteness feature, the agreement that is established between n and the possessor will include the [poss/def] feature, so that n ends up being definite. Consequently, n is spelled out as a suffixed definiteness marker that reflects the gender and number (and case, in the varieties that have overt case) of the possessed noun in addition to definiteness. Since n is definite, all higher heads, including D, will have to be definite as well. In other words, the basis of the connection between possessors and definiteness is the presence of the [poss/def] feature in n. The close relation between possession and definiteness can also be seen in languages outside of Scandinavian. Armenian is an example. This language has possessive suffixes that unambiguously identify the person feature of a phono-
Marit Julien
logically null possessor, as in (14a). The suffix -6 by contrast, can combine with an overt possessor of any person, as indicated in (14b) and (14c). But interestingly, when there is no overt possessor, the -6 suffix can be interpreted as a marker of a third person possessor or as a plain definiteness marker, as (14d) shows. (Examples from Koptjevskaja-Tamm 2003a: 641.) (14) a.
c.
tun-6s b. house-poss.1sg ‘my house’ Petros-i tun-6 d. Petros-gen/dat house-poss ‘Petros’ house’
im tun-6 my house-poss ‘my house’ tun-6 house-poss/def ‘his/the house’
(Arm)
My interpretation of these facts is as follows. The nominal suffixes that we see in (14) are realisations of n. When a covert or overt possessor is present (overt possessors surfacing in a position which is higher than n, as we see), n will agree with the possessor. The -6s suffix in (14a) spells out the number and person of the possessor, in addition to the possessive feature. The -6 suffix, by contrast, is void of person features; it only spells out the possessive feature. Because of this, it is compatible with any possessor, but when there is no overt possessor, it can be taken to indicate a covert third person possessor, since third person is really no person. However, since the possessive feature is nondistinct from a definite feature, -6 can also be interpreted as a pure definiteness marker, with no indication that a possessor is present.4 I conclude that in the morphology and in the syntax, a possessive feature is treated as a kind of definiteness feature, although in the possessor, the interpretation of a possessive feature can be distinguished from that of an ordinary definiteness feature. Another important point is that the agreement relation between n and a pronominal possessor is arguably what licenses the possessor in Scandinavian. Still, the relation does not lead to case marking of the possessor. Instead, the case features of Num, n and the possessor are all valued by the same DP-external element. This indicates that contrary to what Chomsky (2000) suggests, agreement relations involving arguments are not necessarily structural case relations. In other words, as argued by Carstens (2001) and others, structural case is not simply a reflex of agreement. Rather, it appears that casemarking requires that an element that is inherently a case-marker is part of the relation. In the examples currently under discussion, complete agreement between the possessor on the one hand and the projections belonging to the possessee
Possessor licensing, definiteness and case in Scandinavian
on the other is triggered because the pronominal possessor comes with a set of unvalued phi-features. These features get valued by the features that are present in the functional domain of the possessee. In other words, an agreement relation between the possessor and the possessee is inevitable. It is then not possible for the case features of the possessor to stay unvalued – these features too will be valued by the case feature that is present in every head of the possessee’s projection from Num upwards, and the result is that the possessor shares the case of the possessee. This case will have to be licensed from outside of the possessed nominal phrase. One could therefore say that inside the nominal phrase, agreement blocks case assignment instead of being a prerequisite for it. As for the Icelandic example shown in (4), where the possessee is inherently relational and also indefinite, I would claim that the indefiniteness of the larger nominal phrase is a consequence of the inherently relational noun being able to license its possessor without assistance from n. For concreteness I propose that the possessee carries a [poss] feature which agrees with the [poss] feature of the pronominal possessor. I would further take the [poss] feature of the possessee to be a remnant from older stages of the language, where nouns in general could license their possessor without assistance from n (see below). In (most variants of) Mainland Scandinavian, nouns never come with possessor features any more, so that even inherently relational nouns cannot be indefinite if they combine with their possessor directly, hence the partitive PP construction in (5).5 Note that even if it is arguably licensed by the possessed noun, the possessor in (5) still agrees with the functional elements of that noun’s projection in person, number and case. This is what we should expect: the possessed noun itself does not have such features, only a [poss] feature. But the pronominal possessor has unvalued case and phi-features here as elsewhere, and these get valued as usual by the functional elements in the projection of the possessee.
. Nonpronominal possessors When it comes to possessors that are not expressed as pronouns, we have already seen, in (8b) and (9b), that they appear with genitive case in Icelandic, regardless of the case of the possessee. This must mean that nonpronominal possessors do get case inside their host DP in this variety.6 In older varieties of Icelandic, and of Scandinavian more generally, the licensing of the genitive possessor did not trigger definiteness. Moreover, the genitive was a lexical case, which appeared on the objects of certain verbs and
Marit Julien
prepositions, as well as on possessors (see e.g. Faarlund 2004). I take this to mean that the genitive was licensed by those verbs and prepositions, and by the possessed nouns themselves. The situation appears to be very much the same in the language of most speakers of Icelandic even today. That is, nonpronominal possessors come with a genitive case feature which is licensed by the possessee. I take this licensing to take place with the possessor staying in Spec-NP. Whether or not n is involved, definiteness is not triggered, since the genitive case feature is not a definiteness feature, unlike the [poss] feature that is found in pronominal possessors. There are, however, some speakers of Icelandic who spell out a definiteness suffix on nouns that combine with genitive possessors. This indicates that they have developed a grammar where even nonpronominal possessors come with a [poss] feature that agrees with a definiteness feature in n. But whereas the [poss] feature of possessive pronouns is reflected in their possessive form, the [poss] feature of nonpronominal possessors is realised as genitive case. Since nonpronominal possessors do not come with a set of unvalued phi-features, the agreement between these possessors and n gets no other manifestation than the definiteness that is seen in n and the genitive that is seen on the possessor. And since the possessor relation forces the nonpronominal possessor to have genitive case morphologically, there is no room for it to share the case of the possessee. Interestingly, Delsing (1991) shows that the article system in Swedish was introduced at the same time as verbs and prepositions ceased to appear with genitive nominals. Delsing’s interpretation of that fact is that the genitive became a structural case that had to be licensed by the D head above the possessed noun. I think Delsing is essentially right, except that I take the element that became a functional licenser of genitive case to be n. Arguably, exactly the same process has recently begun in Icelandic. In some of those varieties of Scandinavian where n is no longer a case assigner, a postnominal nonpronominal possessor is licit if it is realised as a possessive PP. This is what we saw examples of in (1) and in (2a). Notably, in every variety of Scandinavian that has possessive PPs at all, only one preposition is available for this use. In some Norwegian varieties the possessive preposition is åt (see (2a) and (10b)), which otherwise has a directional meaning, as illustrated in (15a). (15) a.
Ho reis-te åt by-en. she travel-past to town-def ‘She went to (the) town.’
(Nor)
Possessor licensing, definiteness and case in Scandinavian
b. Ho reis-te frå by-en. she travel-past from town-def ‘She left (the) town.’
The contrast between (15a) and (15b) indicates that the preposition åt is meaningful in (15a) in a way that it is not in (2a) and (10b). I conclude that possessive prepositions are functional prepositions. They are the spellout of a possessive feature, which again agrees with a [poss/def] feature in n. Hence, just like possessive pronouns, possessive PPs trigger definiteness on the possessed nominal. The main difference is that possessive Ps have no phi-features that can agree with those of n. Given what I have just said, it is interesting to note that in Bantu languages, possessive prepositions agree in class with the possessed noun and not with the possessor – see the Swahili example in (16), taken from Carstens (2001: 155). (16) a.
kitabu cha mwalimu 7.book 7.of 1.teacher ‘the teacher’s book’ b. *kitabu wa mwalimu 7.book 1.of 1.teacher
(Swa)
According to Carstens (2001) this means that either the relation Agree does not work as suggested in Chomsky (2000, 2001a), or else the possessor is not the complement of the possessive preposition in Bantu. Carstens takes the latter to be true. However, visible agreement between the possessee and the possessive P is exactly what my model predicts, even if P forms a constituent with the possessor, as I am claiming. Now what about case relations in the possessive PP construction? Yadroff & Franks (2001) suggest that functional prepositions are case features that are split off from the nominal phrase and realised separately. The problem with taking this approach to Scandinavian possessive PPs is that the complement of a possessive P can have overt case. We see this in the Faroese examples in (1), where the possessor has dative case. Similar constructions can also be found in Mainland Scandinavian. In (17a) I give an example from Solør Norwegian (my own dialect) as it was spoken a generation or two ago. (17) a.
Detta er romm-e åt jint-om. this is room-def.n.sg to girl-def.pl.dat ‘This is the girls’ room’.
(SolNor)
Marit Julien
b. Je ga de åt jint-om. I gave it to girl-def.pl.dat ‘I gave it to the girls.’
In this variety the preposition åt will also take a complement in the dative when it is not possessive, as in (17b). I conclude that the dative case of the possessor in (17a) is due to the preposition, and that the preposition itself is not the realisation of a case feature. It is the spellout of a possessive feature. Thus, although the possessive feature of the preposition agrees with the corresponding feature in n, there is no case relation between the two elements.
. Prenominal possessors I now turn to a discussion of prenominal possessors in Scandinavian. First, in 6.1, I present constructions with prenominal nonpronominal possessors. In 6.2 I show that these possessors make their containing DPs definite, and I give an explanation for this. In 6.3 I argue that prenominal possessors are in SpecDP in Scandinavian, and in 6.4 I explain why there is no suffixed definiteness marker on the possessed noun when the possessor is prenominal. In 6.5 I look at the case features of prenominal possessors, and finally, in 6.6, I show that my analysis of prenominal possessors is further supported by Solør Norwegian. . Prenominal nonpronominal possessors In Scandinavian varieties other than Icelandic, a nonpronominal possessor can be postnominal only if a preposition is inserted.7 If there is no preposition, the possessor shows up in prenominal position, and it is followed by a possessive -s, as in (18a), or by a pronoun-like possessive element, as in (18b). The latter construction, which I call ‘prenominal possessor doubling’, is found in Norwegian, in Jutlandic dialects of Danish, and also in Dutch and German varieties. The possessive -s is used in Danish and Swedish, and to some extent in Norwegian, although its acceptability depends on the dialect. (18) a.
nabo-en-s svart-e katte neighbour-def.m.sg-poss black-def catF ‘the neighbour’s black cat’ b. nabo-en si svart-e katte neighbour-def.m.sg poss.f.sg black-def catF ‘the neighbour’s black cat’
(Nor)
Possessor licensing, definiteness and case in Scandinavian
The Faroese counterpart of the possessive –s has a more expanded phonological form, as shown in (19). (19) a.
Jogvan-sa(r) bók Jogvan-poss book ‘Jogvan’s book’
b.
pápi min-sara bátur father my-poss boat ‘my father’s boat’
(Far)
It has been argued before that the two constructions in (18) (and the one in (19)) are structurally the same (Fiva 1987; Delsing 1993, 1998; Krause 1999). This appears to be true, at least for the standard varieties of Scandinavian, and I will treat them together in the following.8 . Prenominal possessors trigger definiteness Just like their postnominal counterparts, prenominal possessors trigger definiteness in their host DP in Scandinavian. This is true even of an indefinite possessor. We see this most clearly from the fact that if the possessee is modified by an adjective, the adjective must appear in the definite form. I demonstrate this in (20), using Norwegian examples, but exactly the same holds for other Scandinavian varieties. (20) a.
ei svart(*-e) katte indef.f.sg black-def catF ‘a black cat’ b. mi svart*(-e) katte my.f.sg black-def catF ‘my black cat’ c. nabo-en si svart*(-e) katte neighbour-def.m.sg poss.f.sg black-def catF ‘a neighbour’s black cat’ d. ein nabo si svart*(-e) katte indef.m.sg neighbour poss.f.sg black-def catF ‘a neighbour’s black cat’
(Nor)
In an indefinite DP, as in (20a), the adjective cannot appear in its definite form. But whenever a possessor is present, the definiteness marking of the adjective becomes obligatory, regardless of whether the possessor is pronominal, as in (20b), definite, as in (20c), or indefinite, as in (20d). This means that the possessed DP as a whole is definite. On my view, the definiteness of the possessed nominal phrases in (18), (19) and (20) means that the possessor originates in Spec-NP and has a possessive
Marit Julien
feature that agrees with the definiteness feature in n. This is what leads to the whole possessed DP being definite. However, in those varieties that have any of the constructions shown in (18) and (19), n has lost its ability to license case. Because of this a nonpronominal possessor does not get case inside nP if it does not combine with a preposition. When there is no preposition, licensing of the possessor requires the activation of a functional head Poss, located immediately below D (as proposed by many researchers earlier). To make the following points clearer, I will make reference to the example in (21), whose syntactic structure, according to my analysis, is as shown in (22). (Nor)
(21) ein nabo si katte indef.m.sg neighbourM poss.f.sg catF ‘a neighbour’s cat’ (22) Ü
Ü
Ü
Ü
On the assumption that nP is a phase in the sense of Chomsky (2001b) (a strong phase in the sense of Chomsky 2001a), and case licensing cannot take place across phase boundaries (Chomsky 2001b), it follows that the possessor cannot be case-licensed by Poss if it stays in its base position. It will have to move at least to Spec-nP. However, we see from the above examples that a prenominal possessor in Scandinavian precedes adjectives. Hence, if it is true that it starts out inside nP, it must have moved to an even higher position than Spec-nP. My
Possessor licensing, definiteness and case in Scandinavian
proposal is, as indicated in (22), that a prenominal possessor in Scandinavian sits in Spec-DP. As for the element that follows immediately after a prenominal possessor – that is, the possessive -s or its pronominal counterpart – I take it to be a realisation of the Poss head. Since it belongs to the extended projection of the possessee, the Poss head agrees with the possessee in gender and number, and it also shares the possessive/definiteness feature. If only the possessive feature of Poss is spelled out, we get the possessive -s, or in Faroese, -sa/-sar/-sara. The pronoun-like element seen in (18b) and (21) makes the phi-features of Poss visible in addition to the possessive feature.9 In either case the spellout of Poss will be the only overt manifestation of the possessive feature within the larger DP as a whole. . Prenominal possessors are in Spec-DP My reason for assuming that Scandinavian prenominal possessors are in SpecDP is the following. In definite DPs in Scandinavian, it appears to be the case that the D-projection needs to have phonologically overt material in it, and this material must have a nominal feature (see Julien 2002). In a DP where only the noun and its suffixes are visible, as in (23a), the whole nP has arguably moved to Spec-DP. I see this as the consequence of D attracting n, which agrees with D after D has had its features valued by n. Instead of moving on its own, n piedpipes the whole nP. But if there is an adjective (or a numeral) between nP and D, movement of nP is blocked. On my analysis, the adjective creates a defective intervention effect, since it agrees with D just like n does but lacks the nominal feature that is required of elements in the D-projection. The solution is to spell out D itself as a prenominal determiner, as in (23b). Note that in Norwegian, Swedish, and Faroese, the head noun will still have a suffixed definiteness marker. This is the so-called ‘double definiteness’ phenomenon.10 (23) a.
a’.
katt-a b. cat-def.f.sg ‘the cat’ [DP [nP katt-a] D nP] b’.
den svart-e katt-a def.sg black-def cat-def.f.sg ‘the black cat’ [DP [D den] [svart-e [nP katt-a]]]
(Nor)
Now when the DP contains a prenominal possessor, there is no prenominal determiner, and the possessor precedes the adjectives. If it is true that the DPlevel must have phonologically overt material in it, this must then be where the possessor is located.
Marit Julien
More direct evidence for this is found in Danish, where a prenominal pronominal possessor can be followed by a determiner in colloquial speech, as shown in (24a). Curiously, the determiner may not appear if there is no adjective present, hence the contrast between (24a) and (24b). (24) a.
min (den) my.sg def.sg ‘my black cat’ b. min (*den) my.sg def.sg ‘my cat’
sort-e kat black-def cat
(Dan)
kat cat
If the prenominal determiner is a realisation of the D head, which I will take to be the case, it follows that at least pronominal possessors in Danish move as high as Spec-DP. It does not then seem unreasonable to assume that other Scandinavian prenominal possessors move to Spec-DP as well, given that there is no indication that the constructions in question are very different syntactically. As for the question of why it is the possessor that moves up to Spec-DP, once it has made it to Spec-nP, and not the whole nP, let us consider the situation that arises when the possessor has moved to Spec-nP, driven by its case relation with Poss. The possessor is then in fact closer to D than n is, if we define closeness in terms of c-command, as in Pesetsky & Torrego (2001). Since the possessor agrees with D just as much as n does, and the possessor is also nominal, it is now the possessor that will be attracted to Spec-DP. There is one striking difference between nP-movement to Spec-DP and possessor movement to Spec-DP, however. While adjectives (and numerals) block movement of the nP, the possessor bypasses all adjectives and numerals on its way up. As far as I can see, this must be a consequence of the case relation that holds between Poss and the possessor. Since adjectives and numerals do not take part in this relation, the possessor is allowed to move across them, attracted by Poss. It is even possible that the possessor touches down in SpecPossP, as indicated in (22), before it moves on to its final destination in SpecDP. In Standard Danish and Standard Swedish, possessors are always prenominal, as I have already mentioned. That is, nonpronominal possessors appear with a possessive -s, whereas pronominal possessors, which themselves spell out a possessive feature, appear DP-initially without any accompanying markers. A Standard Danish pronominal possessor was shown in (24). Its Standard
Possessor licensing, definiteness and case in Scandinavian
Swedish counterpart is shown in (25a), and in (25b) I give an example of a Standard Swedish nonpronominal possessor. (25) a.
min (svart-a) katt my.sg black-def cat ‘my (black) cat’ b. en granne-s (svart-a) katt a neighbour-poss black-def cat ‘a neighbour’s (black) cat’
(Swe)
Given what was said above, it is easy to account for the nonpronominal possessor in (25b): since n is not a case licenser, and no possessive preposition is available, the possessor must be licensed by Poss, and consequently, it ends up in Spec-DP. A harder question is why pronominal possessors are also prenominal in Danish and Swedish. Note that they trigger definiteness on adjectives, which indicates that they agree with n, just like their postnominal counterparts in other Scandinavian varieties, and presumably, they start out in Spec-NP. My proposal is that the Poss head obligatorily appears in all possessed nominal phrases in Standard Danish and Standard Swedish. Hence, even pronominal possessors are attracted by Poss and will consequently also be the constituent that moves to Spec-DP. . The absence of the suffixed definiteness marker A noteworthy property of prenominal possessor constructions in Scandinavian is that the possessed noun has no suffixed definiteness marker, not even in the ‘double definiteness’ varieties.11 In this respect prenominal possessors contrast with postnominal possessors, which regularly go with suffixed definiteness markers, except in Icelandic, as we have seen. The question is then why there is no suffixed definiteness marker on the possessed noun when the possessor is prenominal. I believe it has to do with nP being a phase and with the movement of the possessor through Spec-nP. Note that elements that are never inside nP at any stage of the derivation, such as adjectives or determiners, do not block the realisation of n – see (23b), for example. However, adjectives and determiners agree with n and c-command n, just like the prenominal possessor does from its surface position. Moreover, Poss is routinely spelled out when a preceding possessor is also spelled out. Thus, being c-commanded by an agreeing element does not in itself prevent heads in the nominal projection from being spelled out. It seems that the effect of prenominal possessors on the realisation of n is a
Marit Julien
consequence of their local relation to n at the stage when the nP is completed. I take it that the decision concerning which elements in n will eventually get a phonological realisation is made at this stage (which is different from the decision of which elements will move out of nP). With the possessor sitting in the same projection as n and agreeing with n, only one of these elements needs to be spelled out. Hence, as long as the possessor gets a phonological realisation, n does not get one, and the consequence is that the possessed noun does not have a suffixed definiteness marker.12 It is interesting in this connection to consider nominal phrases with phonologically empty possessors, which are not uncommon in Scandinavian in cases where the possessor can be bound by an antecedent outside the possessed nominal phrase. I give an example in (26). (26) Honi skaka-de på [(ei ) huvud-et (ei )]. she shake-past on head-def.n.sg ‘She shook her head.’ or ‘She shook the head.’
(Swe)
Whether the possessor is postnominal or prenominal in this construction (there is the possibility that a phonologically empty possessor stays in its base position, even in the varieties where overt possessors move obligatorily), it is clearly the case that there is no overt possessor in Spec-nP at any stage of the derivation. And strikingly, the suffixed definite article must appear. This is another indication that the absence of the suffixed definiteness marker in (25a) and (25b) is due to the movement of the overt possessor.13 . Prenominal possessors and case In Scandinavian, possessors in the prenominal possessor doubling construction have no visible case – the varieties that have this construction have lost morphological case on nouns. But in German, these possessors have overt case. In Modern German, they have dative case, as in (27), but at an earlier stage they could appear in the genitive case, as in (28), which is Middle Low German, taken from Koptjevskaja-Tamm (2003a: 667). As for the element in Poss, we see that it shares the case of the possessee. In addition, it is an ordinary nonreflexive possessive pronoun, which follows from the fact that German have no other possessive pronominal elements to choose from. (27) Da ist (der) Anna ihr Wagen. there is def.f.sg.dat Anna her.m.nom carM ‘There is Anna’s car.’
(Ger)
Possessor licensing, definiteness and case in Scandinavian
(28) mein-es Vater-s sein Buch my-m.sg.gen father-gen his.sg.nom book.sg.nom ‘my father’s book’
(MidLowGer)
The genitive in (28) shows that the use of the prenominal possessor doubling construction is not necessarily a consequence of losing the genitive case on possessors. Instead, it is a consequence of losing the capacity to license a nonpronominal possessor inside nP. That is, none of the heads inside nP is a case licenser any more. Unless a preposition is inserted, a nonpronominal possessor can then only be licensed if the Poss head is activated, which triggers movement of the possessor to prenominal position. . An argument from Solør Norwegian Another argument that fronting of possessors has to do with their licensing can be made on the basis of Solør Norwegian. In this dialect, pronominal possessors are normally postnominal. In this position they can be focused, as in (29a).14 Alternatively, the focus can be on some other constituent inside the possessed nominal phrase, for example on the possessee, as in (29b), or on a constituent outside the possessed nominal phrase, as in (29c). (29) a.
Detta er te-kann-a this is tea-pot-def.f.sg ‘This is my teapot.’ b. Detta er TE-KANN-A this is tea-pot-def.f.sg ‘This is my teapot.’ c. DETTA er te-kann-a this is tea-pot-def.f.sg ‘This is my teapot.’
MI .
(SolNor)
my.f.sg mi. my.f.sg mi. my.f.sg
In (30) I show a pronominal possessor in prenominal position. In (30a) we see that a prenominal pronominal possessor can be focused. In (30b) and (30c) we see that this word order does not allow any other constituent than the possessor to be focused. In other words, a pronominal possessor can move to prenominal position only if it is focused. However, as (29) demonstrated, this movement is not obligatory.15 (30) a.
Detta er MI te-kanne. this is my.f.sg tea-potF ‘This is my teapot.’
(SolNor)
Marit Julien
b. *Detta er mi TE-KANNE. this is my.f.sg tea-potf ‘This is my teapot.’ te-kanne. c. *DETTA er mi this is my.f.sg tea-potf ‘This is my teapot.’
But when a nonpronominal possessor appears in prenominal position (in which case we have a prenominal possessor doubling construction, since the possessive -s is not used in this dialect), the focus may be on the possessor, as in (31a) and (32a), on the possessee, as in (31b) and (32b), or on a constituent outside the possessed nominal phrase, as in (31c) and (32c). (31) a.
Detta er a K ARI si this is she Kari poss.f.sg ‘This is Kari’s teapot.’ b. Detta er a Kari si this is she Kari poss.f.sg ‘This is Kari’s teapot.’ c. DETTA er a Kari si this is she Kari poss.f.sg ‘This is Kari’s teapot.’
(32) a.
(SolNor)
te-kanne. tea-potf TEKANNE.
teapotf tekanne. teapotf
Detta er NABO-N si this is neighbour-def.m.sg poss.f.sg ‘This is the neighbour’s teapot.’ b. Detta er nabo-n si this is neighbour-def.m.sg poss.f.sg ‘This is the neighbour’s teapot.’ si c. DETTA er nabo-n this is neighbour-def.m.sg poss.f.sg ‘This is the neighbour’s teapot.’
te-kanne. tea-potf
(SolNor)
TEKANNE.
teapotf tekanne. teapotf
It is thus evident that fronting of nonpronominal possessors has nothing to do with focus. It simply occurs obligatorily when the derivation does not include any element that can license the possessor in its base position – in other words, when the possessive P is not part of the numeration. Regardless of whether it is focused or not, the possessor then moves out of nP and ends up in Spec-DP. It could also be mentioned here that a possessive PP cannot be fronted, in this or other dialects. Neither the whole PP nor the possessor alone, without the preposition, can be moved to the front of its containing DP – see (33b) and (33c). It is also not possible to move the PP to the front of the clause, as
Possessor licensing, definiteness and case in Scandinavian
in (33d). It is not so bad, however, to move only the possessor, without the preposition, to clause-initial position, as in (33e). (33) a.
b. c. d. e.
Je liker itte bikkj-a åt NABO-N . I like not dog-def.f.sg to neighbour-def.m.sg ‘I don’t like the neighbour’s dog.’ *Je liker itte åt NABO-N bikkje/ bikkj-a. I like not to neighbour-def dog.indef/ dog-def *Je liker itte NABO-N bikkje/ bikkj-a åt. I like not neighbour-def dog.indef/ dog-def to *Åt NABO-N liker je itte bikkj-a. to neighbour-def like I not dog-def ?N ABO-N liker je itte bikkj-a åt. neighbour-def like I not dog-def to ‘The neighbour’s dog I don’t like.’
(SolNor)
A focused non-possessive PP, by contrast, may well be extracted out of a containing DP and moved to clause-initial position. Alternatively, the complement of the preposition may move on its own. I show this in (34). (34) a.
Je har lesi mange artikler om NOMINALFRASER. I have read many articles about nominal.phrases ‘I have read many articles about nominal phrases.’ b. Om NOMINALFRASER har je lesi mange artikler. about nominal.phrases have I read many articles ‘About nominal phrases, I have read many articles.’ c. N OMINALFRASER har je lesi mange artikler om. nominal.phrases have I read many articles about ‘nominal phrases, I have read many articles about.’
(SolNor)
And in Icelandic, where nonpronominal possessors are DPs with genitive case, such possessors can be focus fronted inside the possessed DP, as illustrated in (35). Recall that the possessor is licensed in postnominal position, so that moving it to prenominal position has nothing to do with its licensing. (35) a.
köttur(-inn) Jón-s/J ÓN-s cat-def.m.sg.nom Jón-gen ‘Jón’s/Jón’s cat’
b.
J ÓN -s köttur Jón-gen cat(nom) ‘Jón’s cat’
(Ice)
In short, whereas possessive PPs cannot move at all, other possessors may be fronted inside the containing nominal phrase, and other PPs can move to clause-initial position, which is also an option for a possessor embedded under a possessive P. How can this pattern be explained? It seems that there are very
Marit Julien
specific requirements on elements that undergo focus movement inside their containing nominal phrase. In fact, it seems that focus movement of this kind must target an element that agrees with n and thereby also with the higher heads in the functional domain of the possessed nominal. Presumably, if it does not agree, the focused element will not find a landing place higher up in that functional domain. In the case of pronominal possessors and DP possessors, the focused element is also the element that agrees with n. Consequently, these possessors can undergo focus movement inside the possessed nominal phrase. In the case of a focused possessive PP, on the other hand, it is the P that agrees with n whereas the complement of P bears focus. Since the P is not focused, the PP as a whole is not a target of focus movement, and pied-piping of the PP is not allowed. The possessor itself, without the P, does not agree with n and cannot find a landing place inside the possessed nominal phrase – it will have to move out of that phrase altogether. Finally, if a DP-internal PP is not possessive, it does not agree with any element inside the larger DP, and it will not be affected by focus movement inside that DP. It is however free to focus move out of that DP, and the same is true of the DP complement inside the PP.
. The definiteness of possessed DPs My claim that possessed nominal phrases in Scandinavian are necessarily definite contrasts with the widespread view that the definiteness of the possessor determines the definiteness of the possessed DP as a whole, a view put forward as early as Jackendoff (1968). For example, Schoorlemmer (1998) argues that this is the case when the possessor does not co-occur with a determiner. Since prenominal possessors do not normally occur with a determiner in Scandinavian, we might then expect these possessors to determine the definiteness of their containing DPs. But as we see, this is not borne out, and accordingly, Schoorlemmer’s generalisation does not seem to hold. It is however possible to find Scandinavian DPs that appear to be possessed and indefinite. I would nevertheless maintain that these DPs are either not really possessed or else they are definite after all. A DP that is only seemingly possessed has a constituent that is marked with a suffixed -s without being a real possessor. I give an example of this is (36a). Here the s-marked noun trøstens looks like a possessor. Still, the DP as a whole is clearly indefinite: the indefinite determiner eit has the neuter gender of the head noun ord and not the masculine gender of trøstens. But the thing is that trøstens does not have the possessor theta-role. A real possessor can be
Possessor licensing, definiteness and case in Scandinavian
added, as in (36b). Because of this, pseudopossessor is an appropriate term for constituents of the type exemplified by trøstens here. (36) a.
eit trøst-en-s ord indef.n.sg comfort-def.m.sg-s wordN ‘a word of comfort’ b. ditt trøst-en-s ord your.n.sg comfort-def.m.sg-s wordN ‘your word of comfort’ c. det etter-lengta trøst-en-s ord-et def.n.sg after-longed comfort-def-s word-def.n.sg ‘the longed-for word of comfort’
(Nor)
Closer examination reveals that pseudopossessors are very different from possessors syntactically. Quite unlike real possessors, a pseudopossessor in a definite DP is compatible with a preposed determiner and a suffixed definiteness marker on the head noun, as illustrated in (36c). We also see here that a pseudopossessor can be preceded by adjectives.16 In other words, it is located in the adjectival field of its host DP. In addition it is rather like an adjective semantically. Moreover, s-marked pseudopossessors are perfectly natural in those Norwegian dialects that do not use the possessive -s at all. All these facts indicate that the -s that appears on pseudopossessors is not identical to the possessive -s synchronically, although the two elements probably both have developed from an -s that marked genitive case. While the possessive -s is the realisation of a possessive feature, the -s that appears on pseudopossessors functions more like an adjectival derivational marker. Now the example in (37) seems to be a more serious challenge to the view that possessed DPs in Scandinavian are definite. Here we have a possessed DP as the associate of the expletive subject in the presentational construction. The ability to appear in this environment is commonly used as a test for indefiniteness, and one might want to conclude that the indefinite possessor makes the larger DP indefinite in this example. vel-brukt-e skjerf (37) Det ligg ein gamal mann-s it lies indef.m.sg old manM -poss well-worn-def scarfN (Nor) her. here ‘There is an old man’s well worn scarf lying here.’
There is no doubt that we are dealing with a real possessor, since another possessor cannot be added:
Marit Julien
(38) *mitt (ein) gamal mann-s skjerf my.n.sg indef.m.sg old man-poss scarf Intended meaning: ‘my [old man’s] scarf ’
(Nor)
There are however a couple of further facts to note about the possessed DP in (37). Firstly, the indefinite determiner belongs to the possessor, since it has the gender of the possessor and not the gender of the possessee. Secondly, the adjective following the possessor has definite form. The latter fact suggests that the possessed DP as a whole is definite, and the former means that there is no evidence that it is indefinite other than the syntactic environment that it appears in. It has been argued before, by Woisetschlaeger (1983), Lyons (1999), Vangsnes (1999), and others, that the definiteness effect seen in expletive constructions has to do with semantic properties that are distinct from formal definiteness. Woisetschlaeger (1983) notes that a formally definite DP behaves as an indefinite if it denotes a generic concept. More generally, it seems that formally definite DPs are acceptable in the postverbal position of existential constructions if they do not have a specific and inclusive interpretation.17 For the case at hand, the crucial fact seems to be that the possessor does not have a specific reference. If a specific interpretation is forced by adding viss ‘certain’ to the possessor phrase, the expression gets much less acceptable: (39) ?*Det ligg ein viss gamal mann-s vel-brukt-e skjerf her. (Nor) it lies a certain old man-s well-worn-def scarf here
In other words, a nonspecific indefinite possessor, as the one in (37), does not make its host DP semantically definite. The correct generalisation is then that the presence of a possessor in a Scandinavian DP makes that DP formally definite, in the sense that D and other functional heads have a definiteness feature with a positive value. There is an exception to this generalisation, though. We have already seen, in (8b), that for some speakers of Icelandic a nonpronominal and postnominal genitive possessor does not trigger definiteness marking on the possessed noun. In that particular example the possessed DP as a whole is nevertheless definite. But in (40) we see two more examples, which show that possessed Icelandic DPs can also be indefinite. In the absence of marking on the noun it is the indefinite form of the adjective that shows that these DPs are indeed indefinite.
Possessor licensing, definiteness and case in Scandinavian
(40) a.
% slitinn
hatt-ur worn.indef.m.sg.nom hatM -sg.nom manns manM .sg.gen ‘an old man’s worn hat’ b. % slitinn hatt-ur worn.indef.m.sg.nom hatM -sg.nom ‘a worn hat belonging to the teacher’
gamals old.indef.m.sg.gen (Ice)
kennar-ans teacher-def.m.sg.gen
So we see that the presence of a possessor does not in itself make the larger DP definite for all speakers of Icelandic. But note that the definiteness of the larger DP is not determined by the definiteness of the possessor either. In (40a) the possessor is indefinite and in (40b) it is definite, yet the larger DP is indefinite in both examples. Thus, for those speakers who have a grammar where a possessor does not trigger definiteness, a DP with a genitive possessor is free to be definite or indefinite. To avoid misunderstandings I want to point out here that I am not claiming that possessors universally make their containing DPs definite. The definiteness that possessed DPs display in Scandinavian is a consequence of the agreement relation between the possessor and n, on my analysis. In languages where there is no such agreement relation, a possessed DP is not necessarily definite. Italian provides an example: (41) a.
il mi-o car-o amico def. m.sg my-m.sg dear-m.sg friendMASC ‘my dear friend’ b. un mi-o car-o amico indef.m.sg my-m.sg dear-m.sg friendMASC ‘a dear friend of mine’
(Ita)
We see that a pronominal possessor in Italian is preceded by a determiner. If the determiner is the spellout of D, as I am assuming, this suggests that the possessor is sitting in Spec-PossP. Moreover, the determiner can be definite or indefinite. I take this to mean that between the possessor and the heads in the functional domain of the possessed nominal there is no agreement relation involving a possessive feature. The possessor nevertheless agrees with features related to the possessee, but the agreement only involves gender and number. Consequently, the possessor does not trigger definiteness, and since it does not share the definiteness feature of D, it will not be attracted to Spec-DP either. Instead, D itself is spelled out. As for the licensing of the possessor, it is probably taken care of by the Poss head alone.
Marit Julien
In short, we see from Italian that the syntax of possessors can be rather different from what we find in Scandinavian. It is possible that other patterns also exist. Hence, I am not claiming that the analysis I have presented here is necessarily valid outside of Scandinavian. I suspect there are languages that are like Scandinavian as far as possessor syntax is concerned, but I will not take up that discussion here.
. Summary I have argued in this paper that DP-internal possessors in Scandinavian all originate in a low position in the DP, in fact inside NP, and that they have a possessive feature that agrees with a corresponding feature in n. In n, this feature is interpreted as a definiteness feature, which is why possessed DPs are normally definite in this language group. In the case of pronominal possessors, the agreement relation just mentioned leads to licensing but not to case marking. This shows that case is not just a reflex of agreement. I have also tried to show how a number of different possessor constructions found in Scandinavian can all be derived from the same basic syntactic structure. However, the variation in this area is so great that much more space would be needed to do full justice to each construction. What I have given here is a sketch that presents the main features of my analysis of Scandinavian possessors.
Notes * I would like to thank the audience at CGSW 19 for their feedback when I first presented the ideas laid out in this paper. I also thank Marcel den Dikken, Christina Tortora and Øystein Vangsnes for many useful suggestions. . The following abbreviations are used in the glosses: dat = dative, def = definite, f = feminine, g = genitive, indef = indefinite, m = masculine, n = neuter, nom = nominative, pl = plural, poss = possessive, refl = reflexive, sg = singular. . In other words, I do not to treat these constructions as cases of inalienable possession. Support for this view is found in Bonneau, Pica & Nakajima (1999), where it is claimed that inalienable possession requires an animate possessor. . Note that numerals are not generated in the Num projection. Numerals are found much higher up in the DP, above all adjectives, as (2b) illustrates. . Similarly, Amharic has a marker that can be interpreted as a definiteness marker or as a third person masculine singular possessive marker (Girma A. Demeke, p.c.):
Possessor licensing, definiteness and case in Scandinavian
(i)
mäs’haf-u book-def/poss.3m.sg ‘his/the book’
. There exist dialects of Norwegian and Swedish that have the pattern shown in (4). . Literary Faroese is similar, but since this is not how the language is spoken today, I ignore it. . This is true of the standard varieties. In some Northern Swedish dialects one can nevertheless find constructions like those in (i) (see Delsing 2003; Holmberg & Sandström 1996). Here the nonpronominal possessor combines directly with the possessed noun, and the possessor either has dative case, as in (i-a), or no visible case at all, as in (i-b). (i)
a.
b.
hus-e pojk-om house-def.n.sg boy-def.sg.dat ‘the boy’s house’ bil-n präst-n car-def.m.sg priest-def.m.sg ‘the priest’s car’
But note that the possessee obligatory has a suffixed definite article in the dialects in question. This indicates that in these dialects the nonpronominal postnominal possessor is licensed by n. . Some dialects have a possessive -s with a different syntax (see Julien to appear). . In Norwegian, the element that follows the possessor in the prenominal possessor doubling construction is formally identical to the reflexive possessive pronoun. The latter is exemplified in (i). (i)
si. {OK Kari/*eg} er glad i katt-a Kari/ I is fond in cat-def.f.sg poss.f.sg ‘Kari/*I is fond of her (own) cat.’
On my view, the explanation for this similarity is that the same set of features is spelled out in both cases. It does not matter that the syntactic environment is different. Regardless of whether it spells out Poss or sits in a possessor position, the ‘reflexive’ possessive pronoun si spells out a possessive feature, plus gender and number features that ultimately stem from the possessee. Note that the features of the possessor are not represented at all in this element. Hence, when it appears in possessor position, as in (i), it can only refer to a possessor if it is bound, and its antecedent must be third person, since if the antecedent is first or second person the corresponding first or second person possessive pronoun will win the competition for lexical insertion. In some West Jutlandic dialects, nonreflexive possessive pronouns can be used in the prenominal possessor doubling construction. I give an example in (ii). (ii) Jens hans støvler Jens his boots ‘Jens’s boots’
Marit Julien
Here the element that spells out Poss does reflect the features of the possessor – which are third person masculine singular in our current example. Parallel constructions are found in Dutch and German. The reason why a nonreflexive possessive pronoun is used could be that the variety in question has no possessive pronominal element that does not reflect the features of the possessor (Dutch and German) or that the choice between the ‘reflexive’ pronominal element and an ordinary possessive pronoun depends on other factors. In those West Jutlandic dialects where (ii) is accepted, the ordinary possessive pronoun can appear in Poss if the possessor is animate. Otherwise a ‘reflexive’ possessive pronoun is used (see Nielsen 1986). . A more detailed version of the analysis sketched here is given in Julien (2002). . This is true of Norwegian, Faroese, and Standard Swedish. A number of Swedish dialects are exceptional in that prenominal possessors are compatible with suffixed possessed nouns, as in the following example from Delsing (2003: 35) (see also Holmberg & Sandström 1996; Vangsnes 1999): (i)
Per-s hus-et Per-poss house-def.n.sg ‘Per’s house’
I ignore these varieties here and concentrate on the grammar of the main varieties. . I take the restriction against spelling out n and a possessor in Spec-nP to be parameterised. There are dialects without this restriction. For example, Lappträsk Swedish, a dialect spoken in Finland, has constructions like the following (from Vangsnes 1999: 157): (i)
mett te stór hús-e my.n.sg def.sg big house-def.n.sg ‘my big house’
We see here that the suffixed definiteness marker, which I take to be a realisation of n, and the prenominal determiner, which I take to be a realisation of D, both appear despite the presence of the prenominal possessor, which on my analysis moves through Spec-nP and gets spelled out in Spec-DP. I conclude that at least in the nominal domain, this dialect has no restriction on spelling out Spec and head simultaneously. A prenominal possessor and a prenominal determiner can also co-occur in colloquial Danish (see (24a)). Hence, it appears that colloquial Danish has no such restriction either. The reason why there is no suffixed definiteness marker in (24b) is that n is never spelled out in Danish anyway. Even suffixed definiteness markers in Danish are arguably realisations of D (see Julien 2002). . In Norwegian, we find constructions that look similar to (26), but since Norwegian allows overt pronominal possessors to be postnominal, it is likely that the empty possessor in (i) is also postnominal. Hence, n is spelled out as it always is when the possessor is postnominal. (i)
med [vesk-a ei ]. Hoi sving-te she swing-past with bag-def.f.sg ‘She swung her bag.’ or ‘She swung the bag.’
Possessor licensing, definiteness and case in Scandinavian
This example also shows that a phonologically empty possessor is not contingent on a relation of inalienable possession. However, the precise characterisation of the relations that allow the possessor to be phonologically empty remains to be investigated. . Delsing and Egerland (2002) claim that postnominal pronominal possessors are not stressed in Mainland Scandinavian. We see now that this generalisation is not true of Solør Norwegian. Moreover, as far as I have been able to establish it is not true of Norwegian in general. . A possible interpretation of this is that a focus feature high up in the nominal phrase attracts the focused possessor. The question is then what to say about focused possessors that stay in place. I will not go into this discussion here, but I would like to point out that a parallel phenomenon is seen in the clausal syntax of Scandinavian: focused objects, for example, may move to the front of the clause or stay in place. Hence, whatever accounts for the latter fact will probably also explain the behaviour of Solør Norwegian pronominal possessors. . See Koptjevskaja-Tamm (2003b) and Julien (to appear) for more details. . I follow Lyons (1999) and use the term ‘inclusive’ instead of the more common ‘unique’.
References Bonneau, J., P. Pica, & T. Nakajima (1999). A restrictive/non-restrictive distinction in possessive nominals. In K. Shahin, S. Blake, & E.-S. Kim (Eds.), Proceedings of WCCFL 17 (pp. 88–101). Stanford, CA: CSLI Publications. Carstens, V. (2001). Multiple agreement and case deletion: Against φ-(in)completeness. Syntax, 4, 147–163. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (2000). Minimalist inquiries: The framework. In R. Martin, D. Michaels, & J. Uriagereka (Eds.), Step by Step: Essays in minimalist syntax in honor of Howard Lasnik (pp. 89–155). Cambridge, MA: The MIT Press. Chomsky, N. (2001a). Derivation by phase. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 1–52). Cambridge, MA: The MIT Press. Chomsky, N. (2001b). Beyond explanatory adequacy. Ms., MIT. Delsing, L.-O. (1991). Om genitivens utveckling i fornsvenskan. In S.-G. Malmgren & B. Ralph (Eds.), Studier i svensk språkhistoria 2 (pp. 12–30). Acta Universitatis Gothoburgensis, Gothenburg. Delsing, L.-O. (1993). The Internal Structure of Noun Phrases in the Scandinavian Languages. PhD Dissertation, Lund University. Delsing, L.-O. (1998). Possession in Germanic. In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the Determiner Phrase (pp. 87–108). Amsterdam: John Benjamins. Delsing, L.-O. (2003). Syntaktisk variation i skandinaviska nominalfraser. In Ø. A. Vangsnes, A. Holmberg, & L.-O. Delsing (Eds.), Dialektsyntaktiska studier av den nordiska nominalfrasen (pp. 11–64). Oslo: Novus.
Marit Julien
Delsing, L.-O. & V. Egerland (2002). Kinship nouns in possessive constructions in Italian and Scandinavian. Sprachtypologie und Universalienforschung, 55, 103–120. Faarlund, J. T. (2004). The Syntax of Old Norse. Oxford: Oxford University Press. Fiva, T. (1987). Possessor Chains in Norwegian. Oslo: Novus. Giusti, G. (1994). Enclitic articles and double definiteness: A comparative analysis of nominal structure in Romance and Germanic. The Linguistic Review, 11, 241–255. Holmberg, A. & G. Sandström (1996). Scandinavian possessive constructions from a Northern Swedish viewpoint. In J. R. Black & V. Motapanyane (Eds.), Microparametric Syntax and Dialect Variation (pp. 95–120). Amsterdam: John Benjamins. Jackendoff, R. (1968). Possessives in English. In S. R. Anderson, R. Jackendoff, & S. J. Keyser (Eds.), Studies in Transformational Grammar and Related Topics (pp. 25–51). Waltham, MA: Department of English, Brandeis University. Julien, M. (2002). Determiners and word order in Scandinavian nominal phrases. Studia Linguistica, 56, 264–314. Julien, M. (to appear). Nominal Phrases from a Scandinavian Perspective. Amsterdam: Elsevier. Kester, E.-P. (1993). The inflectional properties of Scandinavian adjectives. Studia Linguistica, 47, 139–153. Koptjevskaja-Tamm, M. (2003a). Possessive noun phrases in the languages of Europe. In F. Plank (Ed.), Noun Phrase Structure in the Languages of Europe (pp. 621–722). Berlin: Mouton de Gruyter. Koptjevskaja-Tamm, M. (2003b). A woman of sin, a man of duty, and a hell of a mess: Non-determiner genitives in Swedish. In F. Plank (Ed.), Noun Phrase Structure in the Languages of Europe (pp. 515–558). Berlin: Mouton de Gruyter. Krause, C. (1999). Two notes on prenominal possessors in German. MIT Working Papers in Linguistics, 33, 191–217. Lyons, C. (1999). Definiteness. Cambridge: Cambridge University Press. Nielsen, B. J. (1986). Om pronominet sin i jysk. Danske Folkemål, 28, 41–101. Pesetsky, D. & E. Torrego (2001). T-to-C movement: Causes and Consequences. In M. Kenstowicz (Ed.), Ken Hale: A life in language (pp. 355–426). Cambridge, MA: The MIT Press. Reinhart, T. & E. Reuland (1993). Reflexivity. Linguistic Inquiry, 24, 657–720. Ritter, E. (1993). Where’s gender? Linguistic Inquiry, 24, 795–803. Sandström, G. & A. Holmberg (1994). Adjective incorporation and the syntax of the Scandinavian noun phrase. Department of General Linguistics, University of Umeå, Report, 35, 81–97. Santelmann, L. (1993). The distribution of double determiners in Swedish: Den support in D. Studia Linguistica, 47, 154–176. Schoorlemmer, M. (1998). Possessors, articles and definiteness. In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the Determiner Phrase (pp. 55–86). Amsterdam: John Benjamins. Sigurðsson, H. Á. (1993). The structure of the Icelandic NP. Studia Linguistica, 47, 177–197. Taraldsen, K. T. (1990). D-projections and N-projections in Norwegian. In J. Mascaró & M. Nespor (Eds.), Grammar in Progress (pp. 419–431). Dordrecht: Foris.
Possessor licensing, definiteness and case in Scandinavian
Vangsnes, Ø. A. (1999). The Identification of Functional Architecture. PhD Dissertation, University of Bergen. Vikner, C. & P. A. Jensen (2002). A semantic analysis of the English genitive. Interaction of lexical and formal semantics. Studia Linguistica, 56, 191–226. Woisetschlaeger, E. (1983). On the question of definiteness in ‘An old man’s book’. Linguistic Inquiry, 14, 137–154. Yadroff, M. & S. Franks (2001). The origin of prepositions. In G. Zybatow, U. Junghanns, G. Mehlhorn, & L. Szucsich (Eds.), Current Issues in Formal Slavic linguistics (pp. 69–79). Frankfurt am Main: Peter Lang. Zamparelli, R. (1998). A theory of kinds, partitives and of/z possessives. In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the Determiner Phrase (pp. 259–301). Amsterdam: John Benjamins.
Pronouns are determiners after all* Dorian Roehrs Indiana University
This paper provides new arguments for the claim that pronominal constructions such as us linguists involve pronominal determiners taking a complement (cf. Postal 1966), rather than appositives of some sort (Jespersen 1914; Cardinaletti 1994). Discussing pronouns of different morphological person in combination with nouns of different semantic type (common and proper), the paper generalizes to pronominal determiners Brugè’s (1996) hypothesis that demonstratives are base-generated in an intermediate phrase and move to DP. Furthermore, it provides empirical evidence that pronominal determiners may not only be strong, as traditionally assumed, but also weak and clitic-like, thereby extending Cardinaletti & Starke’s (1999a) influential tripartite division of “intransitive” to “transitive” pronouns.
.
Introduction
Following some suggestions to revise the traditional structural description of noun phrases such as the linguists (e.g. Brame 1981, 1982; Szabolcsi 1983), Abney (1987) more extensively discusses the DP-Hypothesis, according to which the determiner D is a functional head which projects its own phrase DP and takes NP as its complement (for recent overviews of the DP-Hypothesis, see Bernstein 2001; Coene & D’hulst 2003; Longobardi 2001; for typological surveys, see Plank 2003; Rijkhoff 2002): (1)
DP D the
NP linguists
Abney (1987: 281–284) briefly discusses pronominal constructions such as we linguists, which involve a pronoun and a noun. Following Postal (1966), he sug-
Dorian Roehrs
gests that the pronoun is a determiner and generalizes the structure of the DP in (1) to these pronominal noun phrases (although I put the pronoun under D, I will suggest below that pronouns can be in both D and Spec,DP): (2)
DP D we
NP linguists
I will call this proposal “the General DP-Hypothesis”. This extension is far from uncontroversial. Jespersen (1914: 85) suggests that these pronominal DPs are cases of apposition (cf. also Olsen 1991: 37; Diessel 1999: 67), which, in current generative grammar, can be structurally represented by adjunction (e.g. Delorme & Dougherty 1972: 8; Cardinaletti 1994: 203).1 There are two main options (for discussion of appositions, see Lawrenz 1993; also Meyer 1992). Traditionally, phrases such as Peter my friend have been analyzed as “loose” appositives, where the second part my friend involves a phrase (XP). Extending this approach to pronominal DPs, one could claim that we linguists involves the phrase linguists adjoined to another phrase (in the spirit of Emonds 1976): (3) a.
DP DP we
b.
XP linguists DP
D we
YP XP linguists
YP Y
(ZP)
Considering the representations in (3) in more detail, the assumed phrase linguists is adjoined to DP in (3a) and to a lower phrase (YP) in (3b). With regard to the pronoun in (3b), there are two options: on the one hand, the pronoun could be base-generated in D; on the other hand, it could have moved there by N-to-D raising as in Cardinaletti (1994). As the second main option, phrases such as my friend Peter have been taken to be “close” appositives. For this approach, suppose that the head Peter is ad-
Pronouns are determiners after all
joined to the head friend. Extending this type of analysis to pronominal DPs, we obtain the following representations: (4) a.
DP D D we
b.
N linguists DP
D we
NP N1 N1
N2 linguists
Considering (4), the second part linguists can be adjoined to the pronoun in D or to a lower head, for instance, to N. Similar to (3b), the pronoun in (4b) could be base-generated in D or could have moved from a lower position. I will call this general line of investigation “the Appositive Hypothesis”. The main goal of this paper is to argue against the Appositive Hypothesis and in favor of the General DP-Hypothesis. To this end, I bring to bear a number of new arguments, most of which come from German. As a minor and more tentative objective, I extend Cardinaletti & Starke’s (1999a) influential tripartite division of “intransitive” to “transitive” pronouns, that is pronominal determiners, despite the fact that Cardinaletti (1994) and Cardinaletti and Starke (1999b: 278) explicitly deny this. I will provide empirical evidence that pronominal determiners may not only be strong, as traditionally assumed, but also weak and clitic-like. The organization of the paper is as follows: after briefly dismissing the possibility that pronominal DPs involve “close” appositives, Section 2 provides arguments from adjective inflection and Idiot-nouns that “loose” appositives cannot be considered the correct analysis either. Section 3 discusses the combination of pronouns and proper names, deriving more arguments for the General DP-Hypothesis. In Section 4, we turn to some consequences of the current proposal for Cardinaletti (1994). After cataloguing the different types of DPs and commenting on them, I summarize the discussion in Section 6. In the appendix, I briefly discuss singular pronominal DPs. Due to the length of
Dorian Roehrs
this paper, I will not be able to review in detail previous arguments in favor of the General DP-Hypothesis (e.g. Postal 1966; Pesetsky 1978; Abney 1987: 281– 284; Lawrenz 1993: esp. 81–93; Uriagereka 1995: esp. 79–81; Corver & Delfitto 1999: esp. 802–804; Lyons 1977, 1999: esp. 26–30, 141–145; Furuya to appear).
. Pronominal DPs must have a determiner As a first step in the investigation, we check if the pronominal construction is a syntactically freely composed structure and not some type of “fixed” expression (it is not my intention at this point to argue for or against either of the two competing hypotheses, i.e. the General DP or Appositive Hypothesis). To this end, consider three properties that pronominal DPs share with “regular” DPs. First, just as with other DPs, the noun proper can be modified by an adjective and the adjective itself can be modified by an adverb (5a). Second, there is also an ordering requirement on the adjectives (Sproat & Shih 1991). Compare (5b) to (5c): (5) a.
{ich I b. {du you c. *{du you
/ / / / / /
ein} a ein} a ein} a
völlig blöder Idiot totally stupid idiot großes schwarzes deutsches big black German deutsches schwarzes großes German black big
Schwein pig Schwein pig
Third, as illustrated above, pronouns occur with countable animate nouns. In the cases where a mass noun can be used as a term of endearment, a classifier has to be used (cf. Stavrou 2003): (6) a. you *(sweet bit of) honey b. a *(bit of) honey
Having illustrated these three properties, I take the pronominal construction to be a syntactically varying expression (and not some kind of idiom). To narrow down the choices for the correct analysis, consider the two main types of adjunction again. In the introduction, we considered possible syntactic representations of “close” and “loose” appositives. One of the interesting properties of “close” appositives is that they, being heads adjoined to heads, do not allow true modifiers (for more properties, see Lawrenz 1993: Chap. 2; Lindauer 1995: 47; Löbel 1991; Lühr 1991):2
Pronouns are determiners after all
(7) a.
meine my b. meine my
Freundin girl-friend Freundin girl-friend
(*liebe) Maria (dear.strong/weak) Mary KleinMaria Little(no infl) Mary
Note that Klein- ‘little’ in (7b) has no inflection and is presumably part of a complex name. As already illustrated above, the pronominal DP allows inflected adjectives. Assuming the adjective to be part of the adjoined structure, we conclude that these constructions cannot involve “close” appositives. Notice also that with the exception of vocatives and exclamations, adjectives modifying singular countable nouns require the presence of an overt determiner, even in predicative contexts: (8) a.
Meine my b. Meine my
Mutter mother Mutter mother
ist is ist is
(eine) Lehrerin. a teacher *(eine) gute Lehrerin. a good teacher
Since all pronominal DPs allow adjectives, we conclude furthermore that they all must have a determiner. As a “regular” determiner is not present, the question arises how the presence of adjectives is licensed. Several options present themselves: on the one hand, one could claim that due to adjunction, the “regular” determiner has been phonetically deleted in the appositive (9a). On the other hand, one could suggest that the pronoun itself functions as the determiner (9b). Consider the simplified structures:3 (9) a.
Appositive Hypothesis [DP [ pronoun ] [DP determiner adjective noun ]] b. General DP-Hypothesis [DP pronoun [ adjective noun ]]
There is a third basic possibility, as in Cardinaletti (1994), where the adjective is part of the “matrix DP” and only the noun (phrase) is adjoined (see Section 4 for discussion). Rather than attempting to answer the question of what allows the deletion of the “regular” determiner in the adjoined element, I will propose below that the pronoun itself is the determiner.4 In the remainder of this section, I present two arguments that the pronominal construction must have determiners. Besides the general requirement noted above that adjectives need to be licensed by an overt determiner, there is evidence from adjectival inflection and Idiot-nouns that a determiner must be present.
Dorian Roehrs
. Adjectival morphology In this section, I provide an argument that deletion of the “regular” determiner cannot occur at PF. Bearing in mind that adjectives need to be licensed by a determiner, I argue that the pronoun fulfills this function. As such, this section provides an argument for the General DP-Hypothesis. To set the stage for the discussion, consider first some typical “loose” appositives, involving a mass noun and a plural indefinite noun. Both of these noun phrases are modified by an adjective and have a null determiner in D in Germanic (cf. Maria, *(die/eine) nette Studentin ‘Mary, the/a nice student’). What is crucial here is that both adjectives must have a strong ending: (10) a.
Das ist Liebfrauenmilch, sehr süßer / This is Liebfrauenmilch, very sweet(strong) / Wein. wine b. Das sind wir, sehr nette / *netten That is us, very nice(strong) / nice(weak)
*süße sweet(weak)
Mädchen. girls
Turning to the pronominal DP, although the judgments are sharp in the nominative/accusative singular in (11) and in the dative/accusative plural in (13), Bhatt (1990: 154–155) observes that both a strong and a weak ending are possible in the dative singular and the nominative plural in (12) (see also Darski 1979: 200; Duden 1995: 280). These sets of data are illustrated with pronominal constructions in the first person (there is no pronominal construction in the genitive): (11) a.
Ich dummes / *dumme Schwein habe meinen Job I stupid(nom.st) / stupid(wk) pig(n) have my job verloren! lost ‘I (stupid pig) have lost my job.’ b. Sie haben mich dummes / *dumme Schwein erwischt! they have me stupid(acc.st) / stupid(wk) pig(n) caught ‘They have caught me (stupid pig).’
(12) a.
Sie haben mir dummem / %dummen Esel Geld they have me stupid(dat.st) / stupid(wk) donkey(m) money geklaut! stolen ‘They stole money from me (stupid donkey).’
Pronouns are determiners after all
b. Sie haben mir %dummer / dummen Gans Geld they have me stupid(dat.st) / stupid(wk) goose(f) money geklaut! stolen ‘They stole money from me (stupid goose).’ c. Wir dumme / dummen Idioten haben unseren Job we stupid(nom.st) / stupid(wk) idiots have our job verloren! lost ‘We (stupid idiots) have lost our job.’ (13) a.
Er hat uns dumme / *dummen Idioten erwischt! he has us stupid(acc.st) / stupid(wk) idiots caught ‘He has caught us (stupid idiots).’ b. Sie haben uns dummen Idioten Geld geklaut! they have us stupid(dat.st/wk) idiots money stolen ‘They stole money from us (stupid idiots).’
While the strong and weak endings are equally possible in the nominative plural (12c), there seems to be a preference for the strong ending in the dative masculine and for the weak ending in the dative feminine (12a–b). Note in passing that German allows singular pronominal DPs in argument position. As such, it is similar to Warlpiri and Nama (Lyons 1999: 142) but different from many other languages (see the appendix for some discussion of singular pronominal DPs). Chart 1 summarizes the general types of possible adjective endings. Before we turn to the Appositive Hypothesis to see how it fares with these data, consider the basic distribution of adjective endings in German in schematic terms. While the adjective is weak if preceded by the definite determiner der ‘the’ (14a), it is strong if nothing (overt) precedes it (14b). With the indefinite determiner ein ‘a’, we find “mixed” inflection (14c): in the nominative masculine singular and the nominative/accusative neuter singular, the adjective is strong; in the remaining instances, it is weak: (14) a. [ der(strong) adjective(weak) ] b. [ adjective(strong) ] c. [ ein(strong/weak) adjective (weak/strong) ]
The presence of an adjective requires a determiner in the adjoined noun phrase. If a “regular” (in-)definite determiner were covertly present, we would expect only very sharp judgments with the corresponding set of endings (see Duden 1995 for details). In other words, there should be no variation, in contrast to the facts. To illustrate, consider two scenarios where the adjectival endings are
Dorian Roehrs
Chart 1.
nominative accusative dative genitive
singular
plural
strong strong strong/weak –
strong/weak strong strong/weak –
licensed either before the determiner is phonetically deleted or, alternatively, afterwards. If the indefinite determiner ein ‘a’ were present at the time the endings are licensed in the singular (14c), the strong endings in the nominative and accusative and the weak ending in the dative could be explained (the “mixed” inflection). However, the strong ending in the dative would be unaccounted for. Furthermore, assuming a null plural indefinite determiner (14b), the strong adjectival endings in the plural fall into place but not the weak ones. Alternatively, if the definite determiner were present before the adjectival endings are licensed (14a), then we could explain all the weak endings, leaving the strong endings unaccounted for. Turning to the second scenario, where the determiners are deleted before the adjectival endings are licensed (14b), the adjective would not be preceded in its noun phrase and we could explain the strong endings (similar to the regular appositives in (10)). However, the weak endings become a mystery. I conclude that neither deletion of the determiner before nor after licensing of the adjectival ending can provide an account of the full picture summarized in Chart 1 and a different solution is needed. The alternative is to propose that the pronoun itself is the determiner. Postal (1966) argued that the pronoun in the pronominal construction is a kind of determiner (see also, among many others, Abney 1987: 281–284; Bhatt 1990: 151; Corver & Delfitto 1999: esp. 802–804; Furuya to appear; Hudson 1987: 122–123; Jackendoff 1977: 106; Lawrenz 1993: esp. 81–93; Longobardi 1994: 636; Lyons 1977, 1999: esp. 26–30, 141–145; Panagiotidis 2002a: 39–40; Pesetsky 1978; Radford 1993: 77, 109, 1997: 154; Sommerstein 1972; Uriagereka 1995: 79–81). I will call this instance of the determiner the “pronominal determiner” or simply “pronoun”. Turning to the adjectival inflections in more detail, German has a languagespecific rule that regulates the distribution of the strong inflection on determiners and adjectives. This Principle of Monoinflection states that the strong ending is licensed on the first overt “agreeing” category, be it the determiner
Pronouns are determiners after all
or the adjective (for detailed discussion and references, see Roehrs 2002, in progress): (15) Principle of Monoinflection The first overt agreeing category within a noun phrase carries the strong and the second one the weak ending.
In order to account for the data, two additional assumptions are necessary. First, unlike regular determiners, I assume that pronominal determiners may be analyzed by speakers as both agreeing and non-agreeing. Independent evidence for this possibility comes from the determiner-like element manch‘some’, which is provided for comparison below. We start with the instances that allow both a strong and a weak ending (cf. (12)). If the pronoun in the pronominal construction in the plural is analyzed as agreeing, then it will be the first relevant overt element and according to the Principle of Monoinflection, it receives the strong ending (alternatively, “absorbs” it). Similarly for manch- ‘some’: (16) a.
wir netten Leute we(strong) nice(weak) people b. manche netten Leute some(strong) nice(weak) people
In contrast, if the pronoun is analyzed as non-agreeing in the relevant sense, then the adjective will be the first relevant overt item and it gets the strong ending. Again, the same holds for manch- ‘some’:5 (17) a.
wir nette Leute we nice(strong) people b. manch nette Leute some nice(strong) people
Turning to the varying preferences in the dative singular, speakers seem to prefer to analyze mir ‘me’ as non-agreeing in the masculine and as agreeing in the feminine. I hypothesize that this has to do with the endings of the pronoun and their (dis-)similarity to those of the definite determiner. While the final -r in mir does not coincide with -m of masculine dem ‘the’, it does in the feminine with der ‘the’. Consequently, speakers seem to prefer the strong ending on the adjective in the masculine but the weak one in the feminine (see Bhatt 1990: 154 fn. 4 for a similar explanation for the masculine cases):
Dorian Roehrs
(18) a.
mir dumme-m / %dummen Esel me stupid(dat.strong) / stupid(weak) donkey(m) b. de-m *dummem / dummen Esel the(dat.strong) stupid(dat.strong) / stupid(weak) donkey(m)
(19) a.
mi-r %dummer / dummen Gans me stupid(dat.strong) / stupid(weak) goose(f) b. de-r *dummer / dummen Gans the(dat.strong) stupid(dat.strong) / stupid(weak) goose(f)
Next, consider the instances that involve very sharp judgments. First, we turn to the nominative and accusative singular (cf. (11)). Generalizing Brugè’s (1996) work on demonstratives, I propose in Roehrs (2002, in progress) that not only demonstratives but also (in-)definite determiners move from an article phrase (artP) into the DP. Assuming that pronouns are determiners, we can illustrate this for ich dummes Schwein ‘I stupid pig’ as follows: (20)
Ü
Besides Brugè (1996), there are other scholars who have argued for such a low position of the demonstrative (Campbell 1996; Giusti 1997; Panagiotidis 2000, see also Section 4).6 Furthermore, considering the fact that demonstratives and determiners are historically related, it is not surprising that there is overt evidence for the lower determiner position art (e.g. for the Scandinavian languages, see Julien 2002; Vangsness 1999, 2004). The second assumption for explaining the German data is that determiners may move to the DP at different times, that is, before and after the strong ending is licensed. By assumption, determiners in inherently case-marked and/or plural DPs move to the DP before the adjective endings are licensed. Consequently, these pronouns will precede the adjective at the time of licensing, leaving the discussion above intact. In contrast, singular structurally case-marked pronominal determiners move to DP after the strong ending is licensed. Crucially, at the time the strong ending is licensed in (21b), the adjective is the first overt element and gets a strong ending.7
Pronouns are determiners after all
Chart 2.
ACC: uns, euch DAT: uns, euch
weak
strong
-en -en
-e -en
(21) a.
ich dummes Schwein I stupid(strong) pig(n) b. [DP D [artP dummes [artP ich [NP Schwein ]]]] c. [DP ichi +D [artP dummes [artP ichi [NP Schwein ]]]]
Unlike in the cases above, there can be no variation here, induced by the different analyses of the pronoun as agreeing or non-agreeing, as the pronoun is simply following the adjective at the relevant time. This accounts for the lack of the weak ending and the sharp judgments. Turning to the accusative and dative plural (cf. (13)), if it is correct that plural pronouns precede the adjective at the time the strong ending is licensed, then we would also expect variability in these cases. Surprisingly, however, the accusative allows a strong ending only. To explain this, I will provide a functional account. In the context of the present discussion, Darski (1979: 203) observes that the accusative and dative plural are the only instances where the pronouns themselves are not distinguished in their marking as wir ‘we’ and ihr ‘you’ are invariably uns ‘us’ and euch ‘you’. In other words, the adjective with the possibility of varying inflection is the only candidate to bring about a different marking between the two cases. Note in Chart 2 that the adjectival endings in the dative are the same, rendering them ambiguous between a strong and a weak inflection. If a clear morphological difference between the dative and the accusative is to be made, the only possibility is for the accusative adjective to be strong. That this line of reasoning is not implausible becomes apparent when we consider the relatively free word order in German. With the possibility of Scrambling of objects in German, the interpretation of the objects involved is ambiguous. However, the presence of adjectives with the different endings resolves this issue:8 (22) a.
Sie haben uns netten Jungen euch kluge they have us nice(dat.st/wk) boys you smart(acc.st) Mädchen vorgestellt. girls introduced ‘They introduced to us (nice boys) you (smart girls).’
Dorian Roehrs
b. Sie haben uns nette Jungen euch klugen they have us nice(acc.st) boys you smart(dat.st/wk) Mädchen vorgestellt. girls introduced ‘They introduced us (nice boys) to you (smart girls).’
To summarize this subsection, we first considered the Appositive Hypothesis and pointed out that it cannot account for the full set of data, independent of what “regular” determiner is used or when it is phonetically deleted. As an alternative, we proposed that the pronoun itself is the determiner. Assuming that pronouns can be analyzed as agreeing and non-agreeing, that they move from a position below the adjective, and that this movement may occur for different determiners at different times, we accounted for most of the data, both with regard to variation and clear lack thereof. For the remainder of the examples, we proposed a functional account. Having provided an argument against the deletion of the “regular” determiner at PF, and thus against the Appositive Hypothesis, we now turn to Idiot-Nouns. . Idiot-noun phrases require a determiner Above, we pointed out that adjectives need to be licensed by the presence of an overt determiner. However, there are also nouns which need an overt determiner, even in predicative constructions. In this context, Stowell (1991: 50) observes that predicative [+ animate] noun phrases differ with regard to the obligatoriness of the determiner. Relevant for current purposes, consider the following: (23) a.
Er He b. Er He
ist is ist is
*(ein) Idiot *(an) idiot (ein) Arzt *(a) doctor
It is crucial here that the noun Idiot requires the presence of a determiner in German. Consider (24). Unlike Delorme & Dougherty (1972: fn. 6), I consider comma intonation or its absence as essential (cf. Lyons 1977: 10).9 Now, if an appositive follows a full DP such as the proper name Paul or mein Freund ‘my friend’ without a pause and generally falling intonation, it has to be preceded by a determiner: (24) a.
Paul *(der) Linguist kommt. Paul the linguist is coming
Pronouns are determiners after all
b. Paul *(der) Idiot kommt. Paul the idiot is coming c. Mein Freund *(der) Linguist kommt. my friend the linguist is coming d. Mein Freund *(der) Idiot kommt. my friend the idiot is coming
In contrast, if Idiot follows a pronoun without a pause and falling intonation, a definite determiner is not required: (25) a.
Ich I b. ?Ich I c. *Ich, I,
Idiot komme. (idiot) am coming der (Voll-)Idiot komme. the (complete)idiot am coming Idiot, komme. idiot, am coming
Turning to the Appositive Hypothesis first and recalling that Idiot-nouns require a determiner, we face the question as to why the deletion of the determiner is possible with the pronoun in (25a) but not with regular DPs in (24). Answering this non-trivial question can be avoided in the General DPHypothesis. Considering the contrast in (24) and (25a), I suggest that, unlike the proper name and the noun phrase with a possessive pronoun, the pronoun by itself is not a full DP but only part of it, that is, it is the determiner (see Panagiotidis’ extensive work). Recall from above that Idiot-nouns are special in that they require a determiner. Putting both observations together, we can combine the pronoun with Idiot into a full DP, accounting for the difference between (24) and (25a). In other words, I analyze (25a) as a pronominal construction consisting of a pronoun and a complement, (25b) as a DP involving a pronoun and an appositive DP, and (25c) as a DP with a pronoun and an “apparent” appositive NP. The latter example is ungrammatical as Idiot is not licensed by a determiner (there is an intonational pause between the pronoun and Idiot).10 To sum up this subsection, we observed that Idiot-nouns require an overt determiner. Pointing out a difference between the combination of Idiot with regular full DPs vis-à-vis a pronoun, we argued that the pronoun itself is the determiner licensing Idiot. Concluding the entire section, we isolated two cases (adjectives and Idiot-nouns) where a determiner is required to license the second part of the pronominal construction. While an explanation in terms
Dorian Roehrs
of the Appositive Hypothesis raises a number of questions, the General DPHypothesis provides a straightforward account. So far, we have restricted our attention to common nouns. Next, we turn to the combination of a pronoun and a proper name.
. Pronominal DPs involving proper names Going back to work of Kripke’s (e.g. 1971), proper names are traditionally regarded as rigid designators as they, by themselves, denote the same object in all possible worlds. It is thus surprising that some languages allow proper names to co-occur with a determiner-like element: among others (Lyons 1999: 123), Italian, Hungarian, and German: (26) a.
(die) Maria the(nom) Maria ‘Maria’ b. (der) Maria the(dat) Maria
(German)
Presumably, this element is an expletive (Longobardi 1994: 650), an element with no semantics but only a syntactic function. If it is correct that pronouns can be determiners, we expect not only first and second person pronouns to function as determiners, as seen in Section 2, but also the ones of the third person. First, we consider proper names in combination with pronouns of the third person and then with pronouns of the first and second person. . Pronouns as determiners for proper names Interestingly, the Northern Scandinavian dialects and Icelandic allow proper names to be preceded by a pronoun (Delsing 1993: 54–55, 1996: 33). Unlike in the English She, Mary, is nice, no such intonational pause is needed in Norwegian as the pronoun and the proper name form a single prosodic unit (Kari Gade, p.c.): (27) a.
hun Mari she(nom) Mari ‘Mari’ b. henne Mari she(acc) Mari
(Norwegian/Vestfold)
Pronouns are determiners after all
Apparently, these pronouns have grammaticalized into an expletive element.11 Relevant here, I interpret this combination as another argument for the General DP-Hypothesis and I suggest that in as much as the presence of an article is language-specific, so is that of the pronoun. (This argument can be strengthened if it turns out that adjunction to expletives is impossible in general.) Furthermore, considering the possibility of pro-clitic determiners in SwissGerman (d autoban ‘the highway’, k xint ‘the children’; see e.g. Leu 2001: 55), it is not very surprising to find a language that has a weak form of the pronoun as the determiner (data from Holmberg & Sandström 1996: 75; also Delsing 1996: 34): (28) a.
a Lisa she Lisa b. n Erik he Erik
(Northern Swedish)
In fact, Hellan & Platzack (1999) treat both a and n as clitics. If we assume with Cardinaletti & Starke (1999a: 151) that deficient pronouns cannot be modified and we interpret appositives as modifiers, then (28) cannot involve adjunction. I conclude that the pronoun is a determiner. With this discussion in mind, there are two surprising points to consider: first, observing that the definite article and the third person pronoun are typically in complementary distribution, the possibility of a third person pronoun taking an NP complement needs to be commented on. Second, considering the observation that pronominal determiners are strong pronouns, basically already pointed out by Sommerstein (1972), cases involving deficient pronouns are unexpected. We return to these issues in Sections 4 and 5. . Proper names as predicates In the previous subsection, we considered third person pronouns as determiners, arguing that they are grammaticalized forms. With inherent indexicals such as first and second person pronouns, it is less likely that they can become true expletives and so far I have not found any examples. If this holds generally, we have an account for the ungrammaticality of the example in (29a) vis-à-vis (29b) (for a more general discussion, see Roehrs 2004). In order to get a clear contrast, (29a) must not involve a pause between the pronoun and the proper name and the intonation should be generally falling. In contrast, the example in (29b) involves a break and a rising intonation on the proper name:
Dorian Roehrs
(29) a. *{ich / du} Peter I / you Peter b. {ich / du}, Peter
However, there is no absolute ban on the co-occurrence of an inherently indexical pronoun and a proper name. To set the stage for the discussion, note that some proper names such as Willi and Heinz can be used as predicates, usually in the derogatory sense of “idiot” (cf. Duden 1995: 565). As expected from the discussion of Idiot-nouns above, the determiner is obligatory: (30) a.
Du bist vielleicht *(ein) {Willi/Heinz/*Peter/*Joachim}. you are really a {Willy/Heinz/Peter/Joachim} b. Du bist vielleicht *(ein) Idiot!
Crucially, it is these very proper names that can combine with an inherently indexical pronoun. In this case, the proper name cannot be interpreted as a rigid designator but only as a predicative element (an interpretation reinforced by pronoun doubling, for reasons I do not understand). Unlike in Section 3.1, the pronoun does not function as an expletive:12 (31) a.
du Willi (du) you Willy (you) 1. #‘you Willy’ 2. ‘you idiot’ b. *du Peter (du)
Interestingly, the readings of the proper name reverse when the name functions as a “close” appositive: (32) a.
mein Freund Willi 1. ‘my friend Willy’ 2. #‘my friend idiot’ b. *mein Freund Idiot
This difference in interpretation follows if we assume that predicative Willi in (31a) is a complement of the pronoun and referring Willi in (32a) is headadjoined inside the regular DP. So far, this proposal is consistent with the discussion above. In order to make the argument conclusive that Willi in (31a) can only involve complementation, we need to show that “loose” adjunction to a pronoun, as in the Appositive Hypothesis, is not possible in this case.
Pronouns are determiners after all
To begin with, “loose” appositives involving proper names exist independently. Crucially, when a determiner is added to the proper name, the interpretation of the proper name is ambiguous: (33) mein Freund, der Willi my friend the Willy 1. ‘my friend Willy’ 2. ‘my friend the idiot’
In order to allow for both interpretations, I assume that der Willi involves a “loose” appositive (and not a “close” appositive as in (32a)). Besides other differences in interpretation (Löbel 1991: 6), independent evidence for this assumption comes from the fact that “loose” appositives have to follow genitive DPs, here forced by the presence of the determiner. As expected, (34b) is ambiguous but (34a) is not: (34) a.
der Sohn (*der) Willi unserer Nachbarin the son the Willy of our neighbor ‘the son Willy of our neighbor’ b. der Sohn unserer Nachbarin, (der) Willi the son of our neighbor, the Willy ‘the son of our neighbor, Willy/the idiot’
Furthermore, the impossibility of the determiner with the “close” appositive in (34a) fits well with the observation from above that these kinds of adjunctions cannot involve (inflected) adjectives. There are two ways now to obtain the predicative reading: complementation to the first and second person pronoun, as in (31a), and “loose” apposition to regular DPs, as in (33). In order to make an argument for the General DP-Hypothesis, we need to demonstrate that a pronominal determiner does not allow “loose” apposition. This seems to be borne out by the fact that the predicative reading is not (readily) available and the example is simply ungrammatical with a doubled pronoun: (35) du, der Willi (*du) #you, the idiot (you)
To sum up this subsection, we observed that pronominal DPs with inherently indexical pronouns can involve certain proper names. Crucially, these proper names have a predicative reading only. This contrasts with regular DPs where “close” appositives allow only the rigid designator interpretation and “loose” ones are ambiguous. The lack of the “true” proper name reading in the pronominal DP was suggested to follow from complementation of the would-
Dorian Roehrs
be proper name to the non-expletive pronoun. If this is correct, then we derive another argument for the General DP-Hypothesis. To sum up this entire section, I discussed the combination of pronouns of different persons with proper names of different types. This general possibility provides another argument for the General DP-Hypothesis. Before I conclude this section, consider some other variants of the Appositive Hypothesis. First, we could also assume that the pronoun itself is left-adjoined to the second part of the pronominal construction. However, as can be seen from examples (11a) vs. (13a) from Section 2.1, the verb agrees with the pronoun. Furthermore, while ich Idiot ‘I Idiot’ can occur in argumental position in German, the bare countable singular Idiot ‘idiot’ cannot. I conclude that the pronoun is crucial for the agreement facts and the argumenthood of the noun phrase and should not be adjoined to it. It is for these (and other) reasons that this possibility is implausible. Second, in the discussion of the Appositive Hypothesis, we restricted our attention to “close” and “loose” appositives, where the former exhibits adjunction to a head and the latter to a phrase. Interestingly, Chomsky’s (1994: 421) Bare Phrase Structure model allows for adjunction not only to X and XP but also to X´ as at the time a phrase is adjoined to the root of the tree, that root is an XP (and not yet an X´). Let us call this kind of adjunction “intermediate apposition”. If the adjective is part of the adjoined phrase, all the arguments against “loose” apposition will equally apply to “intermediate” apposition rendering this approach untenable (for the general possibility of the adjective being part of the matrix DP, see below). Third, a reviewer observes that as far as we know, “pronoun + noun” combinations are not the same as “pronoun + comma + noun” sequences. In other words, s/he questions the fact that I have related we linguists to traditional “loose” appositives such as Peter my friend (cf. also Lawrenz 1993: 81–93). While I believe that this is a legitimate concern, note that I employed traditional syntactic representations of appositives, which I argue against in this paper. The net result is that these cases should not be structurally collapsed. However, besides complementation and adjunction, there might be a third way of syntactically representing the combination “pronoun + noun”, for instance, some other kind of juxtaposition of the pronoun and the second part of the noun phrase (Himmelmann 1997: 218).13 While I could only speculate about its syntactic representation, this structure should derive syntactic properties that presumably lie between complementation and adjunction (as evidenced by the lack of an intonational break).
Pronouns are determiners after all
In a different context, the same reviewer mentions vocatives and exclamations, which do not have to take a determiner. Inspired by this comment, suppose for concreteness that a vocative/exclamation structure forms the second part of the juxtaposition, giving the sequence “pronoun + vocative/exclamation” (remaining vague about its concrete syntactic representation). While this proposal straightforwardly accounts for the lack of the “regular” determiner and perhaps the typical use of emotive noun complements in singular pronominal DPs in German (see the appendix), this type of approach has to answer the following questions (among others): (i) the juxtaposition analysis must explain the unusual possibility of allowing different morphological cases on the vocative/exclamation (see Section 2.1); (ii) this approach has to explain why the adjective may vary with regard to its inflection when it is part of the pronominal DP but not when part of an “isolated” determinerless vocative/exclamation: arme(*n) Idioten ‘poor(strong) idiots’; (iii) it needs to give an account for the correlation in the co-occurrence of an expletive pronoun with a referential proper name and of a non-expletive pronoun with a predicative proper name (for another semantic issue, see footnote 10). As I know of no such account, I conclude for the time being that this approach is also on the wrong track. A fourth option is discussed by Cardinaletti (1994), where the adjective is in the matrix DP and the noun in an adjoined position (in fact, allowing an analysis of a “close” appositive). Consider this in more detail.
. Cardinaletti (1994) and some consequences Cardinaletti (1994) argues that unlike deficient pronouns, strong pronouns involve more complex DPs, in which the strong pronoun moves from N to D.14 (36) a.
noii we b. due two
due t i two amici friends
(Italian)
Evidence that the pronoun in (36a) has moved comes from the following structures, in which the pronoun follows an adjective, as it does in English: (37) a.
Povero lui! Poor him!
Dorian Roehrs
b. Poveri noi! Poor us!
Cardinaletti analyzes the latter cases as nominals in non-argument function, e.g. exclamations, which are not necessarily introduced by the category D (cf. Longobardi 1994). Returning to (36a), she argues that pronouns move from N to D (and perhaps higher, p. 204). As a consequence of this proposal, Cardinaletti (1994: 203) concludes that the pronominal construction noi linguisti ‘we linguists’ must be treated as involving adjunction of linguisti as noun phrases do not have two head nouns. Furthermore, assuming that deficient, that is weak and clitic-like, pronouns have less complex internal structures and considering the fact that they cannot be modified (e.g. by adjunction), she can derive the fact that pronominal DPs can have strong pronouns only. However, this generalization seems to be empirically wrong. We have seen above that there are both weak and clitic-like pronominal determiners. Let us juxtapose the relevant pronominal DPs for convenience: (38) a.
noi linguisti we(strong) linguists b. hun Mari she(weak) Mary c. a Lisa she(clitic) Lisa
(Italian) (Norwegian) (Northern Swedish)
If we follow Cardinaletti (1994) for (38a), then noi is a strong pronoun. Furthermore, following Cardinaletti and Starke (1999a: 154) in that expletive pronouns are deficient and that weak and strong pronouns in Germanic are usually homophonous, I conclude that hun in (38b) is a weak pronoun. Finally, a in (38c) is a clitic (cf. also Hellan & Platzack 1999) as it is an expletive form but different from the weak/strong pronoun (hun). We have arrived at a tripartite division of “transitive” pronouns, a possibility already hinted at by Holmberg (1999: 264). While Cardinaletti (1998) argues this to be also true for possessive pronouns, Cardinaletti (1994) and Cardinaletti and Starke (1999b: 278) explicitly deny it for “transitive” determiners. Finally, in view of the data above, we need to weaken Sommerstein’s (1972: 204) observation that pronominal determiners always have some degree of stress and cannot be phonetically reduced (cf. also Lyons 1977: 18). Having illustrated that not only strong pronouns can be part of the pronominal DP but also weak and clitic ones, we need to show that this subtype of the Appositive Hypothesis cannot account for the above-mentioned facts either. Although N-to-D raising of the pronoun can also explain the inflec-
Pronouns are determiners after all
tion of the adjective in German (under similar assumptions as those in Section 2.1) and perhaps the possibility of a missing determiner with Idiot-nouns under head-adjunction (i.e. “close” apposition), as far as I can see, this type of proposal cannot account for the different interpretations of proper names (referring vs. predicative), depending on the kind of pronoun. The question arises as to how we can make Cardinaletti’s interesting proposal compatible with the General DP-Hypothesis. If we follow Postal (1966) in that pronouns are determiners and the hypothesis that determiners move from art to D (Roehrs 2002, in progress), then we are not forced to conclude that pronouns exhibit N-to-D raising and consequently pronominal DPs involve adjunction. With the pronoun undergoing art-to-D raising in (36a), we, in fact, arrive at the same proposal that we made to account for the different adjective endings in German (cf. also X-toD raising of pronouns in Progovac 1998: 177). Furthermore, claiming that the pronoun is not a noun (pace Cardinaletti) not only explains the lack of movement of the common noun in (36b) without further assumptions but also the possibilities illustrated above that a pronoun can co-occur with a proper name. Moreover, assuming a complement structure of pronominal DPs removes criticism voiced by Diesing’s (1999: 244) and Uriagereka’s (1999: 299–270) with regard to determining the range of the noun phrase, which is typically provided by the head noun (e.g. Longobardi 1994: 634) and not some higher phrase as in Cardinaletti & Starke. More speculatively, in order to keep Cardinaletti & Starke’s (1999a) general proposal intact, we might suggest that strong, weak, and clitic-like pronouns involve differently complex structures inside the matrix DP such that strong and weak pronouns are phrases in Spec positions and clitics are in D (for a similar proposal for “transitive” demonstratives and historically related determiners, see Roehrs & Sapp 2004, who extend work of van Gelderen’s on the development of complementizers).15
. Discussion It has often been observed that definite determiners and third person pronouns are in complementary distribution in languages where they differ: while pronouns are “intransitive” (they / *they linguists), determiners are “transitive” (the linguists / *the). In order to explain this, it has been proposed that these elements are related to one another with differences in the details of that relation (e.g. Postal 1966; Sommerstein 1972; Abney 1987: 282; Lyons 1977: 14,
Dorian Roehrs
1999: 27, 303; Rauh 2003). This complementary distribution follows straightforwardly if one assumes that the relevant elements are in the same position in the syntactic representation and one of the two is the suppletive form of the other in the corresponding syntactic context. Let us make this intuition of contextual “competition” more concrete. Recall from Section 4 that we extended Cardinaletti & Starke’s (1999a) tripartite division to “transitive” pronouns. Suppose with Cardinaletti & Starke (1999a) that there are three corresponding structural representations: strong (pronominal) determiners are phrases, weak (pronominal) determiners are deficient phrases, and clitic-like (pronominal) determiners are heads. These “transitive” pronouns are in contextual competition with their corresponding demonstratives/articles, where demonstratives are phrases, “deficient” demonstratives are deficient phrases, and articles are heads.16 If this is correct, then in a language without articles, we can expect a third person pronoun to appear where we would usually find only articles in other languages. Interestingly, Furuya, to appear, discusses cases where Japanese allows third person pronouns in the plural (39a). These pronouns are case-less (39b) (her (7c) and (19)): (Japanese) kare-ra sensei he-pl teacher ‘they teachers’ b. *[watashi-tachi-{ga/no} sensei-ga ] heikinjumyo-ga nagai. I-pl-{nom/gen} teacher-nom average span-nom long ‘The average lifespan of us teachers is long.’
(39) a.
In order to determine the structural category of these pronominal determiners, two facts are suggestive: first, while these pronouns have no case-morpheme, regular demonstratives such as kono ‘this’ have been argued to consist of a stem and a genitive case morpheme ko-no (Fukui & Speas 1986: 134–135); second, Japanese pronouns seem to be compositional with regard to number in that they have a stem morpheme and a number morpheme kare-ra (‘he-pl = they’). I will tentatively conclude that these pronominal determiners are deficient phrases. We are now ready to provide a catalogue of the pronominal structures, we have discussed so far (the columns sort the examples with regard to person of the pronouns, the rows divide the data with regard to the tripartite division of pronouns and the different head nouns) (Chart 3). Some remarks are in order here. Starting with the second column, first and second person pronouns have no “competition” from (third person) demonstratives. Hence, noi linguisti ‘we linguists’ and du Willi ‘you idiot’ (under certain conditions) are always possible. Next, according to our interpretation of
Pronouns are determiners after all
Chart 3. first/second person strong pronoun com. noun noi linguisti ‘we linguists’ (demonstrative) prop. name du Willi ‘you Willy (= idiot)’
third person (them linguists)
weak pronoun com. noun watashitachi sensei ‘we teachers’ karera sensei ‘they teachers’ (“def.” demon.) prop. name hun Mari ‘(she) Mary’ clitic pronoun (article)
com. noun prop. name
a Lisa ‘(she) Lisa’
the data, Japanese pronouns seem to be weak (thanks to Kaori Furuya, p.c., for helping me clarify this point). Furthermore, a reviewer as well as Marcel den Dikken and Christina Tortora remind me that the Romance languages have deficient first and second person pronouns. Thus, pronominal DPs of this kind should in principle be possible, although I have not come across any examples yet (the gaps in Chart 3). Turning to the third column, strong pronouns of the third person are in competition with regular demonstratives and besides the well-known exception of them linguists (see Déchaine & Wiltschko 2002: 421–422), they do not seem to exist.17 The possibility of Norwegian hun Mari ‘(she) Mary’ with an expletive pronominal determiner is parallel to proper names preceded by a regular expletive determiner, as in German die Maria ‘(the) Mary’, although both determiners presumably differ with regard to their structural size. Northern Swedish a Lisa ‘(she) Lisa’ is parallel to dialectal Norwegian Kari-a ‘Kari (-DEF)’ (Delsing 1993: 54, 1996: 37) and it seems to be a dialect-specific feature which option is chosen. The reason that free-standing (expletive) pronominal determiners are possible at all presumably has to do with the absence of corresponding free-standing “deficient” demonstratives and articles. The same lexical absence allows Japanese third person pronouns to take an NP complement. More generally then, as far as I can see at this point, languages differ in what lexical choices they make available for transitive determiners, consequently (dis-)allowing the corresponding transitive pronouns.18, 19
. Conclusion This paper provided new arguments in favor of the General DP-Hypothesis. After briefly dismissing the possibility of “close” appositives, we provided arguments from adjective inflection and from Idiot-nouns that “loose” appositives
Dorian Roehrs
cannot be considered the correct analysis of the pronominal construction. As an alternative, the General DP-Hypothesis fared better in all respects. Further evidence for this proposal came from the discussion of the combination of pronouns and proper names. Next, we turned to some consequences of the current proposal for Cardinaletti (1994) and Cardinaletti & Starke (1999a). Finally, I catalogued the different types of DPs and commented on them. At the very least, I hope to have convinced the reader that pronominal DPs are an exciting topic of investigation that can shed light on a number of linguistic issues. If on the right track, this discussion showed that pronouns are determiners after all and not spell-outs of DP (e.g. Weerman & Evers-Vermeul 2002).
Appendix: Singular pronominal DPs It is often noted that unlike plural pronominal DPs, their singular counterparts are impossible in argument position in many languages (for a syntactic account of this asymmetry in English and Japanese, see Furuya to appear). However, these singular DPs can be “rescued” by an embedded plural noun (data due to Rex Sprouse, p.c.): (40) a. *You {generation / idiot} have messed this all up! b. You generation of idiots have messed this all up!
(For a formal account of agreement-attraction, as in (40b), see den Dikken 2001.) Notice that this asymmetry is also apparent in German in that pronominal constructions in the singular are typically provided using emotive nouns such as Idiot ‘idiot’ or names for animals (Section 2.1). Although these combinations are perfectly grammatical, other common nouns seem to be awkward in “out-of-the-blue” contexts, as in (41a). However, Rauh (2004) extensively discusses conditions that license these examples; for instance, the use of stress in an appropriate context in (41b) (stress is indicated by capital letters; her example (12)): (41) a.
??Du
Linguist hast ihnen geholfen. You linguist has them helped ‘You (linguist) helped them.’
Pronouns are determiners after all
b. Wenn noch nicht einmal du LINGUIST die neue if still not even you linguist the new Rechtschreibung beherrschst, wer sollte es dann tun? spelling master who should it then do ‘If even you linguist have not mastered the new spelling rules, who can?’
Leaving English open, Rauh (2004) provides a pragmatic account within Gricean conversational maxims. Interestingly and in a different context, Baker (2001) concludes that “the pragmatic values of particular constructions do not seem to be defined by Universal Grammar, but emerge out of the system of a particular language taken as a whole” (p. 434). If this is more generally true, then the difference between English and German may lie in their different pragmatics.
Notes * Parts of this paper were presented at the 19th Comparative Germanic Syntax Workshop at the CUNY Graduate Center, New York City, in June 2004. I thank the audience for helpful questions and comments. Furthermore, I am very grateful to two reviewers and, especially, to my editors Marcel den Dikken and Christina Tortora for numerous detailed comments on both presentation and content. Thanks are due to Rumiko Sode for pointing out Furuya, to appear, and making this paper available to me. All errors are my own. . Throughout this paper, I am using the term ‘pronominal DP’ to mean a pronoun followed by an overt noun and not a pronoun followed by a null/empty element (see, among many others, Panagiotidis 2002a, b, 2003a, b). . The following abbreviations will be used throughout this paper: no infl = no inflection, nom = nominative case, acc = accusative, dat = dative, gen = genitive, m = masculine, n = neuter, f = feminine, sgl = singular, and pl = plural. . Besides the phonetic deletion of the determiner, there are two other options for (9a): (i) we could imagine a process of syntactic “pruning” of the DP-level à la Ross (1966). While I think that this is exceptionally possible, it should not be allowed in general and I will not consider this possibility here further; (ii) a reviewer points out that vocatives/exclamations do not have determiners and, as such, (9a) would not have a determiner to begin with. I argue against this possibility at the end of Section 3. . When we consider superlatives licensed in French by a determiner and in German by a pronoun, we find apparent evidence for the possibility of the deletion of a determiner, as shown in (9a). First, French provides overt evidence that the superlative of post-nominal adjectives requires an independent determiner to license it (i-a) (cf. Kayne 2004). If this is true, one of the determiners of the pre-nominal superlative in (i-b) must be elided. Following Longobardi’s (2001) discussion that French is the most restricted language in that it
Dorian Roehrs
always takes a determiner in argument position, I propose to elide the determiner inside the DegP in (i-b) (cf. Baker 2003: 123 fn. 18): (i)
a.
b.
la fille [DegP la plus gentille] the girl the most nice ‘the nicest girl’ fille la [DegP la plus jolie] the most beautiful girl the ‘the most beautiful girl’
(French)
With this in mind, consider a case in German where a pronoun can license a superlative: (ii) a. b.
mit with mit with
den [DegP the uns [DegP us
den the den the
ältesten] oldest ältesten] oldest
Kindern children Kindern children
If we treat (ii-a) on a par with (i-b), then (ii-b) involves deletion of a determiner, a possibility claimed to be available for the Appositive Analysis in (9a). However, I will assume that similar to “positive” adjectives (Julien 2002: 269–270), superlatives are also in a Spec position in the extended projection of the noun phrase and, as such, are not parallel to appositives. Consequently, the question remains as to what would allow deletion of the determiner in (9a). . All my commentators wonder if the analyses in (16b) and (17b) straightforwardly extend to the cases in (16a) and (17a). In other words, they ask if there is any direct evidence that the pronouns in the (a)-examples may be agreeing and non-agreeing. Before providing some evidence for that, I consider a detailed suggestion made by Marcel den Dikken and Christina Tortora that avoids this assumption. My editors suggest a possible link between the English constructions many men / many a man and the variable agreement facts in the pronominal DP wir nette(n) Leute ‘we nice people’. They propose the following structures, where many and wir may vary with regard to their position in the DP-level (see also Sections 4 and 5). Depending on their position, different agreement relations hold. With regard to German, they suggest that an overt definite determiner in D brings about a weak ending while a null D results in a strong inflection: (i)
a. b.
(ii) a. b.
[DP Ø [D’ many [ . . . men ]]] [DP many [D’ a [ . . . man ]]] [DP Ø [D’ wir [ netten(weak) Leute ]]] [DP wir [D’ Ø [ nette(strong) Leute ]]]
One advantage is that this structural approach does not have to posit the assumption that speakers may analyze pronouns as agreeing and non-agreeing. While this is a very intriguing possibility, it seems unclear how to derive the varying preferences with the singular dative pronoun where the morphological form of the pronoun itself seems to be of importance (see below). Putting it in different positions will probably not be enough as we still have to refer to its (non-)agreeing morphological form when the masculine dative pronoun is
Pronouns are determiners after all
in D in (18) and the feminine dative one is in Spec,DP in (19). I turn to an argument that indicates that the pronoun itself may be agreeing and non-agreeing. Gallmann (1996: 284, 1998) discusses the German morphological condition “double or nothing”, according to which a suffix on the head noun Dirigent ‘conductor’ must be licensed by adjectival agreement on a second element and vice versa: (iii) a. b.
Orchester ohne [dies-en Dirigent-*(en)] orchestra without this(acc) conductor Orchester ohne [Dirigent-(*en)]
Interestingly, a pronoun can but does not have to co-occur with an overtly inflected noun in this case: (iv) a. b.
(?) mich
Idiot me(acc) idiot mich Idiot-en
The alternation in (iv) follows straightforwardly if the pronoun may be agreeing and nonagreeing. . There is also some overt evidence in English and German. As noted by Schütze (2001: 215 fn. 14), a lower pronoun is sometimes possible if the noun phrase functions as an argument: (i)
Lucky us / *we linguists have to explain our profession to everyone.
If we follow Longobardi (1994: 620) in that arguments always project a DP, then we conclude that the DP-level in (i) is present. We can assume then that the pronoun moves to license the DP at LF, which, under Longbardi’s assumptions, is generally allowed for the Germanic languages. Overt evidence for a lower artP in German comes from non-proximal demonstratives, intensified by da ‘there’: (ii) [DP dasi schöne [artP ti Bildk da [NP tk von Maria]]] that nice picture there of Mary First, I assume with Bernstein (1997) that the demonstrative and its intensifier are basegenerated in the same phrase (here artP pace Bernstein). Second, the demonstrative moves from Spec,artP to Spec,DP and the head noun partially raises (cf. Haider 1993: 30). If so, the intensifier da intervening between head noun and the thematic object provides overt evidence for a lower position of artP in German. . This discussion is a straightforward extension of the proposal for the (reanalyzed) “mixed” paradigm involving ein-words ‘a-words’, given in Roehrs (2002, in progress). . Marcel den Dikken and Christina Tortora point out that this solution involves transderivational as well as transparadigmatic comparison and, as such, fits rather poorly with the formal approach chosen in this paper. Furthermore, they note that this type of ambiguity is generally tolerated in languages (e.g. Dutch), often disambiguated in non-morpho-syntactic ways. While I agree with my editors on all counts (cf. also Panagiotidis 2002a: 117), note in this respect that Vater (1991: 18–19) observes that in rare cases, a syntactic resolution of ambiguity is possible (e.g. insertion of a determiner to disambiguate the grammatical function of two object mass nouns in German). One could speculate that the morphological
Dorian Roehrs
disambiguation discussed in the text has been grammaticalized on the adjective. Crucially for my purposes, the main point of this section (i.e. the explanation of the weak adjective endings in terms of pronouns functioning as determiners) remains unaffected. Note that Lawrenz (1993: 88) independently recognizes the importance of the weak ending but does not provide an analysis of the full set of data. . For instance, the difference in intonation (indicating presumably different syntactic domains) provides a straightforward account for the following contrast: (i)
a. *they linguists b. they, linguists
Delorme and Dougherty (1972: fn. 5) notice the peculiarity of the third person pronoun and assign “?” to both (i-a) and (i-b). Later, they give (i-a) as fully grammatical (their (36f)). I will follow Déchaine & Wiltschko (2002: 421–422) and Rauh (2003) in treating (i-a) as ungrammatical, which becomes apparent if there is no pause and falling intonation between the pronoun and the second element. . A qualification is in order here. The grammaticality judgment in (25c) is given with regard to direct “reflexive” interpretation of Idiot referring back to the speaker. With Idiot! felicitous in use with an addressee or with reference to a third party, the example improves if one interprets Idiot as a kind of “guise” of the speaker, perhaps on a par with him in constructions such as Bill likes him, grammatical in some circumstances. In this case, the interpretation would be indirect such that the speaker would be equated with the “guise” at some level. . The pronoun determiner can also have non-expletive functions, for instance, when a young person called Mary is contrasted with an older person of the same name, as in (i-a). Kari Gade, p.c., informs me that the regular definite determiner sounds archaic to her (i-b). For completeness’ sake, the plural form is given in (i-c): (i)
a.
b. c.
hun unge Mari she young Mary ‘the young Mary’ den unge Mari de to Mari-e-ne the/they two Mary-pl-DEF ‘the two Marys’
. Female names with the same property are Trulla and Suse. Furthermore, there seems to be some dialectal variation as to which names allow the predicative reading (a point I will not investigate here). Finally, note in passing that there is some overlap with semi-suffixes involving proper names (data from Duden 1995: 507–508): (i)
a.
b.
du Heini du you Heini you ‘you scoundrel’ unser Sanitäts-heini our first-aid guy
Pronouns are determiners after all
c. *du Peter du you Peter you d. Nörgel- peter nagging person . An indication for this might be Pesetsky’s (1978) observation that the Romance languages require another element for the pronominal DP to be felicitous: (i)
nous autres linguistes we (other) linguists
(French)
There are a number of issues here. It is not clear to me if this is a dialectal feature (as hinted at in Pesetsky’s footnote 9 for French and the different judgments of Italian noi linguisti ‘we linguists’ given in Pesetsky on page 358 and Cardinaletti 1994, see below) or if this has to do with the argument status of the pronominal DP as claimed by Lyons (1999: 142), where autres is required in argument use. Finally, note that Spanish is very different in that the pronominal DP is brought about by a regular (third-person) DP and first/second person agreement on the verb (see Lyons 1999: 144). Clearly, in order to be able to interpret these data properly, more research is needed here to get a clearer empirical picture. . For the discussion of pronominal nouns in Japanese and Thai, see Panagiotidis (2002a, b). . In this respect, consider an example involving a vocative, discussed in Cardinaletti (1994). She states that the determiner cannot be realized, as in (i-a). At first glance, it is not entirely clear why under my assumptions, the determiner is not possible in the lower position. Note, however, that this might be a language-specific fact as Norwegian has a lower determiner without an upper one (Svenonius 1993: fn. 12; data from Kari Gade p.c.): (i)
a. b.
Povero (*il) professore! poor the professor du, gamle sko-en min, skal ikke kastes you old shoe.DEF my shall not be thrown out ‘You, my old shoe, shall not be thrown out.’
(Italian) (Norwegian)
In order to capture the language difference in (i), we could employ Cardinaletti & Starke’s (1999a) general framework, in this paper extended to “transitive” pronouns. Suppose that articles are heads, either clitics (il) or suffixes (-en). As a light element, clitic il surfaces in the D-position. As the DP level is not present in vocatives in (i-a), il cannot surface. Unlike the Italian clitic il, Norwegian -en is a suffix that can attach only to nouns. This selectional restrictiveness is a typical property of affixes and explains the lower position of the determiner. Strong pronouns, on the other hand, are more flexible in their distribution. Either they appear in a lower position in non-argumental noun phrases or in DP in argumental ones ((ii-a) is due to a reviewer): (ii) a. b.
Fortunati noi lucky we Noi fortunati We lucky
linguisti! linguists linguisti possiamo leggere molti libri. linguists can read many books
(Italian)
Dorian Roehrs
Finally, to be fair, Cardinaletti (1994) provides some more arguments for her analysis, which, due to the length of this paper, I cannot attend to here. . For the time being, I remain vague as to what a “deficient” demonstrative is. I will just point out that this intermediate category is not surprising given the fact that articles typically develop out of demonstratives. . In fact, them linguists may not be an exception at all (for they stones, see Radford 1993: 109). Replying to Diesing’s (1999: 243) and Holmberg’s (1999: 263–264) comments on Cardinaletti & Starke (1999a: 218 fn. 2), Cardinaletti & Starke (1999b: 283–288) go to great lengths to show that bare them is a simple demonstrative, similar to German der. That it is not a personal pronoun may also be gleaned from the possibility of combining them with an inanimate common noun as in them pants. I thus take this suggestion to be on the right track. Furthermore, inspired by Diesing’s and Holmberg’s comments on the different binding possibilities of demonstratives (Condition C) vis-à-vis personal pronouns (Condition B), we can check these relations for the clear cases of the pronominal constructions, the ones in the first and second person. Now, if it is correct that the pronominal construction involves pronominal determiners as part of R-expressions, then we expect Condition C violations. However, although I perceive a clear contrast between bare pronouns and pronominal DPs, it is not as strong as one would expect under a Condition C violation (following standard assumptions, I take bare German pronouns of the first and second person to be ambiguous between anaphors and pronominals, but see Reuland 2001: 463–465): (i)
a. b.
(ii) a. b.
{Sie / Wir} haben uns (, Linguisten,) im Spiegel gesehen. They / We have us (, linguists,) in.the mirror seen. {Sie / ?? Wir} haben uns Linguisten im Spiegel gesehen. {Sie / Wir} wußten, daß wir (, Linguisten,) unschuldig sind. They / We knew that we (, linguists,) innocent are. {Sie / ?? Wir} wußten, daß wir Linguisten unschuldig sind.
Speculatively, this lack in strength of the contrast is perhaps due to the parallel but grammatical true appositives. Whatever the exact interpretation of the facts in (i) and (ii) is, I take the contrast between the bare pronoun (in combination with an appositive) in the (a)examples and the pronominal DPs in the (b)-examples as another argument for the General DP-Hypothesis. . For possible candidates to fill the gaps in Chart 3, see Himmelmann’s (1997: 215) discussion of some Australian pronominal systems. . There two other ways that could account for the complementary distribution of definite determiners and third person pronouns: first, one could assume that “intransitive” pronouns undergo N-to-D raising (cf. Zwarts 1994; Panagiotidis 1998; Rauh 2004). However, we showed above that certain third person pronouns can take a complement (cf. also Panagiotidis 2002a: 37). Second, if the competing elements turn out to be in different positions, Marcel den Dikken and Christina Tortora point out that the complementary distribution would also follow from the Doubly filled DP filter (e.g. Abney 1987: 271). Considering occasional “violations” of this filter in the clause and regular DPs (see e.g. Bavarian German cited in Haegeman 1994: 383 and Finland Swedish cited in Santelmann 1993: fn.
Pronouns are determiners after all
19, respectively), this suggestion makes the interesting prediction that non-appositional sequences such as [DP pronoun [D’ determiner [NP noun ]]] should occasionally occur. As far as I am aware, this possibility has not been reported yet.
References Abney, S. (1987). The English Noun Phrase in its Sentential Aspect. PhD Dissertation, MIT. Baker, M. C. (2001). The natures of nonconfigurationality. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory (pp. 407–438). Malden, MA: Blackwell. Baker, M. C. (2003). Lexical Categories. Verbs, nouns and adjectives [Cambridge Studies in Linguistics 102]. Cambridge: Cambridge University Press. Bernstein, J. B. (1997). Demonstratives and reinforcers in Romance and Germanic languages. Lingua, 102, 87–113. Bernstein, J. B. (2001). The DP Hypothesis: Identifying clausal properties in the nominal domain. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory (pp. 536–561). Malden, MA: Blackwell. Bhatt, C. (1990). Die syntaktische Struktur der Nominalphrase im Deutschen [Studien zur deutschen Grammatik 38]. Tübingen: Narr. Brame, M. (1981). The general theory of binding and fusion. Linguistic Analysis, 7(3), 277– 325. Brame, M. (1982). The head-selector theory of lexical specification and the nonexistence of coarse categories. Linguistic Analysis, 10(4), 321–325. Brugè, L. (1996). Demonstrative movement in Spanish: A comparative approach. University of Venice Working Papers in Linguistics, 6(1), 1–61. Campbell, R. (1996). Specificity operators in Spec, DP. Studia Linguistica, 50(2), 161–188. Cardinaletti, A. (1994). On the internal structure of pronominal DPs. The Linguistic Review, 11, 195–219. Cardinaletti, A. (1998). On the deficient/strong opposition in possessive systems. In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the Determiner Phrase (pp. 17–53). Amsterdam: John Benjamins. Cardinaletti, A. & M. Starke (1999a). The typology of structural deficiency: A case study of the three classes of pronouns. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 145–233). Berlin: Mouton de Gruyter. Cardinaletti, A. & M. Starke (1999b). Reply to Comments: Responses and Demonstratives. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 273–290). Berlin: Mouton de Gruyter. Chomsky, N. (1994). Bare phrase structure. MIT Occasional Papers in Linguistics, 5, 1–48. Coene, M. & Y. D’hulst (2003). Introduction: The syntax and semantics of noun phrases. Theoretical background. In M. Coene & Y. D’hulst (Eds.), From NP to DP. Volume 1: The syntax and semantics of noun phrases (pp. 1–33). Amsterdam: John Benjamins. Corver, N. & D. Delfitto (1999). On the nature of pronoun movement. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 799–861). Berlin: Mouton de Gruyter.
Dorian Roehrs
Darski, J. (1979). Die Adjektivdeklination im Deutschen. Sprachwissenschaft, 4, 190–205. Déchaine, R.-M. & M. Wiltschko (2002). Decomposing pronouns. Linguistic Inquiry, 33(3), 409–442. Delsing, L.-O. (1993). The Internal Structure of Noun Phrases in the Scandinavian Languages. A comparative study. PhD Dissertation, University of Lund. Delsing, L.-O. (1996). Nominalfrassyntax i skandinaviska dialekter. Nordica Bergensia, 9, 24–74. Delorme, E. & R. C. Dougherty (1972). Appositive NP Constructions: we, the men; we men; I, a man; ETC. Foundations of Language, 8, 2–29. Diesing, M. (1999). Comments on Cardinaletti and Starke: The typology of structural deficiency. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 243–247). Berlin: Mouton de Gruyter. Diessel, H. (1999). Demonstratives. Form, function, and grammaticalization. Amsterdam: John Benjamins. Dikken, M. den (2001). Pluringulars, pronouns and quirky agreement. The Linguistic Review, 18, 19–41. Duden (1995). Die Grammatik, Vol. 4. Mannheim: Dudenverlag. Emonds, J. E. (1976). A Transformational Approach to English Syntax. New York: Academic Press. Fukui, N. & M. Speas (1986). Specifiers and Projection. MIT Working Papers in Linguistics, 8, 128–172. Furuya, K. (to appear). Us linguists. Proceedings of the Workshop on Altaic in Formal Linguistics. MIT. Gallmann, P. (1996). Die Steuerung der Flexion in der DP. Linguistische Berichte, 164, 283– 314. Gallmann, P. (1998). Case underspecification in morphology, syntax and the lexicon. In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the Determiner Phrase (pp. 141–175). Amsterdam: John Benjamins. Gelderen, E. van (2004). Economy, innovation, and prescriptivism: From Spec to Head and Head to Head. Journal of Comparative Germanic Linguistics, 7, 59–98. Giusti, G. (1997). The categorial status of determiners. In L. Haegeman (Ed.), The New Comparative Syntax (pp. 95–123). London: Longman. Haegeman, L. (1994). Introduction to Government & Binding Theory (2nd ed.). Oxford: Blackwell. Haider, H. (1993). Deutsche Syntax – Generativ. Tübingen: Narr. Hellan, L. & C. Platzack (1999). Pronouns in Scandinavian languages: An overview. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 123–142). Berlin: Mouton de Gruyter. Himmelmann, N. P. (1997). Deiktikon, Artikel, Nominalphrase. Zur Emergenz syntaktischer Struktur. Tübingen: Niemeyer. Holmberg, A. (1999). Comments on Cardinaletti and Starke: The typology of structural deficiency. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 263–266). Berlin: Mouton de Gruyter. Holmberg, A. & G. Sandström (1996). Vad är det för särskilt med nordsvenska nominalfraser? Nordica Bergensia, 9, 75–89.
Pronouns are determiners after all
Hudson, R. A. (1987). Zwicky on heads. Journal of Linguistics, 23, 109–132. Jackendoff, R. S. (1977). X’ Syntax: A study of phrase structure. Cambridge, MA: The MIT Press. Jespersen, O. (1914). A Modern English Grammar. Part II. London: George Allen & Unwin. Julien, M. (2002). Determiners and word order in Scandinavian DPs. Studia Linguistica, 56, 264–315. Kayne, R. (2004). Some preliminary comparative remarks on French and Italian definite articles. Ms., New York University. Kripke, S. A. (1971). Identity and necessity. In M. K. Munitz (Ed.), Identity and Individuation (pp. 135–164). New York: New York University Press. Lawrenz, B. (1993). Apposition. Begriffsbestimmung und syntaktischer Status. Tübingen: Narr. Leu, T. (2001). A sketchy note on the article-modifier relation. Evidence from Swiss-German which sheds light on Greek determiner spreading. Generative Grammar in Geneva, 2, 55–69. Lindauer, T. (1995). Genitivattribute. Eine morphologische Untersuchung zum deutschen DP/NP System. Tübingen: Niemeyer. Löbel, E. (1991). Apposition und das Problem der Kasuszuweisung und Adjazenzbedingung in der Nominalphrase des Deutschen. In G. Fanselow & S. Felix (Eds.), Strukturen und Merkmale syntaktischer Kategorien (pp. 1–32). Tübingen: Narr. Longobardi, G. (1994). Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry, 25(4), 609–665. Longobardi, G. (2001). The structure of DPs: Some principles, parameters, and problems. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory (pp. 562–603). Malden, MA: Blackwell. Lühr, R. (1991). Adjazenz in komplexen Nominalphrasen. In G. Fanselow & S. Felix (Eds.), Strukturen und Merkmale syntaktischer Kategorien (pp. 33–50). Tübingen: Narr. Lyons, C. (1977). The demonstratives of English. Belfast Working Papers in Language and Linguistics, 2(5), 1–23. Lyons, C. (1999). Definiteness. Cambridge: Cambridge University Press. Meyer, C. F. (1992). Apposition in Contemporary English. Cambridge: Cambridge University Press. Olsen, S. (1991). Die deutsche Nominalphrase as Determinansphrase. In S. Olsen & G. Fanselow (Eds.), DET, COMP und INFL. Zur Syntax funktionaler Kategorien und grammatischer Funktionen (pp. 35–56). Tübingen: Niemeyer. Panagiotidis, P. (1998). The internal structure of English pronouns. In S. Lambropoulou (Ed.), The 12th International Symposium on Theoretical and Applied Linguistics. Proceedings, Vol. 1 (pp. 84–94). Thessaloniki: Aristotle University of Thessaloniki. Panagiotidis, P. (2000). Demonstrative determiners and operators: The case of Greek. Lingua, 110, 717–742. Panagiotidis, P. (2002a). Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. Amsterdam: John Benjamins. Panagiotidis, P. (2002b). Pronominal nouns. In H. J. Simon & H. Wiese (Eds.), Pronouns – Grammar and Representation (pp. 183–203). Amsterdam: John Benjamins.
Dorian Roehrs
Panagiotidis, P. (2003a). Empty Nouns. Natural Language & Linguistic Theory, 21(2), 381– 432. Panagiotidis, P. (2003b). One, empty nouns, and θ-assignment. Linguistic Inquiry, 34(2), 281–292. Pesetsky, D. (1978). Category switching and so-called pronouns. In Papers from the Fourteenth Regional Meeting (pp. 350–361). Chicago Linguistic Society. Plank, F. (Ed.). (2003). Noun Phrase Structure in the Languages of Europe [Empirical Approaches to Language Typology, EUROTYP 20–7]. Berlin: Mouton de Gruyter. Postal, P. M. (1966). On the so-called pronouns in English. In F. Dinneen (Ed.), Nineteenth Monograph on Language and Linguistics (pp. 177–206). Washington, DC: Georgetown University Press. Progovac, L. (1998). Determiner phrase in a language without determiners. Journal of Linguistics, 34, 165–179. Radford, A. (1993). Head-hunting: On the trail of the nominal Janus. In G. G. Corbett, N. M. Fraser, & S. McGlashan (Eds.), Heads in Grammatical Theory (pp. 73–113). Cambridge: Cambridge University Press. Radford, A. (1997). Syntactic Theory and the Structure of English. A minimalist approach. Cambridge: Cambridge University Press. Rauh, G. (2003). Warum wir Linguisten “euch Linguisten”, aber nicht “sie Linguisten” akzeptieren können. Linguistische Berichte, 196, 389–424. Rauh, G. (2004). Warum “Linguist” in “ich/du Linguist” kein Schimpfwort sein muß. Linguistische Berichte, 197, 77–105. Reuland, E. (2001). Primitives of binding. Linguistic Inquiry, 32(3), 439–492. Rijkhoff, J. (2002). The Noun Phrase. Oxford University Press. Roehrs, D. (2002). The boundary between morphology and syntax: Determiners Move into the determiner phrase. IULC Working Papers Online, 2, 1–46. Roehrs, D. (2004). Person restrictions inside the DP. Ms., Indiana University. Roehrs, D. (in progress). The Morpho-Syntax of the Germanic Noun Phrase: Determiners move into the determiner phrase. PhD Dissertation, Indiana University. Roehrs, D. & C. D. Sapp (2004). Different word orders in the early north Germanic DP: The rise of the suffixal article. Paper presented at WECOL, University of Southern California, Los Angeles. Ross, J. R. (1966). A proposed rule of tree-pruning. Mathematical linguistics and automatic translation, Harvard Computation Laboratory, Report No. NSF-17, Cambridge. [Reprinted in 1969. D. A. Reibel & S. A. Schane (Eds.), Modern Studies in English: Readings in transformational grammar (pp. 288–299). Englewood Cliffs, NJ: PrenticeHall Inc.] Santelmann, L. (1993). The distribution of double determiners in Swedish: Den Support in D0 . Studia Linguistica, 47(2), 154–176. Schütze, C. T. (2001). On the nature of default case. Syntax, 4(3), 205–238. Sommerstein, A. H. (1972). On the so-called definite article in English. Linguistic Inquiry, 3, 197–209. Sproat, R. & C. Shih (1991). The cross-linguistic distribution of adjective ordering restrictions. In C. Georgopoulos & R. Ishihara (Eds.), Interdisciplinary Approaches to Language. Essays in honor of S.-Y. Kuroda (pp. 565–593). Dordrecht: Kluwer.
Pronouns are determiners after all
Stavrou, M. (2003). Semi-lexical nouns, classifiers, and the interpretation(s) of the pseudopartitive construction. In M. Coene & Y. D’hulst (Eds.), From NP to DP. Volume 1: The syntax and semantics of noun phrases (pp. 329–353). Amsterdam: John Benjamins. Stowell, T. (1991). Determiners in NP and DP. In K. Leffel & D. Bouchard (Eds.), Views on Phrase Structure (pp. 37–56). Dordrecht: Kluwer. Svenonius, P. (1993). Selection, adjunction, and concord in the DP. Studia Linguistica, 47(2), 198–220. Szabolsci, A. (1983). The possessor that ran away from home. The Linguistic Review, 3, 89– 102. Uriageraka, J. (1995). Aspects of the syntax of clitic placement in western Romance. Linguistic Inquiry, 26(1), 79–123 Uriagereka, J. (1999). Commentary on the typology of structural deficiency. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 267–271). Berlin: Mouton de Gruyter. Vangsnes, Ø. A. (1999). The Identification of Functional Architecture. PhD Dissertation, University of Bergen. Vangsnes, Ø. A. (2004). Rolling up the Scandinavian noun phrase. Paper presented at the 27th GLOW Colloquium in Thessaloniki, Greece. Vater, H. (1991). Determinatien in der DP. In S. Olsen & G. Fanselow (Eds.), DET, COMP und INFL. Zur Syntax funktionaler Kategorien und grammatischer Funktionen (pp. 15– 34). Tübingen: Niemeyer. Weerman, F. & J. Evers-Vermeul (2002). Pronouns and case. Lingua, 112, 301–338. Zwarts, J. (1994). Pronouns and N-to-D movement. In M. Everaert, B. Schouten, & W. Zonneveld (Eds.), OTS Yearbook 1994 (pp. 93–112). Utrecht: Utrecht University.
Index
A accomplishment 188, 191, 193, 194, 201 accusative 85, 109, 161–163, 180, 256–258, 260, 261, 275 achievement 188, 191, 193 acquisition 3, 41–43, 45, 49, 51–53, 56, 58, 60–62, 66–70, 72, 161 L1 (first language) – 2, 3, 41, 42, 50–54, 58, 60, 62, 65–67, 69, 70 L2 (second language) – 2, 3, 41–43, 45–47, 49, 51–54, 57–60, 62–70, 72 L3 (third language) – 42, 43, 62, 66, 68–70 activity 188, 189, 191, 194, 198 adjacency 3, 4, 79, 91, 93, 96, 97, 99, 101, 102, 105, 106, 109 Afrikaans 17, 115, 128, 133, 134, 136, 138, 139, 141, 143, 144, 147–149, 161, 175, 179, 180, 182 Agree 3, 8, 102, 115, 119–124, 133, 136, 147–150, 192, 224, 226, 229 agreement 7, 8, 12, 13, 30, 31, 33, 34, 56, 61, 98–100, 102, 106, 107, 110, 115, 116, 121, 122, 135, 136, 138–140, 143, 147, 211, 217, 218, 222, 224–229, 243, 244, 268, 276, 277, 279 adjectival –/inflection see “strong agreement/inflection” and “weak agreement/inflection” – attraction 274 long-distance – 115, 116, 119 strong –/inflection 8, 254, 256–261, 269, 276
weak –/inflection 8, 255–261, 276, 278 Aktionsart see “aspect” Amharic 244 anaphor 159–162, 164, 165, 178–182, 280 complex –/reflexive 5, 159–162, 164, 165, 177, 180–182 simplex –/reflexive 160, 164, 175, 179 weak –/reflexive 4, 161, 162, 164, 169, 173–175, 177–180 anticausative 5, 158, 165, 171, 178, 180, 182, 183 apokoinou 14, 21–23, 33, 35, 36 apposition/appositive 8, 252–255, 257, 262–264, 266–270, 276, 280 “close” – 252–255, 266–269, 271, 273 “loose” – 252–254, 256, 266–268, 273 Appositive Hypothesis 253–255, 257, 262–264, 266, 268, 270 arity operations 158, 161, 163, 180, 183 Armenian 225 article phrase (artP) 8, 260, 261, 277 aspect 6, 7, 156, 166, 181, 187–191, 197–199, 202, 204, 209, 210, 213 external/presentational – 188, 197 internal – (Aktionsart) 188–191, 198, 199 asymmetry 12, 18, 115, 117, 118, 144, 274 asyndetic coordination 36 Avoid Pronoun Principle 127, 128
Index
B backgrounding 19 Bangla (Bengali) 3, 4, 79, 80, 90, 93–97, 103, 105, 106, 109, 111 Bare Phrase Structure 268 C cartography 147 Case Resistance 91 causative 171, 173–175, 178, 180, 182 Celtic 32 Chinese 193, 212 clause union 3, 79–81, 83, 85, 89, 92–94, 96, 97, 99, 103, 105, 106, 109, 110 clitic 9, 29, 32, 36, 37, 165, 248, 251, 252, 265, 270–273, 279 coercion operator 194 competing grammars 147 complementizer 13, 15–18, 34, 80, 95–98, 102, 107, 109, 132 complex verb 43, 44, 59, 67, 72 Complex Inversion 37 conditional 24, 27, 36, 49 – inversion 36 conjunction 2, 20, 21, 46, 47, 71 – -induced inversion 20 connective 45–47, 64, 65, 71 control 3, 79, 81, 82, 84–87, 97, 99, 149, 173, 183, 218 counterfactual 24, 25 cycle 2, 15–18, 21, 23, 26, 27, 31, 34–36 cyclic(ity) 28, 122, 124 D Danish 126, 191, 217, 223, 230, 234, 235, 246 colloquial – 234, 246 dative 85, 86, 128, 147, 148, 181, 221, 229, 230, 236, 244, 245, 256–259, 261, 275–277 deficient (article/demonstrative/pronoun) 265, 269, 270, 272, 273, 280, 281 definiteness 7, 8, 128, 217–219, 221–233, 235, 236, 240–244, 246 demonstrative 27, 260, 273, 277, 280
determiner 35, 109, 160, 219, 222, 223, 233, 234, 240–243, 246, 251, 252, 254–260, 262, 263, 265–267, 269, 271, 273, 275–279, 281 diachronic/diachrony 5, 115, 128, 135–138, 140, 141, 144 ditransitive 104, 110, 148 Dutch 3, 5, 6, 12, 15–21, 23, 24, 26–28, 34, 35, 41, 71, 91, 97, 101, 107, 110, 115, 128, 129, 132–134, 138–144, 147, 155, 156, 158–164, 169, 173, 175–177, 179–183, 191, 193, 196, 197, 200, 211, 230, 246, 277 colloquial – 34, 35 – dialect of Strijen 34 – dialect of Amsterdam 17 Early Modern – 20 Heerlen – 5, 6, 183 Middle – 138 Obachsberg – 183 E empty operator 24–26 English 2, 4–6, 12, 17–20, 26, 35, 41–43, 53–56, 58, 60–70, 72, 101, 110, 126, 129, 131, 133, 134, 136, 138–141, 143, 144, 147, 149, 155, 156, 181, 187–197, 199–202, 209–212, 218, 264, 269, 274–277 – dialects of Northern Ireland 18 Middle – 20, 139, 140 Old – 139 EPP-feature 98, 121, 123, 127, 129–133, 136, 146–148 eventive verb 6, 187, 188, 190, 191 exclamation 269 expletive (Expl) 4, 5, 14, 115–128, 130–133, 135–147, 149, 150, 182, 241, 242, 264–266, 269, 270, 273 transitive – construction (TEC) 145
Index
vP- – (vP-Expl) 4, 118, 135, 136, 138, 140, 141, 143–145, 149 Extension Condition 123 Externalize Theme 166 extracyclicity 27–29, 36 extradependence 27–29, 37 extraposition 79, 81, 83, 84, 86, 91, 92, 96, 102, 103, 105, 109 F Faroese 136, 144, 217, 219, 229, 231, 233, 245, 246 feature-checking 115, 120 focus 21, 22, 29, 37, 52, 71, 72, 108, 111, 148, 159, 160, 177, 178, 181, 198, 217, 237–240, 247 – particle 22, 159, 160 – -sensitive adverb 29, 37 French 20, 35, 37, 155, 165, 166, 176, 200, 201, 203, 205, 206, 210, 213, 275, 276, 279 Old – 20 Frisian 34, 161, 179, 180, 182 functional preposition 229 G General DP-Hypothesis 252–256, 265, 267, 268, 271, 273, 274, 280 generic 6, 14, 156, 187, 188, 203–207, 209, 213, 242 – operator 156, 204, 205 – sentence 6, 156, 187, 188, 203–207, 209, 213 genitive 7, 106, 109, 221, 222, 225, 227, 228, 236, 237, 239, 241–244, 256, 258, 267, 272, 275 German 2–5, 8, 16, 19, 20, 29, 36, 37, 41–72, 79–81, 91, 93, 94, 96–98, 101–107, 109, 110, 117, 126–129, 132–137, 139, 143–145, 147–149, 155–169, 171, 173, 176, 177, 180, 181, 183, 230, 236, 246, 253, 254, 257, 258, 260–262, 264, 268, 269, 271, 273–277, 280
Early Modern High – 20 Middle High – 20 Middle Low – 236 Old High – 137 Siberian Mennonite Low – 20 Germanic 1, 4, 5, 7, 9, 11, 14, 15, 17, 19, 20, 30, 32, 33, 37, 41, 43, 68, 71, 106, 108, 115, 116, 125, 128, 130, 131, 133, 136, 137, 140, 142, 144, 147, 155, 158, 164, 181, 191, 193, 209, 256, 270, 275, 277 (Continental) West- – 15, 18, 94 North – 15 Greek 32, 37, 131, 155, 166, 181 H Head Final Filter (HFF) 101, 110 Hindi 109, 111 Human Sentence Processing Mechanism (HPSM) 81, 86, 88, 104, 106 Hungarian 35, 264 I i-within-i condition 170 Icelandic 4, 7, 16, 35, 117, 118, 126, 127, 129, 132–135, 137–140, 143–145, 148, 149, 217, 218, 220–222, 224, 227, 228, 230, 235, 239, 242, 243, 264 imperative 36, 54 imperfect(ive) 6, 156, 193, 197, 212 inchoative 171, 177, 184 Individual Level Predicate 6, 187, 188, 201, 209, 213 Indo-Aryan 79, 91, 93, 109 infinitival marker 91 infinitive 3, 4, 18, 35, 59, 79–82, 84, 85, 87, 88, 92, 93, 97, 100, 104, 107, 109, 210 coherent – 3, 79–90, 92, 93, 96, 97, 100, 103–105 non-coherent – 3, 79–86, 88, 89, 92, 93, 96, 97, 99, 103, 109 interlanguage 41–43, 47, 49, 52, 53, 55, 58, 59, 62, 64–70 intraposition 81, 82, 84, 86, 87
Index
Italian 52, 58, 131, 133, 134, 166, 197, 243, 244, 264, 269, 270, 279 J Japanese 109, 272–274, 279 K Kannada 182 Korean 53, 58, 66, 193 L left dislocation 49 Left-to-Right Constraint 81, 88 linker 2, 11–13, 15–18, 24, 32, 33, 35 Locative Inversion 147 M Mainland Scandinavian (MSc) 29, 34, 126, 133, 134, 136, 138, 140, 143, 164, 211, 220, 227, 229, 247 Merge 2, 11, 12, 14, 15, 20, 32–34, 116–122, 124–126, 131–133, 135, 136, 143, 144, 146, 148, 149 middle 5, 6, 155–158, 160, 163–166, 168, 169, 178–183, 236 morphological impoverishment 138, 140 morphological merger 98 morphology 2, 4, 11, 30, 31, 66, 70, 101, 102, 135, 136, 138–141, 156, 211, 226, 256 inflectional – 4, 11, 138–140, 161 Move 7, 13, 14, 34, 71, 94, 95, 115, 116, 120, 121, 124, 143, 144, 173, 205, 207, 212, 232, 234, 236–240, 247, 251, 260, 262, 270, 271 N Nama 257 narrative inversion 24, 25, 35 negation 22, 83, 84, 86–90, 109 Negative Polarity Item (NPI) 21, 22 nominative 106, 109, 126, 133, 148, 161, 162, 221, 244, 256–258, 260, 275
nonargument 145, 155, 158, 160–162, 165, 166, 168, 169, 180–182 nonfinite verb 35, 59, 61, 67–69 Norwegian 29, 37, 50, 72, 126, 191, 207, 217, 219, 221, 228–231, 233, 237, 241, 245–247, 264, 270, 273, 279 Solør – 229, 230, 237, 247 O Optimality Theory (OT) 31 P part-whole relation 218, 220, 221 passive 97, 125, 126, 130, 132, 134, 137–142, 145, 148, 156, 158, 165, 171, 184, 185 impersonal – 126, 130, 138–142, 156, 168, 169, 171, 175–177, 179, 182, 183 perfect(ive) 6, 23, 60, 66, 67, 95, 188–190, 197, 209–212 Persian 109 phase 4, 7, 115, 119, 120, 122, 123, 146, 148, 149, 223, 232, 235 strong – 7, 223, 232 piedpiping 121, 133, 134, 142, 240 head- – 133–136, 138 – of PP 240 spec- – 133, 134, 136, 140, 142 Portuguese 52, 58, 193 positional (dependent) marking 11–14, 16–18, 20–22, 26, 27, 30–35 [poss] feature 227, 228 possession 218, 225, 244, 247 alienable – 218 inalienable – 218, 244, 247 possessor 7, 8, 109, 217–247 nonpronominal – 217, 219, 221, 222, 227, 228, 230, 232, 234, 235, 237–239, 242, 245 pronominal – 7, 8, 217–19, 221, 222, 224–228, 234, 235, 237, 240, 243, 244, 245, 247
Index
postnominal – 7, 217–219, 221, 222, 224, 228, 230, 231, 235–237, 239, 242, 245–247 prenominal – 4, 7, 8, 157, 167, 172, 174, 182, 217, 219, 223, 230–240, 245, 246 pseudo – 241 presentational aspect see “aspect” presentational (clause/construction) 125, 126, 138, 140, 141, 149, 241 Principle of Monoinflection 258, 259 processability 42, 51, 58, 68 processing 52, 80, 81, 87, 88, 104–106, 110, 111 progressive 6, 188–190, 194, 195, 197, 212 pronominal 7, 8, 95, 127, 135, 149, 161, 182, 217–219, 221, 222, 224–228, 231, 233–237, 240, 243, 244, 246, 247, 251–260, 263–265, 267–276, 279, 280 – possessor see “possessor” pronoun 8, 15, 22, 36, 83, 84, 93, 109, 127, 128, 175, 179, 180, 224, 225, 236, 245, 246, 251–253, 255, 256, 258, 259, 261–273, 275–278, 280, 281 “intransitive” – 251, 253, 271, 280 possessive – 225, 236, 245, 246, 263 strong – 8, 253, 264, 269–273, 279 “transitive” – 8, 251, 253, 270–273, 279 weak –/pronominal determiner 9, 15, 162, 251, 253, 265, 270–273 proper name 262–269, 271 psych-movement 174, 177 Q quotative inversion 14, 18–20, 33, 35 R re-positioned zu 80, 106, 107 reanalysis 79, 88, 89, 105 reciprocal 163, 164, 183 reference time 189, 190, 199, 202, 203
reflexive
5, 7, 36, 155, 156, 158–170, 173–175, 177–183, 244–246, 278 complex – see “anaphor” inherent – 5, 36, 155, 158–168, 170, 176–180, 182, 183 simplex – see “anaphor” weak – see “anaphor” relational noun 218, 220, 227 restructuring 3, 80, 83, 96, 105 resumptive 45, 48–50, 57, 62 Right Roof Constraint 110 S scope 17, 83, 84, 86–90, 125, 130, 148, 213 scrambling 83, 84, 91, 103, 109, 148, 149, 261 sentence connecting adverb 28 simplex verb 43, 53, 56, 61 Simplicity 81, 88 sisterhood 11, 30, 34, 121 small clause 32 South Asian languages 109 speech act adverbial 28 speech time (interval) 187, 196, 198 stative 6, 187, 188, 190, 194, 201, 203–209, 212, 213 Status (checking) 4, 97, 100, 103 strong – agreement/inflection see “agreement” – feature 31, 118, 129, 146 – phase see “phase” – pronoun see “pronoun” Stylistic Inversion 35 subject-verb inversion 35, 49, 52, 57, 62, 63 suffixed definiteness marker 7, 224, 225, 230, 233, 235, 236, 241, 246 Swahili 229 Swedish 2, 3, 20, 41–50, 53–58, 60–62, 64, 65, 67–72, 126, 138, 139, 141, 144, 191, 217, 228, 230, 233–235, 245, 246, 265, 270, 273, 280 colloquial spoken – 46
Index
Early Modern – 20, 139 Lappträsk – 246 Northern – 245, 248, 265, 270, 273 Old – 20 T telic 191, 198 tense 6, 7, 12, 13, 30, 34, 56, 61, 71, 129, 156, 187–191, 193, 195–200, 202, 203, 209–212 Future Perfect – 209, 210 Simple Future – 199, 200 Simple Past – 199, 200 Simple Present – 6, 187, 188, 190, 191, 193, 195, 196, 199–201, 203, 209, 211, 212 terminative 169, 170 topic 11, 21, 24–27, 52, 71, 118, 137 – drop 24, 25 transfer 2, 3, 41–43, 52–54, 58, 60, 65, 66, 68–70, 122 transitive 8, 145, 147, 148, 158, 160, 166, 168–180, 182, 251, 253, 270–273, 279 – -expletive construction (TEC) see “expletive” – pronoun see “pronoun” transitivizing prefix 170 Turkish 53, 58, 66, 109 U unaccusative 125, 137, 145, 148, 156, 157, 165, 166–177, 180 unergative 155–157, 165, 167–169, 171, 176, 177, 180, 182, 183 Unlike Category Condition (UCC) 108 Unlike Feature Constraint (UFC) 108 UTAH 166, 181 V V1 (Verb First) 2, 12, 14, 23–26, 33, 34, 36, 45, 55, 57, 61, 71, 79, 97, 104 V2 (Verb Second) 1–3, 5, 11–14, 16–21, 23, 24, 26–37, 41–49, 51–63,
66–71, 79, 80, 97, 104, 118, 134, 137, 138, 148, 149 embedded – 16, 34, 134, 148 non-subject-initial –/clause 43, 45, 54, 55, 56, 61–64, 66, 68 subject-initial –/clause 15, 16, 22, 26, 44, 56 symmetrical – 148 V3 (Verb Third) 2, 3, 14, 23, 26–29, 33, 37, 42, 43, 45–47, 49–51, 53–58, 62–66, 68, 71, 72 – -adverb 51 V4 (Verb Fourth) 57, 58 valency-reduction morpheme 158, 169 verbal complex 23, 94, 100, 109 verb raising 18, 94, 129, 130 Very Short Duration (VSD) 6, 187, 194–196, 198, 199, 209, 211, 212 Vestfold 264 vocative 269, 279 Vorfeld 44–46, 71 vP-expletive (vP-Expl) see “expletive (Expl)” vP-raising 5, 131–134, 142, 144, 148, 149 vulnerability of the C-domain 42, 51 W Warlpiri 12, 257 weak – agreement/inflection see “agreement” – anaphor/reflexive see “anaphor” – feature 31 – pronoun/pronominal determiner see “pronoun” West Jutlandic 245, 246 wh-copying 118 wh-expletive 117, 118 wh-question 17, 18, 25, 57 Y yes/no-question 17, 18, 24, 25 Yiddish 16
In the series Linguistik Aktuell/Linguistics Today the following titles have been published thus far or are scheduled for publication: 87 JULIEN, Marit: Nominal Phrases from a Scandinavian Perspective. Expected October 2005 86 COSTA, João and Maria Cristina FIGUEIREDO SILVA (eds.): Studies on Agreement. Expected December 2005 85 MIKKELSEN, Line: Copular Clauses. Specification, predication and equation. viii, 212 pp. Expected September 2005 84 PAFEL, Jürgen: Quantifier Scope in German. ca. 354 pp. Expected December 2005 83 SCHWEIKERT, Walter: The Order of Prepositional Phrases in the Structure of the Clause. 2005. xii, 338 pp. 82 QUINN, Heidi: The Distribution of Pronoun Case Forms in English. 2005. xii, 409 pp. 81 FUSS, Eric: The Rise of Agreement. A formal approach to the syntax and grammaticalization of verbal inflection. xii, 323 pp. + index. Expected October 2005 80 BURKHARDT, Petra: The Syntax–Discourse Interface. Representing and interpreting dependency. viii, 256 pp. + index. Expected October 2005 79 SCHMID, Tanja: Infinitival Syntax. Infinitivus Pro Participio as a repair strategy. 2005. xiv, 251 pp. 78 DIKKEN, Marcel den and Christina M. TORTORA (eds.): The Function of Function Words and Functional Categories. vii, 292 pp. 77 ÖZTÜRK, Balkız: Case, Referentiality and Phrase Structure. 2005. x, 268 pp. 76 STAVROU, Melita and Arhonto TERZI (eds.): Advances in Greek Generative Syntax. In honor of Dimitra Theophanopoulou-Kontou. 2005. viii, 366 pp. 75 DI SCIULLO, Anna Maria (ed.): UG and External Systems. Language, brain and computation. 2005. xviii, 398 pp. 74 HEGGIE, Lorie and Francisco ORDÓÑEZ (eds.): Clitic and Affix Combinations. Theoretical perspectives. 2005. viii, 390 pp. 73 CARNIE, Andrew, Heidi HARLEY and Sheila Ann DOOLEY (eds.): Verb First. On the syntax of verb-initial languages. 2005. xiv, 434 pp. 72 FUSS, Eric and Carola TRIPS (eds.): Diachronic Clues to Synchronic Grammar. 2004. viii, 228 pp. 71 GELDEREN, Elly van: Grammaticalization as Economy. 2004. xvi, 320 pp. 70 AUSTIN, Jennifer R., Stefan ENGELBERG and Gisa RAUH (eds.): Adverbials. The interplay between meaning, context, and syntactic structure. 2004. x, 346 pp. 69 KISS, Katalin É. and Henk van RIEMSDIJK (eds.): Verb Clusters. A study of Hungarian, German and Dutch. 2004. vi, 514 pp. 68 BREUL, Carsten: Focus Structure in Generative Grammar. An integrated syntactic, semantic and intonational approach. 2004. x, 432 pp. 67 MIŠESKA TOMIĆ, Olga (ed.): Balkan Syntax and Semantics. 2004. xvi, 499 pp. 66 GROHMANN, Kleanthes K.: Prolific Domains. On the Anti-Locality of movement dependencies. 2003. xvi, 372 pp. 65 MANNINEN, Satu Helena: Small Phrase Layers. A study of Finnish Manner Adverbials. 2003. xii, 275 pp. 64 BOECKX, Cedric and Kleanthes K. GROHMANN (eds.): Multiple Wh-Fronting. 2003. x, 292 pp. 63 BOECKX, Cedric: Islands and Chains. Resumption as stranding. 2003. xii, 224 pp. 62 CARNIE, Andrew, Heidi HARLEY and MaryAnn WILLIE (eds.): Formal Approaches to Function in Grammar. In honor of Eloise Jelinek. 2003. xii, 378 pp. 61 SCHWABE, Kerstin and Susanne WINKLER (eds.): The Interfaces. Deriving and interpreting omitted structures. 2003. vi, 403 pp. 60 TRIPS, Carola: From OV to VO in Early Middle English. 2002. xiv, 359 pp. 59 DEHÉ, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002. xii, 305 pp. 58 DI SCIULLO, Anna Maria (ed.): Asymmetry in Grammar. Volume 2: Morphology, phonology, acquisition. 2003. vi, 309 pp.
57 DI SCIULLO, Anna Maria (ed.): Asymmetry in Grammar. Volume 1: Syntax and semantics. 2003. vi, 405 pp. 56 COENE, Martine and Yves D’HULST (eds.): From NP to DP. Volume 2: The expression of possession in noun phrases. 2003. x, 295 pp. 55 COENE, Martine and Yves D’HULST (eds.): From NP to DP. Volume 1: The syntax and semantics of noun phrases. 2003. vi, 362 pp. 54 BAPTISTA, Marlyse: The Syntax of Cape Verdean Creole. The Sotavento varieties. 2003. xxii, 294 pp. (incl. CD-rom). 53 ZWART, C. Jan-Wouter and Werner ABRAHAM (eds.): Studies in Comparative Germanic Syntax. Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000). 2002. xiv, 407 pp. 52 SIMON, Horst J. and Heike WIESE (eds.): Pronouns – Grammar and Representation. 2002. xii, 294 pp. 51 GERLACH, Birgit: Clitics between Syntax and Lexicon. 2002. xii, 282 pp. 50 STEINBACH, Markus: Middle Voice. A comparative study in the syntax-semantics interface of German. 2002. xii, 340 pp. 49 ALEXIADOU, Artemis (ed.): Theoretical Approaches to Universals. 2002. viii, 319 pp. 48 ALEXIADOU, Artemis, Elena ANAGNOSTOPOULOU, Sjef BARBIERS and Hans-Martin GÄRTNER (eds.): Dimensions of Movement. From features to remnants. 2002. vi, 345 pp. 47 BARBIERS, Sjef, Frits BEUKEMA and Wim van der WURFF (eds.): Modality and its Interaction with the Verbal System. 2002. x, 290 pp. 46 PANAGIOTIDIS, Phoevos: Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. 2002. x, 214 pp. 45 ABRAHAM, Werner and C. Jan-Wouter ZWART (eds.): Issues in Formal German(ic) Typology. 2002. xviii, 336 pp. 44 TAYLAN, Eser Erguvanlı (ed.): The Verb in Turkish. 2002. xviii, 267 pp. 43 FEATHERSTON, Sam: Empty Categories in Sentence Processing. 2001. xvi, 279 pp. 42 ALEXIADOU, Artemis: Functional Structure in Nominals. Nominalization and ergativity. 2001. x, 233 pp. 41 ZELLER, Jochen: Particle Verbs and Local Domains. 2001. xii, 325 pp. 40 HOEKSEMA, Jack, Hotze RULLMANN, Víctor SÁNCHEZ-VALENCIA and Ton van der WOUDEN (eds.): Perspectives on Negation and Polarity Items. 2001. xii, 368 pp. 39 GELDEREN, Elly van: A History of English Reflexive Pronouns. Person, Self, and Interpretability. 2000. xiv, 279 pp. 38 MEINUNGER, Andre: Syntactic Aspects of Topic and Comment. 2000. xii, 247 pp. 37 LUTZ, Uli, Gereon MÜLLER and Arnim von STECHOW (eds.): Wh-Scope Marking. 2000. vi, 483 pp. 36 GERLACH, Birgit and Janet GRIJZENHOUT (eds.): Clitics in Phonology, Morphology and Syntax. 2001. xii, 441 pp. 35 HRÓARSDÓTTIR, Thorbjörg: Word Order Change in Icelandic. From OV to VO. 2001. xiv, 385 pp. 34 REULAND, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. xii, 255 pp. 33 PUSKÁS, Genoveva: Word Order in Hungarian. The syntax of Ā-positions. 2000. xvi, 398 pp. 32 ALEXIADOU, Artemis, Paul LAW, Andre MEINUNGER and Chris WILDER (eds.): The Syntax of Relative Clauses. 2000. vi, 397 pp. 31 SVENONIUS, Peter (ed.): The Derivation of VO and OV. 2000. vi, 372 pp. 30 BEUKEMA, Frits and Marcel den DIKKEN (eds.): Clitic Phenomena in European Languages. 2000. x, 324 pp. 29 MIYAMOTO, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 2000. xiv, 232 pp. 28 HERMANS, Ben and Marc van OOSTENDORP (eds.): The Derivational Residue in Phonological Optimality Theory. 2000. viii, 322 pp.
A complete list of titles in this series can be found on the publishers website, www.benjamins.com