Repairs: The Added Value of Being Wrong 9781614510796, 9781614510802

Table of contents :
Acknowledgements
Contributors
1 Introduction
2 Semantic competition over morphological representations. A case study from Slavic
3 Repairs for Reasoning
4 Generic rescue: Argument alternations and the monotonicity condition
5 Prepare and Repair: On pre-emptive strikes and post-hoc patches
6 Repair-driven verb movement in English locative inversion
7 Linearisation as repair
8 Repairing resumptive structures, or: How faulty is the Lexicon?
9 That-trace effects and resumption - How Improper Movement can be repaired
10 Passives of reflexive verbs: The repair of a Principle A violation
Subject index

Citation preview

Patrick Brandt and Eric Fuß (Eds.) Repairs

Interface Explorations

Editors Artemis Alexiadou and T. Alan Hall

Volume 27

Repairs

The Added Value of Being Wrong

Edited by Patrick Brandt and Eric Fuß

ISBN 978-1-61451-080-2 e-ISBN 978-1-61451-079-6 ISSN 1861-4167 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliografische Information der Deutschen Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. © 2013 Walter de Gruyter, Inc., Boston/Berlin Cover image: iStockphoto/Thinkstock Typesetting: PTP-Berlin Protago-TEX-Production GmbH, Berlin Printing: Hubert & Co. GmbH & Co. KG, Göttingen ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Acknowledgements A serious and often enough irreparable mistake consists in not having acknowledged someone (or something) who (or which) one should have acknowledged. The editors feel that it would be an even bigger mistake not to acknowledge at all (and more acknowledgements can be found in the initial footnotes or in the appendices to the individual contributions). They wish to express their gratitude for the professional support and kind help coming from the Deutsche Gesellschaft für Sprachwissenschaft (DGfS), during the 31st meeting of which a workshop “Repairs” was held in the city of Osnabrück that was a sine qua non among the conditions that had to be met. Given those moments’ richnesses vis-a-vis their fading presence, it is a matter of course that less than the biggest part of the structure of the event could be preserved in the mapping to the present product (nor should it have been). The editors are grateful to the authors for their enthusiasm and patience, to the peer reviewers for their indispensable help, and to the editing and production team at De Gruyter for their careful attendance. The following alphabetic but otherwise unsorted list of co-operators may be hoped to strike a balance between transparent recognition and personal confidentiality. Artemis Alexiadou (U Stuttgart), Olga Borik (U Autònoma Barcelona), Jennifer Carlson (De Gruyter), Regine Eckardt (U Göttingen), Emily Farrell (De Gruyter), Günther Grewendorf (U Frankfurt), Fabian Heck (U Leipzig), Daniel Hole (HU  Berlin), Wolfgang Konwitschny (De Gruyter), Cécile Meier (U Frankfurt), Sonja Müller (U Bielefeld), Andreas Pankau (U Frankfurt), Beatrice Primus (U Köln), Angelika Wöllstein (IDS Mannheim), Lara Wysong (De Gruyter). All collective and individual efforts notwithstanding, there is prone to be mistakes, errors and misunderstandings remaining. Hopefully, these can be repaired, to the extent that they aren’t don’t cares, or forgiven. P.B. E.F.

Contributors Josef Bayer General and Germanic Linguistics Department of Linguistics University of Konstanz Universitätsstraße 10 Box D 191 78457 Konstanz Germany Marcel den Dikken Linguistics Program CUNY Graduate Center 365 Fifth Avenue New York, NY 10016-4309 USA Mojmír Dočekal Department of Linguistics and Baltic Languages Masaryk University Arna Nováka 1/1 602 00 Brno Czech Republic Bernhard Fisseni Fachbereich Mathematik Universität Hamburg Bundesstraße 55 20146 Hamburg Germany

Holden Härtl Universität Kassel Institut für Anglistik/Amerikanistik Kurt-Wolters-Straße 5 34125 Kassel Germany Joost Kremers Universität Göttingen Seminar für Deutsche Philologie Käte-Hamburger-Weg 3 37073 Göttingen Germany Ivona Kucerova Linguistics & Languages Togo Salmon Hall 608 McMaster University 1280 Main Street West Hamilton, Ontario, Canada L8S 4M2 Martin Salzmann University of Leipzig Department of Linguistics Beethovenstraße 15 04107 Leipzig Germany Hans-Christian Schmitz Fraunhofer FIT Schloss Birlinghoven 53754 Sankt Augustin Germany

viii   

   Contributors

Florian Schäfer Institut für Linguistik/Anglistik Universität Stuttgart Keplerstraße 17 70174 Stuttgart Germany

Volker Struckmeier Institut für Deutsche Sprache und Literatur I Albertus Magnus Platz Universität zu Köln 50923 Köln Germany

Contents Acknowledgements Contributors  vii

1

2

3

 v

Patrick Brandt and Eric Fuß  1 Introduction Mojmir Dočekal and Ivona Kucerová Semantic competition over morphological representations. A case study from Slavic  31 Hans-Christian Schmitz and Bernhard Fisseni Repairs for Reasoning  55

4

Holden Härtl Generic rescue: Argument alternations and the monotonicity condition  95

5

Marcel den Dikken Prepare and Repair: On pre-emptive strikes and post-hoc patches  131

6

Martin Salzmann Repair-driven verb movement in English locative inversion

7

Joost Kremers Linearisation as repair

8

Volker Struckmeier Repairing resumptive structures, or: How faulty is the Lexicon?

9

 155

 207

 237

Josef Bayer and Martin Salzmann That-trace effects and resumption – How Improper Movement can be repaired  275

Florian Schäfer 10 Passives of reflexive verbs: The repair of a Principle A violation Subject index

 365

 335

Patrick Brandt and Eric Fuß

Introduction Grammatical structures connect systems of thought and reasoning and systems of articulation and perception. It appears obvious that the specific conditions of these systems hardly fit each other. The quality of what needs to be done seems so different, even if we do not know much about the representation of thought, independently of how it might translate into natural language expressions. We know of syntactic structures that they are built recursively with very restricted resources  – across languages, grammars appear to use a small set of lexical categories, an even smaller set of functional vocabulary and a small set of eversimilar restrictions and rules that condition or manipulate the structure-building process. How can such a limited system be useful for exchange about what we call reality and possibility, in all its richness, diversification and complication?

1 Relating sound and meaning In a most general setting, we can think of different systems (cognitive modules in the sense of Fodor 1983) that are engaged in relating sound and meaning as languages that each manipulate their specific vocabulary in a specific fashion in the sense of Carnap (1931: 435): “In order to characterize a particular language, one has to specify its vocabulary and its syntax, that is, the words that occur in it and the rules according to which sentences can be built from these words and according to which such sentences can be transformed into other sentences of that same language or a different language (so-called rules of inference and rules of translation).”

Viewing thus the connection between sound and meaning as, essentially, a translation problem, it seems clear that translating from one system into the other is by no means a straightforward matter. In theories putting the burden of the problem on the role of different levels of representation (e.g., Head-driven Phrase Structure Grammar (Pollard and Sag 1987), Lexical Functional Grammar (Bresnan 1982) or, more recently, Williams’ 2003 Representation Theory), it is maybe most obvious that there can be no simple homomorphism (i.e., a structure-preserving mapping) between different levels, as this would render the different levels superfluous indeed. Taking the basic distinction between syntax and semantics (i.e., the language in which we state truth conditions) as a starting point as well as

2   

   Patrick Brandt and Eric Fuß

leitmotif for what is to follow, Dowty (1979: 29–31) discusses the fundamental fact that if syntactic and semantic structures were completely isomorphic, i.e., the translation would consist in a structure-preserving mapping from one structure into the other, nothing would speak against interpreting the syntactic structures directly (i.e., with respect to a model). If there is a level of semantic representations, then these representations will have their own structures that will only be partially in harmony with the structures of syntax (cf. already Jackendoff 1972), or, more strongly and as Kracht (2007: 47) puts it: “[W]e need a definition of meaning that bans all mentioning of syntactic structure; it is not the task of semantics to state in which way things are put together in the syntax.”

Whatever we want to say about how we get to different levels of representation and how we get to relate them, it seems fair to say that it is exactly the point of having different levels of representation that each one of them takes care of different aspects of grammatical information in ways that fit the purpose. It is clear that the number of levels assumed will have a huge impact on one’s theory,¹ as well as, interdependently, the leeway assumed to be allowed in the translation between levels (that may be modeled, e.g., by making translation partial or by allowing flexibility in the choice of symbols translated from or into). Regardless of such questions, it appears most economical if not even necessary to assume that the representations at certain levels will be derived in certain ways and subject to certain restrictions. Let us look in more detail at theories putting the burden of the translation problem on the supposedly strictly derivational nature of grammar, as such theories form the base of the contributions in this volume. In Government and Binding (henceforth: GB) theory (Chomsky 1981 and subsequent work), it was assumed that the module responsible for thematic roles interacted with the modules responsible for Case assignment and with the module regulating referential dependencies. Similarly in more recent approaches in the wake of GB theory, different parts of the clausal spine take care of interdependent but separate pieces of grammatical information: the verbal projection (VP or vP) constitutes the domain responsible for the coding of argument structure, the tense projection (TP) constitutes the domain responsible for the coding of temporal relations as well as predication and agreement, the complementizer projection (CP) constitutes the domain responsible for linking propositional contents to the discourse. In these approaches, we are really faced with a double translation problem: For one thing, certain parts of the structure have to be handed to the semantics, i.e., a representation of truth conditions must be found for them into which they get translated. For another thing, and quite crucially, certain parts of the mate-

Introduction   

   3

rial in the derivation have to be “passed up”, i.e., kept in memory to be combined with material that is to be added later. Some questions that arise are: how can they be passed up, and in particular, why were they further down to begin with, if they could not be interpreted there? Or were they, and do they get interpreted again? And if so, how could this be, if the different cycles were doing so different things?

2 Subject to basic conditions Let us ask what would be plausible properties of and hence conditions on representations of sound and meaning. Clearly, on the sound side, the sign will have to be pronounced or demonstrated in some fashion. In the case of sound languages, there is a need to linearize the signs, i.e., translate constituent structure into a linearly ordered string representing sound patterns. But when we look at sign languages, there appear to be looser conditions on linearization, as here the option exists of realizing different signs simultaneously more easily. The sign modality is all decisive for the conditions that we want to impose on the sound side. Kremers (this volume) argues exactly that simultaneous realization instantiates the best or default case that is deviated from only if necessary, as in the case of sound languages as opposed to (good parts of) sign languages. Even if the amount of possible variation that we observe at the sound end may seem at first sight discouraging, linguists have succeeded in developing suitable means of describing sounds or signs systematically and have found amazing regularity within the systems responsible for their articulation. Looking at the meaning end, we expect to find little variation as all human languages should be capable of developing the same expressive power. But we cannot be sure that our representations of meanings are right, as we cannot look at meanings directly. Still, solid generalizations have emerged here as well; assuming a traditional picture with predicate-logical representations containing quantifiers and variables, it seems to hold that there must be no vacuous quantification, and looking from the other end, each variable present in a representation fit to express truth conditions will have to be assigned a value eventually in some way or other. Such generalizations could be said not to constitute facts about meanings, though, but rather about the language that we use to express meanings as a matter of convention. Eventually, there seems to be no choice but to try and find out about the systematicity of meanings via the systematicity of the use of certain types of signs that we see realized in certain environments in the spirit of Kühnhold (1973) or Levin (1993), who demonstrate impressively that the possi-

4   

   Patrick Brandt and Eric Fuß

bility of manipulating sound representations in certain fashions relates to certain aspects of the meanings of the signs so coded. Something that is hard to doubt is that meanings have to do with truth conditions, i.e., the conditions under which language users would be ready to associate certain types of situations or states of affairs with certain types of signs, in particular, sentences expressing propositions that may be true or false. What we would be ready to call true or false may, following tradition, be assumed to fall into one of two classes formally, namely, predication or quantification. Disregarding the latter for the moment (but cf. below section 7) and regarding predication, Quine and Strawson reason as follows: “Predication joins a general term and a singular term to form a sentence that is true or false according as the general term is true or false of the object, if any, to wich the singular term refers.” (Quine 1960: 96) “A term may be thought of as a principle of collection of other terms. It may be said to collect just those terms such that when it is assertively tied to any one of them, the result is not just a significant, but also a true, proposition.” (Strawson 1959: 167)

Depending on how we want to conceive of predication, additional possible levels of representation emerge on the meaning side, such as topic and focus (or theme and rheme or foreground and background) structure; the obvious and complex question is what is really needed – i.e., what levels should be assumed and what has to be said about the “talk” between them – to capture generalizations made with respect to the observable language data. A promising candidate to relate to predication is what Strawson (1959: 178–179) uses as a grammatical criterion to identify predicates in the sign: “But it is, I think, unquestionable that these examples show, to speak metaphorically, a kind of effort on the part of language to keep, or seem to keep, in line two criteria for something being predicated, or appearing as a predicate: the grammatical criterion, according to which that which is predicated is introduced by a part of the sentence which carries assertive symbolism; and the category criterion according to which only universals, or complexes containing universals, never particulars simpliciter, can be predicated.”

When we ask what Strawson’s grammatical criterion of carrying “assertive symbolism” amounts to in the observable data, we find that it is what is generally called “subject-predicate agreement” by linguists. This subject-predicate agreement is related to other phenomena that constitute prominent properties of sentences, in particular, that of being tensed, which is in turn crucial for the assignment of Case. Case assignment may go along with displacement, which is modeled in current generative theory with an EPP feature, the acronym recalling

Introduction   

   5

the old “extended projection principle”, a.k.a. the generalization that sentences must have (syntactic) subjects. In current minimalist syntactic theory, agreement and a spurious EPP feature that is made responsible for displacement is pretty much all there is left to the machinery, so it is not surprising that the more syntax-oriented papers in this volume in some way or other all make reference to it, and some very centrally so (cf. the papers by den Dikken, Bayer and Salzmann, Salzmann, and Schäfer).

3 Perfectionism vs. the need for repairs Regardless of (or maybe actually rather because of) fundamental questions very much pending, it has become a prominent question as to what extent the grammar constitutes an optimal or perfect solution to the translation problem. According to the “Strong Minimalist Thesis” (SMT, Chomsky 2001 and subsequent work), “[…] language is an optimal solution to interface conditions that the faculty of language (FL) must satisfy; that is, language is an optimal way to link sound and meaning, […]” (Chomsky 2008: 2). More specifically, the grammar is “perfectly designed” in the sense that “it contains nothing more than what follows from our best guesses regarding conceptual, biological, physical necessity” (Boeckx 2006: 4). Thus, the workings and the output of the computational system are constrained by considerations of derivational and representational economy (see also Chomsky 1991, 1993 for early proposals along these lines): Operations performed by the computational system of language reduce to the bare minimum required to construct hierarchical syntactic structures that can be interpreted at the interfaces to other cognitive systems in terms of sound and meaning (Merge as the only structure-building operation, locality of movement/internal Merge, small derivational cycles/phases etc.). Likewise, representations generated by the computational system may not contain superfluous elements or, more generally, elements that cannot be interpreted at the interfaces to other cognitive systems (Principle of Full Interpretation, fewest symbols, small cycles/phases/ spell-out domains). The hope is that the reductionist character of this perspective not only leads to an elegant and at the same time rigorous theory of Universal Grammar but also makes available deeper explanations, in the sense that certain properties of Universal Grammar need no longer be treated as theoretical primitives, but can be attributed to so-called “3rd factor effects” such as principles of efficient computation and independent conditions imposed at the interfaces to other cognitive systems (cf., e.g., Chomsky 2005, Chomsky 2007).

6   

   Patrick Brandt and Eric Fuß

Thus, under current minimalist assumptions (i.e, the SMT) there should be no room for repairs, that is, additional operations that apply to ensure interface compatibility. However, there are well known basic facts suggesting that the structures assembled by the computational system may seem unfit for straightforward interpretation at the interfaces. On the one hand, syntactic structures may contain symbols without meaning, for example, pleonastic elements such as expletives and INFL-supporting auxiliaries: (1) * (There) is someone in the room. (2) We * (do) not leave. On the other hand, there are cases where the interpretation of an expression seemingly fails to be connected (in any transparent way) to visible symbols in the syntactic representation. E.g., middle constructions are generally interpreted in a modal and generic fashion although there is no element in the structure signaling this.² (3) The book reads well. ‘The book can be easily read.’ Such mismatches between form and meaning seem to call for additional mechanisms that ensure that examples like (1)–(3) are correctly interpreted (or, rendered interpretable at all). Relevant proposals in the literature typically differ with respect to the point in the derivation where the rescue operation takes place. In the case of pleonastic elements, for example, it is sometimes assumed that the problematic elements are part of the syntactic derivation, but deleted during the covert syntax, prior to the LF-interface (cf. e.g. Chomsky 1986 on expletive replacement). Alternative proposals have it that elements that lack meaning are not part of the syntactic derivation at all, but are inserted late in the PF branch of the computation as some form of last resort (cf., e.g., Chomsky 1991, as well as Lasnik 2000 on do-support). Concerning modal as well as generic interpretations that are not transparently coded, it has become custom to posit silent operators at LF that achieve the interpretive effect, usually in terms of quantification over variables that are traditionally used to restrict the interpretation function, i.e., variables over times or worlds (on which more below in section 5). The alleged absence of repairs is intimately connected to another minimalist presumption, namely, that we do not expect the workings of the computational system to exhibit any sort of flexibility or optionality – in any given derivation, the set of syntactic operations should reduce to the minimal number of computational steps that are necessary to meet interface requirements; crucially, this is

Introduction   

   7

usually taken to imply that these operations are carried out obligatorily (“crashproof syntax”, Frampton and Gutmann 2002: 1): “The Minimalist Program is guided by the idea that the syntactic system of the language faculty is designed to optimally meet design specifications imposed by the interface systems. […] an optimal derivational system, at least from a computational point of view, is a system that generates only objects that are well-formed and satisfy conditions imposed by the interface systems.”

However, there are some indications that the “bare minimum” is neither always sufficient nor necessary to guarantee interpretability at the interfaces. An important observation (going back to Ross 1969) suggesting that the workings of the computational system are possibly less “perfect” than expected involves cases where an island violation is rescued by deleting the problematic part of the structure at PF: (4) a. * She kissed a man who bit one of my friends, but Tom does not realize which one of my friends she kissed a man who bit. b.

She kissed a man who bit one of my friends, but Tom does not realize which one of my friends.

(5) a. * I believe the claim that he bit someone, but they don’t know who I believe the claim that he bit. b.

I believe the claim that he bit someone, but they don’t know who.

(4) and (5) show that island violations can be remedied by sluicing (both sets of examples involve extraction from a complex NP; in (4), which one of my friends originates in a relative clause; in (5) who is extracted from an NP-internal complement clause). This possibility seems to suggest that the syntactic component is capable of generating otherwise illicit structures as long as the syntactic output can be patched up by relevant post-syntactic operations (i.e., ellipsis, see Fox and Lasnik 2003 for further discussion). Further examples of so-called “repair by ellipsis” come from cases where an otherwise obligatory movement operation can be suspended if the violating structure undergoes deletion at PF (Lasnik 1999, Merchant 2001, Lasnik and Park 2003, Craenenbroeck and den Dikken 2006, Bošković 2011, den Dikken, this volume). For example, Lasnik (1999) discusses cases where head movement fails to apply when the relevant parts of the structure are subject to ellipsis, as in (6) (VP-ellipsis/pseudogapping):

8   

   Patrick Brandt and Eric Fuß

(6) a.

You will [VP2 believei [AgroP Bobj [VP1 ti tj ]]].

b. * You will [VP2 [AgroP Bobj [VP1 believe tj ]]]. c.

You might not believe me but you will [VP2 [AgroP Bobj [VP1 believe tj ]]].

Lasnik assumes that this effect can be explained if we assume that strong features (which normally cause the derivation to crash at PF) can remain unchecked in cases where the violating structure is eliminated by PF deletion (cf. however den Dikken, this volume, for an alternative account casting some doubt on the notion that PF operations can mend “narrow-syntactic improprieties”). In a similar vein, Merchant (2001) argues that (normally obligatory) EPPdriven movement of the subject to SpecIP/TP can be suspended in cases of sluicing: (7) a. * Which Marx brother did she say that [[a biography of _ ] will appear this year]? b.

A biography of one of the Marx brothers will appear this year – guess which!

(7a) illustrates a familiar Subject Condition effect: Extraction from a complex NP in subject positon is banned. In contrast, the very same operation seems to be available in (7b), where the remainder of the wh-clause is elided. Under the assumption that the island status of subjects is linked to the derived positon (while extraction is possible from the VP-internal base position of subjects of unaccusatives and passives), the acceptability of (7b) can be attributed to the absence of EPP-driven subject movement to SpecIP/TP. Again, this implies that the imperfect output of a “sloppy” syntactic derivation (in that case involving an unchecked EPP feature) can be saved by the application of post-syntactic repair operations (VP ellipsis).³, ⁴ These brief considerations already suggest that operations triggered by the need to “patch up” the (eventual) syntactic output are in fact much more frequent than traditional wisdom would have us believe.

4 Amending improprieties This volume aims at exploring the breadth of repair phenomena, adopting the perspective that repairs are devices that are routinely and virtuously employed by the grammar to solve the translation problem. In particular, we want to pursue

Introduction   

   9

a notion of “repair” according to which repairs not only serve to rescue faulty derivations, but can as well be put to service to economically code interpretations that are difficult or perhaps even impossible to express transparently. Reinhart (2006) argues for such a productive or creative understanding of repairs as special or marked operations that are allowed to apply if and only if they are needed to achieve interpretations that could not be achieved using minimal means. Taking the coding of focus for illustration, the minimal strategy here consists in assigning stress to the most deeply embedded constituent, i.e., determining the stress on the basis of the independently needed constituent structure alone (cf. Cinque’s 1993 “nuclear stress rule”, henceforth NSR). This “neutral focus” or “information focus” leads to a range of possible focus interpretations, namely, the ones corresponding to the constituents that contain the main stress in the sentence. This can be evidenced by different types of wh-questions to which a sentence with neutral focus may serve as a felicitous answer: (8) Lucy is building a DESK. Possible corresponding questions: (9) a. What’s the matter?

(IP focus)

b. What’s Lucy doing?

(VP focus)

c. What’s Lucy building?

(Object focus)

While it may be good enough on most occasions, neutral focus does not cover certain special focus interpretations; e.g., an interpretation according to which the verb alone is in focus is not in the set of possible foci derived from the nuclear stress rule. To achieve this special interpretation, the focus must be “shifted” – stress can no longer be determined on the basis of the syntax but must be assigned independently (and idiosyncratically) to the V node, cf. (10). (10) Lucy is BUILDING a desk. Corresponding question: (11) What’s Lucy doing with a desk? It is worth stressing that different languages may employ different strategies to overcome certain problems. Staying with accent placement and focus interpretation, Neeleman and Reinhart (1998) point out that languages like Dutch

10   

   Patrick Brandt and Eric Fuß

or German that allow scrambling would have the object in a position where it does not receive the syntax-determined stress, i.e. the stress that is assigned by Cinque’s NSR. (12a) is a structure where accent placement is on the direct object, whereas (12b) is a structure (base-generated according to Neeleman and Reinhart, or maybe derived by moving the object (or other material) further up in the structure). (12) a. … weil Lucy grad ’n TISCH baut because Lucy just a table builds ‘since Lucy is just building a table’ b. … weil Lucy ’n Tisch grad BAUT because Lucy a table just builds It thus appears that the choice of repair strategy is restricted by other properties (parameters) of the languages where it applies. English cannot use scrambling and has to resort to applying a special stress assignment operation (“stress shift”). Dutch and German have a syntactic choice and use that to derive the special interpretation (cf. Krifka 1998 for pertaining discussion). On Reinhart’s approach, the system first tries the regular (best possible, easiest) way, which is also sufficient in the general case. Only if the regular way does not deliver the intended interpretation may another way be tried which is more costly in involving extra operations. For further examples and discussion of economy-based types of repair, making use of the notion that additional operations are warranted if they serve to code interpretations which otherwise wouldn’t be available, cf. the contributions by Bayer and Salzmann, Struckmeier, and Schäfer. We may wonder whether from this economy-based perspective, the term “repairs” is not misleading in that it suggests that something which is faulty or damaged in some way is restored to a previously ideal state. To the extent that something was never and could never be ideal to begin with, “amends” is perhaps a better term. In any event, it appears useful to distinguish between what might be called genuine repairs (rendering derivations interpretable) and what might be called “creative repairs” or “amends” that make available interpretations that could not be coded using standard means.

Introduction   

   11

5 Toward a typology of repairs In a “temporal” dimension,⁵ we can distinguish between repairs that remedy an existing damage and thus apply post-hoc and repairs that take precautions against potential translation problems that arise at later stages. Assuming a standard model of grammar where the syntactic component is taken to mediate between a set of other cognitive systems, including other modules of grammar (i.e., the lexicon, as well as the articulatory-perceptual and conceptual-intentional systems), there are two basic types of repairs then along the “temporal” dimension (adopting the terminology of den Dikken, this volume):⁶ 1.

2.

“Pre-emptive strikes”: An operation (which does not have any moduleinternal motivation/trigger) is carried out during the syntactic computation to warrant interpretability at the interfaces to the post-syntactic cognitive systems. “Post-hoc patches”: In cases where the syntactic computation generates an output that does not meet the interpretative needs of the post-syntactic systems, post-syntactic operations (at the interfaces or beyond) apply to ensure interpretability (or make available certain interpretations/phonological realizations that differ from the expected meaning/spell-out linked to a certain syntactic structure).

Taken together with the traditional separation of at least a syntactic (broadly: expression-related) and a semantic (broadly: interpretation-related) level, the picture that emerges is at least as complex as in (13); there are different types of patch-ups dependent on whether the repair operation is triggered by requirements of the articulatory-perceptual system (A-P; in more recent minimalist work (e.g. Chomsky 2008), this is often referred to as the S(ensori)-M(otor) system) or conceptual-intentional (C-I) system:⁷ (13) a. Pre-emptive strikes/A-P b. Post-hoc patches/A-P c. Pre-emptive strikes/C-I d. Post-hoc patches/C-I In this typology, the above-mentioned “standard” cases of “repair by ellipsis” fall under (13b) (Post-hoc patches/A-P), since they typically involve the deletion (at PF) of unchecked features the presence of which would lead the derivation

12   

   Patrick Brandt and Eric Fuß

to crash at the interface to the articulatory-perceptual system (but see below for a discussion of recent work where (residual) EPP effects are analyzed in terms of symmetry-breaking movement triggered by requirements of the C-I interface). Looking at the literature, we find the following examples instantiating the four possible combinations: Pre-emptive strikes/A-P: In early minimalist work (e.g., Chomsky 1993, 1995), syntactic movement is conceived of as a last-resort mechanism that serves to remove (formal morphosyntactic) features which cannot be interpreted at the interfaces to the syntax-external systems and thus would lead to a crash of the derivation. This concept of movement is actually quite close to the notion of preemptive strikes introduced above (note, however, that in contrast to the perspective taken here, the apparent “repair” character of movement was not deemed an asset; rather, movement was characterized as an “imperfection”, cf. Chomsky 2000, 2002 for discussion). Other precursors to the notion of A-P-driven preemptive strikes include approaches where (overt) syntactic movement is taken to satisfy requirements imposed at the PF interface, e.g., analyses that treat the EPP as a PF requirement (cf. e.g. Holmberg 2000), analyses of head-movement as triggered by the need satisfy some version of Lasnik’s Filter/The Stray Affix Filter (cf. e.g. Harley 2004), or symmetry-breaking movement that is triggered by the need to make linearization (based on asymmetric C-command, Kayne 1994) possible (cf. Moro 2000; see also below for more detailed discussion). Post-hoc patches/A-P: The most prominent proposals instantiating this type of repair involve analyses where the output of the syntactic computation is modified by post-syntactic operations to ensure interpretability at A-P, in particular approaches to verbal inflection in terms of affix-hopping (i.e., post-syntactic lowering of inflectional heads to V; cf. Chomsky 1957, Halle and Marantz 1993, Bobalijk 1995, Lasnik 2000; see also Embick and Noyer 2001), analyses of clitic placement in terms of PF reordering/Prosodic Inversion (Bonet 1991, Halpern 1992), and post-syntactic theories of linearization, more generally (cf. e.g. Epstein et al. 1998 and Richards 2004, 2007 on the linearization of abstract syntactic structures based on the notion of symmetric c-command, which requires that a subset of the c-command relations established in the syntax must be ignored at PF). Of interest in this connection are also cases of “repair-by-syncretism”, where a conflict between morphosyntactic features can be remedied by the use/insertion of a form that is underspecified for the relevant distinctions (cf. Albright & Fuß 2012). A well-known example comes from free relatives in German (Riemsdijk 2006, Fuß, Grewendorf and Groat 2012). Similar to English, the wh-pronoun introducing a free relative clause must meet selectional requirements of both the matrix clause and the free relative itself. As a result, mismatches between the case carried by the wh-pronoun and the case required/assigned by the matrix

Introduction   

   13

predicate typically lead to ungrammaticality (for exceptions cf. Vogel 2001 and Fuß, Grewendorf and Groat 2012), cf. (14b). However, (15) shows that mismatches in abstract case features can be resolved by case syncretism (the nominative and accusative forms of the neuter wh-pronoun was are identical in German): (14) a.

Ich stelle ein, wen mir der Personalchef empfiehlt. I hire who.acc me the staff manager recommends ‘I am going to hire who(ever) the staff manager recommends to me.’ (matrix: acc/FR: acc)

b. * Ich stelle ein, wer den Personalchef überzeugt. I hire who.nom the.acc staff manager convinced ‘I am going to hire who(ever) convinces the staff manager.’ (matrix: acc/FR: nom) (15) Ich nehme, was dich überzeugt hat. I take what.nom you.acc convinced has ‘I will take what has convinced you.’ (matrix: acc/FR: nom) Pre-emptive strikes/C-I: In current minimalist theorizing, syntactic movement is often taken to be motivated by the need to keep an element in the derivation until it reaches a position where it can receive an interpretation. This is exactly the job of generalized EPP features (now sometimes called “edge features”), which are used to implement movement to the left edge of derivational cycles (“phases”, vP and CP) to keep material accessible to operations in the next higher phase. Maybe because this is not very satisfactory, the question was raised whether no deeper explanation could be found. Most recently, the minimalist quest for extrasyntactic causes of syntactic movement has led to a line of thinking where movement is taken to repair “instable” syntactic structures (cf. Moro 2008; Chomsky 2007, 2008, 2012; Ott 2011), which result from merging phrase-structurally identical elements: (16) a. Merge (X, Y) → {X, Y} b. Merge (XP, YP) → {XP, YP} (16a) represents the first step of each derivation (i.e., merging two lexical items), while (16b) illustrates an operation combining two internally complex elements, e.g., merging a (subject) DP with a projection of v to create a specifier of vP. By hypothesis, “symmetric” structures of the type in (16) create interpretability prob-

14   

   Patrick Brandt and Eric Fuß

lems at the C-I interface, since their syntactic category (the “label”) cannot be determined unequivocally by the relevant (syntactic) algorithm. No such problems arise with asymmetric structures of the type {X, YP} where one element is a lexical item, which is automatically selected as the projecting category. Importantly, it is assumed that the kind of asymmetry required by the labeling algorithm can be restored by syntactic movement: (17) {XP (…) {H {XP, YP}}} Movement of one of the two phrasal elements in (17) is taken to have a symmetry-breaking effect: Under the assumption that only complete chains are visible to syntactic operations, the tail of a discontinuous syntactic object (i.e., of a non-trivial chain) cannot participate in labeling processes. As a result, the asymmetry between XP and YP is restored, and the relevant syntactic object can be labeled by inspecting the relevant properties of YP. Chomsky (2008, 2012) considers the possibility that the notion of symmetry-breaking movement might provide a deeper explanation for (residual) EPP effects (Merge of DPsubj and vP creates an instable structure, which requires movement of the subject to SpecTP). Crucially for our purposes, the gist of this approach is very close to what we call pre-emptive repairs driven by requirements of the C-I interface in that the relevant movement operations lack a proper syntactic trigger, but are solely motivated by the need to remedy a structure which would lead to translation problems at the interfaces. Looking toward the border between what is traditionally called semantics (truth conditions) and pragmatics (felicity conditions) respectively, we can directly observe what we might call pre-emptive strikes in what falls traditionally under certain figures of speech, like e.g. meiosis, litotes, and more generally in cases of politeness where saying what could be said would constitute what Brown and Levinson (1987) call a face threat to one of the interlocutors. Consider, e.g., (18) and (19). (18) Otto was only very slightly drunk. (19) Otto was not unamused. The statements in (18) and (19) give rise to what Horn (1989) calls R(elevance)implicatures after Grice (1975), which are, essentially, invitations to the hearer to strengthen the literal meaning of the expression in a special fashion (cf. as well below). Strengthening is already what the hearer does in the most general case, as the sign in its literal meaning generally grossly underdetermines the actually interpreted content. The point is that in the case of R-implicature, what is said is

Introduction   

   15

less informative vis-a-vis its elaborateness than what could have been said (and presumably less elaborately so) already from the start so the strenghtening that would normally have taken place is pre-empted or “forced down” by placing the beginning bets lower, so to speak. One may ask whether the eventual effect is not one of more massive strengthening, really, as the reader may test by subtracting “only”, “very”, and “slightly” one after the other and see what happens to the meaning of (18). Interestingly, R-implicatures are considered to be more cultural/conventional as compared to the better understood Q(uantitiy) implicatures (cf. Horn 1989, ch. 3.3.1). Post-hoc patches/C-I: Recall what we said about meanings above: we do not know how they are represented. But we can talk about relations between meaning representations such as entailment very meaningfully and thus give order to at least parts of our meanings (even if we do not know what they are themselves really). If we use as a lingua franca some common logic language, say a version of predicate logic with quantification over sets and in particular indexical elements such as times or possible worlds, then examples of post-hoc patches at the meaning end are legion. We can say that every example of symbolism in the language that expresses truth conditions that is not matched in some fashion by symbolism in the syntactic derivation, i.e., what gets translated into sound (or sign) is an example of a post-hoc patch in this sense. Returning to the example in (3) above, a typical analysis of the modal (or generic) reading of a middle construction these days would consist in saying that it is as a whole associated with a silent modal (or generic) operator that may be written in front of a certain formula, derived, hopefully, in the normally applying way that satisfies compositionality (however thought of, but reflecting Frege’s insight that we are able to derive the meaning of complex expressions from the meanings of their parts and the structure relating them). Using silent (i.e., non-traceable) operators to make truth conditional representations do what they are supposed to do has become very much practice in many a domain, including not transparently coded modality or genericity, as well as distributivity.⁸ When we write something into the semantics that is not visible in the syntax, then this is a post-hoc patch at the C/I interface. Looking again at the border between semantics and pragmatics, one could argue that the part of meaning theory that is described by Grice’s (1975) theory of implicatures is nothing but a post-hoc patch, there to explain how we can manage to arrive at actually usable, i.e., restrictive enough interpretations on the basis and in the face of the logically possible literal meanings that are often too many as well as too indeterminate to be really helpful in conversation. Landman (2000) and Chierchia (2004) argue forcefully that certain (quantity-based) impli-

16   

   Patrick Brandt and Eric Fuß

catures may embed, i.e., can be calculated in tandem with truth-conditions (but cf. Geurts 2010). If they do, then they instantiate cases of grammaticized repair. Up to now, we have focused on repair operations driven by the interaction of different levels/modules, i.e., inter-module repairs. In addition, repairs may also be triggered by requirements of the module where they take place (see also den Dikken, this volume, for discussion). The latter might be called intra-module repairs. A typical example of intra-module repair are dissimilation effects in phonology, morphology, and syntax (cf. Leben 1973, Goldsmith 1979 on phonemebased haplology, Neeleman and van de Koot 2006 on (morpo-)syntactic haplology, Nevins 2012 on the general concept of morpho-phonological dissimilation). To repeat, a typology of repair operations will always depend on one’s model’s architecture; we have concentrated here on the well-established syntaxsemantics divide (broadly: expression vs. interpretation) in a model respecting really only one derivational level; it goes without saying that with additional levels there come additional interfaces and hence additional loci for repairs.

6 OT and Coercion The notion of “repair” is occasionally employed in optimality theoretic (henceforth: OT) frameworks (Prince and Smolensky 1993), and one may wonder in how far “repair” as understood here differs from what is called “coercion” in the literature. Let us try, briefly, to differentiate between these concepts and then turn to specific notions of repair as employed in this volume. In OT, the notion of repair is commonly used to describe a situation where a candidate becomes acceptable although it has properties that are ruled out by the grammar in other contexts (i.e., lead to a fatal violation of a given constraint under normal circumstances). Thus, in OT, repairs typically have a last-resort character: A certain representation/derivation which is normally not tolerated by the grammar is exceptionally ruled in because it serves to avoid a violation of a higher ranked constraint (which is violated by other potential candidates in the competition). Typical examples include the analysis of do-support in English (Speas 1995, Grimshaw 1997), or the so-called IPP (Infinitivus Pro Participio) effect in German and Dutch verb clusters (Schmid 2005; see Hinterhölzl 2009 for an alternative non-OT analysis that also makes use of the notion of repair). As we see it, the fundamental difference between what one might call – we believe: misleadingly – “repair” in OT frameworks is that in OT, there is no absolute notion of and therefore no fundamental opposition between “perfect” and “defect” or “good” and “bad” or “usable” and “unusable”. Instead, in OT, one

Introduction   

   17

always looks for the best – or least bad – candidate among a range of competing candidates none of which is typically “perfect” or “defect” (or “good” or “bad” etc.) in absolute terms. In particular, nothing is actively done in OT to remedy particular problems or improve problematic structures; you go with the problems if there is no better alternative. In contrast, repairs as viewed here are grammatical procedures that are called upon to overcome problems that are due to conditions that have to be met unconditionally. It is perfectly possible, of course, to assume a “mixed” model of grammar that is in part absolute in terms of the conditions that it enforces and that in part involves the notion of competition between alternative candidates. Contributions in this volume that explicitly make reference to competition are Kucerova and Dočekal’s analysis of (im)perfective morphology in Slavic and Romance as well as Salzmann’s analysis of locative inversion. “Coercion” is a term that is more intimately connected with “repair” as understood here; in fact, it constitutes a special case of repair, namely repair towards or at the level of semantic/pragmatic representations. While there are different understandings of the term,⁹ it appears safe to say that coercion involves the idea that the semantics of expressions that do not genuinely fit are made to fit each other. To cite Lauwers and Willems (2011: 1219): “[A]t the basis of coercion, there is a mismatch (cf. Francis and Michaelis 2003) between the semantic properties of a selector (be it a construction, a word class, a temporal or aspectual marker) and the inherent semantic properties of a selected element, the latter being not expected in that particular context.”

While “coercion” in its most typical uses appears to be a genuinely semantic/ pragmatic notion, “repair” is not committal as respects the locus where it applies. The term coercion is narrower in other respects as well, as comes out when we look at two domains that have been central for the development of the concept, namely, changing NP types (“type shifting” as proposed in Partee and Rooth 1983 and Partee 1986) and changing aspectual types (“aspectual shift”, cf. Verkuyl 1972 and much work since). The idea of shifting NP types is originally motivated by coordination facts as in (20): (20) Poldi and every other player ran. In (20), the NP “Poldi” has the basic type of an ordinary individual (type “e(ntity)”), while the NP “every other player” must have a higher type, corresponding to the set of sets that contain all the other players (type (e,t),t).¹⁰ Under independently warranted assumptions (in particular, the idea that conjunction has a unique general meaning, cf. Gazdar 1980), NPs must be of the same seman-

18   

   Patrick Brandt and Eric Fuß

tic type so they can be turned into one semantic object. The solution lies in shifting “Poldi” to the same type as “every other player”, i.e., interpreting an ordinary individual as a set of sets as well (namely, the set of all the sets having Poldi as a member). It should be noted that this is semantically harmless in that it does not matter, truth-conditionally, whether we look at Poldi as an ordinary individual or as the set of properties that he has (= the sets that have him as a member).¹¹ Note as well that it is impossible to go the other way around: there is no way of interpreting “every other player” as an ordinary individual. The NP shifting type of coercion then is restricted by being semantically harmless, for one thing. Second, the target of coercion corresponds to a type that is independently present in the language (e.g., that of a generalized quantifier). Third, and as pointed out in Partee (1986), the operations yielding the required coerced types as well may be considered “natural” in corresponding to operations independently available and expressible in the language where shifts apply. As we pointed out above, repairs need not be semantically harmless at all but in the case of “creative repairs” or “amends” à la Reinhart indeed may furnish interpretations that would not be otherwise available at all. It follows that repairs need not be “natural” e.g. in the sense of NP type shifting operations: the output of a repair need not correspond to anything that is independently available in the language, i.e., could be coded by independent alternative means. Let us look next at examples of bona fide “aspectual coercion” as a bridge to specific notions of repair as presented in this volume. It appears that adverbs that test, supposedly, for the distinction between states and changes of states (events) do coerce the eventuality they apply to into the aspectual type that they are supposed to test for, if the “original” type is the wrong one; consider (21). (21) Poldi ran in ten minutes. Without the change of state indicating frame adverbial “in ten minutes”, the proposition in (21) expresses a state (cf. for relevant tests Vendler 1957, Verkuyl 1972 and Dowty 1979). Put together with the frame adverbial, however, we interpret it as a change of state. Let us assume for concreteness that we model this by adding a presupposition that (ten minutes) before a time where Poldi ran, there was a time where he did not run. There is a change of state interpretation now as is called for by the frame adverbial. Coercing the other way around, i.e., coercing from a change of state (event) into a state appears possible as well, if less easy.¹² (22) (#) Poldi ran away for ten minutes.

Introduction   

   19

In (22), the proposition without the durative adverbial expresses a change of state (event). Put together with the durative adverbial, we force it into a state (that appears to consist in iterated events). Keeping aside the exact nature of the meaning of (22), let us assume that in analogy to the coercion from state to change of state as just sketched that going from a change of state to a simple state involves getting rid of the presupposition that codes the “pre state” of the change of state. Recognizing that adding presuppositions falls under the general rubric of meaning strenghtening, i.e., moving to more specific or restrictive meanings as ubiquitously observed and discussed in the realm of the semantics/pragmatics divide and the theory of implicatures more specifically, erasing a presupposition falls under the opposite rubric, that of meaning weakening or moving from more specific or restrictive meanings to less specific or restrictive ones.¹³ Recalling the discussion of more or less natural coercion above, we may note that although neither way of coercion is semantically harmless, the effects of the respective coercions appear to be akin at least to processes that are independently established in pragmatics, and may perhaps make their way into the grammar from there; we should note that the question as to what extent different types of coercion are part of pragmatics or semantics proper is very much an open one still.

7 Added values of being wrong In what follows, we give a brief overview of the papers collected in this volume, highlighting the way the notion of repair is put to use to provide a fresh and better understanding of phenomena at the interfaces between different modules of grammar and/or grammar-external cognitive systems. Concentrating on a set of poorly understood phenomena in aspectual interpretation, Kucerova and Dočekal (“Semantic competition over morphological representations. A case study from Slavic”) address unexpected asymmetries in the interpretation of Romance versus Slavic aspectual morphology. They propose that imperfective and perfective morphology in Spanish and Czech respectively carry the same denotations but are associated with different presuppositions. They argue that Czech perfective morphology carries an “activity presupposition” that necessitates use of the perfective form in change of state contexts. Absence of this presupposition as in the case of Czech imperfective morphology may be in conflict with requirements imposed by, e.g., habituality operators in the structure and may then lead to a repair which consists in shifting the interpretation to the domain of worlds, leading thus to a modal ability reading. The repair here thus provides a shortcut to an interpretation that could have been coded more trans-

20   

   Patrick Brandt and Eric Fuß

parently, but arguably only less economically so. An important aspect of Kucerova and Dočekal’s paper is that different aspects of meaning are represented at distinguished levels (assertive vs. presupposed meanings), and the repair in their approach consists in “pushing” offending meaning aspects from one domain (times) to another (worlds). The fundamental role that the commitment to particular levels plays in explaining hitherto poorly understood linguistic phenomena becomes maybe even more obvious in Schmitz and Fisseni’s paper (“Repairs for reasoning”). Schmitz and Fisseni concentrate on cases of “modal enrichment” as exemplified in (23) (23) It is five past three, but my watch is five minutes ahead. Interpreted literally, (23) conveys that it is factually five past three, and that my watch is ahead in such a way that it displays ten past three. However, and as the authors show experimentally, hearers do not interpret (23) literally but rather as conveying that it is in fact three o clock. Schmitz and Fisseni argue that the first sentence is enriched by adding a modal operator that for (23) says something like “According to my watch, it is five past three”. Asking for the reasons behind (or triggers for) modal enrichment, Schmitz and Fisseni discuss alternative approaches: According to a repair line of reasoning, modal enrichment takes place ‘post hoc’ in order to meet Gricean maxims (“relevance”) or in order to conform to a given script of deriving and conveying information (“reasoning”). According to a third approach, however, (23) does not involve any repair at all. Instead, this approach has it that the first conjunction is essentially presented as “quoted”; the second conjunct ensuingly breaks the quotation context, such that the information it contains is attributed to the speaker. According to this approach, there is no repair but instead there is a slot for a modal operator from the start that is provided by rhetorical structure: “the speaker does often and not solely relate his own positions but ‘puts on stage’ different positions that he attributes to different sources which are called ‘discourse entities’”. While Schmitz and Fisseni’s results speak in favor of the reasoning approach according to which a repair does indeed take place, they do not eventually exclude that there may be no repair at all as on the rhetorical structure approach. On the way toward deciding the issue, the authors point out that there is a great demand in establishing different types of modal enrichment experimentally and modeling them formally, in particular as concerns their interaction. Even if the question of which levels exactly should be assumed can by no means be said to be decided, the question of interactions between repairs at different levels certainly has a place on the research agenda already, as shown in

Introduction   

   21

Härtl’s contribution (“Generic rescue: Argument alternations and the monotonicity condition”). Härtl argues that complement drop (a serious foul concerning the expression of argument structure) can be repaired by genericity, a phenomenon that is typically modeled by assuming a silent operator that is introduced into the derivation at a later stage (cf. above section 5). Looking from the bottom up, it looks as if the grammar constructs something that is actually unusable but still keeps it in the derivation, expecting that the structure can and will be amended further up. It is not, however, that anything goes – in particular, Härtl proposes a “monotonicity condition” stating that removing truth-conditionally relevant material from lexical representations is disallowed, unconditionally. Only if monotonicity is observed may improprieties pertaining to the expression of argument structure be repaired in certain contexts, prominently those that facilitate a generic interpretation of propositional contents. Recall that within derivational generative theories, different parts of the clausal spine code different types of information, where the lower parts of the tree (VP/vP) are responsible for the coding of ordinary argument structure (i.e., relations between “ordinary individuals”), whereas the higher parts (TP and CP) are responsible for the coding of apparently more sophisticated information like tense and mood, as well as for establishing the link to the discourse (cf. above section 1). Taking Härtl’s cases, we may wonder whether certain improprieties that are not and maybe cannot be repaired locally can be compensated for “later”, i.e., higher up in the tree, by an independently existing operation that may itself have repair-nature (cf. section 5 above). More succinctly, “playing foul below” in the realm of argument expression (complement drop) might necessitate “playing foul above” by inserting a certain type of operator (a generic operator in Härtl’s cases). Note that such questions are reminiscient of older prominent issues; in the case at hand, we feel reminded of Ross’ (1973) “penthouse principle”, stating that “the rules are different if you live in the penthouse”, read “more is allowed at the top”. In analogy, it appears to be true that it is easier to tamper in the more discourse-affine regions of the tree than it is to tamper in the more embedded regions. At the current state of research, however, we can do little more than note that it is time to systematically explore questions of repair interaction.¹⁴ According to Marcel den Dikken (“Prepare and repair: On pre-emptive strikes and post-hoc patches”), Lasnik’s (1999) seminal cases of “repair by ellipsis” do not instantiate genuine cases of repair. Recall that under Lasnik’s analysis, PF deletion of the verb that has stayed in situ eliminates uninterpretable features that would otherwise have led to a crash of the derivation. In contrast, den Dikken proposes an account based on the assumption that functional heads marked for ellipsis cannot trigger movement operations. As a result, the verb does not raise and undergoes PF-deletion together with the VP. In this way, marking for deletion

22   

   Patrick Brandt and Eric Fuß

preempts the necessity of syntactic movement to begin with. Den Dikken then discusses cases of ellipsis in predicate inversion constructions. Here, movement of the subject of an inverted predicate to the left of the predicate would undo the information-structural effect of predicate inversion unless the raised predicate disappears from the PF string. Den Dikken concludes that there are clear cases of level-internal repairs (“PF repair by PF ellipsis”), while it is questionable whether PF operations can solve problems of the narrow-syntactic computation. Still, it seems that the offending PF form is the result of a particular syntactic derivation. A different outlook on related inversion constructions is presented in Martin Salzmann’s contribution (“Repair-driven verb movement in English locative inversion”). Taking an Optimality Theoretic perspective on repairs, Salzmann provides a new approach to locative inversion in English based on the interaction of ranked, violable constraints. Building on the idea that non-feature driven syntactic movement is possible if this is the only way to satisfy a high ranked syntactic constraint (Heck and Müller 2000, 2007), Salzmann argues that in English locative inversion, V-to-T movement across the subject (which stays in situ) is exceptionally licensed since it serves to satisfy a higher-ranked informationstructural constraint which requires a focus marked constituent to occupy the rightmost position in its clause. Interestingly, it appears that the notion of repairs in Optimality Theory requires us to think of repairs primarily in terms of preemptive strikes, in the sense that the syntactic component generates a structure with special properties (which otherwise would lead to a fatal violation) which is “later” ruled in since it satisfies a higher-ranked constraint in the OT component of grammar. The contribution by Joost Kremers (“Linearisation as repair”) deals with interface problems relating to the linearisation of syntactic structures. It is usually taken for granted that the mapping of syntactic structures to PF is governed by conditions ensuring a total, non-overlapping linear ordering of syntactic terminal nodes. Kremers claims, however, that (total) linearity is not a universal, but rather a modality-specific property of language, as shown by synchronicity effects in sign languages, where different elements (e.g., a predicate and its modifying adverb) can be expressed simultaneously. According to Kremers, simultaneous articulation is not an exception, but rather the minimal strategy to express (terminal) nodes that are syntactic sisters: “Only if simultaneous realization is not possible, linearization is applied as a repair strategy.” This entails that syntax underdetermines linear order (Epstein et al. 1998, Richards 2004), which calls for repair in cases where “3rd factors in language design” (Chomsky 2005) – namely the serial character of the A-P interface in spoken languages – render simultaneous articulation impossible.

Introduction   

   23

Volker Struckmeier (“Repairing resumptive structures, or: How faulty is the Lexicon?”) also discusses problems pertaining to the realization of abstract syntactic structures. In contrast to Kremers, however, the repair strategy he proposes is not triggered by properties of the A-P interface, but by gaps in the inventory of lexical items that languages have at their disposal. Struckmeier discusses special cases of resumptive relative clauses in German, where a 3rd person relative pronoun combines with a 1st or 2nd person resumptive prominal subject. He then proposes that this apparent contradiction results from a repair strategy which compensates the lack of an appropriate lexical element (i.e. a 1st/2nd person relative pronoun). Assuming a model of grammar where the syntactic computation operates on abstract (syntactico-semantic) feature bundles, which are then realized by late, post-syntactic insertion of lexical material (Halle & Marantz 1993), Struckmeier suggests that in the problematic cases, a special insertion procedure applies (“Repair Vocabulary Insertion”), which leads to articulation of a single feature bundle at different positions in the tree: The relative operator is spelledout at the head of the movement chain (in SpecCP), while the phi-features are realized by inserting a 1st/2nd person subject pronoun at the tail. From this point of view, we might say that repair operations facilitate the expression of an interpretation that would otherwise – due to limitations of the lexicon – not be available. Another case for repair-driven resumption is made in the contribution by Josef Bayer and Martin Salzmann (“That-trace effects and resumption as the result of improper movement”). Bayer and Salzmann provide evidence that thattrace-effects in English and German and resumption in cases of long distance relativization in Swiss German are actually reflexes of one and the same underlying restriction. More precisely, they argue that long-distance A’-movement is available only for potentially contrastive elements, which is attributed to the circumstance that intermediate complementizers bear a feature uContrast and thus cannot be crossed by sentence topics, since this would instantiate some form of Improper Movement. The traditional observation that the that-trace effect is confined to subjects is attributed to the fact that subjects typically are default topics. It is then shown that different languages make use of different (repair) strategies to facilitate extraction of subjects such as complementizer deletion and the Adverb Effect in English, or scrambling in German. In the second part of the paper, Bayer and Salzmann discuss resumption patterns in relative constructions of Alemannic. Focusing on obligatory resumption in cases of long relativization, they argue that the insertion of an additional pronoun can be used to repair cases of Improper Movement where a relative operator would otherwise have to undergo illicit (long) extraction over dass ‘that’. Similar to den Dikken’s contribution, the problem causing the repair is basically information-structural in nature, in that long A’-movement of topics would lead to a chain linked to conflicting

24   

   Patrick Brandt and Eric Fuß

information-structural categories. Importantly, the nature of the repair operation is very much in the spirit of Reinhart’s work: Resumption makes available an interpretation (long extraction of a topic) which cannot be coded by the syntax without further ado due to independent (locality) restrictions. A related perspective on repairs as a means to facilitate the expression of meanings which cannot be coded otherwise is presented in the contribution by Florian Schäfer (“Passives of reflexive verbs: The repair of a Principle A violation”). Dealing with issues raised by passivized reflexive verbs in Icelandic and German, Schäfer argues that the problems this construction creates for case and binding theory can be overcome by a language-specific repair operation of Default Agreement. More precisely, it is shown that under an Agree-based approach to case and binding, Default Agreement not only provides a formal antecedent for the anaphor, but also establishes the syntactic preconditions for the phonological realization of accusative case on the reflexive pronoun. From that point of view, Default Agreement can be characterized as a syntactic repair operation (a pre-emptive strike available in some languages) which satisfies needs of both C-I and A-P: At C-I, the Agree relation between the anaphor and the T-head affected by the operation of Default Agreement is interpreted as a binding relation; at A-P, Default Agreement on T leads to the realization of dependent case, i.e., accusative on the reflexive pronoun. We hope that the original work presented in this collection will help foster the view that the regular handling of translation problems within the grammatical system and at its interfaces is constitutive for the expressivity and economy of language. As is sometimes the case in life, things repaired may turn out better than they were before.

Notes 1 Cf. for lucid exposition the introduction in Francis and Michaelis (2003). 2 Cf. Brandt (2009) for a proposal to derive the modal and generic interpretation of middles in terms of repair. 3 For a wider range of applications of “rescue by PF deletion” (including (lack of) that-trace effects, the lack of intervention effects with null arguments and traces, and the observation that barriers lose their island status if headed by a trace), cf. Bošković (2011). 4 Traditional cases of repair at the LF-interface (to satisfy principles like Full Interpretation) include e.g. deletion of (intermediate) chain links to create chains that are legitimate LFobjects, cf. Lasnik and Saito (1984), Chomsky (1991). 5 The scare quotes around “temporal” are there because this is a metaphorical way of speaking; the idea is simply that certain operations are ordered with respect to each other or depend on each other.

Introduction   

   25

6 N.B.: 1. is actually very close to the general characterization of movement in recent versions of the MP, see below. 7 Note that there is a further logical possibility, namely repairs that are triggered by requirements imposed by the lexicon, in the sense that the application of a special repair operation is triggered by lexical quirks, such as the absence of appropriate lexical items to spell out a certain structure. See Struckmeier, this volume, for some discussion. 8 Early and groundlaying work includes Lawler (1972), Lewis (1975), Carlson (1977), Kratzer (1978) and Massey (1976). 9 We should note that in the last few years, the notion of coercion has been extended to many a domain, “at the risk of becoming fuzzy and overworked” (Lauwers and Willems 2011: 1231). 10 Cf. Barwise and Cooper (1981) on the theory of generalized quantification and Partee (1986) for the motivation of different types of NP denotations. 11 In Montague’s (1973) work on the Proper Treatment of Quantification, the strategy consists in “generalizing to the worst case”, so all NPs are given here the type of sets of sets. 12 Cf. Brennan and Pylkkänen (2008) and references therein for attempts at differentiating types of coercion according to the processing cost they give rise to. 13 This is also possible, but taken to be conventional rather than calculable using Grice’s conversational maxims. Irony and metaphor are possible cases in point (cf. Grice 1975), as well as indirect speech and politeness strategies more generally (cf. Horn 1989 : 194–197). 14  Cf. again Brandt’s (2009) analysis of the modal and generic interpretation of middle constructions, according to which a blindly applying repair consisting in modalization leads to a new problem that needs to be repaired in turn by generalization.

8 References Albright, Adam and Eric Fuß. 2012. Syncretism. In: Jochen Trommer (ed.), The Morphology and Phonology of Exponence, 236–287. Oxford: Oxford University Press. Barwise, Jon and Robin Cooper. 1981. Generalized quantifiers and natural language. Linguistics and Philosophy 4: 159–219. Bobaljik, Jonathan. 1995. Morphosyntax: the syntax of verbal inflection. Doctoral dissertation, MIT. Boeckx, Cedric. 2006. Linguistic Minimalism: Origins, Concepts, Methods, and Aims. Oxford: Oxford University Press. Bonet, Eulalia. 1991. Morphology after syntax. Pronominal clitics in Romance. Doctoral dissertation, MIT. Bošković, Željko. 2011. Rescue by PF deletion, traces as (non)interveners, and the that-trace effect. Linguistic Inquiry 42.1: 1–44. Brandt, Patrick. 2009. Generische Möglichkeit in Medialkonstruktionen. In: Abraham, Werner and Elisabeth Leiss (eds.), Modalität. Epistemik und Evidentialität bei Modalverb, Adverb, Modalpartikel und Modus, 79–100. Tübingen: Stauffenburg. Brennan, Jonathan and Liina Pylkkänen. 2008. Processing events: behavioral and neuromagnetic correlates of aspectual coercion. Brain and Language 106: 132–143. Bresnan, Joan (ed.). 1982. The Mental Representation of Grammatical Relations. Cambridge: MIT Press.

26   

   Patrick Brandt and Eric Fuß

Brown, Penelope and Stephen Levinson. 1987. Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Carlson, Gregory. 1978. Reference to kinds in English. PhD dissertation University of Massachusetts, Amherst. Carnap, Rudolf. 1931. Die physikalische sprache als universalsprache der wissenschaft. In: Erkenntnis II: 431–465. Chierchia, Gennaro. 2004. Scalar implicatures, polarity phenomena, and the syntax/pragmatics interface. In: Adriana Belletti (ed.), Structures and Beyond, 39–103. Oxford: Oxford University Press. Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Chomsky, Noam. 1991. Some notes on economy of derivations and representations. In: Robert Freidin (ed.), Principles and Parameters in Comparative Grammar, 417–454. Cambridge, MA.: MIT Press. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In: Kenneth Hale and Samuel J. Keyser (eds.), The View from Building 20, 1–52. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The Minimalist Program, Cambridge, MA: MIT Press. Chomsky, Noam. 2000. Minimalist inquires: the framework. In: Roger Martin, David Michaels, and Juan Uriagereka (eds.), Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik, 89–156. Cambridge, MA: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In: Michael Kenstowicz (ed.), Ken Hale. A Life in Language, 1–52. Cambridge Mass.: MIT Press. Chomsky, Noam. 2002. An interview on minimalism. In: Adriana Belletti and Luigi Rizzi (eds.), Noam Chomsky: On Nature and Language, 92–161. Cambridge: Cambridge University Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In: Adriana Belletti (ed.), Structures and Beyond. The Cartography of Syntactic Structures, Vol. 3, 104–131. Oxford: Oxford University Press. Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36.1: 1–22. Chomsky, Noam. 2007. Approaching UG from below. In: Uli Sauerland and Hans Martin Gärtner (eds.), Interfaces + Recursion = Language?, 1–29. New York: Mouton de Gruyter. Chomsky, Noam. 2008. On phases. In: Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizarreta (eds.), Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud, 133–166. Cambridge, MA: The MIT Press. Chomsky, Noam. 2012. Problems of projection. To appear in Lingua. Cinque, Guglielmo. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24: 239–297. Craenenbroeck, Jeroen van and Marcel den Dikken. 2006. Ellipsis and EPP repair. Linguistic Inquiry 37: 653–64. Dowty, David. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel. Embick, David and Rolf Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32: 555–595. Epstein, Samuel D., Erich M. Groat, Ruriko Kawashima, and Hisatsugu Kitahara. 1998. A Derivational Approach to Syntactic Relations. Oxford: Oxford University Press. Fodor, Jerry A. 1983. The Modularity of Mind. Cambridge, Mass.: The MIT Press.

Introduction   

   27

Fox, Danny and Howard Lasnik. 2003. Successive cyclic movement and island repair: The difference between Sluicing and VP Ellipsis. Linguistic Inquiry 34: 143–154. Fox, Danny. 2002. Antecedent Contained Deletion and the copy theory of movement. Linguistic Inquiry 34: 63–96. Frampton, John and Sam Gutmann. 2002. Crash-proof syntax. In: Samuel D. Epstein and T. Daniel Seely (eds.), Derivation and Explanation in the Minimalist Program, 90–105. Malden, MA: Blackwell Publishing. Francis, Elaine and Laura Michaels. 2003. (ed.) Mismatch: Form-Function Incongruity and the Architecture of Grammar. Stanford: CSLI. Fuß, Eric, Günther Grewendorf and Erich Groat. 2012. Free relatives. Ms., University of Frankfurt. Gazdar, Gerald. 1980. A cross-categorial semantics for coordination. Linguistics and Philosophy 3: 407–409. Geurts, Bart. 2010. Quantity Implicatures. Cambridge: Cambridge University Press. Goldsmith, John A. 1979. Autosegmental Phonology. New York: Garland. Grimshaw, Jane. 1997. Projection, heads, and optimality. Linguistic Inquiry 28: 373–422. Halle, Morris and Alec Marantz. 1993. Distributed Morphology and the pieces of inflection. In: Samuel J. Keyser and Kenneth Hale (eds.), The View from Building 20, 111–176. Cambridge: The MIT Press. Halpern, Aaron. 1992. Topics in the placement and morphology of clitics. Doctoral dissertation, Stanford University. Harley, Heidi. 2004. Head movement and conflation. In: Keir Moulton and Matthew Wolf (eds.), Proceedings of NELS 34, 239–254. Amherst, Mass.: GLSA. Heck, Fabian and Gereon Müller. 2000. Repair-driven movement and the local optimization of derivations. Ms, University of Leipzig. Heck, Fabian and Gereon Müller. 2007. Derivational optimization of wh-movement. Linguistic Analysis, 33(1–2), 97–148. Hinterhölzl, Roland. 2009. The IPP-Effect, phrasal affixes and repair strategies in the syntaxmorphology interface. Linguistische Berichte 218: 191–215. Holmberg, Anders. 2000. Scandinavian stylistic fronting: how any category can become an expletive. Linguistic Inquiry 31: 445–483. Horn, Laurence R. 1989. A Natural History of Negation. Stanford: CSLI Publications. Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge: MIT Press. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, Mass.: The MIT Press. Kracht, Marcus. 2007. The emergence of syntactic structure. Linguistics and Philosophy 30: 47–95. Kratzer, Angelika. 1978. Semantik der Rede. Königstein: Scriptor. Krifka, Manfred. 1998. Scope Inversion under the rise-fall contour in German. Linguistic Inquiry 29: 75–112. Kühnhold, Ingeburg. 1973. Präfixverben In: Hugo Moser (ed.), Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Erster Hauptteil: Das Verb, 141–362. Düsseldorf: Schwann. Landman, Fred. 2000. Events and Plurality. Dordrecht: Kluwer. Lasnik, Howard and Myung-Kwan Park. 2003. The EPP and the subject condition under sluicing. Linguistic Inquiry 34: 649–660. Lasnik, Howard. 1999. On feature strength. Three minimalist approaches to overt movement. Linguistic Inquiry 30.2: 197–217.

28   

   Patrick Brandt and Eric Fuß

Lasnik, Howard. 2000. Syntactic Structures Revisited. Contemporary Lectures on Classic Transformational Theory. Cambridge, Mass.: MIT Press. Lasnik, Howard and Mamoru Saito. 1984. On the nature of proper government. Linguistic Inquiry 15: 235–289. Lauwers, Peter and Dominique Willems. 2011. Coercion: Definition and challenges, current approaches, and new trends. In: Linguistics 49: 1219–1235. Lawler, John. 1973. Studies in English Generics. Ph.D. dissertation University of Michigan. Leben, William. 1973. Suprasegmental Phonology. Ph.D. dissertation, MIT. Levin, Beth. 1993. English Verb Classes and Alternations. Chicago: University of Chicago Press. Lewis, David. 1975. Adverbs of quantification. In: Edward L. Keenan (ed.), Formal Semantics of Natural Language, 3–15. Cambridge: Cambridge University Press. Massey, Gerald. 1976. Tom, Dick, and Harry, and all the king’s men. American Philosophical Quarterly 13: 89–107 Merchant, Jason. 2001. The Syntax of Silence. Oxford: Oxford University Press. Montague, Richard. 1973. The proper treatment of quantification in ordinary English. In: Jaakko Hintikka, Julius Moravcsik and Patrick Suppes (eds.), Approaches to Natural Language, 221–242. Dordrecht: Reidel. Moro, Andrea. 2000. Dynamic Antisymmetry. Cambridge, MA: MIT Press. Moro, Andrea. 2008. Rethinking symmetry: a note on labelling and the EPP. Ms., Università Vita-Salute San Raffaele/ Harvard University. Müller, Gereon. 2000. Elemente der optimalitätstheoretischen Syntax. Tübingen: Stauffenburg. Neeleman, Ad and H. van de Koot. 2006. Syntactic haplology. In: Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax, vol IV, 684–710. Oxford: Blackwell. Neeleman, Ad and Tanya Reinhart. 1998. Scrambling and the PF Interface. In: Miriam Butt and Wilhelm Geuder (eds.), The Projection of Arguments: Lexical and Compositional Factors, 309–353. Stanford: CSLI. Nevins, Andrew. 2012. Haplological dissimilation at district stages of exponence. In: Jochen Trommer (ed.), The Morphology and Phonology of Exponence, 84–116. Oxford: Oxford University Press. Ott, Dennis. 2011. Local instability. The syntax of split topics. Ph.D. dissertation, Harvard University. Partee, Barbara. 1986. Noun phrase interpretation and type-shifting principles. In: Jeroen Groenendijk, Dick de Jongh, and Martin Stokhof (eds.), Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, 115–143. Dordrecht: Foris. Partee, Barbara and Mats Rooth. 1983. Generalized conjunction and type ambiguity. In: Rainer Bäuerle, Christoph Schwarze and Arnim von Stechow (eds.), Meaning, Use, and Interpretation of Language, 361–383. Berlin: de Gruyter. Pollard, Carl and Ivan Sag. 1987. Information-based Syntax and Semantics. Volume 1: Fundamentals. Stanford: CSLI. Prince, Alan and Paul Smolensky. 1993. Optimality Theory: Constraint interaction in Generative Grammar. Rutgers University for Cognitive Science: RuCCS Technical Report No.2. Quine, Willard van Orman. 1960. Word and Object. Cambridge: MIT Press. Reinhart, Tanya. 2006. Interface Strategies. Cambridge: MIT Press. Richards, Marc. 2004. Object shift and scrambling in North and West Germanic: A case study in symmetrical syntax. PhD thesis, University of Cambridge.

Introduction   

   29

Richards, Marc. 2007. Dynamic linearization and the shape of phases. Linguistic Analysis 33.3–4: 209–237. Riemsdijk, Henk van. 2006. Free relatives. In: Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax, Vol. II, 338–382. Oxford: Blackwell. Ross, John Robert. 1969. Guess who? In: Robert I. Binnick, Alice Davison, Georgia M. Green, and Jerry L. Morgan (eds.), Papers from the Fifth Regional Meeting of the Chicago Linguistic Society, 252–286. Chicago: University of Chicago, Chicago Linguistic Society. Ross, John Robert. 1973. The penthouse principle and the order of constituents. In: Claudia Corum, Cedric Smith-Stark and Ann Weiser (eds.), You Take the High Node and I’ll Take the Low Node, 397–422. Chicago: CLS. Schmid, Tanja. 2005. Infinitival Syntax. Infinitivus Pro Participio as a Repair Strategy. Amsterdam: John Benjamins. Speas, Margaret. 1995. Generalized control and null objects in Optimality Theory. In: Jill Beckman, Laura Walsh-Dickie, and Suzanne Urbanczyk (eds.), Papers in Optimality Theory, 637–653. Amherst, Massachussetts: UMass Occasional Papers in Linguistics 18. Strawson, Peter. 1959. Individuals: An Essay in Descriptive Metaphysics. London: Methuen. Vendler, Zeno. 1957. Verbs and times. The Philosophical Review 66: 143–160. Verkuyl, Henk. 1972. On the Compositional Nature of the Aspects. Dordrecht: Reidel. Vogel, Ralf. 2001. Case conflict in German free relative constructions: An Optimality Theoretic treatment. In: Gereon Müller, Wolfgang Sternefeld (eds.), Competition in Syntax, 341–375. Berlin: Mouton de Gruyter. Williams, Edwin. 2003. Representation Theory. Cambridge: MIT Press.

Mojmír Dočekal and Ivona Kučerová

Semantic competition over morphological representations. A case study from Slavic Abstract: This paper investigates some cross-linguistic differences in aspectual interpretations of Slavic and Romance languages. We propose that certain readings cannot be determined compositionally from the syntactic representation but instead they result from semantic competition over available morphological representations. We argue that there are certain last resort interpretations, namely, type-shifting and shifts to possible worlds, that may be used to rescue a derivation that would not otherwise be interpretable. In turn, the study supports the view of the grammar as featuring a component that is driven by competition, at least on the LF branch of the derivation.

1 Introduction This paper¹ investigates some cross-linguistic differences in aspectual interpretations of Slavic and Romance languages. We propose that certain readings cannot be determined compositionally from the syntactic representation but instead they result from semantic competition over available morphological representations. We argue that there are certain last resort interpretations, namely, type-shifting and shifts to possible worlds, that may be used to rescue a derivation that would not otherwise be interpretable. Crucially, such a rescue strategy is limited only to some structural environments. Consider the sentence in (1) from Czech. As we can see, an utterance with an imperfective verb modified by an in-adverbial lacks an episodic reading.² (1) # Petr četl Vojnu a mír za dvě hodiny. Petr.nom read.impf War and Peace in two hours ‘Petr was reading War and Peace in two hours.’ The structure is not ungrammatical though. In fact it has an – often unnoticed – bounded ability reading, highlighted by the context in (2). The relevant reading may be paraphrased as ‘There was a time in the past in which Petr was able to finish reading War and Peace in two hours and in fact he did it at least once’. Crucially, for this reading to be judged true it is sufficient if there was only one

32   

   Mojmír Dočekal and Ivona Kučerová

verifying instance of the event in the actual world. It suffices if other instances of the event were possible; they don’t need to have been instantiated. (2) Když Petr studoval rychločtení, tak četl Vojnu a mír when Petr studied speed-reading then read.impf War and Peace za dvě hodiny. in two hours ‘When Petr took a course in speed-reading, he was reading War and Peace in two hours./… he was able to read War and Peace in two hours.’ Interestingly, a parallel construction in Romance, here exemplified by Spanish (3), may have a non-episodic reading as well. The truth-conditions are stronger, though, in that one verifying instance of the event is not sufficient for the utterance to be judged true. This becomes apparent once the utterance is followed by something like ‘but she did it only once’. While this is a plausible continuation of (2), it yields a contradiction if preceded by (3). The Spanish reading is thus best characterized as a habitual reading which implies the ability reading of (2) but is stronger in its truth-conditional requirements.³ (3) Frida ensayaba el libreto en una hora. Frida rehearsed.impf the libretto in one hour ‘Frida used to rehearse the libretto in one hour.’ Spanish (Cipria and Roberts, 2000, p. 307, (15b)) The observed pattern is puzzling for two reasons. Typically, imperfective morphology is ambiguous between some form of a generic reading (which the habitual reading is an instance of) and an episodic reading. It is not immediately clear why the episodic reading is blocked in case the imperfective verb is modified by an in-adverbial. What is even more puzzling is that the Czech pattern lacks even the habitual reading found in Romance and instead it instantiates an ability reading with an actualization requirement. We argue that in order to account for the attested readings we need to consider differences in temporal anchoring properties of these two families of languages and differences in the presuppositional content of their aspectual morphology. We provide evidence that Romance and Slavic have different presuppositions associated with perfective and imperfective morphology. Concretely, the Czech perfective morphology presupposes existence of an event followed by a hiatus⁴ (we will call this presupposition the Activity presupposition). Following Heim’s Maximize Presupposition principle we argue that the Czech perfective morphology may be used only for telic events (either achievements, or accomplishments)

Semantic competition over morphological representations   

   33

as they satisfy the hiatus requirement. Crucially, the Romance perfective morphology lacks the Activity presupposition which affects the mapping between imperfective events and (a)telic events in these languages. As for the ability reading attested in Czech, we argue that the reading results from a a conflict between an existential assertion and the lack of presupposition associated with the imperfective morphology. Since temporal anchoring cannot be resolved in the actual world, the grammar opts for a systematic repair strategy, namely, a shift to possible worlds, similar to what happens in the interpretation of counterfactuals. In this case, the ability reading arises as a resolution of the conflict as the ability reading allows for the existential assertion to be non-redundant even in the context when the presupposition is satisfied. In particular, the existential assertion induces a requirement of one verifying instance for the truth-conditions of the utterance to come true. Other events are modalized, i.e., realized only in possible worlds, resulting in the ability reading. As we will see, this repair strategy is not restricted to Aspect but can be found in other places of the grammar where there is some compositional mismatch. Suggestive evidence comes from the modal interpretation of middles. The paper is organized as follows. First, we summarize differences between Romance and Slavic aspect (sec. 2.1.) and introduce our assumptions about the aspectual semantics for these two groups of languages (sec.  2.2.). Section 2.3. refines our semantics by introducing presuppositional differences. Section  3. applies these conclusions to the ability reading by introducing first the habitual reading (sec.  3.1.) and then restricting the Czech reading to only one verifying instance (sec. 3.2.).

2 Romance versus Slavic Aspect 2.1 The (a)telicity mapping puzzle We follow Giorgi and Pianesi (2001b, among others) in that Perfectivity/Imperfectivity and Telicity/Atelicity need to be treated as two distinct semantic phenomena.⁵,⁶ The (im)perfective distinction may be characterized with respect to the reference time and the event time (Klein, 1994), while the difference between Telicity and Atelicity may be characterized either in terms of homogeneity of the event (where only Atelicity is homogeneous), or in the number of the events (Higginbotham, 2000) (where Atelicity consists of exactly one event).⁷ Crucially, while the perfective/imperfective distinction is a matter of lexical denotation, the telic/ atelic distinction may arise through semantic and pragmatic inferences.

34   

   Mojmír Dočekal and Ivona Kučerová

If we assume that the perfective and imperfective morphology cross-linguistically receives the same interpretation,⁸ then we expect – everything else being equal – that the mapping between the (im)perfective morphology and the (a)telic interpretation should not vary across languages.⁹ However, this prediction is not borne out. In Romance (at least in Spanish and Italian), the imperfective morphology is always interpreted as atelic and the perfective morphology is ambiguous between telic and atelic, while in the Slavic languages (at least in Czech and Russian) the mapping is reversed, i.e., the imperfective morphology may be telic or atelic but the perfective morphology is telic. The differences are schematized in (4)–(5). Examples demonstrating them are given in (6)–(9).¹⁰ (4) The mapping between the morphology and its interpretation in Romance: Morphology

Interpretation

Imperfective Perfective

Atelic Telic

(5) The mapping between the morphology and its interpretation in Slavic: Morphology

Interpretation

Imperfective Perfective

Atelic Telic

(6) Spanish Imperfecto is always atelic (Cipria and Roberts,  2000, p. 304–305, (10–11)) a. Corría petróleo por las cañerías. flow.3sg.impf oil through the pipes Oil flowed/was flowing through the pipes.’

atelic/*telic

b. Corrían 3000 litros de petróleo por las cañerías. flow.3pl.impf 3000 liters of oil through the pipes ‘3000 liters of oil flowed through the pipes.’ atelic/*telic (7) Spanish Préterito can be telic or atelic (Cipria and Roberts, 2000, p. 305, (12– 13)) a. Corrió petróleo por las cañerías. flow.3sg.pf oil through the pipes ‘Oil flowed through the pipes.’

atelic

Semantic competition over morphological representations   

b. Corrieron 3000 litros de petróleo por las cañerías. flow.3pl.pf 3000 liters of oil through the pipes ‘3000 liters of oil flowed through the pipes.’

   35

telic

(8) Czech Imperfective can be atelic or telic a. Tenhle obraz maloval Lada. this painting painted.impf Lada ‘Lada painted this picture.’

telic

b. Tenhle obraz maloval Lada, ale bohužel ho nedokončil. this painting painted.impf Lada but unfortunately him not-finished ‘Lada started painting this picture but unfortunately he didn’t finish it.’ atelic (9) Czech perfective can be only telic a.

Tenhle obraz namaloval Lada. this painting painted.pf Lada ‘Lada painted (and finished) this picture.’

telic

b. # Tenhle obraz namaloval Lada, ale bohužel ho this painting painted.pf Lada but unfortunately him nedokončil. not-finished # ‘Lada finished painting this picture but unfortunately he didn’t finish it.’ *atelic The mapping differences are puzzling. One could argue that the denotation of the imperfective and perfective morphology varies across languages, or the difference in the mapping needs to be attributed to a third factor. The latter is the route we will take in this paper: in particular, we will argue for a uniform denotation of Perfect and Imperfect and we will attribute the difference in the mapping to differing presuppositions associated with the aspectual morphology. We will provide evidence that the Czech perfective morphology carries a presupposition that is not present in the Spanish Perfect. Because of the presence of the presupposition the mapping is subject to semantic competition parallel to semantic competition independently proposed for other morphological features carrying presuppositions, such as determiners, ϕ-features, or number marking (Heim, 1991; Sauerland, 2002, 2003; Sauerland et al., 2005). As we will see, the

36   

   Mojmír Dočekal and Ivona Kučerová

difference in the presuppositional content of the aspectual morphology is one of the crucial ingredients of the Czech ability reading.

2.2 The semantics of the perfective and imperfective morphology In order to argue that a third factor is responsible for the mapping differences observed in the previous section, we need to first establish that the denotation of the aspectual morphology does not interestingly differ between Slavic and Romance. In the following sections we will look mostly at Czech and Spanish but as far as we were able to establish the same findings carry over to Russian and Italian and they possibly distinguish Slavic and Romance languages in general. One of the crucial differences between these languages concerns temporal anchoring of events and its grammatical realization. As Giorgi and Pianesi (2001a) observed, Romance imperfectives lack temporal anchoring. As the examples in (10) and (11) show, Romance imperfective utterances are judged as infelicitous unless a temporal anchoring is lexically provided. Situational anchoring is not sufficient. In contrast, situational temporal anchoring is sufficient for the Czech imperfective morphology. For instance, (12) would be felicitous in a context in which the speaker has a small daughter Frida, you entered their office and saw their desk covered by bits and pieces of a chewed apple. Spanish (11) uttered at the very same context would still be judged as infelicitous. (10) a. # Mario mangiava una mela. Mario ate.impf an apple ‘Mario ate an apple.’ b.

Alle tre Mario mangiava una mela. at three Mario ate.impf an apple Italian; (Giorgi and Pianesi, 2001a, (3–4))

(11) # Frida (se) comía una manzana. Frida SE ate.impf an apple ‘Frida ate an apple.’ (12) Frida jedla jablko. Frida ate.impf apple ‘Frida ate an apple.’

Spanish

Czech

Semantic competition over morphological representations   

   37

Consequently, since the grammatical realization of Imperfect in Romance lacks temporal anchoring, the grammatically Past imperfective may combine with a future-oriented adverb in Spanish, (13), and Italian, (14). On the other hand, since the Czech imperfective structure includes a temporal anchoring, the temporal interpretation of the grammatical tense and the temporal adverb must coincide, as exemplified by (15). (13) Frida se iba mañana. Friday SE left.impf tomorrow ‘Friday intended to leave tomorrow. (But in the end she didn’t)’

Spanish

(14) Mario partiva domani. Mario left.impf tomorrow ‘Mario had the intention/was committed to leave tomorrow.’ Italian (Giorgi and Pianesi, 2001a, (7a)) (15) * Marie odjížděla zítra. Marie left.impf tomorrow

Czech

The differences in temporal anchoring are important for our understanding of the cross-linguistic differences in the interpretation of the aspectual morphology. As the examples in (16) demonstrate, Spanish Imperfect is ambiguous between progressive,¹¹ habitual and intentional reading (Cipria and Roberts, 2000, among others). (16) Possible meaning of Spanish Imperfecto (Cipria and Roberts, 2000, p. 300, (2)) a. Ibamos a la playa cuanda nos encontramos go.1pl.impf to the beach when recpr meet.1pl.pret con Miguel. with Miguel ‘We were going to the beach when we ran into Miguel.’ progressive b. Ibamos a la playa los domingos. go.1pl.impf to the beach on Sundays ‘We went/used to go to the beach on Sundays.’

habitual

38   

   Mojmír Dočekal and Ivona Kučerová

c. Hasta ayer, íbamos a la playa de vacaciones, pero until yesterday go.1pl.impf to the beach on vacation but hoy Pepa dijo que no hay dinero para eso. today Pepa say.3sg.pret that not there is money for that ‘Up until yesterday we were going to the beach on vacation, but today Pepa said that there is no money for that. intention in the past However, since the intentional reading arises only if there is a discrepancy between the time of the event and the aspectual morphology, i.e., only if the imperfective morphology lacks a temporal anchoring (Giorgi and Pianesi, 2001b), the only two readings to consider are the progressive and the habitual reading. The Czech imperfective at first sight behaves differently from its Spanish counterpart. However, it appears plausible that the seeming differences are caused by the more complex morphological formation of Czech aspectual forms that brings in an additional meaning component.¹² Once we separate the additional morphology, similarities between the two languages emerge. As can be seen in (17), Czech morphologically simplex imperfectives are ambiguous between a progressive and a habitual reading.¹³ (17) Only morphologically simple imperfectives are ambiguous between progressive and habitual: a. Jeli jsme na pláž, když jsme potkali Michala. driven aux.1pl to beach when aux.1pl met Michal ‘We were driving to the beach when we ran into Michal.’ progressive b. O nedělích jsme často jeli na pláž. on Sundays aux.1pl often driven to beach ‘Often on Sunday we drove to the beach.’

habitual

The aspectual semantics of Romance and Slavic languages thus does not seem to be entirely different. Consequently, it should be plausible to define a common lexical entry for the imperfective and perfective morphology in Slavic and Romance. For the purposes of this paper, we define the denotation of the aspectual semantics within event semantics. We follow the intuition that the imperfective aspect corresponds to a situation seen from the inside but a situation described by the perfective aspect is seen from the outside (completed). From this point of view, aspect connects event time with reference time (Klein, 1994).¹⁴ Spelling out possible relations between event time and reference time allows us

Semantic competition over morphological representations   

   39

to distinguish two basic semantic relations: INCLUDES and INCLUDED. We can make a further step and define the denotation of the aspectual morphemes in terms of INCLUDES and INCLUDED as well, as in (18) (e is an event or a state, τ(e) stands for a time trace of the event). (18) Lexical entries for the aspectual morphemes [to be modified]: a. [[ perfective ]] =  λPλt∃e.τ(e)⊆t & P(e)

≈ INCLUDES

b. [[ imperfective ]] =  λPλt∃e.t⊆ τ(e) & P(e)

≈ INCLUDED

The INCLUDES/INCLUDED semantics captures the difference between perfective and imperfective but it does not say anything about its relation to (a)telicity. For the semantics of (a)telicity we follow Giorgi and Pianesi (2001b). Under their analysis, telic events consist of two separate events: activity (process), e1, and result, e2. Consequently, telic events are not homogeneous. As we can see in (19) and (20), the formalization in (18) allows us to derive both the atelic and telic interpretation of the Czech imperfective morphology. This is a welcome result because we know that the imperfective morphology is indeed attested both with the telic and the atelic interpretation. (19) Atelic interpretation of imperfective: a. Petr včera četl tu knihu. Petr yesterday read.impf that book ‘Yesterday Petr read the book.’ b. ∃e∃x[read(e) ∧ Agent(Petr,e) ∧ t(e) ∧ Theme(x,e) ∧ book(x)] (20) Telic interpretation of imperfective: a. Petr včera četl tu knihu. Petr yesterday read.impf that book ‘Yesterday Petr read/finished reading the book.’ b. ∃  ∃x [read ∧ Agent (Petr, ) ∧ Theme (x,) ∧ book (x)] Note that (20) under its telic interpretation still remains semantically imperfective; intuitively, it means that the process part of the event holds before the reference time. More formally, e2 holds at a time which is included in yesterday while e1 extends before the time of yesterday.

40   

   Mojmír Dočekal and Ivona Kučerová

While more careful empirical work is called for here, we believe that it is legitimate to pursue the hypothesis that the difference between the Romance and the Slavic aspectual mapping does not lie in the semantics of Perfect/Imperfect per se but that it must be attributed to something else. The goal of the next section is to figure out what the additional factor is.

2.3 Perfective v. imperfective morphology and their presuppositions We argue that perfective and imperfective do not differ only in the relation of the reference time and the event time. In Czech the perfective morphology imposes a precondition on the context that is not present in the case of the imperfective morphology. In particular, we argue that the perfective morphology presupposes the existence of the beginning of the event, roughly the first homogeneous part of telic events (e1). Consequently, since only the beginning of the event is presupposed, it is entailed that the event is not homogeneous. If this was a homogeneous event, presupposition of a subevent would hold for the whole event which is not a case here. Consequently, since this is not a homogeneous event, there is a hiatus between the first and subsequent event(s).¹⁵ We will call this presupposition Activity presupposition.¹⁶ In contrast, the Czech imperfective has no such presupposition. That we deal with a presupposition and not with another type of inference can be shown by projection properties of the Activity presupposition, namely, it projects under negation and under a question operator, cf. (21)–(22). The expected properties are found in other structural environments as well, for example, the Activity presupposition does not project from the antecedent of a conditional, as in (23).¹⁷ (21) Activity inference survives under negation only in perfective: a. Jan nedopsal knihu. Jan.nom neg-wrote.pf book.acc ‘Jan didn’t finish writing a book.’ Jan started writing a book b. Jan nepsal knihu. Jan.nom neg-wrote.impf book.acc ‘Jan didn’t write a book.’ Jan started writing a book

Semantic competition over morphological representations   

   41

(22) Activity inference survives in questions only in perfective: a. Dopsal Jan knihu? wrote.pf Jan.nom book.acc ‘Did Jan finish writing a book?’ Jan started writing a book b. Psal Jan knihu? wrote.impf Jan.nom book.acc ‘Did Jan write a book?’ Jan started writing a book (23) Activity presupposition of the antecedent does not project a. Pokud Jan dopsal knihu, tak si ho Marie if Jan.nom wrote.pf book.acc then refl him Marie vezme. gets-married ‘If Jan finished writing a book, Marie will marry him.’ Jan started writing a book b. Pokud Jan psal knihu, tak si ho Marie if Jan.nom wrote.impf book.acc then refl him Marie vezme. gets-married ‘If Jan wrote a book, Marie will marry him.’ Jan started writing a book In contrast, neither Spanish perfective nor imperfective carry such a presupposition, as shown in (24) and (25). (24) There is no Activity presupposition projection in Spanish questions: a. ¿Escribia ayer Maria un poema? wrote.impf yesterday Maria a poem ‘Did Maria write a poem yesterday?’ Maria started writing a poem b. ¿Escribió ayer Maria un poema? wrote.pf yesterday Maria a poem ‘Did Maria write a poem yesterday?’ Maria started writing a poem

42   

   Mojmír Dočekal and Ivona Kučerová

(25) There is no Activity presupposition projection under negation in Spanish: a. Ayer Maria no escribia una poema. yesterday Maria not wrote.impf a poem ‘Yesterday Maria didn’t write a book.’ Maria started writing a poem b. Ayer Maria no escribio una poema. yesterday Maria not wrote.pf a poem ‘Yesterday Maria didn’t write a book.’ Maria started writing a poem Even though the Czech imperfective morphology does not carry the Activity presupposition, it is still compatible with its content, as can be seen in (26). However, this is not a presupposition since it can be cancelled, as shown in (27). (26) Imperfective may have an Activity inference a. (A museum guide standing in front of a painting:) b. Tenhle obraz maloval Lada. this picture.acc painted.impf Lada.nom ‘Lada painted this picture.’ Activity inference (27) Imperfective does not need to have an Activity inference a. (A teacher about a picture one student chose to copy for his art class:) b. Petr maluje tenhle obraz. Ale ještě si Petr.nom paints.impf this picture.acc but not-yet refl nekoupil ani barvy. not-bought.pp even colors ‘Petr is supposed to paint this picture. But he even has not bought colors yet.’ no Activity inference We argue that in order to capture the Czech facts, the lexical entry of the Czech perfective must be enriched by the Activity presupposition.¹⁸ We state the Activity presupposition in terms of a homogeneous part of an event. The lexical entry thus requires a proper part of the whole event (e′) such that the whole event is the terminative counterpart of e′ and for all time intervals of e′ the predicate P

Semantic competition over morphological representations   

   43

holds. There is no such presupposition for the imperfective morphology. The final lexical entries for the Czech perfective and imperfective morphology are given in (28). (28) Lexical entries for the Czech aspectual morphemes [final]: a. [[ perfective ]] =  λPλt∃e: ∃e’(ter(e’) = e) ∧∀t’⊆τ(e’)(P(e’)) . τ(e)⊆t & P(e) b. [[ imperfective ]] =  λPλt∃e. t⊆ τ(e) & P(e) Now we are finally in the position to address the asymmetry in the usage of the perfective and imperfective morphology in Czech. We argue that the asymmetry is a result of semantic competition. Whenever competing morphological items differ with respect to presupposition α and if the given context satisfies α, then the item presupposing α must be used. This principle is known as the Maximize Presupposition principle and has been first suggested in Heim (1991), following Hawkins (1991) and other work in lexical pragmatics (see also Sauerland 2002, 2003; Sauerland et al. 2005; Heim 2008).¹⁹ Consequently, whenever the activity part is presupposed, the perfective morphology must be used. Since the imperfective morphology is compatible with asserting the activity event but does not presuppose the activity event, the Czech imperfective morphology can be either telic or atelic. In this way, the presuppositional differences together with the Maximize Presupposition principle give us the asymmetry in the usage of the Czech aspectual morphology without affecting our compositional semantics. Presumably, the asymmetry in the Romance morphology is caused by the Romance imperfective morphology that poses more requirements on the context than its perfective counterpart. We leave the question of Spanish presuppositions for future research.²⁰ To summarize, it seems reasonable to assume that the lexical denotation of Perfect and Imperfect is the same in Czech and Spanish but the languages differ when it comes to presuppositions associated with their aspectual morphology. As a result, the usage of the aspectual morphology in Czech is not identical to the usage of the aspectual morphology in Spanish. In the following section we will investigate whether the presuppositional difference might also be behind the habitual v. ability reading differences.

44   

   Mojmír Dočekal and Ivona Kučerová

3 The ability versus the habitual reading We are now in a position to address the question of the Czech bounded ability reading attested for sentences like (29), repeated below. As we have seen, the question has two subparts: First, how come the in-adverbial blocks the episodic reading typically associated with the imperfective morphology. Second, how come there is no habitual reading either and instead the only attested reading is the bounded ability reading. Note that the first subquestion holds for Romance as well, thus the answer should reflect the common properties of the Slavic and Romance imperfective morphology. In contrast, the second subquestion is specific to Czech (or Slavic) and thus it is likely to follow from the presuppositional differences between the languages. (29) Petr četl Vojnu a mír za dvě hodiny. Petr.nom read.impf War and Peace in two hours ‘Petr was reading War and Peace in two hours.’ ‘Petr used to read War and Peace in two hours.’ ‘Petr was able to read War and Peace in two hours.’

# episodic # habitual ability

Subsection 3.1. addresses the first part of the question by looking in detail at how the structure of (29) gets compositionally interpreted. This will help us clarify at which point of the derivation the problem with the episodic reading emerges and it will make it easier to see the logically plausible options for repairing the structure. The actual process responsible for the difference between the Romance habitual and the Czech bounded ability reading will be discussed in section 3.2.

3.1 Temporal adverbs and reversal of event time and reference time Let’s consider how exactly the structure of (29) gets compositionally interpreted. In the first step of the derivation, we merge V, read, with the object, War and Peace. The resulting phrase, VP, is aspectually underspecified: the only semantic information comes from the lexical semantics of the verb. In this particular case, the VP can receive either an accomplishment or an activity interpretation. In the next derivational step, the adverbial in two hours is merged.²¹ For concreteness, we assume that in-adverbials, in contrast to for-adverbials, are functions that take two events as their argument: process and telos (Higginbotham, 2000). For foradverbials one event suffices. If we apply this semantics to our example, after the adverbial in two hours is merged, the structure receives the accomplishment

Semantic competition over morphological representations   

   45

interpretation: if a one-event interpretation were selected, the structure would not be interpretable. Furthermore, we assume that the event time of the accomplishment is linked to the reference time. For accomplishments, the event time should be included in the reference time, otherwise the result subevent would be outside of the reference time of the event. In the next step, imperfective Asp° is merged. The imperfective aspect requires the event time to include the reference time. However, the accomplishment semantics of vP gives us the opposite relation. Unless the structure is further modified by something that can reverse the relation of the reference time and the event time, the episodic reading is predicted to be ungrammatical. As we have seen in section 1, this prediction is indeed borne out both in Romance and Slavic. Note that if the problem lies in the relation of the event time and the reference time, it follows that the structure could be rescued if it were possible to reverse the relation of the two relevant times. As we see in (30) this is indeed possible: the habitual reading is an instance of such a reversal (Dowty, 1979, among others).²² (30) Imperfective verbs may combine with in-adverbials a. When John was in a better shape, he was running a marathon in two hours. b. After John took a summer course in speed-reading, he was reading War and Peace in two hours. The question is how exactly the reversal arises. There are two hypotheses to consider. The reversal might arise either by an insertion of a type-shifting (habitual) operator (Dowty, 1979; de Swart, 1998, 2000; van Geenhoven, 2005; Boneh and Doron,  2008, among others), or it could be an instance of aspectual coercion (Dowty, 1979, and much later work). Under the type-shifting operator hypothesis, a habitual operator can freely combine both with imperfective and perfective verbs.²³ If it combines with an imperfective verb, the resulting reading is habitual, as in (30). According to the coercion hypothesis, imperfective is inherently ambiguous between a progressive and a habitual reading. The semantics of an in-adverbial should be compatible only with the habitual reading. If imperfective combines with an in-adverbial, the progressive is impossible. Consequently, the habitual reading is the only remaining possibility. This hypothesis thus predicts that if imperfective is ambiguous, we should always get the habitual reading. In other words, while according to the type-shifting hypothesis the habitual reading is a result of a special operation and as such might be further restricted, the coercion hypothesis predicts that the habitual reading should always be an

46   

   Mojmír Dočekal and Ivona Kučerová

option. As can be seen in Czech (31), the availability of the habitual (or ability) reading is in fact limited. Thus the prediction made by the coercion hypothesis, unlike the prediction made by the type-shifting hypothesis, is not borne out. (31) # Když měl Petr lepší auto, tak jel do Prahy za dvě hodiny. when had Petr better car, then drove to Prague in two hours ‘When Petr had a better car, he was driving to Prague in two hours.’ Importantly, both hypotheses share the assumption that what looks like an ability reading is an entailment of the habitual reading. This cannot be correct because as we have already seen the Czech ability reading is compatible with the ‘only once’ interpretation which is impossible in the case of a true habitual or a frequentative reading. (32) In Czech, once-adverbial may modify the ability reading but not the frequentative reading: a. * Petr jednou objev-OVA-l na zahrádce plevel. Petr once discover-freq-past.pf at garden crabgrass ‘* Petr once used to discover crabgrass in his garden.’ b.

Po letním kursu rychločtení Petr jednou četl after summer course of-speed-reading Petr once read.impf Vojnu a mír za dvě hodiny. War and Peace in two hours ‘After taking a summer course in speed-reading, Petr was once reading War and Peace in two hours.’

Interestingly, a reading very similar to the Czech bounded ability reading appears outside of the aspectual domain, namely, in a certain type of reflexive constructions resembling English middles, as in (33). Unlike English middles, this construction requires (at least) one verifying instance and the possibility of pluralization of the event.²⁴ (33) V tomhle stroji se káva praží dobře. in this machine coffee roasts well ‘Coffee roasts well in this machine.’ Even though this construction is often characterized as generic or habitual, it differs from both of them in that it requires a verifying instance (unlike true generics) and in that one verifying instance is sufficient (unlike habituals).

Semantic competition over morphological representations   

   47

Despite these difficulties we would like to suggest that the Czech ability reading is indeed compatible with the type-shifting operator hypothesis, i.e., we side with van Geenhoven (2005) and others in that both in Slavic and Romance the pluralization of events arises via a habitual operator. We argue, however, that in order to obtain the relevant reading we need to consider the presuppositional content of the aspectual morphology as well.

3.2 Maximize Presupposition and the Habitual reading To repeat, we propose that in order to account for the ability reading in Czech we need to adopt a version of the type-shifting operator hypothesis. For concreteness, we will use the definition of a habitual operator “ᅻ” given in (34).²⁵ (34) (after van Geenhoven 2004, p. 158, (60)) ᅻtV (x) at t = 1  ⇐⇒∃t′(t ′⊆ t ∧ V (x) at t ′∧ number(t ′) > 1 ∧∀t ′(t ′⊆ t ∧ V (x) at t ′→ ∃t ′′(t ′′⊆ t∧(t ′′ > t′∨t ′′ < t ′)∧V (x) at t ′′∧∃t ′′′(t ′ < t′ ′′ < t ′′∨t′ > t ′′′ > t ′′∧ ¬ V (x) at t ′′′)))) According to this definition, for an event to be pluralized, there must be a hiatus between iterated instances of the event. This non-overlap requirement introduces boundedness of the iterated event. Consequently, since pluralized events are bounded, the iterated event is compositionally combinable with an in-adverbial. We can go one step further though. Notice that the operator asserts the existence of a hiatus. Therefore, there must be an activity preceding the hiatus. We argue that this is the semantic source of the actuality requirement of habitual readings. In other words, the actuality follows from pluractification of the event and not from the aspectual semantics per se (see Piñón 2003 for pertaining discussion). Furthermore, the hiatus assertion and the semantics of in-adverbials limit this type of pluralization to telic events. Even though this is a good result we still face the question of why there is only one verifying instance needed in Czech while the other instances of the event may be shifted to possible worlds, i.e., why the reading attested in Czech is an ability reading instead of the habitual reading attested in Romance. We would like to suggest that the ability reading results from a conflict between the hiatus assertion (by the habitual operator) and the lack of the Activity presupposition of the Czech imperfective morphology.²⁶ The reasoning goes as follows. If the speaker uses the imperfective morphology felicitously, the speaker were not in a position to presuppose the Activity

48   

   Mojmír Dočekal and Ivona Kučerová

presupposition. This follows from the Maximize Presupposition principle. If the speaker were in a position to presuppose the Activity presupposition, she would have to use the perfective morphology. More precisely, given our semantics, if the speaker uses the imperfective morphology, she was not in a position to presuppose the existence of a hiatus. One might ask why the hiatus cannot be accommodated. To answer this, let us look at the reasoning components, namely, assertion, entailment and presupposition: (35) a. Assertion: There exists a hiatus²⁷ between two distinct time intervals (∃t ′∃t ′′∃t ′′′(t ′ < t ′′′ < t ′′∨ t ′ > t ′′′ > t ′′∧ ¬ V (x) at t ′′′) ∧ V (t ′) ∧ V (t ′′)) b. Entailment: If there is a hiatus, there is a time interval preceding the hiatus c. Presupposition: none (a hiatus is not presupposed) By the Gricean reasoning, the hearer concludes that if the speaker were in a position to assert the existence of a hiatus and yet did not see herself in a position to presuppose the hiatus, then it follows that the speaker didn’t believe that she was in a position to presuppose the hiatus: (36) a. Assertion: There exists a hiatus between two distinct time intervals (∃t ′∃t ′′∃t ′′′(t ′ < t ′′′ < t ′′ ∨ t ′ > t ′′′ > t ′′ ∧ ¬V (x) at t ′′′) ∧ V (t ′) ∧ V (t ′′)) b. Entailment: If there is a hiatus, there are two time intervals surrounding the hiatus c. Presupposition: ¬Believe(hiatus) By the epistemic step, i.e., secondary strengthening of the presupposition justified by the experthood of the attitude holder, it follows that the speaker believed that the hiatus may not be presupposed (or accommodated) (cf. Sauerland 2004; Chemla 2008): (37) a. Assertion: There exists a hiatus between two distinct time intervals (∃t ′∃t ′′∃t ′′′(t ′ < t ′′′ < t ′′ ∨ t ′ > t ′′′ > t ′′ ∧ ¬V (x) at t ′′′) ∧ V (t ′) ∧ V (t ′′)) b. Entailment: If there is a hiatus, there are two time intervals surrounding the hiatus c. Presupposition: Believe ¬ (hiatus)

Semantic competition over morphological representations   

   49

Since the hiatus is asserted but cannot be presupposed for the actual world, it follows that there is exactly one verifying instance required by the assertion of the hiatus. The process of pluralization of the event is not presupposed in the actual world and consequently is free to shift to possible worlds. As a result, the hearer is free to interpret the pluralization of events as being instantiated in possible worlds as long as there is one instance of the hiatus in the actual world. Romance is different in that the relevant presupposition is not present in the Romance imperfective morphology. Consequently, the speaker is free to accommodate the presupposition of the hiatus. Since the hiatus can always be accommodated, there is no need to shift the hiatus to possible worlds. We argue that the shift to possible worlds is not only unnecessary, in fact it is impossible. We assume this restriction follows from economy of interpretation which allows additional LF operations only if the relevant semantic interpretation is not available otherwise (Fox, 1995, 2000; Reinhart, 1995, 2006). Consequently, if the presupposition of the hiatus cannot be shifted to possible worlds, a regular habitual reading becomes obligatory. So far we concentrated only on morphologically simplex imperfective verbs. Interestingly, morphologically complex imperfective verbs in Czech may be formed by prefixes whose lexical entry carries the Activity presupposition. The current proposal makes the following prediction. If such an imperfective verb is modified by an in-adverbial, the competition with the perfective morphology will not be relevant anymore and consequently we expect to get the Romance type of habitual reading. This prediction is indeed borne out as can be seen in (39). (38) Petr (*jednou) rozmalovával pokoje za Petr once finished-the-first-phase-of-painting.impf rooms in hodinu. hour ‘Petr used to finish the first part of painting rooms (for example, the first coat) in an hour.’ To conclude, the presented case study makes two theoretical points. First, we argued that in order to understand aspectual semantics and the morphologysemantics mapping in the aspectual domain, we must take into account the presuppositional content of the aspectual morphology. For instance, the denotation of perfective and imperfective in Slavic and Romance languages may be identical but the usage of the aspectual morphology differs because of distinct presuppositions associated with the morphology in different languages. Even though we differ in the actual implementation, a similar point has been made, for example, by Grønn (2005) and Tatevosov (2011).

50   

   Mojmír Dočekal and Ivona Kučerová

Second, as the obligatory status of the bounded ability reading in Czech shows, there are systematic repair strategies that may resolve compositional conflicts by inserting a free type-shifting operator or by shifting the interpretation to possible worlds. Even though more work needs to be done to fully understand the crosslinguistic differences in the aspectual domain, we believe our study supports the view of the grammar as featuring a component that is driven by competition, at least on the LF branch of the derivation.

Notes 1 We would like to thank Klaus Abels, Boban Arsenijević, Hana Filip, Danny Fox, Sabine Iatridou, Roni Katzir, Nathan Klinedinst, audiences at the Czech Formal Grammar 2009 conference, IJN/CNRS Paris in April 2009, UCL Linglunch in May 2009 and FASL 19 at Cornell in May 2009. We would like to thank all our Czech friends and family members who keep an eye on our Czech judgements. Special thanks go to Luisa Martí for her patience while clarifying Spanish judgments and to Anna Moro for her help with Italian and Romance in general. Last but not least, the authors are happy to acknowledge that IK was financially supported by the UK Arts and Humanities Research Council (grant 119403) and MD was financially supported by GAČR (grant 405/09/0677). 2 We approximate the intended reading by the English progressive. 3 We approximate the habitual reading by the English ‘used to’ construction. 4 If predicate P holds at time t and t′′ and there is time t′ such that t < t′ < t′′, and P doesn’t hold at t′, we call such time t′ hiatus. 5 While for Giorgi and Pianesi (2001b) there is a connection between the type of event and (a)telicity, some authors, for example, Bertinetto (2001), deny even this connection. Since the relation between the event type and (a)telicity is not directly relevant to our puzzle we will leave the details aside. 6 In Slavic Telicity usually coincides with Perfectivity. For example, in Czech almost all morphologically perfective verbs formed by perfectivizing prefixes are telic. The only exception are verbs prefixed by pro- and po-. 7 But see, for example, Filip (2008) for the view that having a terminating point is a property of telic events. In particular, it is a result of a conventional implicature and as such it doesn’t belong to the core semantic meaning of Aspect. 8 In section 2.2. we will define the lexical entry of the perfective morphology as the INCLUDES relation and the lexical entry of the imperfective morphology as the INCLUDED relation. 9 Later in the paper (sec. 2.3.) we will argue that in fact everything is not equal. 10 The examples are not entirely parallel because of non-trivial differences in the information structure realization and the usage of definite articles. We minimize the differences by sticking to definite descriptions in the Czech examples. 11 Note that the Romance progressive is not semantically identical to the English progressive (Bonomi, 1997). 12 Czech imperfective and perfective verbs may be modified by a large number of prefixes and infixes that encode rather specific lexical meanings usually described under the label

Semantic competition over morphological representations   

   51

‘Aktionsart’. In the rest of the paper, we will mostly abstract from this additional morphology but will shortly return to it in the end of section 3.2. 13 More complex morphological forms have only the habitual reading, as seen in (i). This is because verbs morphologically derived from simple imperfectives are either perfective or habitual. (i) More complex imperfectives may have only the habitual reading: a. # Jezdívali jsme na pláž, když jsme potkali Michala. driven aux.1pl to beach when aux.1pl met Michal ‘We were driving to the beach when we ran into Michal.’ #progressive b. O nedělích jsme jezdívali na pláž. on Sundays aux.1pl driven to beach ‘We drove/used to drive to the beach on Sundays.’ habitual 14 According to Dowty (1979) and Landman (1992), among others, at least some aspectual phenomena must be treated within intensional semantics. In this paper we mostly stick to a characterization of (Im)Perfectivity in terms of extensional semantics (in line with Klein 1994 and Paslawska and von Stechow 2003) as most of the data we examine do not call for intensional treatment. Reference to modality will become necessary once we turn to the ability reading attested in Czech. 15 The relation to the hiatus will be relevant for the habitual v. ability reading. 16 Even though we use the notion of semantic presuppositions here, we are not entirely convinced that this is the correct characterization. It is plausible that this type of precondition is a secondary presupposition derived from the event representation of perfective, along the lines of the analysis of soft presupposition triggers in Abusch (2010). 17 Notice that even if the presuppositions are secondary presuppositions, the projecting properties clearly show that we deal with some form of presupposition and not other type of inference. 18 We simplify here. Czech perfectives are mostly formed by some additional morphological material (prefix or infix), they can never be simplex. Consequently, it is not clear whether it is the additional morphology or the perfective structure per se that carries the presupposition. An anonymous reviewer brought to our attention that, for instance, in Bulgarian, the imperfective/ perfective distinction is cumulatively realized within the Tense affix as in Spanish and interestingly the available readings are parallel to Spanish as well. It is likely that a complete empirical picture of Slavic and Romance aspectual differences will have to take into account the exact type of morphological distinctions and will have to consider the role of morphological markedness in a fuller detail than it can be done in this paper. We will thus leave to further empirical investigation whether all perfective formations in Czech and in Slavic in general behave in this way or whether further refinement is needed. 19 We would like to thank Roni Katzir for his suggestion to use the Maximize Presupposition principle to account for the morphological asymmetry. 20 One must still be careful about the lexical denotation of various Slavic prefixes though. That imperfective does not carry the Activity presupposition does not mean that the imperfective morphology cannot combine with a prefix which carries a presupposition. The question of the presuppositional content of various aspectual prefixes cannot be fully addressed in this paper but it will become relevant for the discussion of predictions made by the current proposal toward the end of section 3.2. 21 Let’s assume for concreteness that the adverb adjoins to vP.

52   

   Mojmír Dočekal and Ivona Kučerová

22 Similarly, perfective verbs may combine with a for-adverbial combination receiving a frequentative reading (Zucchi and White, 2001; van Geenhoven, 2004, 2005, among others). (i) Perfective verbs may combine with for-adverbials (from van Geenhoven 2004) a. John discovered crabgrass in his yard for 6 weeks. (Dowty, 1979, p. 63) b. I discovered water under my sink for a month. (Partee, p.c.) c. The prospectors struck oil for two weeks. (Mittwoch, 1991, p. 79) 23 If the operator combines with a perfective verb, the resulting reading is frequentative. For the purposes of this paper we put the frequentative reading aside. 24 See also Brandt (2009) for an argument that the modal interpretation of German middles stems from repairing a contradiction that is derived in the compositional process. 25 Any habitual operator that iterates events in a non-overlapping fashion would work as well. 26 An alternative would be to posit a silent modal operator. Such an operator could introduce the boundedness properties as well (see Bhatt 1999; Hacquard 2006; Mari and Martin 2008, among others, for pertaining discussion). The disadvantage of this alternative is that one would still need to restrict the operator to Slavic. 27 See fn. 4.

4 References Abusch, Dorit. 2010. Presupposition triggering from alternatives. Journal of Semantics 27: 37. Bertinetto, Pier Marco. 2001. On a frequent misunderstanding in the temporal-aspectual domain: The ‘Perfective = Telic confusion’. In: Carlo Cecchetto, Gennaro Chierchia, and Maria Teresa Guasti (eds.), Semantic Interfaces. Reference, Anaphora and Aspect, 177–210. Stanford, CA: CSLI Publications. Bhatt, Rajesh. 1999. Covert modality in non-finite contexts. Doctoral Dissertation, University of Pennsylvania. Boneh, Nora, and Edit Doron. 2008. Habituality and the habitual aspect. In: Susan Rothstein (ed.), Theoretical and Crosslinguistic Approaches to the Semantics of Aspect, 321–347. Amsterdam: John Benjamins. Bonomi, Andrea. 1997. The progressive and the structure of events. Journal of Semantics 14: 173–205. Brandt, Patrick. 2009. Generische Möglichkeit in Medialkonstruktionen. In: W. Abraham and E. Leiss (eds.), Modalität. Epistemik und Evidentialität bei Modalverb, Adverb, Modalpartikel und Modus, volume 77 of Studien zur deutschen Grammatik, 79–100. Tübingen: Stauffenburg. Chemla, Emmanuel. 2008. An Epistemic Step for Anti-Presupposition. Journal of Semantics 25: 141–173. Cipria, Alicia and Craige Roberts. 2000. Spanish imperfecto and pretérito: Truth conditions and Aktionsart effects in a situation semantics. Natural Language Semantics 8: 297–347. Dowty, David. 1979. Word Meaning and Montague Grammar. Dordrecht: D. Reidel. Filip, Hana. 2008. Events and maximalization. In: Susan Rothstein (ed.), Theoretical and Crosslinguistic Approaches to the Semantics of Aspect, 217–256. John Benjamins. Fox, Danny. 1995. Economy and scope. Natural Language Semantics 3: 283–341. Fox, Danny. 2000. Economy and Semantic Interpretation. Cambridge, Mass.: MIT Press.

Semantic competition over morphological representations   

   53

van Geenhoven, Veerle. 2004. For-adverbials, frequentative aspect, and pluractionality. Natural Language Semantics 12: 135–190. van Geenhoven, Veerle. 2005. Atelicity, pluractionality, and adverbial quantification. In: Henk Verkuyl, Henriette de Swart, and Angelike van Hout (eds.), Perspectives on Aspect, 107–125. Dordrecht: Springer. Giorgi, Alessandra and Fabio Pianesi. 2001a. Imperfect dreams: The temporal dependencies of fictional predicates. Probus 13: 31–68. Giorgi, Alessandra and Fabio Pianesi. 2001b. Ways of terminating. In: Carlo Cecchetto, Gennaro Chierchia, and Maria T. Guasti (eds.), Semantic Interfaces: Reference, Anaphora and Aspect. Stanford: CSLI Publications. Grønn, Atle. 2005. Presuppositional variance and aspectual meaning. In: Discourse Domains and Information Structure, 11–20. Edinburgh: ESSLI. Hacquard, Valentine. 2006. Aspects of modality. Doctoral Dissertation, Massachusetts Institute of Technology. Hawkins, John A. 1991. On (in)-definite articles: implicatures and (un)grammaticality prediction. Journal of Linguistics 27: 405–442. Heim, Irene. 1991. Artikel und Definitheit. In: Arnim von Stechow and Dieter Wunderlich (eds.), Semantik: Ein internationales Handbuch der zeitgenössischen Forschung, 487–535. Berlin: Mouton de Gruyter. Heim, Irene. 2008. Features on bound pronouns. In: Daniel Harbour, David Adger, and Susana Béjar (eds.), Phi-theory: Phi Features across Interfaces and Modules, 35–56. Oxford: Oxford University Press. Higginbotham, James. 2000. On events in linguistic semantics. In: J. Higginbotham, F. Pianesi, and A. Varzi (eds.), Speaking of Events, 49–79. New York: OUP. Klein, Wolfgang. 1994. Time in Language. New York: Routledge. Landman, Fred. 1992. The progressive. Natural Language Semantics 1: 1–32. Mari, Alda and Fabienne Martin. 2008. Tense, abilities and actuality entailment. In: Paul Dekker Maria Aloni and Floris Roelofsen (eds.), Proceedings of the Sixteenth Amsterdam Colloquium, volume 16, 151–156. Amsterdam: ILLC/Department of Philosophy University of Amsterdam. Mittwoch, A. 1991. In defence of Vendler’s achievements. Journal of Linguistics 6: 71–85. Paslawska, Alla and Arnim von Stechow. 2003. Perfect readings in Russian. In: Artemis Alexiadou, Monika Rathert, and Arnim von Stechow (eds.), Perfect Explorations, 307–362. Mouton de Gruyter. Piñón, Christopher. 2003. Being able to. In: G. Garding and M. Tsujimura (eds.), WCCFL 22 Proceedings, 384–397. Somerville, MA: Cascadilla Press. Reinhart, Tanya. 1995. Focus – the PF interface. Manuscript University of Utrecht. Reinhart, Tanya. 2006. Interface Strategies. Optimal and Costly Computations. Cambridge, MA: MIT Press. Sauerland, Uli. 2002. The present tense is vacuous. Snippets 6. Sauerland, Uli. 2003. A new semantics for number. In: R. Young and Y. Zhou (eds.), Proceedings of SALT 13, 258–275. Cornell University, Ithaca, NY: CLC Publications. Sauerland, Uli. 2004. Scalar implicatures in complex sentences. Linguistics and Philosophy 27: 367–391. Sauerland, Uli, Jan Andersen, and Kazuko Yatsushiro. 2005. The plural is semantically unmarked. In: Stephan Kepser and Marga Reis (eds.), Linguistic Evidence – Empirical,

54   

   Mojmír Dočekal and Ivona Kučerová

Theoretical, and Computational Perspectives, 413–434. Berlin, Germany: Mouton d’Gruyter. de Swart, Henriëtte. 1998. Aspect shift and coercion. Natural Language and Linguistic Theory 16: 347–385. de Swart, Henriëtte. 2000. Tense, aspect and coercion in a cross-linguistic perspective. In: Miriam Butt and Tracy Holloway King (eds.), Proceedings of the Berkeley Formal Grammar Conference. CSLI Online Publications [http://cslipublications.stanford.edu/LFG/5/bfg00/ bfg00deswart-abs.html]. Tatevosov, Sergei. 2011. On the modal meaning of Slavic perfective. Talk Presented at the 20th Annual Meeting of FASL at MIT. Zucchi, A., and M. White. 2001. Twigs, sequences, and the temporal constitution of predicates. Linguistics and Philosophy 24: 223–270.

Hans-Christian Schmitz and Bernhard Fisseni

Repairs for Reasoning* Abstract: We describe and experimentally investigate phenomena of modal enrichment, that is, phenomena in which a recipient non-literally interprets an utterance by creating and applying a modal operator. We give competing explanations for these phenomena – namely an explanation according to which modal enrichment is a repair procedure for making the utterance match a script of information processing vs. an explanation according to which modal enrichment is triggered by rhetorical structure.

1 Introduction Let us consider the following example, discussed by Schmitz (2011): (1) A: Wie spät ist es? (What time is it?) B: Es ist 5 nach 3, meine Uhr geht aber 5 Minuten vor. (It’s 5 past 3, but my watch is 5 minutes fast.) When you ask test subjects what time it is according to B’s answer, by far most of them will reply that it is (exactly) 3 o’clock, although this contradicts the literal meaning of the answer sentence. Schmitz  (2011) performed such experiments several times. The results of one experiment are given in Table 1¹ – the data proved to be reliable in repitition. In addition, Schmitz asked the subjects how they arrived at the non-literal interpretation, and they replied that they interpreted the answer in the sense of “By my watch it’s 5 past 3, but my watch is 5 minutes fast” which entails that it is (exactly) 3 o’clock.

exp. 1:

15 : 00

15 : 05

36 (85.7 %)

6 (14.3 %)

Table 1: What time is it according to B’s answer? (42 test subjects)

The example has not been made up. The answer was originally uttered by Benedikt Löwe (at that time a logician at the University of Bonn who did not intend to become a provider of linguistic examples). When asked what he meant, he

56   

   Hans-Christian Schmitz and Bernhard Fisseni

answered that he meant that it was exactly 3 o’clock. Thus, the test subjects grasped the intended meaning, they interpreted Benedikt’s answer correctly. How did they do so? Obviously, they expanded the literal meaning of the answer sentence by deriving a modal operator such as ‘by the speaker’s watch’ and applying it to the first conjunct. Then they inferred the answer to A’s question from the non-literal interpretation of the first conjunct and the literal interpretation of the second conjunct. Schmitz  (2011) calls meaning expansion by creation and application of a modal operator a form of enrichment, namely ‘modal enrichment’. He argues that modal enrichment is a conventionalized operation of expanding the truthconditional content of sentences and gives several examples to prove that modal enrichment can take place in communication. Moreover, he shows that under certain conditions it is possible that modal enrichment is licensed for the recipient although not intended by the speaker. In such a situation, a misunderstanding can be avoided if the speaker explicitly blocks the enrichment of his utterance. Schmitz (2011) argues that in German this blocking function is fulfilled by the adverb “tatsächlich” (in fact). The present paper can be seen as a sequel to Schmitz (2011): instead of elaborating on the fact that modal enrichment occurs and the question of how it can be avoided, we will focus on the issue of why recipients perform operations of modal enrichment. We will experimentally investigate some phenomena in closer detail, in particular the initial watch-example (1), in order to expose under which conditions modal enrichment takes place and under which conditions it does not. The derived data will be the basis for explanations of the phenomena. In particular, we will discuss two approaches in closer detail and show how these can be evaluated. According to the first approach (Reasoning), the literal interpretation of (1) does not match a presupposed script of deriving and conveying information. It therefore has to be repaired by modal enrichment. According to the second approach (Rhetorical Structure), modal enrichment is not a repair procedure but rather a primary (close to hard-wired) way of interpreting information. We elaborate upon this approach with respect to current theories of polyphony. The outline of the paper is as follows: in section 2 we will report on experimental investigations in modal enrichment in order to prove the reliability and strength of the effect and determine under which circumstances it can be expected. In section 3, we will investigate the impact of the conjunction “aber” (but) on modal enrichment. In section 4, we will elaborate upon the mentioned approaches for explaining modal enrichment and discuss their evaluation. Finally, in section 5, we will sum up our findings.

Repairs for Reasoning   

   57

2 Modal enrichment In this section, we will discuss experimental data regarding the interpretations of some key examples, in particular variations of the initial watch-example (1).

2.1 Experimental settings All our experiments took place in a classroom setting: the examples were projected onto a wall and read aloud twice. The test subjects were then asked a question – after the presentation of example (1), e.g., they were asked what time it is according to B’s answer. They noted down their answers on paper. In most experiments, we did not predefine a list of answers to choose from, that is, the questions were not multiple choice questions (exceptions will be discussed below). The experiments took place between 2005 and 2009 at the Universities of Bonn, Frankfurt am Main and Duisburg-Essen. The test subjects were mainly undergraduate students in computational linguistics, linguistics or German studies. It can be assumed that all of them were naïve regarding the issues investigated, meaning that they were not linguistically biased. All test subjects were native speakers of German, except in the first experiment mentioned in the Introduction. In the experiments, we investigated the interpretation of German sentences in dialogue contexts. In the present paper we also provide English translations of the examples, although these were not provided during the experiments. We assume that the observed phenomena of modal enrichment also occur with the English translations; the experimental data, however, are only valid for the German examples.

2.2 Data We start with the original watch-example (1). Schmitz (2011) tested this example several times with different test groups: in all experiments the vast majority of subjects modally enriched the answer sentence. In the original example, “aber” (but) appeared as an adverb. The tendency towards a non-literal interpretation persisted if “aber” was changed into a conjunction between the two answer clauses (example 2a), if it was replaced with “und” (and, example (2b)) or if it was completely left out (example (2c)). The results of respective experiments are given in Table 2.²

58   

   Hans-Christian Schmitz and Bernhard Fisseni

exp. 2a (48 test subjects): exp. 2b (22 test subjects): exp. 2c (48 test subjects):

15 : 00

15 : 05

15 : 10

41 (85.4%) 18 (81.8%) 39 (81.2%)

6 (12.5%) 3 (13.6%) 8 (16.6%)

1 (2.1%) 1 (4.6%) 1 (2.1%)

‘?’

Table 2: What time is it according to B’s answer?

(2) A: Wie spät ist es? (What time is it?) a. B: Es ist 5 nach 3, aber meine Uhr geht 5 Minuten vor. (It’s 5 past 3, but my watch is 5 minutes fast.) b. B: Es ist 5 nach 3, und meine Uhr geht 5 Minuten vor. (It’s 5 past 3, and my watch is 5 minutes fast.) c. B: Es ist 5 nach 3. Meine Uhr geht 5 Minuten vor. (It’s 5 past 3. My watch is 5 minutes fast.) These experiments show, firstly, that the data are reliable and, secondly, that modal enrichment is not an effect of the junctor “aber” (but) alone. There is no significant difference in the interpretations of the examples (2a–2c). Nevertheless, we think that the occurrence of “aber” might still contribute to modal enrichment. We will discuss this, empirically not well-founded, hypothesis in section 3. The test subjects derived a non-literal interpretation of B’s answer. The question arises whether we have to assume the existence of a conventionalized operation of modal enrichment to properly derive this interpretation or whether we can just assume that the non-literal interpretation is the result of some non-conventionalized ‘modulation’ of the answer meaning. The idea behind modulation, as we understand it here, is that a given utterance is somehow – it does not matter much how – made compliant with conversational maxims.³ Of course, a conventionalized operation can serve to make an utterance compliant with conversational maxims, too. However, it exists independently of the maxims, and it can also be applied without reaching compliance. Thus, the assumption of such an operation becomes plausible if it can be shown that test subjects reliably apply it even in the interpretation of utterances that cannot be made compliant with the conversational maxims. To solve the issue, we let test subjects interpret the undoubtedly odd answers of examples (3).

Repairs for Reasoning   

   59

(3) A: Wie spät ist es? (What time is it?) a. B: Es ist 5 nach 3. Meine Uhr geht 3 Stunden vor. (It’s 5 past 3. My watch is 3 hours fast.) b. B: Es ist 5 nach 3. Meine Uhr geht 6 Stunden vor. (It’s 5 past 3. My watch is 6 hours fast.) c. B: Es ist 5 nach 3. Meine Uhr geht 9 Stunden vor. (It’s 5 past 3. My watch is 9 hours fast.) Assuming that a conventionalized operation of meaning enrichment exists, we expect that the subjects apply this operation and derive the interpretation that it is in fact five past 12, five past 9 or five past 6, respectively (15 : 05 – 3/6/9 hours). We expect a clear tendency towards a non-literal interpretation in each example, although the answer is not made compliant with the conversational maxims. (It still seems anomalous even if interpreted non-literally.) If the non-literal interpretation was derived by free modulation we would not expect such a clear result. Since the answer cannot be made fully adequate, we would expect the test subjects not to show a clear tendency towards non-literal interpretation. Rather, we expected them to interpret the answers literally, to reply that they just could not properly interpret them or to derive different non-literal interpretations. The results of the experiment are given in Table 3: as can be seen, there is a clear tendency towards the answers predicted by the hypothesis that modal enrichment is a conventionalized operation.⁴ Even ridiculous answers like that of example (4) are interpreted in a foreseeable manner: (4) Dialog im Oktober 2007 (dialogue taking place in October 2007): A: Was für ein Wochentag war der 11. Januar? (Which weekday was January 11th?) B: Der 11. Januar war ein Mittwoch, aber mein Kalender ist von 2005. (January 11th was a Wednesday, but my calendar is from 2005.)

exp. 3a: exp. 3b: exp. 3c:

– 3/6/9 hours

+ 3/6/9 hours

literal (15 : 00)

other

14 (82.4%) 14 (82.4%) 13 (76.5%)

1 (5.9%)

1 (5.9%) 2 (11.8%) 1 (5.9%)

1 (5.9%) 1 (5.9%) 3 (17.6%)

Table 3: What time is it according to B’s answer? (17 test subjects)

60   

   Hans-Christian Schmitz and Bernhard Fisseni

We asked test subjects which weekday January 11th was according to B’s answer.⁵ (The experiment took place in July 2007. In fact, January 11th 2007 was a Thursday, January 11th 2005 was a Tuesday.) A number of subjects’ first reaction upon being asked was to act amused, confused or show mild signs of aggression. Still, as can be seen from the results given in Table 4,⁶ they interpreted the example in accordance with the hypothesis that they would apply an operation of modal enrichment. (If January 11th was a Wednesday in 2005 then it must have been a Friday in 2007: two years have 730 days, 730 modulo 7 (a week) is 2. Thus, two days must be added. Two days after a Wednesday is a Friday.)⁷

exp. 4

– 2 days (Mon)

+ 2 days (Fri)

literal (Wed)

other

‘?’

7 (14.9%)

23 (48.9%)

8 (17.0%)

3 (6.4%)

6 (12.8%)

Table 4: Which weekday was Jan 11th according to B’s answer? (47 subjects)

Both experiments show that subjects have an operation of modal enrichment available and that they can foreseeably apply it even in the interpretation of odd examples.

2.3 Discussion Let us discuss the experiments and our interpretation of the results: an opponent might object that the results do not prove the existence of a conventionalized operation. It might still be that the subjects chose their interpretations by freely modulating the answer meanings in order to get the best interpretation possible. This objection, however, presupposes that there is a ‘best’ interpretation of the examples and, further, that the non-literal interpretation of (4), e.g., is ‘better’ than the literal one. We do not find this very convincing. It can still be objected that examples (3) and (4) are far from natural. We would not expect such answers in real world situations, and if we were confronted with them we would rather ask what the speakers means than just ‘understand’ him. Most probably, the subjects reacted as if the experiments were meant as tests not of linguistic examples but their intelligence. (Of course, we told them there were no wrong answers and that we were not testing them but the examples. However, even if they believed us they may still have tried to be ‘right’.) While this is true, it does not affect our interpretation. Even if the subjects reacted like in an intelligence test, they proved that they can perform modal enrichment. It does not matter for our argument that they performed this opera-

Repairs for Reasoning   

   61

tion in a non-natural setting since the original example (1) shows that they do not only perform it non-natural settings. (Reminder: the original example is a real world example.) The important point is that they performed this operation but they did not reach compliance with conversational maxims.⁸ Our opponent may concede that the original example is a real-world example, but he may insist that the experimental situation is not – not even for the first experiments. The interpretation of B’s answer takes place in a lab situation, without reference to an action context in which the interpreter for some reason has to know what time it is. It might be that in such a context where the information on the time is crucial, B’s answer is interpreted differently. When we first heard the original answer we needed to know the time – that is, we interpreted the answer in a context as named by our opponent – and still we understood it properly. So, at least we were able to enrich the answer ‘in the wild’. Nevertheless, it is a justified demand to test this with a broader group of subjects.

2.4 Robustness and commitment To meet the critique of our potential opponent, we performed the following experiment: we gave sheets with the text of example (5) to test subjects. We asked them to read the text and choose one of the options O-1 to O-6. (The numbering of the options from O-1 to O-6 was not on the original sheets, we added it here to be able to refer to each option individually.) (5) Anweisung: Lesen Sie bitte folgenden Text. Bitte wählen Sie dann die Fortsetzung, die Ihres Erachtens am besten passt: „Sie sind auf dem Weg zum Bahnhof, um einen Zug um 15:07 zu nehmen und haben den Bahnhof beinahe erreicht. Dieser Zug ist, wie Sie wissen, stets pünktlich; er verspätet sich nie. Eigentlich haben Sie einen Zeitpuffer eingeplant, aber Sie sind ein wenig ins Träumen geraten und wissen nicht, wie viel Uhr es ist und ob Sie Ihren Zug schon verpasst haben. Daher sprechen Sie einen Passanten an, der Ihnen antwortet: ‚Es ist 10 nach 3, meine Uhr geht aber 5 Minuten vor.‘ “ (Instruction: Please read the following text. After that, please choose the response which is, according to your intuition, most appropriate: “You are on your way to the train station to take a train at 3 : 07 and have nearly reached the station. The train is, as you know, always on time. It is never late. According to your plan, you should have a time buffer, but you started to daydream on the way and now, you are not sure what time it is and whether you have

62   

   Hans-Christian Schmitz and Bernhard Fisseni

already missed your train. Thus, you ask a passer-by, who answers: ‘It’s 3 : 10, but my watch is 5 minutes fast.’” Was tun Sie nun? (What do you do now?) – O-1: Sie bedanken sich bei dem freundlichen Passanten und gehen zügig los, damit Sie den Zug noch erreichen. (You thank the friendly passer-by and start to walk fast to reach the train.) – O-2: Sie bedanken sich bei dem freundlichen Passanten; anschließend gehen sie fluchend nach Hause. (You thank the friendly passer-by; then, you go home cursing.) – O-3: Bevor Sie zum Bahnhof gehen, versichern Sie sich noch einmal bei dem freundlichen Passanten, dass Sie noch genug Zeit haben. (Before you go to the station, you make sure that you have enough time, asking the friendly passer-by again.) – O-4: Sie rüsten sich innerlich für den Nachhauseweg, versichern sich aber noch einmal bei dem freundlichen Passanten. (You prepare to go home, but ask the friendly passer-by again.) – O-5: Sie fragen den freundlichen Passanten: „Entschuldigen Sie, wie viel Uhr ist es jetzt bitte tatsächlich?“ (You ask the friendly passer-by: ‘I’m sorry, what time is it now, really?’) – O-6: Sie fragen einen anderen Passanten, dessen Uhr vielleicht genauer geht. (You ask another passer-by, whose watch may be more exact.) The six options the subjects had to choose from can be (partially) ordered on a scale: 1. We assume that a subject chooses O-1 if he chooses a non-literal interpretation of the answer due to modal enrichment. 2. Option O-3 demands a non-literal interpretation, too. However, we assume that the subjects who choose this option are not completely sure whether the non-literal interpretation is right. 3. We assume that the subjects who choose options O-5 and O-6 are indecisive about which interpretation is right. 4. Option O-4 demands a preference for the literal interpretation. 5. Finally, option O-2 demands a clear decision in favour of the literal interpretation.

Repairs for Reasoning   

   63

We ran the experiment in two trials. The results of both trials are given in Table 5.⁹ As can be seen, even in the action context a very clear majority of test subjects chose the interpretation derived by modal enrichment.¹⁰

O-1

O-2

trial 1 (50 subj.) trial 2 (30 subj.)

44 (88%) 29 (96.7%)

1 (3.3%)

total (80 subj.)

73 (91.3%)

1 (1.3%)

O-3

O-4

O-5

2 (4%)

4 (8%)

2 (2.5%)

4 (5%)

O-6

Table 5: Answers of experiment 5 (two trials)

Let us take stock: the experiments show that recipients can perform operations of modal enrichment and that in our examples it can be foreseen that they do. Thus, a speaker who makes an utterance as in our examples can expect that his utterance will be interpreted non-literally. He can anticipate modal enrichment. Is the speaker even committed to the non-literal interpretation of his utterance? To answer this question we performed a further experiment. We constructed a situation in which the standard watch-answer is given (as we know, it can be expected that it will be interpreted non-literally). Then it turns out that the non-literal interpretation of this answer is false while the literal interpretation is right. In the example, the recipient (Felix) interprets the answer non-literally and then complains that he was misled. We asked the test subjects whether he is right in complaining and, thus, whether the speaker (Benedikt) is committed to the non-literal meaning of his answer. We presented the test-subjects with a two-sheet questionnaire. The text of example (6a) was on the first sheet. The subjects had to read the text and then interpret and note down what time it is according to Benedikt’s answer. After that, they had to read the text of example (6b) that was on the second sheet and they had to decide whether Felix’ complaint was justified. To this end, we gave them three options to choose from. (Again, the numbering of the options from O-1 to O-3 was not on the original sheets. We added it here to be able to refer to each option individually.) (6) a. Felix (F) und Benedikt (B) spazieren durch einen Park, der ungefähr fünf Minuten Fußweg vom Bahnhof entfernt ist, und unterhalten sich. Felix möchte einen Zug nehmen, der um Punkt 16 : 15 h vom Bahnhof abfährt. Der Zug ist stets pünktlich. Benedikt weiß davon.

64   

   Hans-Christian Schmitz and Bernhard Fisseni

F: Wie spät ist es? B: Es ist 10 nach 4. Meine Uhr geht 10 Minuten nach. (Felix (F) and Benedikt (B) are on a walk in a park, from which it is a walk of about five minutes to the station; they are having a conversation. Felix wants to take a train that starts at 4 : 15 sharp from the station. The train is always on time. Benedikt knows about this. F: What time is it? B: It’s 10 past 4. My watch is 10 minutes late.) Wie spät ist es der Antwort zufolge? (What time is it according to Benedikt’s answer?) b. Felix vermutet, den Zug nicht mehr erreichen zu können, und beschließt (durchaus verärgert) einen späteren Zug zu nehmen. Nach 5 Minuten kommen Felix und Benedikt an einer Uhr vorbei, die 16 : 15 h anzeigt. Felix beschwert sich bei Benedikt. F: Was redest Du denn für einen Quatsch! Es ist ja erst 4 Uhr 15! B: Warum? Ich habe vorhin gesagt, es sei 10 nach 4. (Felix assumes that he will not be able to reach the train and decides (quite annoyed) to take a later train. Ten minutes later, Felix and Benedikt pass a clock, which shows 4 : 15. Felix complains to Benedikt. F: You are talking complete nonsense! It’s only a quarter past 4! B: Why? I’ve said that it’s 10 past 4.) Beschwert sich Felix bei Benedikt zu Recht? (Is Felix justified in complaining to Benedikt?) – O-1: Benedikt hat Felix getäuscht. Felix musste ihn so verstehen, dass es bereits 20 nach 4 war. Felix ärgert sich zu Recht. (Benedikt has misled Felix. Felix had to understand Benedikt’s words to mean that it was already 20 past 4. Felix is justified in being angry.) – O-2: Benedikts Äußerung war vielleicht missverständlich, aber durchaus richtig. Felix hätte nachfragen können. (Benedikt’s statement might have been mistakable, but was in principle right. Felix could have asked.) – O-3: Benedikt hat die Uhrzeit korrekt genannt. Dass Felix ihn falsch verstanden hat, ist Felix’ Fehler.

Repairs for Reasoning   

   65

(Benedikt gave the time correctly. It is Felix’s mistake that he misunderstood him.) As in the previous experiment, the options can be ordered on a scale, namely from O-1 (misleading, fraud) via O-2 (mistakable, but in principle right) to O-3 (acceptable, correct). The results are given in Table 6. In each column, the numbers of subjects who chose a certain interpretation are given: 20% of the subjects interpreted Benedikt’s answer literally, 80% (74% + 6%) chose a non-literal interpretation. The 16 : 00-interpretations (6%) might have been due to modal enrichment and a miscalculation, that is, an error.¹¹ In the rows, the numbers of choices for one of the options O-1 to O-3 are given: according to 50% of the subjects, Benedikt was committed to the non-literal interpretation of his answer (O-1), only 4% found the answer completely acceptable (O-3). The second percentage within the brackets refers to the ratio of the subjects who evaluated Felix’ complaint in the respective way (row) and the subjects who chose the respective interpretation (column) – e.g., 67.6% of those who chose the non-literal interpretation 16:20h (column) found that Felix was right in complaining to Benedikt (row: O-1).

16:00

16:10

16:20

total

O-1 O-2 O-3

1 (2%, 33,3%) 2 (4%, 66.7%)

10 (20%, 100%)

25 (50%, 67.6%) 12 (24%, 32.4%)

25 (50%) 23 (46%)  2 (4%)

total

3 (6%)

10 (20%)

37 (74%)

Table 6: experiment 6 (50 test subjects)

Two thirds of those who chose the foreseeable non-literal interpretation 16 : 20 agreed with the complaint to Benedikt and made him responsible for the misunderstanding; and even those subjects who interpreted the answer literally found it mistakable. We are not surprised that those who interpreted the answer as meaning that it was 16 : 00 – most probably due to a calculation error – refrained from complaining (or agreeing to a complaint). The data show that under circumstances a speaker not only can foresee a non-literal interpretation but must foresee it and accommodate his utterance accordingly. He can be accused of noncooperativness if he does not; it is not just on the recipient to assure that the right non-literal interpretation is chosen.

66   

   Hans-Christian Schmitz and Bernhard Fisseni

2.5 Enrichment beyond the watch Until now, we have shown very clear data on modal enrichment for the intial watch-example and variations of it. One might suspect that this is a singular example insofar that only this example or examples of this very particular type demand modal enrichment. To prove that this is not the case, Schmitz (2011) has already provided examples. Further examples for modal enrichment, not reported so far, include (7), among others. (In contrast to the intial watch-examples, these are not real-world examples, but have been made up.) (7) a. A: Was wird Peter wohl sagen, wann er kommt? (When does Peter say he will arrive?) B: Er kommt um 3 Uhr, er verspätet sich aber aus Prinzip immer um eine Stunde. (He will arrive at three o’clock, but as a matter of principle he will be one hour late.) b. A: Wann kommt eigentlich Peter? (By the way, when will Peter arrive?) B: Er kommt um 3 Uhr, er verspätet sich wie immer aber um eine Stunde. (He will arrive at three o’clock, but as always he will be one hour late.) We presented these example to test subjects and asked them when they thought Peter would arrive. The results are given in Table 7: the clear majority expected Peter to arrive at 4 o’clock, although literally understood, B says that he will arrive at 3 o’clock.¹² We asked subjects how they arrived at their expectations. Those who expected Peter to arrive at 4 o’clock replied that they interpreted B’s answers in the sense of ‘Peter will say that he will arrive at 3 o’clock, but as a matter of principle/as always he will be one hour late’. They concluded that in fact Peter will arrive at 4 o’clock. That is, they performed an operation of modal enrichment by deriving a modal operator ‘Peter said that’ and applying it to the first answer clause. There is a peculiarity in example (7a): A asks for the content of Peter’s utterance rather than for his time of arrival. Conventionally, however, the latter is the underlying purpose of asking the question.¹³ And, as (7b) shows, we can also omit the reference to speaking with Peter and virtually get the same interpretations.

exp. 7a (17 subjects): exp. 7b (10 subjects)

15 h

16 h

2 (11.8%) 3 (30%)

15 (88.2%)  7 (70%)

Table 7: When do you think Peter will be here?

Repairs for Reasoning   

   67

A final example that proves that various forms of modal enrichment – that is, enrichment with various form of modal operators – are possible, is (8). We did not collect experimental data on the interpretation of this example but leave it to the intuitions of the reader. (We would expect to pay 5 €.) (8) A: Wieviel kostet das Buch? (How much is the book?) B: 10 €, aber es gibt 50 % Rabatt. (10 €, but there is a discount of 50%.)

2.6 Examples without enrichment We have seen that subjects perform operations of meaning enrichment with various examples. The following experiment raises the issue whether modal enrichment is really a kind of ‘repair’, that is, an operation that is not applied by default but only if something goes wrong: (9) A: Wie spät ist es? (What time is it?) B: points to a clock according to which it is 15:10h. Es ist 5 nach 3, meine Uhr geht aber 5 Minuten vor. (It’s 5 past 3, but my clock¹⁴ is 5 minutes fast.) Example (9) is identical to the initial watch-example except that we add that B points to a clock that displays 15:10. We drew a picture of such a clock on a blackboard and pointed to it while reading the example to the test subjects. Again the subjects had to indicate what time it is according to B’s answer. We now expected that most subjects would interpret the answer literally and take its second clause to be an explanation why B’s answer differs from the display of the clock. The results are given in Table 8: in fact, most of the subjects interpreted the answer literally, but also nearly half of them did not (though, only one test person came to the interpretation expected if modal enrichment were demanded). That is, the results are not very clear, and it should make us wonder why so many test subjects still chose a non-literal interpretation.

exp. 9:

15 : 00

15 : 05

15 : 10

15 : 15

‘?’

1 (7.1%)

8 (57.1%)

2 (14.3%)

1 (7.1%)

2 (14.3%)

Table 8: What time is it according to B’s answer? (14 test subjects and native speakers of German)

68   

   Hans-Christian Schmitz and Bernhard Fisseni

Nevertheless, modal enrichment does not ‘always’ take place. Schmitz  (2011) proves that, firstly, the replacement of the conjunctions – so far: “aber” (but) or “und” (and) – can change the interpretations significantly: the vast majority of test subjects interpreted example (10a) literally, and subjects were undecided about the interpretation of example (10b) – about half of them interpreted it literally and about half expanded its meaning. Secondly, a strong accent on “ist” (is) like in example (10c) triggers the literal interpretation. Some might explain this as an effect of a so-called ‘verum-focus’. Thirdly, by inserting “tatsächlich” (in fact) modal enrichment is blocked, so that example (10d) is interpreted literally. (10) a. Es ist 5 nach 3, obwohl meine Uhr Minuten vorgeht. (It’s 5 past 3, although my watch is 5 minutes fast.) b. Es ist 5 nach 3, denn meine Uhr geht 5 Minuten vor. (It’s 5 past 3, because my watch is 5 minutes fast.) c. Es IST 5 nach 3, meine Uhr geht aber 5 Minuten vor. (It IS 5 past 3, but my watch is 5 minutes fast.) d. Tatsächlich ist es 5 nach 3, meine Uhr geht aber 5 Minuten vor. (In fact, it’s 5 past 3, but my watch is 5 minutes fast.) The fourth – and for our concern here: most interesting – way of avoiding modal enrichment is to change the order of the answer clauses. Example (11a) is a modification of the initial watch-example which we already discussed above (cf. (2c)); example (11b) is derived from (11a) by changing the order of the answer clauses: (11) A: Wie spät ist es? (What time is it?) a. B: Es ist 5 nach 3. Meine Uhr geht 5 Minuten vor. (It’s 5 past 3. My watch is 5 minutes fast.) b. B: Meine Uhr geht 5 Minuten vor. Es ist 5 nach 3. (My watch is 5 minutes fast. It’s 5 past 3.) We performed an experiment by letting two groups of test subjects interpret the examples and note down what time it is according to the answers provided. (Each test group interpreted only one example.) The results are given in Table 9: while example (11a) was, as we have already seen, mainly interpreted non-literally, (11b) was interpreted literally by most subjects. The results exhibit a significant

Repairs for Reasoning   

   69

difference between the examples which can be traced back to the order of the answer clauses – the two-sided Fisher-test for exact data yields a p-value of NoLexMvt). This interaction will mainly be relevant for unergatives (see below for cases where verb movement applies to keep adverbials in focus). The following table shows the competition for an unaccusative verb. Here, the candidate (73c) without verb movement wins; verb movement leads to a violation of NoLexMvt without having any repair function and is therefore blocked:³³

*! *!

*

Repair-driven verb movement in English locative inversion   

   185

S V  Loc

? b. [tp Loc [asp V [vp SU tv tLoc]]]

Loc V S

  c.  [tp Loc [asp [vp SU V tLoc]]]

Loc S V

*!

  d. [top Loc [tp SU [asp [vp tsu V tLoc]]]]

Loc S V

*!

  e. [top Loc [tp e [asp [vp SU V tLoc]]]]

Loc S V

No  Lex Mvt

  a.  [tp SU [asp [vp tsu V Loc]]]

Subject Case

(74) Loc = top; SU/Ag = foc

Align Focus

EPP

With unergative verbs, however, the candidate with verb movement wins:³⁴

*

*

*!

*!

*

*

*

Here, a violation of NoLexMvt is tolerated because it helps satisfy AlignFocus. In section 6.3 I argued that movement of unaccusative verbs sometimes takes place, namely in those cases where the verb precedes an adverb that marks the left boundary of VP. The question is what triggers verb movement in this case. I repeat example (55) for convenience: (75) And as contemporary literary critics (among others) have shown, [with interpretation] comes often uncomfortable complexity. Little is known about the function of such constructions. I will provisionally propose that verb movement takes place so that the adverb can be in focus. Clearly, since the adverb is not clause-final, such examples will always incur at least one violation of AlignFocus. This is where the gradient nature of AlignFocus becomes relevant: Without verb movement there would be two violations of AlignFocus. Therefore, it is still preferred to move the verb and thereby violate NoLexMvt to prevent a second violation of AlignFocus (in the previous tableaux I have ignored multiple violations of AlignFocus):³⁵, ³⁶

  a. [tp SU [asp Adv [vp V [sc tSU Loc]]]]

S Adv V Loc

Subject Case No  Lex Mvt

(76) Loc = top; SU/th = foc; Adv = foc

Align Focus

   Martin Salzmann

EPP

186   

**!***

? b. [tp Loc [aspV Adv [vp tV [sc SU tLoc]]]] Loc V Adv S

*

*

  c. [tp Loc [asp Adv [vp V [sc SU tLoc]]]] Loc Adv V S

**!

*

*

8.4 What the approach captures The major advantages of the present account are the following: First, the generalized TP analysis helps accommodate (possibly non-locative) adjuncts. Second, the fact that the EPP is taken to hold explains the impossibility of *Died John (cf. (45)). Third, verb movement accounts for the clause-final position of unergative subjects in LI. The approach also captures the subject and topic properties found in the construction; I will discuss these in more detail since they require some explanation.

8.4.1 Capturing the topic properties Since topics are assumed to occupy Spec, TP, some of the facts can no longer be related to the structural position of topics (i.e. TopP/adjunction to TP), i.e. for instance, one can no longer rule out LI in certain contexts because the relevant structural position is lacking. This is unproblematic for the semantic properties (2.2.1) and also for the topic island effects (2.2.2) since intervention does not require a distinct structural position. But in the case of non-finite contexts (2.2.3), this becomes problematic: If the locative is a topic, nothing should prevent it from moving to Spec, TP. Fortunately, the non-applicability of LI in non-finite clauses (but recall fn. 6) can be subsumed under a more general restriction on root transformations. Even though the precise semantic conditions are still unclear (cf. Heycock 2006 for an overview), root transformations are generally said to be ruled out in non-assertive clauses (Hooper and Thomson 1973). Since non-finite clauses are non-assertive, the impossibility of LI is expected; the constraint TopicFirst simply cannot be satisfied in non-finite clauses.³⁷

Repair-driven verb movement in English locative inversion   

   187

8.4.2 Capturing the subject properties Apparent raising to subject can be reanalyzed as long-distance topicalization of the locative to the matrix Spec, TP. But the question is how exactly such sentences are derived, especially because it is not so obvious what happens to the EPP of the embedded T. There are several proposals that have argued against an EPP in non-finite clauses, cf. e.g. Grohmann et al. (2000) and Boškovic (2002), so that the problem would not arise in the first place. This would certainly work for my purposes, but in case one wants to uphold the EPP in non-finite clauses for independent reasons, it would have to be checked by the locative on its way to matrix Spec, TP – not by virtue of its being a topic, but because it is an accessible goal. This requires the locative to be closer to the intermediate Spec, TP than the subject (or at least equally close). If closeness simply depends on c-command, as in more recent Agree-based approaches, the locative will have to move to Spec, vP of the embedded clause to be closer to Spec, TP than the subject. The derivation of a raising example thus looks as follows (depending on one’s assumptions, there may be an additional movement step via the edge of the vP hosting be):³⁸ (77) [tp [On that hill]1 appears [tp __ 1 to be [vP __ 1 located a cathedral __ 1]]]. A similar derivation obtains if the locative just undergoes topicalization to a posi-tion above the subject, as in the following example with LI based on unergatives where the locative originates in a non-finite clause (Culicover and Levine 2001: 300): (78) ([From this pulpit]1 we heard preach a close associate of Cotton Mather __ 1]. Successive-cyclic movement of the locative via Spec, vP and Spec, TP will be necessary as well: (79) [From this pulpit]1 we heard [tp __ 1 asp+preach [vP __ 1 tv a close associate of Cotton M. __ 1 ]]. As for that-trace effects, given that I do not analyze inverted locatives as subjects, they can no longer be related to a ban on subject extraction across that. This is not necessarily a drawback since it is no longer clear whether that-trace effects should be analyzed as effects that depend on the grammatical relation of the extractee. There is growing evidence that that-trace effects have a different source. Some have argued that there simply is a constraint to avoid adjacency between that and the finite verb (Kathol and Levine 1993: 213, cf. also Postal 2004,

188   

   Martin Salzmann

Bruening 2010: 52). This squares nicely with the observation that LI (like subject extraction) is subject to the adjacency effect, i.e. that that-trace effects can be ameliorated if an adverbial separates complementizer and finite verb: (80) [On which table] were you wondering {whether/if} *(under certain circumstances) might have been put the books that you had bought (Culicover 1993: 98, 104) A similar proposal has recently been made by Salzmann et al. (to appear): We provide empirical evidence from German that the sequence complementizerfinite verb is degraded even if no extraction is involved. We then reduce thattrace effects to a phonological EPP that requires Spec, TP to be filled with phonetic material at PF. A related proposal is put forward in Bayer (2005) and Bayer and Salzmann (this volume), according to whom long wh-extraction is blocked if it takes place from a high topic position (i.e. the position of the sentence topic). Both of these alternatives account for inacceptable instances of locative preposing as in (4) and (80) (without the adverbial): There is either long A’-movement from a topic position (the locative moves via the embedded Spec, TP, which counts as a topic position) or the embedded Spec, TP remains unfilled at PF. Furthermore, these approaches correctly predict extraction to be possible once some other element occupies the subject/topic position as in (80). The derivation of (80) is actually quite interesting: It does involve LI in the embedded clause, but it is not the locative that eventually ends up in the matrix Spec, CP position that undergoes inversion, but rather the sentential adverb. The locative thus extracts from a lower position (and then undergoes successive-cyclic movement): (81) [cp [On which table]1 were you wondering [cp __ 1 {whether/if} [tp under certain circumstances might have been [vP put the books that you had bought __ 1]]]] The extracted locative does not function as a sentence topic and is therefore free to extract. Furthermore, the embedded Spec, TP is overtly filled at PF. This example implies that there is extraction from a locative inversion. The attentive reader will have noticed that this seems to lead to a paradoxical situation given that LI was assumed to erect a topic island (cf. 2.2.2). There is fortunately independent evidence that helps resolve the paradox: Culicover (1993: 98–100, fn. 2) points out that topic islands only obtain with fronted elements, but not with basegenerated sentential adverbs (as in (80)):³⁹

Repair-driven verb movement in English locative inversion   

(82) a.

   189

This is the tree [which]1 [just yesterday] I had tried to dig up __ 1 with my shovel.

b. * What did [to Lee]1 Robin __ 1 give? What remains to be explained is how long-distance A’-movement proceeds in the absence of an overt complementizer; the following pair illustrates this for both unaccusatives and unergatives (Culicover and Levine 2001: 285, 302): (83) [Into which room]1 does Terry claim walked that bunch of gorillas __ 1 ? (84) [From this pulpit]1 I believe preached a series of ravenous Tolstoy scholars __ 1. Given the two alternative explanations of that-trace effects proposed above, such examples appear problematic because there is either long A’-movement of a topic (there is no other element in the embedded clause that could assume that function) or because Spec, TP of the embedded clause remains unfilled. Here we follow Bayer (2005: 245–246) and Bayer and Salzmann (this volume) who propose that what looks like long subject extraction is actually just short extraction with the do-you-think-part functioning as a parenthetical (in the previous examples, the parentheticals would be does Terry claim and I believe). The crucial evidence for this reanalysis comes from cases where the ‘do you think’-part is incompatible with the phrase structure of the sentence (for more evidence, cf. the sources mentioned above):⁴⁰ (85) Who could do you think challenge his version of the accident? (http://query.nytimes.com/mem/archive-free/pdf?res=F10811FD3D5415738 DDDA10894DD405B8485F0D3) The final subject property that needs to be explained is the absence of do-support in questions: (86) [On which wall] hung a portrait of the artist? The key to an understanding of the absence of do-support in these examples is the constraint Op-Spec (Grimshaw 1997: 388), which states that an operator must occupy a specifier position, but not necessarily Spec, CP. For subject wh-questions, fronting the subject to Spec, TP is sufficient to satisfy both the EPP and OpSpec. In LI, AlignFocus forces the subject to stay low. Consequently, a locative wh-phrase

190   

   Martin Salzmann

can land in Spec, TP, thereby satisfying the EPP and OpSpec. Structural economy will then prevent do-support and projection of an additional specifier:⁴¹ (87) [tp [On which wall]1 [vp hung a portrait of the artist __ 1]]?

8.5 Extension to other inversion constructions Let me briefly discuss how this approach to inversion can be extended to other inversion constructions of English (given the complexity of these constructions and the subtle differences between them, what follows should by no means be considered a full-fledged account). Culicover and Winkler (2008b: 629) argue that in Comparative Inversion, the EPP can be violated to satisfy AlignFocus (note that this inversion construction involves contrastive focus, not presentational focus). They argue that the subject position remains empty in examples like the following because the subject is preceded by a non-finite auxiliary: (88) They argue that they produced more readable and better researched reviews and editorials [than __ could have the academics under whose NAMES the papers appeared]. This would conflict with the ranking proposed above where EPP outranks AlignFocus and the data in (45) where a post-verbal subject is not licensed in the absence of material in Spec, TP. However, if – as is usually assumed – there is movement of an empty operator to the spec of than, it is not implausible that it moves there via Spec, TP if the subject is forced to stay low for reasons of focus (VP-ellipsis ensures that the subject is clause-final and makes verb movement across the subject unnecessary; intermediate movement to Spec, vP ensures that the operator is close enough to Spec, TP): (89) [cp Op1 than [tp __ 1 could have [vP __ 1 [vP the academics [vp produced __ 1]]]]] The EPP would thus be satisfied derivationally.⁴² Quotative inversion is similar: The quote/the quote operator moves (via Spec, vP) to Spec, TP (and possibly continues to Spec, CP, cf. Collins 1997: 38) because the subject is forced to stay low. The verb moves to Asp (cf. Suñer 2000: 560) so that the subject is in focus:⁴³

Repair-driven verb movement in English locative inversion   

   191

(90) a. “I’m freezing”, complained John. b. [cp Quote-Op1 [tp __ 1 asp + complained2 [vP __ 1 John t2 __ 1]]] In both cases, the EPP-checking is a side-effect of fronting triggered by A’-related features (e.g. OpSpec). There also seem to be instances of inversion where the fronting is purely EPP-driven, viz. there-insertion and (some cases of) predicate inversion: (91) a. There came a man into the room. b. An excellent doctor is Brian. (den Dikken 2006: 153) If as in Hoekstra and Mulder (1990), den Dikken (2006), Broekhuis (2008) and Hartmann (2008) there is analyzed as a predicate, the two constructions can be unified. The theme/subject and there/the predicate are base-generated in a small clause so that both are equidistant from the subject position and both can in principle move (under the assumption, contrary to den Dikken 2006, that the small clause is not a phase). The inverted structure obtains if the subject has to be in focus. Fronting of there/the predicate is purely EPP-driven – topicality cannot be at stake (at least in those cases of predicate inversion that do not involve a definite noun phrase). Among other things, this accounts for the fact that these types of inversion can occur in non-finite clauses and that aux-inversion is possible, cf. Hartmann (2008: 146–147), den Dikken (2006: 98–102).⁴⁴, ⁴⁵

8.6 Why an approach based on inviolable constraints fails Let me briefly explain why I think that the present approach is superior to conceivable alternatives that do not employ violable constraints. For instance, locative inversion could be handled by a specialized T-head that is endowed with a feature that triggers verb movement. To restrict such heads to locative inversion, they would need another feature that restricts its specifier to the elements occurring in fronted position in LI. Given the non-locative adjunct data in 3.2, a feature [+locative] will not be sufficient. But quite apart from this problem (which the present account is also confronted with to some extent, cf. 9.1 below) and from the fact that verb movement probably does not target T, it remains unsatisfactory that the repair character of verb movement can no longer be expressed. Verb movement appears as arbitrary, the link to the information

192   

   Martin Salzmann

structural function of LI and the fact that there is an entire class of inversion constructions in English is completely lost. One could imagine an approach based on inviolable constraints that takes information structure into account: TP would be a syncretic category hosting case and top features (Zubizarreta 1998). Subject externalization would be taken to be optional, e.g. by means of an optional EPP feature. In the presence of such a feature, we would get the non-inverted structure (SVLoc), in the absence of such a feature inversion would obtain. The trigger for verb movement in the case of unergative verbs would be more difficult to find. One could argue that Procrastinate, which is an economy constraint, can be overruled by an inviolable interface constraint, i.e. AlignFocus. In the absence of verb movement the derivation would then crash due to a violation of AlignFocus. The problem of inviolable constraints becomes visible, however, once there are conflicting requirements, namely when a subject is focused and there is no other element that can satisfy the EPP. In that case, inversion is impossible (Broekhuis 2008: 271): (92) a. * Died John. b.

John died.

(92a) satisfies AlignFocus, but violates the EPP; (92b), however violates AlignFocus, but satisfies the EPP. Since constraints cannot be ranked in a model based on inviolable constraints, such examples cannot be explained (cf. also Broekhuis 2008: 271 for discussion). Rather, constraints must be violable and ranked. Under the ranking EPP >> AlignFocus, the correct result can be derived quite easily:⁴⁶ (93)

SU/Agent = foc

EPP Align Focus

  a. [tp e V [vp SU tV]]

VS

? b. [tp SU [vp tsu V]]

SV

*! *

Subject Case

NoLexMvt

*

*

Repair-driven verb movement in English locative inversion   

   193

9 Open issues 9.1 Overgeneration With the current constraint ranking, the approach overgenerates: Since all that is required to occupy Spec, TP is being a topic of some sort, nothing rules out inversion with objects: (94) a. * That book likes John. b. * John pleases this book. Since AlignFocus keeps the subject low, the object is free to satisfy the EPP (provided it is topical). But the result is ungrammatical. In fact, the overgeneration problem also obtains in the alternative approaches reviewed in section 4 since these also involve topicalization. Following Birner (1994, 1995), Levin-RappaportHovav (1995), Chen (2003) I argue that the restrictions follow from the discourse function of LI and should not be encoded syntactically. LI is always claimed to be used for presentational focus (Bresnan 1994), i.e. used for the introduction of a new referent on the scene (which may then become the topic of the next clause). This function limits both the class of elements occurring in Spec, TP as well as the types of verbs: Since a new referent is introduced on a scene, one needs an element to set the stage. This is preferably done by stage topics, which in turn are prototypically realized by locatives or temporal expressions (cf. also Landau 2010: 119). Given that the function of LI is to introduce a new referent, LI prefers those verbs that do not detract from the newness of the new referent. Birner (1995) calls these verbs “informationally light”. This will, of course, favor certain unaccusatives like verbs of existence or appearance and disfavor externally caused verbs of change of state. Furthermore, this explains the virtual absence of transitive verbs from LI because they are normally not informationally light and because it is usually their direct object that introduces new information.⁴⁷ As for unergatives, Levin-Rappaport-Hovav (1995: 251–260) show that they are most acceptable in LI if the verb is informationally light in this context. This is the case if verb and postverbal DP are mutually predictable, for instance if “the activity or the process that the verb describes is characteristic of the entity the verb is predicated of”. This is quite obvious for the cases in (20) of which I repeat one example for convenience: (95) [On the folds of his spotless white clothing, above his left breast], glittered an enormous jewel. (N. Lofts, Silver Nutmeg, 460)

194   

   Martin Salzmann

To glitter is a characteristic of jewels, and in this combination the verb is informationally light (but see Landau 2010: 122–123 for critical discussion). Admittedly, this may not be the entire story: It is not always fully clear what is meant by presentational and whether all instances of LI can be reduced to this function: Green (1980) gives a large list of functions of inversion, Birner (1994) summarizes the function of LI as “linking relatively unfamiliar information to the prior context via the clause-initial placement of information which is relatively familiar”, a generalization that is criticized by Chen (2003: 11–32).⁴⁸ Related to this is the question whether all fronted elements are really stage topics. For the majority of cases, this is arguably correct, but for cases like (24d/f) and instances where the fronted element undergoes wh-movement (6a), this is at least questionable. Since these issues are essentially orthogonal to the goals of this paper, I will not pursue them any further here.

9.2 Verb movement and particle verbs Marcel den Dikken (p.c.) has drawn my attention to the fact that an account in terms of verb movement is problematic in the light of data involving particle verbs. There are examples of both QI and LI where the focal subject follows the particle: (96) a. “Don’t drop the bricks”, shouted out Trudy to Carl. (Collins and Branigan 1997: 5) b. In the room danced around a group of freaky trolls. (example modeled after den Dikken’s examples)

QI LI

Given that the particles normally do not incorporate into verbs in English, one would expect the subject to precede the particle with only the verb moving across it, contrary to fact. At least on the surface, the LI example seems to suggest an extraposition account (while the data in (96a) with the PP-object in its baseposition show that this cannot be correct for QI). But since extraposition of unergative and transitive subjects is normally taken to be impossible, this seems to be an undesirable solution. However, under the repair-perspective taken here, one could imagine an analysis where extraposition is exceptionally possible to satisfy AlignFocus, i.e. one would be dealing with repair-driven extraposition that is exceptionally possible because it would help satisfy the higher-ranking constraint AlignFocus. Examples like (38) with extraposition from the subject position could still be ruled out since EPP outranks AlignFocus. More gener-

Repair-driven verb movement in English locative inversion   

   195

ally, repair-driven extraposition could thus replace repair-driven verb movement in the analysis of LI. I will not pursue this option here because i) extraposition generally does not work for QI (recall 6.2), ii) because unergative subjects do not always satisfy the prosodic conditions for extraposition (recall 4.2.2) and iii) because it seems unattractive to come up with two completely different solutions for the two inversion constructions. Rather, I will tentatively assume that if the repair approach advocated here is pushed to the limit, one could imagine particle incorporation taking place exceptionally in LI/QI to guarantee that the subject remains in focus; this would be possible if AlignFocus outranks a high-ranked constraint banning incorporation. This would be another case of repair-driven movement, but in this case it would be an instance of non-feature-driven head movement and thus the same kind of repair-driven movement as in Heck and Müller (2000, 2007). Incorporation of the particle is in fact the solution offered in Collins and Branigan (1997). In their account, incorporation has to be stipulated while in the present account it can be made to follow from a more general principle. The viability of an incorporation approach depends on whether the sequence verb + particle can be interrupted by an adverb. Collins and Branigan (1997: 5) claim that it cannot, but mention that such examples are judged grammatical by a reviewer. Roberts (2010: chapter 4.1.3) also claims that the sequence verbparticle can marginally be broken up. Since a full investigation of these empirical facts is beyond the scope of this paper, I have to leave this for future research.

10 Conclusion In this paper I have argued that the analysis of English locative inversion requires a reassessment of important aspects of English syntax. The hitherto little-noticed fact that LI is possible with (possibly non-locative) adjuncts has lead to the reanalysis of the subject position Spec, TP as a generalized TP, a position that is not only targeted by subjects but can also host topics. In LI, Spec, TP can be filled by a non-subject because the subject is independently forced to stay low for reasons of focus. Adjuncts could also be accommodated if they do not target the subject position and the EPP is satisfied either by a null subject or derivationally before subject extraposition as in some alternative approaches. However, it is shown that such alternatives cannot account for LI with unergative verbs. I have argued that the correct surface order can be obtained in these cases if the verb moves across the unergative subject. This verb movement is implemented under a repair-perspective, i.e. it can exceptionally take place to avoid the violation of an even more important constraint. In the case at hand, verb movement occurs

196   

   Martin Salzmann

to allow the subject to remain in focus. This is implemented by an Optimalitytheoretic analysis where AlignFocus outranks NoLexMvt so that a violation of the ban against overt movement of the lexical verb is exceptionally tolerated. In contrast with verb-movement approaches to similar inversion constructions, verb movement does not have to be stipulated in the present account, but follows from the interaction of general, independently motivated constraints.

Notes 1 Earlier versions of this work were presented at the TIN-dag in Utrecht (February 2008), at the University of Konstanz (June 2008), at the GGS in Berlin (May 2008), at the University of Tübingen (July 2008), at the LAGB meeting in Essex (September 2008) and at the repair workshop at the DGfS Osnabrück (February 2009). I am very grateful to the audiences for comments that have lead to a significant improvement of my ideas, in particular: Patrick Brandt, Josef Bayer, Hans Broekhuis, Marcel den Dikken, Eric Fuß, Hans-Martin Gärtner, Jutta Hartmann, Sara Holler, Gereon Müller, Arthur Stepanov, Volker Struckmeier, Ralf Vogel, Susanne Winkler. Additionally, the paper has benefited from detailed comments by the anonymous reviewers and by Patrick Brandt. This work has been supported by a grant from the Swiss national Science Foundation, Nr. PBSK1--119747/1. 2 In the literature one can find additional (alleged) evidence for subjecthood, but many of the diagnostics turn out to be inconclusive, cf. Postal (2004), Bruening (2010) for discussion. As we will see in 8.4.2 below, the diagnostics discussed in this section can also be reanalyzed. 3 A similar effect is found when the preverbal locative is wh-extracted. As pointed out in Chung and Kim (2002: 157, fn. 17), non D-linked wh-locatives are not felicitous (as opposed to the D-linked one in 2.1.3 above): (i) ?? On how many walls hung a portrait of the founder’s family member? 4 Once conclusion is referential, LI becomes possible again: (i) To the same conclusion came those who investigated and analyzed the great work of the Chinese emperor Wu. (www.ufoarea.com/aas_chinesediscovered.html) 5 Importantly, as the example above shows, the preverbal locative blocks extraction of any constituent not just of the postverbal theme. For the latter it has often been claimed that the ban against extraction is related to the fact that it is presentationally focused: Questioning such constituents would lead to pragmatic incoherence, cf. Bresnan (1994: 87–88), den Dikken (2006: 125–126). LI would thus be parallel to presentational-there, where one finds the same effect, cf. Hartmann (2008: 145). The example in the text, however, shows that extraction is independently blocked, which follows if the preverbal locative creates a topic island. I will come back to this issue in 8.4 below. 6 Interestingly, Branigan (2000: 554) gives the following as fully grammatical: (i) The photos [vp showed [tp behind this very hedge to have been hiding Jill and Tonyi] during each otheri’s trials]. This example is additionally supposed to show that covert movement (of the relevant features of the postverbal subject) can lead to new binding relationships. I do not know what causes the

Repair-driven verb movement in English locative inversion   

   197

disagreement between Bresnan (1994) and Branigan (2000) and will continue to assume that LI is impossible in non-finite clauses. Furthermore, Culicover and Levine (2001: 298, ex 26; 300, ex. 31) give a few examples where inversion occurs in gerundives and non-finite complements of perception verbs: (ii) We heard from this pulpit preach [a close associate of Cotton Mather]. Importantly, on their account, what one is dealing with here is heavy inversion, i.e. subject extraposition, cf. 4.2 below. Consequently, according to them, the locatives do not have to be analyzed as occupying Spec, TP, unlike in (i) where an extraposition analysis is unavailable. But since the distinction between light and heavy inversion will be questioned in 4.2.2, data as in (ii) may eventually be equivalent to (i). 7 Den Dikken (2006: 100–102) proposes that the locative is base-generated in topic position and that what moves is a pro-PP. Base-generation may be problematic given the following reconstruction data: (i) Beside each otheri sat two handsome young boysi. (Chung and Kim 2002: 150) See also Broekhuis (2008: 296–297) for further critical remarks on the base-generation analysis. 8 There are two competing definitions of equidistance found in the literature, the one found in chapter 3 of Chomsky (1995) and the newer one in chapter 4 of Chomsky (1995). The crucial difference is that in the earlier definition equidistance holds if the target position and the possible intervener are within the same minimal domain (adapted): β is closer to τ than α unless β is in the same minimal domain as τ. This definition is insufficient for LI because the theme c-commands the locative and is not in the same minimal domain as the target position Spec, TP. Consequently, it should prevent the locative from moving to Spec, TP. Under the newer definition, equidistance also obtains if the goal and the potential intervener are within the same minimal domain (adapted): β is closer to τ than α unless β is in the same minimal domain as a) τ or b) α. This definition works for the structures in (18) since in both cases theme and locative are within the projection of the same head. 9 This is not quite correct for the small-clause approach where it is sufficient that the locatives can be interpreted as predicates. I will come back to this in section 3.3. 10 The derivations in den Dikken (2006) and Broekhuis (2008) are somewhat more complex. In den Dikken the small clause is taken to be a relator phrase: [rp Subject [Rel [Predicate]]]. Since the RP is taken to be a phase, the predicate is blocked from moving since it is not part of the edge domain of RP. Den Dikken (2006: 113–115) proposes that the head of the relator phrase, i.e. the relator, incorporates into a linker head F above the RP. This not only extends the phase up to FP, it also leads to equidistance between the subject and the predicate (movement of the relator extends the minimal domain so that it includes both the landing site and the potential intervener, viz. the subject, den Dikken adopts the older definition of equidistance). As a consequence, the predicate can move across the subject: (i) [fp predicate j [F + relatori [rp DP ti tj ]]] Importantly, in den Dikken’s approach, the inverted and the non-inverted structure do not compete, since they are based on different numerations. The non-inverted structure does not involve an empty pro-predicate but rather a normal locative. In Broekhuis (2008: 275, 278–281) the subject of the small clause is moved to Spec, VP (short object shift). For reasons of equidistance (Broekhuis adopts the newer definition of equidistance), the predicate itself cannot front in LI because it is not within the same minimal domain as the shifted theme. Rather, the entire remnant small clause has to front: (ii) [tp [sc ti Pred]j T [vP v + V [vp DPi tv tj]]]

198   

   Martin Salzmann

11 See Green (1982) for an overview over the uses of the different inversion constructions. LI is indeed more frequent in the literary or scholarly language, but as she points out, this is surely (at least in part) related to the various functions of locative inversion that make it particularly suited for such genres. 12 This, of course, depends on one’s definition of equidistance. The subject in principle blocks movement of the locative under both definitions in fn. 8, because the two are not in the same minimal domain. However, this crucially depends on whether domain-extending head movement is adopted or not. In the earlier definition of equidistance, domain-extending movement is normally adopted while the newer version tries to derive locality effects without it. With unergative verbs, equidistance only obtains if both the newer version of equidistance and domain-extending head movement are adopted: In that case, V-to-v-movement, would extend the minimal domain to include subject and locative so that both would be equidistant from Spec, TP. LI with unergative verbs might thus argue for a new definition of equidistance. There are also argument-structure-based accounts that have modified Bresnans’s (1994) original proposal to the effect that LI with unergatives can also be derived, cf. e.g. Demuth and Mmusi (1997); interestingly, these modifications were mostly motivated by LI in Bantu languages where there is no general restriction to unaccusatives, cf. Salzmann (2011) for an overview. Note that I have presupposed in the previous reasoning that locatives occurring with unergatives are arguments generated within the VP; however, as will be shown in the following section, there is good reason to believe that they are in fact adjuncts. If this turns out to be correct, it is quite probable that at least some of them are projected outside VP (perhaps even higher than the subject). In that case, there would be no equidistance problem even without domain-extending head-movement; however, as the next subsection will show, the adjunct status of the locative will affect movement to Spec, TP (and creates difficulties for argumentstructure-based accounts). 13 This implies that „locative“ inversion is a misnomer. Another interesting type are inversions with verbs taking abstract locatives (Postal 2004: 16–17): (i) On this election may well depend the future of our entire planet. (Postal 2004: 17) (ii) Second, to this structure would apply, optionally, a rule we may call verb second. (Birner 1995: 244) The definition of adjunct I am basing myself on here is a very traditional one. An XP is an adjunct if it does not fill any of the slots provided by the lexical meaning of a verb (predicate). This is independent of the exact semantic roles an XP bears, i.e. locatives and probably a few more of the roles found in this list can in principle act as arguments of a verb. In the examples at hand, however, this is crucially not the case. For instance, manner or instrumental adverbials can be added to any action verb, but they normally do not fill a slot provided by the verb. 14 The force of this argument depends, of course, on one’s analysis of that-trace effects. Under the assumptions discussed in 8.4 below, the (un)grammaticality of (31) would not be surprising. 15 It is usually assumed that LI and the presentational there-construction occur with the same types of verbs. Unergatives with directional complements like walk and come are widely attested with presentational there; these can probably be reanalyzed as unaccusatives, along the lines of Hoekstra and Mulder (1990). To what extent unergatives like those discussed in 3.1 (which are not readily amenable to an unaccusative reanalysis) are acceptable with presentational there is unclear at this point; crucially, if they turn out to be acceptable, the same analytical problems as with LI obtain. See Kuno and Takami (2004), Hartmann (2008) for recent discussion.

Repair-driven verb movement in English locative inversion   

   199

16 Cf. den Dikken (2006: 273, fn. 5) for a critical assessment. 17 Culicover and Levine (2001: 289–290, fn. 8) provide another argument for the difference between Light and Heavy Inversion. They claim that WCO effects do not obtain with Light Inversion while they do with Heavy Inversion, which thus patterns with topicalization: Light Inv (A-mvt) (i) In no dogi’s cage hung itsi collar. (ii) * In no dogi’s cage was hanging on a hook itsi most attractive and expensive collar. Heavy Inv. (A’-mvt) (iii) * In no dogi’s cage itsi most attractive and expensive collar was hanging on a hook. top (A’-mvt) I remain skeptical concerning the force of this argument. As the authors note themselves, the judgments are delicate. Furthermore, topicalization is normally not taken to induce WCO effects (cf. Lasnik and Stowell 1991). Finally, the deviance of the example with Heavy Inversion may be due to extraposition of the bindee to an A’-position. See also Postal (2004: 348, fn. 28) for discussion. 18 This argues against the restructuring account in Culicover and Rochemont (1990: 95) who assume V-to-I-movement. Cf. also Culicover and Levine (2001: 287) for critical discussion. 19 Anders Holmberg (p.c.) has pointed out to me that the topics found in this position in Finnish do not create topic islands. In Mexican Spanish, the blocking effect does obtain with topics but not with fronted experiencers which are taken to front because they are the highest arguments of such predicates and not because of some information structural property. 20 Crosslinguistically, the set of elements that can occur in such a flexible position varies a lot. While German permits just about everything in its flexible Spec, CP position, Finnish is more restricted in that it disallows manner and sentential adverbs. At the same time it is more liberal than English in that it allows DP topics, cf. Holmberg and Nikanne (2002). There seems to be an implicational relationships in that the possibility of DPs topics in that position implies the possibility of stage topics. 21 In Culicover and Winkler (2008a: 42, ad ex. 83), a section not contained in the published version Culicover and Winkler (2008b), the impossibility of QI in compound tenses is not linked to syntactic, but to pragmatic factors: it is simply odd to combine most of the complex tenses with a direct quote. To the extent that a complex tense is pragmatically felicitous in QI, it is also acceptable. 22 There are some exceptional cases where sentential negation is possible with LI. Such cases do not seem to have a presentational function in the narrow sense: (i) Thunderstick is right that with success does not come class, but using Belichick as your example of that is simply a crap argument. (dcatblog.blogspot.com/2006/09/belichick-mangini-handshake.html) Unfortunately, I have not been able to find corresponding examples with unergatives. 23 While generating gerundive and perfect participles in Asp (or moving them to Asp) is relatively straightforward, such movement is much more difficult to motivate for bare infinitives and finite verbs. 24 I thus assume a more classical EPP which cannot be reduced to an EPP-feature of a probe that requires Agree to be followed by internal merge of the goal. It is basically as in Bailyn (2004) with the important difference that in Bailyn it can only be satisfied by A-movement. The constraint EPP in this approach has more or less the same effect as EPP(phi) in Broekhuis (2008), but I take the implementation of the EPP proposed here to be preferable for the following reasons: First, the assumption that predicates check phi-features with T strikes me as problematic since there never is a morphological reflex of this agreement (in present-day

200   

   Martin Salzmann

English). Second, some of the fronted constituents are not predicates. Third, we will find instances of altruistic movement to Spec, TP that do not involve any feature checking other than satisfaction of the EPP (8.4 and 8.5). I will assume in what follows that the EPP is checked derivationally. In 8.4 I will also briefly discuss a representational alternative. 25 Constraints of this type are comparable with the so-called EPP-constraints in Broekhuis (2008). Importantly, such constraints do not imply that there is no Agree. Rather, they are intended to mean that if they are satisfied they force Agree to be followed by internal merge. If the constraint is violated, there only is Agree. Broekhuis (2008) just uses one general EPPconstraint for subject and object case checking. Since I will not be dealing with object case here, SubjectCase is sufficient for my purposes. 26 A formulation of the constraint in terms of linear order is necessary to allow the topic to occupy different positions, i.e. Spec, TP or a higher position. If instead TP is analyzed as a syncretic category that hosts case and top features (cf. e.g. Zubizarreta 1998) and can project either a topic or a subject, one could use a constraint EPPtopic instead that forces topics to front to Spec, TopP. Cf. also fn. 37. 27 Constraints forcing displacement are normally counteracted by economy constraints like *Move. Since in LI the effects of *Move already follow from the interface constraint to be introduced presently, it will be disregarded in what follows. Note that *Move does not play a decisive role in Broekhuis’ (2008) analysis of LI either. There is no single evaluation where a violation of *Move would lead to suboptimality of a candidate. 28 They are various definitions of AlignFocus; I have attempted to use a definition that is as neutral as possible and can be used for general information focus and presentational focus and possibly also for contrastive focus (Culicover and Winkler 2008b), but the latter may also require movement to a position more to the left or a marked accent if left in-situ (cf. Broekhuis (2008: 44, fn. 11)). 29 Den Dikken (2006: 87–88) argues against an information structural approach to LI. See Broekhuis (2008: 299–200) for a rejoinder. 30 While repair-driven movement is triggered by syntactic constraints only in the phenomena discussed in Heck and Müller (2000, 2007), it is triggered by an interface constraint in LI. 31 In 9.2 below we will come across another case of exceptional head movement which is altruistic as well and in contrast to V-to-Asp-movement is arguably not feature-driven. Altruistic movement is a notion often postulated for (certain instances of) German scrambling where displacement of a phrase frequently does not take place to check features of its own, but rather to allow another phrase to be in focus. Still, scrambling also often involves fronting of presuppositional/topical material so that it cannot be considered purely altruistic (cf. e.g. Fanselow 2003 for discussion). Repair-driven verb movement in LI is thus eventually quite different from scrambling. 32 To technically rule out non-feature-driven movement, one needs a higher-ranked constraint that penalizes such movement, cf. Heck and Müller (2007). An alternative consists in adopting the Derivations and Evaluations framework by Broekhuis (2008) which only allows featuredriven movement operations. 33 Since the relative ranking of SubjectCase and NoLexMvt cannot be determined based on the data from LI, I will assume a tie for reasons of simplicity; to avoid that candidate b can emerge as optimal, the tie would have to be a conjunctive local tie, cf. Müller (2000: 212f.). I have adopted a structure for unaccusatives where the verb embeds a SC. If there were instead a vP on top of a VP (Collins 1997), there would be an instance of V-to-v movement. Following

Repair-driven verb movement in English locative inversion   

   201

Broekhuis (2008: 210–211) I do not take this to be a violation of NoLexMvt but of some other constraint penalizing head movement, e.g. *StrayFeature that requires amalgamation of the formal features of functional heads with their root (Broekhuis 2008). A distinction between this type of movement and v-to-T movement is needed anyway to derive differences between Germanic languages and to explain why in languages like English only movement of the lexical verb but not of auxiliaries is blocked. This distinction has the additional advantage that the marked character of verb movement in LI is expressed more directly: by a single violation of NoLexMvt. 34 For ease of representation I have omitted the VP-layer. Due to obligatory V-to-v movement, all candidates will have a violation of *StrayFeature. I have put the trace of the locative inside vP, but with adjuncts as in 3.2 (cf. also fn. 12) the base-position arguably has to be posited outside vP (this also holds, of course, for adjuncts with unaccusatives). 35 I suspect that focus is also relevant for cases of verb movement like the following discussed in Chomsky (1995, chapter 4): (i) He reads often to his children. 36 A particularly recalcitrant case are inversion constructions with a clause-final adverb, as discussed in Culicover and Levine (2001: 288, 292): (i) Into the room walked Robin slowly. Such examples differ from canonical examples of LI in that the inverted subject cannot be prosodically prominent (den Dikken 2006: 272, fn. 2). The question then is why the subject can remain low. I must admit that I do not fully understand the information structure of this construction. It seems to be the case that the adverb is focused, and judging by the examples one can find in the literature the inverted subject also seems to have a presentational function. One can therefore argue that it is also AlignFocus which keeps the subject low. Moving the subject to Spec, TP would lead to an additional violation of AlignFocus so that leaving it in its base-position is preferred. The alternative order (ii) Into the room walked slowly ROBIN. requires a strong accent on Robin, arguably because it undergoes Heavy NP-shift (cf. den Dikken 2006: 127–130 on extraposition in LI). This is thus another possibility for the subject to be clause-final. Importantly, the adverb is not focused in such examples. The choice between (i) and (ii) is thus probably related to information structure. A similar case is discussed in Culicover and Rochemont (1990) and Broekhuis (2008: 288–291): (iii) Into the room came John nude. (iv) Into the room nude came John. Here, the choice depends on whether the depictive is part of the focus (iii) or not (iv). 37  The necessity of an overt complementizer in embedded LI (2.2.4) and the incompatibility of aux-inversion with LI (2.2.5) are more difficult to account for without reference to an additional structural layer. So far, I have been assuming that topic fronting is technically implemented without feature checking. One could, however, posit optional topic features on T to trigger fronting of the locative (or merger of certain adjuncts). Once there is such a feature on T, LI can be adequately restricted: matrix verbs can be restricted to select either complement clauses as CPs (with that) or TPs (without that); in the latter case, selection would have to be restricted to TPs without a topic feature. Similarly, a C-head specified for a yes-no question could be specified to select TPs without a topic feature. In the same vein, one could prevent LI in nonfinite contexts by a ruling out topic features on non-finite T. If regular topics (that precede the subject) occupy a (second) specifier of T, the same feature-based restrictions can be used.

202   

   Martin Salzmann

Instead of pursuing a purely feature-based account, one could try to uphold a structural account within the set of assumptions made so far (that would eventually be almost indistinguishable in its empirical consequences): One could take TP to be a co-projection of T and Top, i.e. a syncretic category (Zubizarreta 1998), or, similarly, to be a matching projection (Haider 1988). In both cases, the EPP would be satisfied once one specifier is filled. The impossibility of LI in non-finite clauses would be due to the absence of a TopP layer; the impossibility of subject-auxinversion could be due to the fact that a C-head specified for [yes-no] cannot select TopPs, and the necessity of an overt complementizer with embedded LI would follow if it is assumed that verbs can select either CPs (complement clauses with that) or TPs (complement clauses without that), but not TopPs. On the incompatibility of LI with subject-auxiliary inversion, cf. also Bruening (2010). 38 Culicover and Levine (2001: 288, 297–302) claim that “raising” and long A’-movement more generally is only possible in Heavy Inversion, but not with Light Inversion. Their argument is based on examples involving adverbs where apparently the subject has to be placed clausefinally: (i) Into the room appeared to be walking *Robin slowly/slowly a very large caterpillar. Since it is not obvious to me in what sense subjects in examples like (77) should count as heavy and since the basic distinction between Light and Heavy Inversion has been shown to be empirically untenable in Holler and Hartmann (to appear), I will continue to assume that there is just one type of LI and that long-distance A’-movement is in principle always possible. Cf. also Rizzi and Shlonsky (2006: 359, fn. 4) for clear cases of long A’-movement with a light postverbal subject. 39 According to Culicover (1993: 98, fn. 1), focal fronted elements apparently also lead to an alleviation of that-trace effects: (i) Robin met the man whoi Leslie said that [to KIM]j __ i had given the money __ j. Why no A’-intervention obtains here, is unclear to me. Perhaps this is due to the focal nature of these „topicalized“ XPs. At any rate, the amelioration in (i) is not unexpected under accounts where that-trace effects reduce to a ban on the sequence that-Vfin or a phonological EPP (under the assumption that to Kim can occupy Spec, TP). In contrast with the literature, Culicover and Levine (2001: 302–303) claim that no topic island obtains in LI. Extraction is only taken to be degraded in Heavy Inversion. This may be related to the claim in Culicover (1996: 453–454) that there are no topic islands at all, at least not under certain focal conditions. The degradedness is instead related to processing factors. Since a full evaluation of all these proposals is beyond the scope of this paper, I leave this for further research. 40 Note that the two alternative explanations of that-trace effects require different solutions for the data in (77) and (78). Under the approach based on a ban on topic extraction, the data are unproblematic since there are no sentence topics in non-finite clauses so that there will be no extraction from a topic position; given this, movement of the locative via Spec, TP is unproblematic. Under a phonological EPP , (77) and (78) are problematic since the embedded TP is not overtly filled at PF. Under such an approach it seems inevitable to discard the EPP for non-finite clauses. 41 In Grimshaw (1997) the operator can also be in Spec, vP if there is no material above it. Since I have posited an EPP for TP, the operator has to front at least up to TP. But in principle it would be possible to reformulate the EPP in a relational way so that it can also be satisfied if Spec, vP is occupied. The notion of EPP would then be very close to the notion of pole in Gutierrez-Bravo (2002, 2007).

Repair-driven verb movement in English locative inversion   

   203

42 As the reader will have noticed, these facts are incompatible with a phonological EPP holding at surface structure. 43 Note that Collins (1997) and Suñer (2000) basically have to stipulate verb movement for these constructions while in the case at hand verb movement follows from more general principles. 44 For detailed discussion of predicate inversion, also concerning differences between the type discussed here and inversion with APs, cf. den Dikken (2006: 152–160) and den Dikken (this volume). 45 Questions arise for there-insertion with unergative verbs and more generally if a definition of closeness is adopted that solely relies on c-command. If the subject is closer than there (e.g. if there originates within VP and the subject within vP or if the subject simply c-commands there), there would first have to move to Spec, vP to be close enough to Spec, TP. It is unclear, though, what might drive this movement given that I am assuming that there-fronting is purely EPP-driven. 46 Note that the same result obtains when both subject and locative are focal. In that case, a violation of AlignFocus cannot be avoided, cf. Broekhuis (2008: 273). 47 Culicover and Levine (2001: 308) and Birner (1995: 243) list a number of cases with transitive verbs, but they are all idiomatic and may therefore count as informationally light. 48 Landau (2010: 125–126) argues that the notion presentational focus overgenerates and claims that LI is restricted by a locative feature on the preposed constituent. But as discussed extensively in this paper, the notion “locative” is insufficient, even if used in an extended sense.

11 References Alexiadou, Artemis and Elena Anagnostopoulou. 2001. The subject-in-situ generalization and the role of Case in driving computations. Linguistic Inquiry 32: 193–231. Bailyn, John Frederick. 2004. Generalized Inversion. Natural Language and Linguistic Theory 22: 1–49. Bayer, Josef. 2005. Was beschränkt die Extraktion? Subjekt – Objekt vs. Topic – fokus. In: Franz Josef D’Avis (ed.), Deutsche Syntax: Empirie und Theorie, 233–257. Göteborg: Acta Universitatis Gothoburgensis. Bayer, Josef and Martin Salzmann. This volume. That-trace effects and resumption. How Improper Movement can be repaired. Birner, Betty J. 1994. Information status and word order: An analysis of English inversion. Language 70: 233–259. Birner, Betty J. 1995. Pragmatic constraints on the verb in English inversion. Lingua 97: 233–256. Bošković, Željko. 2002. A-movement and the EPP. Syntax 5: 167–218. Branigan, Philip. 2000. Binding effects with covert movement. Linguistic Inquiry 31: 553–557. Bresnan, Joan. 1994. Locative inversion and the architecture of Universal Grammar. Language 70: 72–131. Broekhuis, Hans. 2008. Derivations and Evaluations. Berlin: Mouton de Gruyter. Bruening, Benjamin. 2010. Language-particular syntactic rules and constraints: English locative inversion and do-support. Language 86: 43–84.

204   

   Martin Salzmann

Chen, Rong. 2003. English Inversion. Berlin: Mouton de Gruyter. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Massachusetts: MIT Press. Chung, Chan and Jong-Bok Kim. 2002. Lexical and constructional constraints in English locative inversion. In: Mary Andronis, Erin Debenport, Anne Pycha and Keiko Yoshimura (eds.), Proceedings of the 38th Annual Meeting of the Chicago Linguistic Society, 145–158. Chicago Linguistics Society: Chicago Collins, Chris. 1997. Local Economy. Cambridge, Massachusetts: MIT Press. Collins, Chris and Phil Branigan. 1997. Quotative inversion. Natural Language and Linguistic Theory 15: 1–41. Coopmans, Peter. 1989. Where stylistic and syntactic processes meet: Locative inversion in English. Language 65: 728–751. Costa, Joao. 1998. Word order variation: A constraint based approach. Ph.D. dissertation, Leiden University. Culicover, Peter. 1993. The adverb effect: evidence against ECP accounts of the that-t effect. In: Amy J. Schafer (ed.), Proceedings of the 23rd Annual Meeting of the North East Linguistic Society, 97–110. GLSA, University of Massachusetts, Amherst. Culicover, Peter W. 1996. On distinguishing A’-movements. Linguistic Inquiry 27: 445–463. Culicover, Peter W. and Robert Levine. 2001. Stylistic inversion in English: A reconsideration. Natural Language and Linguistic Theory 19: 283–310. Culicover, Peter W. and Michael Rochemont. 1990. English Focus Constructions and the Theory of Grammar. Cambridge: Cambridge University Press. Culicover, Peter and Susanne Winkler. 2008a. English focus inversion constructions. Ms. OSU and University of Tübingen. Culicover, Peter and Susanne Winkler. 2008b. English focus inversion. Journal of Linguistics 44: 625–658. Demuth, Katherine, and Mmusi, Sheila. 1997. Presentational focus and thematic structure in comparative Bantu. Journal of African Languages and Linguistics 18: 1–20. Diesing, Molly. 1990. Verb movement and the subject position in Yiddish. Natural Language and Linguistic Theory 8: 41–79. Dikken, Marcel den. 2006. Relators and Linkers. Cambridge, Massachusetts: MIT Press. Fanselow, Gisbert. 2003. Free constituent order: A minimalist interface account. Folia Linguistica 37: 191–232. Green, Georgia M. 1980. Some wherefores of English inversion. Language 56: 582–601. Green, Georgia M. 1982. Colloquial and literary uses of inversions. In: Deborah Tannen (ed.), Spoken and Written language: Exploring Orality and Literacy, 119–153. Norwood, NY: Ablex. Green, Georgia M. 1985. The description of inversions in Generalized Phrase Structure Grammar. In: Mary Niepokuj, M. VanClay, V. Nikoforidou and D. Feder (eds.), Proceedings of the 11th Annual Meeting of the Berkeley Linguistics Society, 117–145. Berkeley, Berkeley Linguistics Society. Grimshaw, Jane. 1997. Projection, heads, and optimality. Linguistic Inquiry 28: 373–422. Grohmann, Kleanthes K., John Drury and Juan Carlos Castillo. 2000. No more EPP. In: Roger Billerey & Brook Danielle Lillehaugen (eds.), Proceedings of the 19th West Coast Conference on Formal Linguistics, 153–166. Somerville, MA: Cascadilla Press. Gutierrez-Bravo, Rodrigo. 2002. Structural markedness and syntactic structure: A study of word order and the left periphery in Mexican Spanish. Ph.D. dissertation, University of California, Santa Cruz.

Repair-driven verb movement in English locative inversion   

   205

Gutiérrez-Bravo, Rodrigo. 2007. Prominence scales and unmarked word order in Spanish. Natural Language and Linguistic Theory 25: 235–271. Haider, Hubert. 1988. Matching projections. In: Anna Cardinaletti, Giulelmo Cinque and Giuliana Giusti (eds.), Constituent Structure. Papers from the 1987 GLOW Conference, 101–121. Dordrecht: Foris. Hartmann, Jutta M. 2008. Expletives in Existentials. English there and German da. Utrecht: LOT. Heck, Fabian and Gereon Müller. 2000. Repair-driven movement and the local optimization of derivations. Ms, University of Leipzig. Heck, Fabian and Gereon Müller. 2007. Derivational optimization of wh-movement. Linguistic Analysis, 33(1–2): 97–148. Heycock, Carolyn. 2006. Embedded root phenomena. In: Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax, 174–209. Oxford: Blackwell. Hoekstra, Teun and René Mulder. 1990. Unergatives as copular verbs; locational and existential predication. The Linguistic Review 7: 1–79. Holler, Sara, and Jutta Hartmann. To appear. Locative inversion in English: Implications of a rating study. In: Sam Featherston and Britta Stolterfoht (eds.), Empirical Approaches to Linguistic Theory: Studies in Meaning and Structure, 241–265. Berlin: Mouton de Gruyter. Holmberg, Anders and Urpo Nikanne. 2002. Expletives, subjects, and topics in Finnish. In: Peter Svenonius (ed.), Subjects, Expletives, and the EPP, 71–105. Oxford: Oxford University Press. Hooper, Joan and Sandra Thompson. 1973. On the applicability of root transformations. Linguistic Inquiry 4: 465–497. Kathol, Andreas and Robert D. Levine. 1993. Inversion as a linearization effect. In: Amy J. Schafer (ed.), Proceedings of the 23rd Annual Meeting of the North East Linguistic Society, 207–221. GLSA, University of Massachusetts, Amherst. Kuno, Susumu, and Ken-ichi Takami. 2004. Functional Constraints in Grammar. On the Unergative-Unaccusative Distinction. Amsterdam: John Benjamins. Landau, Idan. 2010. The Locative Syntax of Experiencers. Cambridge, Mass.: MIT Press. Lasnik, Howard and Timothy Stowell. 1991. Weakest crossover. Linguistic Inquiry 22: 687–720. Levin, Beth and Malka Rappaport Hovav. 1995. Unaccusativity. At the Syntax-Lexical Semantics Interface. Cambridge, Mass.: MIT Press. Matushansky, Ora. 2006. Head movement in linguistic theory. Linguistic Inquiry 37: 69–109. Müller, Gereon. 2000. Elemente der optimalitätstheoretischen Syntax. Tübingen: Stauffenburg. Postal, Paul M. 1977. About a “nonargument” for raising. Linguistic Inquiry 8: 141–154. Postal, Paul M. 2004. Skeptical Linguistic Essays. Oxford: Oxford University Press. Prince, Alan and Paul Smolensky. 1993. Optimality Theory: Constraint interaction in Generative Grammar. Rutgers University for Cognitive Science: RuCCS Technical Report No.2. Rizzi, Luigi and Ur Shlonsky. 2006. Satisfying the Subject Criterion by a non subject: English locative inversion and Heavy NP Shift. In: Mara Frascarelli (ed.), Phases of Interpretation, 341–361. Berlin: Mouton de Gruyter. Roberts, Ian. 2010. Agreement and Head Movement. Clitics, Incorporation, and Defective Goals. Cambridge, Mass.: MIT Press. Salzmann, Martin. 2011. Towards a typology of locative inversion – Bantu, perhaps Chinese and English – but beyond? Language and Linguistics Compass 5: 169–189. Salzmann, Martin, Jana Häussler, Markus Bader and Josef Bayer. To appear. That-trace effects without traces. An experimental investigation. In: Stefan Keine and Shayne Sloggett

206   

   Martin Salzmann

(eds.), Proceedings of the 42nd Annual Meeting of the North East Linguistic Society. GLSA, University of Massachusetts, Amherst. Schachter, Paul. 1992. Comments on Bresnan and Kanverva’s “Locative inversion in Chichewa: A case study of factorization in grammar”. In: Timothy Stowell and EricWehrli (eds.), Syntax and the Lexicon, 103–110. New York/San Francisco/London: Academic Press. Suñer, Margerita. 2000. The syntax of direct quotes with special reference to Spanish and English. Natural Language and Linguistic Theory 18: 525–578. Winkler, Susanne and Edward Göbbel. 2008. Focus in comparative inversion constructions. Handout DEAL II, Leiden. Zubizarreta, Maria Lisa. 1998. Prosody, Focus, and Word Order. Cambridge, Mass: MIT Press.

Joost Kremers

Linearisation as repair* Abstract: This paper discusses the question whether linearisation can be considered a case of repair in the phonological system. Although standard approaches to linearisation assume that linear order is total and that association lines may not cross (Partee et al.’s 1993 Exclusivity and Nontangling Conditions), sign language data suggest that these conditions are not absolute. Sign languages (and to a lesser extent spoken languages) have the ability to express multiple meaningful elements simultaneously. Since it is the phonological form of the elements in question that determines whether simultaneous realisation is possible, only the phonological system can determine whether two elements need to be linearised or whether they can be realised simultaneously. The sheer existence of simultaneity suggests that simultaneous realisation is the preferred option, whereas linearisation applies only to repair structures that cannot be realised simultaneously. Linearisation cannot take place without at least some syntactic information being accessible. It is argued that this access is limited to the absolute minimum: merge order and headedness. Other syntactic information is not available to the linearisation process. For the data discussed here, this is indeed sufficient.

1 Introduction Since Chomsky (1995), the assumption that syntactic structures are purely hierarchical and do not encode linear order has become dominant in generative syntactic theory. On this assumption, some procedure is needed to derive linear order from hierarchical structure. In the literature, there are roughly two kinds of approaches to linearisation. On the one hand, there are approaches inspired by Kayne’s (1994) LCA. These assume a strict correlation between asymmetric c-command and linear order, which means that elements higher in the tree precede elements lower in the tree. On the other hand, there are approaches that assume ordering parameters in some form or other. Although fundamentally different in certain ways, both types of approaches also share two basic assumptions. These are the Exclusivity Condition and the Nontangling Condition, as formulated in Partee et al. (1993: 442):¹

208   

   Joost Kremers

(1) a. The Exclusivity Condition: In any well-formed constituent structure tree, for any nodes x and y, x and y stand in the precedence relation P, i.e., either 〈x,y〉 ∈ P or 〈x,y〉 ∈ P, if and only if x and y do not stand in the dominance relation D, i.e., neither 〈x,y〉 ∈ D nor 〈x,y〉 ∈ D. b. The Nontangling Condition: In any well-formed constituent structure tree, for any nodes x and y, if x precedes y, then all nodes dominated by x precede all nodes dominated by y. In the present paper, I shall refer to two alternative conditions that for practical purposes can be considered equivalent. First, Kayne (1994) assumes the principle of Totality: (2) Totality: Given a tree K and the set T of terminals in K: ∀x,y (x,y ∈ T ∧x ≠ y | x < y∨y < x). Here, a , the author C the Mary every book C he writes sofort chaufft immediately buys ‘the author such that Mary immediately buys every book he writes’ (CNPC) b. de Sänger, won i mi fröi, < wänn mer * (en) im Fernseh the singer C I me be.happy when one him on TV bringt> brings ‘the singer such that I am happy when they show him on TV’ (adjunct island) (59) a. * [Wele Autor]1 chauft d Marie < jedes Buech, wo t1/ er schriibt>? Which author buys the Mary every book C he writes lit.: ‘Which author does Mary buy every book that writes?’ (CNPC) b. * [Wele Sänger]1 fröisch di, < wänn mer t1/en im Fernseh which singer be.happy.2s you when one him on TV bringt>? brings lit.: ‘Which singer are you happy when they show on TV?’ (adjunct island) Gaps and resumptives are thus in complementary distribution.

6.2 Explaining the distribution of resumptive pronouns The distribution of resumptives is governed by two factors: Resumptives amnesty locality violations, cf. (58). This also covers cases like (57d) since PPs are strong islands in German and its varieties. Dative resumptives, however, require a different explanation because extraction of dative objects is possible. Dative resumptives are the reflex of a realizational constraint operative in most German varieties

That-trace effects and resumption – How Improper Movement can be repaired   

   305

that requires the overt realization of oblique case, cf. Bayer et al. (2001), Salzmann (2006b, 2008, 2009b) for details. Resumptives, thus, act as a last resort, occurring only when gap-derivations fail. Since subjects and direct objects are realized by non-oblique cases, they do not have to be expressed overtly; resumptives are therefore not necessary. This statement is less innocuous than it sounds and will be qualified in the next section.

6.3 Resumption involves base-generation While gap relatives can be straightforwardly analyzed as involving movement, the syntax of resumption has been subject to some controversy lately.⁵¹ While base-generation was the default until 1990, several linguists have proposed movement accounts since then, e.g. Pesetsky (1998), Aoun et al. (2001), Boeckx (2003), Bianchi (2004). While a movement account, or one based on Agree, may be adequate for languages where resumption is sensitive to locality (cf. e.g. Boeckx 2003, Alexopoulou 2006), any movement account is confronted with serious difficulties once resumption is not sensitive to locality, as in ZG, cf. (58). Movement accounts of resumption have to resort to rather unorthodox mechanisms, or constraints, to make movement out of islands possible (e.g. resumption as LF-movement as in Demirdache 1991, locality as a PF-constraint, cf. Pesetsky 1998, or movement without Agree as in Boeckx 2003). These mechanisms complicate the grammar of locality in unmotivated ways, or fail altogether, cf. Salzmann (2008: 105–108, 2009a: 33–39, 2011, 2012) for detailed argumentation. Under base-generation, the distribution of (non-dative) resumptives follows straightforwardly since islands and PPs bar extraction and consequently can only be bridged by means of binding as in base-generation. One frequent argument in favor of a movement analysis of resumption are reconstruction effects under resumption (e.g. Aoun et al. 2001). Such effects can indeed be found in ZG, even reconstruction into islands (Salzmann 2008, 2009a, 2011, 2012). However, we do not think that reconstruction should always be taken as a waterproof diagnostic for movement because the relationship between movement and reconstruction has generally turned out to be imperfect: There are instances of reconstruction without prior movement, and there are instances of movement without reconstruction, cf. the discussion in Salzmann 2008, 2009a, 2011, 2012. Under base-generation, the locality effects fall out nicely; to account for reconstruction under resumption (and thus base-generation), alternative mechanisms are available such as the NP-ellipsis theory of resumption, cf. Guilliot & Malkawi (2006). We therefore opt for a base-generation analysis of resumption in ZG. Note that this means that gap and resumptive relatives involve very different derivations.

306   

   Josef Bayer and Martin Salzmann

We are now in a position to be somewhat more precise about the syntax of subject and direct object relativization. In the previous subsection, we remarked rather casually that resumptives are not necessary because they are realized by nonoblique cases. However, since the grammar provides a mechanism that can produce resumptive structures, and since no obvious syntactic constraint prevents resumptives for subjects and direct objects, we must assume that resumptive derivations converge for these relations as well. Given that only gap derivations are grammatical in these environments, we can conclude that they block resumptive derivations. For present purposes it is sufficient to know that movement is preferred in case both movement and base-generation converge; see Salzmann (2009a/c, 2012) where this preference is linked to a translocal constraint in ZG that prefers internal merge over external merge. Before we turn to a precise implementation of movement and base-generation derivations, we need to look at one puzzle in the distribution of resumptives that will motivate the feature specifications on relative operators and relative complementizers.

6.4 The puzzle: long-distance relativization So far, the distribution of resumptive pronouns in ZG is straightforward: they only occur as a last resort when movement derivations fail. Before this background it is rather surprising that resumptives become obligatory once relativization crosses a sentence boundary: in long-distance relativization, resumptives become obligatory for all grammatical relations, including subjects and direct objects (van Riemsdijk 1989, 2008, Salzmann 2006b): (60) a. d Frau, won i gsäit han, dass * (si) immer z spaat chunt the woman C I said have.1s that sh always too late comes ‘the woman who I said is always late’ emb. SU b. es Bild, won i vermuete, dass * (es) niemert cha zale a picture C I suspect that it no.one can pay ‘a picture that I suspect nobody can afford’

emb. DO

Given the factors governing the distribution of resumptives that we have identified above this is unexpected. Oblique Case is not at stake, and locality seems to be an unlikely candidate given that corresponding long wh-extractions (of nontopics) are fully grammatical (resumption is not an option in regular wh-movement, cf. 6.7 below):

That-trace effects and resumption – How Improper Movement can be repaired   

   307

(61) a. [Wele Maa]1 häsch gsäit, dass geschter t1 choo isch? which man have.2s said that yesterday come is ‘Which man did you say came yesterday?’ b. [Wele Maa]1 häsch gsäit, dass s Rägeli t1 küsst hät? which man have.2s said that the Regula kissed has ‘Which man did you say that Regula kissed?’ But if locality is not at stake, what else prevents a movement derivation in longdistance relativization?⁵²

6.5 Relative operators cannot check uContrast – resumption as repair We would like to propose that long-distance movement in relativization fails for the same reason that long-distance movement of (sentence) topics fails in whmovement: The topical restriction of a relative operator has to move via SpecCP into the matrix clause to amalgamate with the operator. In wh-movement, this does not always happen because German provides ways of extracting from a position below TopP. In relativization, we submit, this is not possible because the restriction of (specific, cf. below) relative operators is inherently topical and therefore always targets SpecTopP on its way to the final landing site (on the topicality of the relative pronoun, cf. Bresnan & Mchombo 1987, Bianchi 2004, and Lambrecht 1994: 129–130 who argues that the referent of the head NP enters an aboutness relation with a proposition via the relative pronoun).⁵³ Due to its semantics, the restriction of relative operators is incompatible with iContrast. As a consequence, if on its path it passes a C with uContrast, it will be unable to value that feature so that the derivation eventually crashes. A base-generation derivation, however, converges because the intermediate C does not contain any attracting features and consequently no uContrast. The short-distance/long-distance asymmetry (57a/b) vs. (60) is again due to the fact that the topic restriction does not pass through a head with uContrast in short extraction. Given the semantics of relative clauses where alternatives do not play a role, we assume that no uContrast is involved in the checking operations involving the highest head of the matrix clause and the operator. As a consequence, what is crucial is uContrast on the intermediate complementizer dass.⁵⁴ As with wh-movement, we assume that the intermediate C in long-distance relativization is endowed with an EPP-feature that triggers successive-cyclic A’movement and a subfeature uContrast. Regular relative operators are inherently

308   

   Josef Bayer and Martin Salzmann

specified as iOp while the restriction inherently bears iTop and is thus incompatible with iContrast. Schematically, a potential long-distance derivation looks as follows (we omit the matrix TopP below wo for reasons of space).⁵⁵ (62) * [ForceP Op Force[FinP restriction1 wo [vp [cp t1 C [TopP t1 Top [vP [vp t1 V] v] ]]]]] iOp[x] uOp[x]

iTop[y]

EPP

uContrast[ ] uTop[y] EPP

EPP

EPP

As in wh-movement, the operator is directly inserted into ForceP to check uOp of Force. The restrictive NP originates in the theta-position and then moves to TopP. For the restriction to amalgamate with Op, it has to move into the matrix clause via Spec, CP. And this is again where the violation occurs: This movement step violates (34): uContrast of the intermediate C will remain unvalued and the derivation crashes. The reason why local relativization is grammatical is simply the absence of an intermediate C and therefore the absence of the feature uContrast that could cause problems for the movement of the restriction. The derivation for local relativization looks as follows: (63) [ForceP Op Force [FinP restriction1 wo [TopP t1 Top [vP [vp t1 V] v] ]]] iOp[x] uOp[x] EPP

iTop[y]

uTop[y] EPP

EPP

While the operator is directly inserted into ForceP, the restriction moves via TopP to FinP to amalgamate with Op.⁵⁶ The base-generation derivation is similar to the movement derivation in major respects, the major difference being that the restriction can be realized by means of a phonologically independent element, viz. a resumptive pronoun, basically as already illustrated for wh-movement in (38). As a consequence, the restriction does not have to undergo movement to the matrix clause to amalgamate with Op. Thus, there is no movement via the embedded SpecCP whatsoever. Since the intermediate C does not bear any attracting features and thus no uContrast, the problematic configuration in (34) does not arise. A base-generation derivation then looks as follows (as in (38), the resumptive moves to SpecTopP to check the topic feature):⁵⁷ (64) [ForceP Op1 Force [FinP wo [vp [cp C [topP pron1 Top [vP t1 v]]]]]] iOp[x] uOp[x] EPP

iTop[z] uTop[z]

That-trace effects and resumption – How Improper Movement can be repaired   

   309

6.6 Additional evidence In this subsection we will show that our proposal has coverage beyond the classical core examples and the comparative data considered so far.

6.6.1 Amount relatives and comparatives: abstraction over a degree One of the crucial ingredients of our analysis is that the restriction of relative operators is incompatible with the feature uContrast due to its inherent topicality. However, this does not hold for all operators. Amount relatives involve abstraction over a degree, and the restriction consequently cannot target SpecTopP – just like in amount wh-questions (19), it is inherently incompatible with iTop. As a consequence, the restriction should be able to move via SpecCP into the matrix clause to amalgamate with Op. We therefore expect long-distance relativization with amounts to be possible. This prediction is borne out: There are no resumptives with amount relatives: (65) di 100 Kilo, won er gsäit hät, dass er (*si/*das) uf d Waag bringt the 100 kilos C he said has that he them/that on the scale brings ‘the 100 kilos he said he weighs’ This means that there is long-distance relativization in ZG, but only in one very restricted area, namely where a relative operator (i.e. its restriction) is incompatible with topicality. The derivation for (65) can be sketched as follows: (66) [ForceP Op

Force [FinP restriction1 wo [tp [vP [cp t1 dass [TopP XP [vP t1 V]]]]]]]

iOp[x] uOp[x] EPP

iContrast[y]

EPP

EPP uContrast[y]

The restriction originates in the embedded clause, moves to SpecCP to check uContrast of dass. Subsequently, it undergoes FF to SpecFinP (recall from the beginning of section 4.2 that the feature iContrast does not require the constituent to be actually used contrastively, it merely makes it eligible for contrastive use). Finally, the operator is merged in ForceP, checking uOp of Force. All features are checked, and the derivation converges. A similar observation can be made for comparatives whose syntax is similar to that of relatives in that they also involve a silent operator and require resumptives in oblique positions (Salzmann 2006b: 375):

310   

   Josef Bayer and Martin Salzmann

(67) a. Es sind [mee Patiente] choo als (*sie) de Tokter hät chöne there are more patients come than (them) the doctor has could behandle. treat ‘There came more patients than the doctor could treat.’ acc b. Es sind [mee Lüüt] choo als *(ine) de Tokter there are more people come than (they.dat) the doctor hät chöne Medikamänt verschriibe. dat has could medicine prescribe ‘There came more people than the doctor could prescribe medicine for.’ Once we go across a sentence boundary, we find the same pattern as with amount relatives: Since the restriction of the degree operator is incompatible with topicality, it will not pass through SpecTopP. Therefore, it can check uContrast of the intermediate C. As a consequence, since long-distance movement is possible, we do not find resumption (as in (19), a silent stage topic, or perhaps the temporal adverbial hüt ‘today’, occupies SpecTopP in the embedded clause): (68) Es sind mee Patiente choo, als de Tokter tänkt, There are more patients come than the doctor thinks dass (*si) hüt chönd behandled werde. that they today can treated become ‘There came more patients than the doctor thinks can be treated today.’

6.6.2 Standard German While ZG has a means to overcome the incompatibility of the restriction of relative operators with iContrast, viz, resumption, the standard language does not. As a consequence, we expect long-distance relativization to be impossible. This indeed seems to be the case. Though one occasionally finds such examples in textbooks, most speakers reject long-distance relativization (cf. Lühr 1988: 77, Andersson & Kvam 1984): (69) a. ?? Das is einer der1 ich glaube, dass t1 das schaffen wird. This is one who.nom I believe that this manage will ‘This is a guy who I think will manage this.’

That-trace effects and resumption – How Improper Movement can be repaired   

   311

b. ?? der Mann, den1 ich glaube, dass Maria t1 liebt the man who.acc I believe that Mary loves ‘the man who I believe Mary loves’ Since the restriction, which obligatorily bears iTop, cannot value uContrast of dass, the derivation eventually crashes. All speakers prefer an alternative construction which involves local extraction in the matrix clause and binding between the relative pronoun and a coreferential pronoun in the dependent clause. This construction was termed resumptive prolepsis in Salzmann (2006b):⁵⁸, ⁵⁹ (70) der Mann, von dem1 ich glaube, dass Maria ihn1 liebt the man of who.dat I believe that Mary him loves ‘the man of whom I believe that Mary loves him’ Since the offending element is the intermediate C dass, we expect long relativization to be possible once the complement lacks dass, as is the case in infinitival complements. This prediction is indeed borne out. To make sure that we are not dealing with a monoclausal structure (where a CP-layer would be absent so that one would be dealing with short wh-movement), we deliberately use a nonrestructuring predicate: (71) das Buch, das1 ich erwartet habe t1 geschenkt zu bekommen the book which I expected have.1s given to get ‘the book I expected to be given as a present’ Since there is no dass, there is no uContrast that could cause a problem for the movement of the restriction.⁶⁰ By the same token, we expect long-distance relativization by movement from non-finite clauses in ZG to be possible. This prediction is borne out, a resumptive is not obligatory:⁶¹ (72) s [Buech], won i d Susi überredt ha (s) für mich z chauffe the book C I the Susie convinced have.1sg it for me to buy ‘the book that I convinced Susie to buy for me’

6.7 Why resumption is barred in wh-movement Since ZG is a language that makes productive use of resumptives one may wonder what happens under long-distance wh-movement. Like the standard language,

312   

   Josef Bayer and Martin Salzmann

ZG employs wh-operators that leave behind gaps, and resumption is not possible (cf. also Weber 1964: 304): mäinsch, dass de Hans t1/*em1 ghulffe hät? (73) Wem1 who.dat think that the John helped has ‘Who do you think John helped?’ But the question is what happens if an aboutness topic is to be extracted. Can resumptives come to the rescue? The answer is mixed. In principle, wh-movement is incompatible with resumption in ZG and Alemannic more generally. The reason for this is that base-generation requires Case-unmarked operators (cf. Salzmann 2011: 203–208 for additional discussion), but the overt operators found in whmovement and topicalization are normally Case-marked as e.g. in (73) so that resumption is not an option. There is, however, an alternative construction with wh-operators and topics that is compatible with resumption. In this construction, which Salzmann (2006a) termed A’-splits, the operator appears in a Caseunmarked or default nominative form while Case-information is realized in the theta-position by means of resumption (in case of oblique relations additionally governed by a preposition). ZG only distinguishes a direct and an oblique Case (dative), the direct one being used for both subjects and direct objects. The direct Case can be shown to play a double role in that it may also count as Caseunmarked and appear in A’-splits with the resumptive realizing nominative/accusative. The first triple illustrates this for wh-movement:⁶² (74) a. [Wele Maa]1 häsch gsäit, dass t *(en1) geschter gsee häsch? which man have.2s said that you him yesterday seen have.2s ‘Which man did you say that you saw yesterday?’ b. [Wele Maa]1 häsch behauptet, dass t *(em1) es Buech ggëë which man have.2s claimed that you he.dat a book given häsch? have.2s ‘To which man did you claim that you had given a book?’ c. [Weli Frau]1 häsch behauptet, dass t i d Schuel bisch which woman have.2s claimed that you in the school are mit *(ere1)? with her ‘With which woman did you claim that you went to school?’

That-trace effects and resumption – How Improper Movement can be repaired   

   313

The following examples illustrate the same point with topicalization (Salzmann 2006b: 376, fn. 297): (75) a. [Die Frau]1 hett i nie tänkt, dass *(si1) mi nett findt. this woman had.subj.1sg I never thought that she me nice finds lit.: ‘This woman I would have never thought likes me.’ b. [Dää Maa]1 glaub i, dass t *(em1) nöd wettsch this man know.1sg I that you he.dat not would.like.1sg im Tunkle begägne. in.the dark meet ‘This man I do not think that you would like to meet in the dark.’ c. [Die Frau]1 chan I scho verschtaa, dass t mit * (ere1) this woman can.1sg I certainly understand that you with her wettsch go tanze. want.2sg go ce ‘This woman I can certainly understand that you would like to go dancing with.’ A’-splits, like resumptive relatives, are insensitive to locality as the following example shows where the resumptive is the complement of a preposition embedded within a PP (Salzmann 2006b: 376, fn. 297): (76) [Wele Maa]1 häsch gsäit, dass d < mit de Schwöschter von which man have.2sg said that you with the sister of em1 > i d Schuel bisch? he.dat in the school are lit.: ‘Which man did you say that you went to school with a sister of?’ The derivation of A’-splits is as in (38). Given the fact mentioned in 6.3 above that gap derivations block resumptive derivations, the apparent optionality between movement and resumption in wh- and topicalization constructions comes as a surprise. Our approach to long-distance movement, however, provides a solution: The intermediate C differs in movement vs. base-generation to an extent that one can assume that different reference sets are involved. While it has EPP and uContrast under movement, it lacks such features under base-generation. This difference is not found in local extraction where movement and base-generation are based on the same lexical items and where the functional heads do not differ in their feature specification.⁶³ We can

314   

   Josef Bayer and Martin Salzmann

thus avoid the optionality problem because A’-splits do not compete with movement derivations. Long-distance relativization is different, there is no optionality because the movement derivation is ruled out for independent reasons (the topicality of the restriction), only base-generation converges. We would finally like to note that the syntax of A’-splits can be extended to constructions involving fronted R-pronouns as in the following example: (77) Da1 kän i < käne, wo öppis de1voo verschtaat. > there know I no.one C something there.of understands ‘This issue I don’t know anyone who understands.’ Since R-pronouns do not need Case, they can be base-generated directly in the final landing site and be linked to a resumptive (the pronominal part of the pronominal adverb).⁶⁴

6.8 Resumption in relativization from V2 complements As discussed in 5.2, wh-movement of (sentence) topics from V2-complements is grammatical in standard German and is best analyzed as involving parenthesis. All there is is local extraction so that, as a consequence, no uContrast is involved that could cause problems for the movement of the restriction. It is therefore interesting to look at the situation in Zurich German relativization. One does find cases of (putative long-distance) relativization into a V2-complement. Interestingly, a resumptive is obligatory: (78) a. vuwäge miim Suh, won i gmeint haa, *(er) seig tod because.of my son C I thought have he be dead ‘because of my son, who I thought was dead’ Stalder (1819: 274) b. de Maa, won i gsäit han, ich heg *(en) im Tram gsee the man C I said have.1sg I have.subj him in.the tram seen ‘the man that I said I saw in the streetcar’ Since the relative marker wo(n) certainly belongs to the matrix clause, the strings i gmeint haa/i gsäit han would have to be analyzed as a parenthetical. However, neither do such strings ever occur in parenthetical function nor would the remaining sentence be grammatical since relative clauses require verb-final structure, which is incompatible with the V2 order of the complement. As a consequence, such examples do not involve parenthesis. The presence of resumptives shows

That-trace effects and resumption – How Improper Movement can be repaired   

   315

that movement is not possible; the question is just why. One obvious possibility is to assume that the embedded C bears an EPP-feature with uContrast as a subfeature. This would make extraction of the restriction of specific relative operators (which abstract over individuals) impossible, in accordance with (78) (for (78b) one could additionally argue that the embedded SpecCP is occupied so that a topic island blocks extraction). But this predicts that amount relatives should be possible from V2-complements. This prediction is not borne out, amount relativization into a V2-complement is ungrammatical and cannot be saved by resumption: (79) * di 2 Liter Champagner, won i gsäit han, (si) seged geschter the 2 liters champagne C I said have.1sg they be.subj yesterday trunke worde drunk become ‘the two liters of champagne that I said were drunk yesterday’ Since resumption is generally incompatible with semantic types other than (Salzmann 2006b), the impossibility of resumption in (79), which involves abstraction over degrees, is actually expected. But the impossibility of movement is surprising if the embedded C involves an EPP-feature. Since long-distance extraction is obviously not an option, it is arguably best to analyze V2-complements as ForcePs whose head cannot attract items for intermediate A’-movement (since V2-clauses are root clauses, the attracted wh-phrase reaches a scope position and is therefore frozen for further A’-movement). Consequently, V2-complements do not allow for long-distance extraction at all. Of course, this fact was documented for wh-movement and topicalization long ago (cf. Müller 2010 for an overview). What is interesting about the ZG facts is that resumption in V2-complements is possible.⁶⁵ Indirectly, these facts can also be taken as evidence in favor of the parenthetical analysis of wh-extraction from V2-complements. Concerning (78), since no long-distance movement dependency is possible with V2-complements and since a parenthetical analysis is ruled out, base-generation is the only option in the case of relativization. The case can be made even stronger with the observation that relativization into a V2-complement is only acceptable if the highest specifier of the embedded complement is filled. Consider the following sentence, which is a variant of (78b): (80) * de Maa, won i gsäit han, heg i (en) im Tram gsee the man C I said have.1sg have.subj I him in.the tram seen ‘the man that I said I saw in the streetcar’

316   

   Josef Bayer and Martin Salzmann

Since no extraction is involved and since a parenthetical analysis is also impossible, the prefield of the embedded V2-complement trivially has to be filled. The same obtains with A’-splits into V2-clauses (recall from above that in A’-splits the operator appears in a default nominative Case form while Case information is realized by the resumptive): SpecForceP has to be filled (we should mention that such examples are marked for most speakers, but the contrast is nevertheless clear):

(81) a.

[Wele Maa]1 häsch gsäit, de Hans heg *(em1) geschter which man have.2sg said the John has.subj he.dat yesterday ghulffe? helped ‘Which man did you say John helped yesterday?’

b. * [Wele Maa]1 häsch gsäit, heg (em1) de Hans geschter which man have.2sg said has.subj he.dat the John yesterday ghulffe? helped ‘Which man did you say John helped yesterday?’ The ungrammaticality of (81b) can be explained as follows: A parenthetical analysis is theoretically possible since häsch gsäit can occur as a parenthetical and the wh-operator is compatible with the V1 clause heg (em) de Hans geschter ghulffe. This would imply that (81b) only involves short extraction. However, A’splits are independently impossible in local wh-movement: (82) * [Wele Maa]i häsch emi geschter ghulffe? which man have.2sg he.dat yesterday helped ‘Which man did you help yesterday?’ The reason for this is simple: Since a movement derivation is possible in this context, it will block the A’-split/resumptive derivation – as in the relativization of local subjects/direct objects (cf. 6.3). Since a parenthetical analysis (with local extraction) is blocked by economy (81b), the only remaining possibility involves a real V2-complement. However, since long-distance extraction is not possible from V2-complements, base-generation remains as the sole option. In that case, however, SpecForceP has to be filled, since V2 clauses require the highest specifier to be overt. Consequently, only the version with a filled specifier is grammatical, (81a).

That-trace effects and resumption – How Improper Movement can be repaired   

   317

7 Conclusion In this article we have argued in favor of a new perspective on that-trace effects. We have shown that that-trace effects are not peculiar to subjects, but obtain more generally if a sentence/aboutness topic undergoes long-distance extraction. Since in English, the subject usually coincides with the topic, that-trace effects appear to be subject-related. Data from German, however, show that similar effects obtain with non-subjects that are extracted from the highest position in the middle field. The fact that that-trace effects are much less visible in German can be related to the greater flexibility of the German clause: Since it arguably does not have a fixed subject position, but just a high topic position, other elements than the subject can occupy the topic position thereby making extraction of the subject from a lower position possible. In this case, the subject does not figure as the aboutness topic, and it can undergo long-distance movement. The impossibility of long-distance movement of sentence topics was subsumed under an independently established constraint barring long-distance movement of noncontrastable elements; we took this to be due to a feature uContrast located on the intermediate complementizer dass. The concrete implementation involves the following ingredients: Operator and restriction are taken to be independent syntactic objects that undergo independent checking operations and that have to be amalgamated under adjacency at PF. While Op is directly merged in the matrix scope position (ForceP), the restriction originates in the argument position and moves into the matrix clause to amalgamate with Op. The degradedness of long-topic movement then comes from the fact that the intermediate C endowed with an edge/EPP-feature is taken to involve a feature uContrast as a subfeature. uContrast is incompatible with topical elements. Whenever a topical element – in our case the restriction functioning as the sentence topic of the embedded clause – moves via SpecCP, it will fail to check uContrast of C and the derivation crashes. Extraction is possible once some other element functions as the sentence topic and the restriction can bear iContrast. Extraction of amount phrases provides independent evidence for this approach: Since their restriction is inherently nontopical, there will be no movement via TopP; the restriction can bear iContrast, value the corresponding feature on the intermediate C and move into the matrix clause to amalgamate with Op. In the final part of the article we showed that our approach to long-distance movement has wider coverage. It accounts for a hitherto puzzling fact about Zurich German relativization: While subjects and direct objects are relativized by means of movement in local relativization, they require resumptives as soon as a sentence boundary is crossed. Since resumptives normally occur only as a last resort when gap derivations fail, long-distance relativization by movement

318   

   Josef Bayer and Martin Salzmann

must be ungrammatical. We have linked this to the topical nature of the restriction of relative operators that makes them incompatible with uContrast on the embedded C. As opposed to the standard language, ZG can resort to an alternative mechanism to establish the A’-dependency, viz. resumption where the restriction is realized by an independent element and therefore does not have to undergo movement into the matrix clause for purposes of amalgamation. While resumption is a well-known repair strategy to bridge opaque syntactic domains like islands, the case we are documenting here is remarkable because resumption/ base-generation serves to avoid an information structural conflict: Since under base-generation intermediate Cs do not have any attracting features, uContrast is automatically missing as well. Consequently, there is no risk of there being an unvalued uContrast. Thus, the derivation converges.

8 Acknowledgments Parts of this work have been presented over the years at various conferences and workshops by the authors individually but also jointly, most notably at the repair workshop at the DGfS 2009 in Osnabrück (February 2009) and at the Swiss Workshop in Generative Grammar (SWIGG), Neuchâtel (April 2009). We wish to thank the respective audiences for discussion. We have benefited from discussions with Markus Bader, Ellen Brandner, Peter Culicover, Nomi Erteschik-Shir, Marcel den Dikken, Werner Frey, Alessandra Giorgi, Jane Grimshaw, Jana Häussler, Uli Lutz, Amanda Pounder, Ur Shlonsky, Luigi Rizzi, Yvonne Viesel. None of them must be held responsible for any shortcomings of our analysis. Comments by Patrick Brandt and an anonymous reviewer helped us greatly to improve the quality and readability of this article. Thanks to Annika Nitschke for editorial help. For support of this research we thank the former Konstanz SFB 471 and the Swiss National Science Foundation (grant nr. PBSK-1--119747/1).

Notes 1 We will, of course, comment on the special role of contrastive topics, cf. section 4.2 below. 2 Some of the results have been published in Bayer (2005). 3 There has been a long debate whether the locatives in Locative Inversion are subjects (at least at some point of the derivation). See Bruening (2010) for a recent overview and strong arguments against the subject status of the inverted locatives; see also Salzmann (this volume).

That-trace effects and resumption – How Improper Movement can be repaired   

   319

4 One should critically add here that the approach is close to indistinguishable from Rizzi’s (1990) approach in which the wh-subject in the intermediate SpecCP was taken to turn the zero complementizer into a proper head governor for the subject trace via spec-head agreement. 5 We hope to be realistic in our assumption that objects or adverbs like unfortunately would under no theoretical assumptions qualify for satisfaction of the subject criterion. 6 Locality in the sense of their constraint Attract Closest X (ACX), cf. Pesetsky & Torrego (2001: 369). 7 In their footnote 20, pp. 407–208, they suggest that such an extension could be achieved. 8 See Grewendorf (1988), Haider (1983) and also Müller (1995) as well as the dissenting voice of Fanselow (1987). See Featherston (2005) and Kiziak (2010) for a more recent overview. 9 We note here that Featherston finds a more general subject/object asymmetry. His experiments do not probe into the highest clausal position. Therefore, his results cannot be compared with ours directly. 10 We make the plausible assumption that in the case of ambiguity the grammar computes only derivations which converge and leaves crashing derivations aside. Stranded quantifiers and the like restrict the choice between different trace positions. In fact, stranding is always better when the quantifier and therefore the trace is in a lower position. (i) Wer glaubst du, dass (?*alles) 1933 (√ alles) in Hamburg (√ alles) in der Regierung war? who believe you that all 1933 in Hamburg in the government was ‘Who all do you believe was 1933 in Hamburg in the government?’ We will return to similar cases shortly. 11 Consider the contrast in (i) involving quantifiers taken from Bayer (2012): (i) Hat denn in Venedig { jeder/ keiner } einen Hund? has DENN in Venice everybody nobody a dog ‘Does {everybody/nobody} in Venice have a dog?’ (ii) * Hat {jeder / keiner} denn in Venedig {jeder/keiner} einen Hund? One reviewer finds (ii) “impeccable”. This is unexpected because cursory internet searches reveal that there are hardly any occurrences of (fast) jeder (almost every), niemand (no one) nur XP (only XP) immediately preceding the particle denn in the relevant sense. 12 We are aware of a complication that we cannot explain so far. In German, DP-splitting in the topic position, as defined by discourse particles, tends to be degraded under local extraction, too i.e. extraction which does not cross an overt C. (i) Was2 würde der Peter denn t2 für Bücher kaufen? what would the Peter prt for books buy ‘What kind of books would Peter buy?’ (ii) ?? Was2 würde [t2 für Bücher]1 denn der Peter t1 kaufen? If the position above the modal particle is reached by scrambling, the deviance of (ii) and (12b)/ (13b) can be explained as a freezing effect. If scrambling is taken not to involve movement (as e.g. in Fanselow 2001), the degradedness of these examples can be linked to the specificity of the DPs in question: Specific DPs are often assumed to be islands for extraction. See Meinunger (2000) for extensive discussion of such cases and fn. 37 below. 13 Notice that there is also the broader notion of ‘discourse topic’. There can be more than one discourse topic, and definite descriptions or pronominals referring to identified discourse referents may be distributed over various positions of a clause. ‘Aboutness topics’, on the other hand, appear to be formally defined. There is only one aboutness topic per sentence, and, as Frey (2004; 2006) has shown, this topic seems to occupy a fixed high position in the clause

320   

   Josef Bayer and Martin Salzmann

which can be identified by its position relative to high sentence adverbs or certain modal particles. See also section 4 below. 14 Two precursors are worth mentioning here: (a) Bennis (1986: 225) suggests an Empty Presupposition Condition (EPC) which implies that extraction (not necessarily of a subject) takes place from a position which is preceded by presuppositional material. Since topics are typically part of the presupposition, the EPC translates into the claim that the highest position is occupied by some kind of topic, and that extraction must proceed from a position below the topic. (b) Meinunger (2000: 185) turns to the topic position directly. His Generalized Specificity Condition (GSC) states that topics are islands. Neither the EPC nor the GSC refer to C, but it would be surprising if these generalizations were independent of the that-t filter. 15 Since Reinhart (1981), a popular worry has been that quantified DPs such as every student cannot be topics because there is no entity or set which these DPs would refer to. Accordingly, Gilbert complained but not Every student complained would have a sentence/aboutness topic. This consequence violates elementary intuitions and must be built on some misanalysis. The solution lies in the fact that quantifiers operate on sets, and once the quantifier is stripped off, these sets make plausible topics. A sentence like Every student complained is then about a set of (discourse-relevant) students. See also Gundel (1999). The same applies to whphrases, as Koster (2003) points out, or negative indefinites (under no circumstances etc.) This insight squares with the general insight from formal semantics that quantified DPs need to be decomposed in one way or the other. See Erteschik-Shir (1999). Endriss and Hinterwimmer (2008) suggest analyzing indefinites as “indirect aboutness topics”. Although they do not deal with quantified DPs proper, one can assume that the restrictive part of a quantified DP conforms to their notion of an indirect aboutness topic. We will return to decomposition in section 4. 16 Cf. Salzmann (this volume) for arguments that topics can also target SpecTP in English, thereby satisfying the EPP. 17 See Bayer and Suchsland (1998) for discussion. 18 See Haider (2000) for discussion of word order in connection with functional heads. Various researchers (e.g. den Dikken 2007) observed a contrast in Dutch between sentences with and without verb inversion as in: (i) a. … dat gelachen werd (ok for many) that laughed was b. … dat er gelachen werd (ok for everyone) (bad for everyone) c. … * dat werd gelachen werd d. … dat er werd gelachen werd (ok for many) To find a similar contrast in German is hampered by a number of factors, but it appears as if the core constraints are not really different. German has a rule of inversion if two or more verbs are followed by the auxiliary haben. Consider (ii). (ii) a. … dass gelacht werden können hätte (strange for an independent and irrelevant reason) that laughed become can had b. … ?? dass hätte gelacht werden können hätte c. … dass damals hätte gelacht werden können hätte Interestingly, once inversion has applied, i.e. the auxiliary precedes VP, the position to the immediate right of C cannot remain empty. As seen in (iic), (iib) improves to full grammaticality as soon as some topical material is inserted between dass and hätte. One reviewer denies the degradedness of (iib). We would therefore like to refer the reader to an experimental

That-trace effects and resumption – How Improper Movement can be repaired   

   321

investigation of constructions like (iib) in Salzmann, Häussler, Bader & Bayer (to appear), in which the phonological adjacency of C and the finite verb turns out to be strongly disfavored by native speakers. 19 Consider the ill-formedness of (i) (i) * Wie viel Champagner glaubst du, dass {er/ das} getrunken wurde? how much champagne believe you that it/ that drunk was Intended: ‘How much champagne do you believe was drunk?’ 20 Note that these facts argue against PF-solutions to that-trace effects as in Ackema & Neeleman (2004), Ackema (2010), and Kandybowicz (2006). Let us add that one of our German speaking reviewers claims to not get the contrast in (20) that we and others see here quite strongly. 21 Recall here what we said about operators in topic position in note 15. In the next section we will discuss how the “topic part” of the wh-expression can be distinguished from the “operator part”. Our proposal concerns only the topic part, of course. 22 The original generalization is from Bhatt’s (1999) account of Kashmiri clause structure. Cf. Frey (2006) for cases of FF of object-es. Thus, in German the position in question cannot be limited to subjects. FF targets what Cardinaletti (2004) identifies as a “weak subject position”. As can be expected, it also includes quirky subject-like elements like the dative in the following German example. (i) Mir gefällt das Buch. (derived from: (dass) mir das Buch gefällt) me.dat pleases the book.nom ‘I like the book.’ 23 Exactly the same holds for English: (i) * Unfortunately1, Carl said [t1 [that [t1 no one helped the old man]]] 24 Unlike in Hungarian, focus in German materializes inside VP or in a VP-related lower functional FocP. Thus, the movement seen in ET or in contrastive fronting must not be confused with focus checking. By the time ET or contrastive fronting apply, focus checking has already applied. An early observation of this difference can be found in Behaghel (1932). ET/contrastive fronting target the highest position of a split CP and thus seem to be discourse-related. 25 One reviewer doubts that the contrastive (or corrective) DP the FBI is an aboutness topic. He or she infers this from the fact that it is focal. However, Krifka, to who this reviewer refers, says in (2008b: 267–268) explicitly, and in our view correctly, that they are “aboutness topic(s) that contains a focus, which is doing what focus always does, namely indicating an alternative”. Our reviewer identifies in our example the killing of the president as the aboutness topic. But this would follow only if the aboutness topic is identified with old information. As Krifka points out, such identification is not justified. Consider his example (i) A: What do your siblings do? B: [My [SIster]Focus]Topic [studies MEDicine]Focus, and [my [BROther]Focus]Topic is [working on a FREIGHT ship]Focus Although B’s answer contains foci, it is about the siblings. So we have a case of “focus-intopic”. 26 To see this, consider the statement The king of France visited Konstanz University. It suffers from a presupposition failure because the sentence is about a non-existent entity. The sentence Konstanz University was visited by the king of France has a topic which refers successfully. By checking all the visitors of Konstanz University, once can see that the set contains no person like that. Thus, it can be determined that this sentence is simply wrong.

322   

   Josef Bayer and Martin Salzmann

27 See also Krifka (2008a,b), for whom contrastive topics can be regarded as constituents whose denotations serve the function of both ‘addressation’ and ‘delimitation’. 28 Unless we say otherwise, we always mean ContrastA. ContrastA excludes ContrastT, which may appear as a subfeature of a topic as in (31). 29 Of course, there can be more than one topic. The point is that only one of them serves as the aboutness topic. The others are topics by virtue of being discourse referents familiar from previous discourse. To the best of our knowledge, discourse topics can move to Top – actually a “topic field” – or stay in some scrambling position behind. Consider the optionality of movement in the second clause of the dialogue in (i) where denn delimits the topic field. (i) A: Karl hat einen BMW gekauft. Karl has a BMW bought ‘Karl bought a BMW.’ B: Hat er [den Wagen] denn [den Wagen] bar bezahlen können? has he the car PRT the car cash pay could ‘Could he pay the car in cash?’ Sentence B is invariably about Karl. So Karl is the aboutness topic; den Wagen, referring to the car Karl has bought, is a discourse topic. 30 The ban on Improper Movement requires that types of movement must not be mixed. If A’-movement has applied to some DP, this DP cannot undergo A-movement afterwards, cf. Chomsky (1973) as a starting point. For a detailed treatment which adds scrambling to the inventory of movement types cf. Müller & Sternefeld (1993), for a very recent implementation, cf. Müller (2012). 31 Note that our analysis has implications for the verb-second constraint: Normally it is assumed that only one constituent can appear in front of the finite verb. In our case, we actually have two constituents, at least at LF. Crucially, though, at PF only one element remains. This could imply that the verb-second constraint is actually a PF-constraint. Given the complexity of this issue we leave a discussion of the consequences for further research. The same question arises for negative indefinites. Cf. Penka (2007: 117–119) for discussion. But note that this assumption accords well with the observations in Meinunger (2006) that verb second movement is subject to prosodic restrictions. 32 This analysis may raise questions w.r.t. the Case morphology since in this derivation only the restriction is assigned Case while the operator remains Case-less. This seems to be at odds with the tendency in German for Case morphology to be only visible on the operator and much less so on the restriction. Note though, that the operator is not simply to be equated with D and the restriction with NP. Rather, we assume that the restriction is a full DP so that the case assigned to it will eventually be realized on the spelled-out D-part that results after amalgamation. 33 Note that while the EPP-features on heads like Force, Fin or Top basically function as movement diacritics that indicate that feature valuation must be followed by internal merge, the EPP-feature on intermediate C-heads is independent of uContrast, it just requires that some constituent is moved to the edge domain of C, checking/valuation of unvalued features (such as uContrast) is not a precondition. This difference should be borne in mind given that we make no terminological distinction. 34 Even if one were to allow for the possibility that the restriction undergoes formal fronting to the matrix FinP, uContrast of the intermediate C would still remain unvalued. Additionally, one has to avoid that the operator starts out in the intermediate SpecCP as it could check uContrast of C so that the derivation should converge, contrary to fact. We will provisionally assume that

That-trace effects and resumption – How Improper Movement can be repaired   

   323

for semantic reasons the operator can only be inserted into positions where it can check its uOp-feature (furthermore, it seems reasonable to assume that edge-features on phase-heads can only be checked by means of internal merge). 35 Resumption under wh-movement is strongly degraded in German (cf. also section 6.7 below) but somewhat more acceptable in English. Resumptives are marginally available for object extraction as well. This may seem surprising since topicality cannot be at stake. Rather, what seems to be the case is that base-generation is an independent option which is optional for objects but necessary for subjects to avoid a that-trace effect. The fact that one reviewer finds (37) deviant is not surprising. Repair by resumption avoids the that-t effect but may induce other sacrifices; it is well-known that resumptive pronouns in English are not a fully grammatical device (but rather should be classified as so-called intrusive resumptives, cf. Chao & Sells 1983). Nevertheless, it cannot be denied that resumption is used to avoid that-trace violations. Witness the following examples. (i) Who do you think that he/she is the best Popstar Dancer? http://forums.denden.co.uk/viewtopic.php?f=1&t=205917&start=25 [05.05.2012] (ii) Who do you think that he will win in the final of Wimbledon 2009?? http://www.fanpop.com/spots/tennis/picks/results/262913/who-think-will-win-finalwimbledon-2009 [05.05.2012] (iii) Who do you think that he will save those who blasphemed him or those who love him? http://www.thebereans.net/forum2/archive/index.php/t-50913.html [05.05.2012] 36 We assume, as is standard, that such movement triggering features are in principle optional (for instance, they must be absent when there is no A’-dependency). If they are not present, long-distance movement will not be an option and a grammatical result with an A’-dependency is only possible if a language allows for resumption. 37 To see that scrambling inside VP does not hamper subsequent A’-movement, consider the following example in which one can assume that the DO has been scrambled to the left of IO as shown in (ii). (i) Was glaubst du, dass der Chef wohl alles der Stiftung vermacht hat? what believe you that the boss prt all the.dat foundation donated has ‘What all do you think that the boss has donated to the foundation?’ (ii) Was1 glaubst du, [cp t1 dass der Chef wohl [vP … [t1 alles]2 der Stiftung t2 vermacht hat]]? 38 See the proposal that movement must not apply string-vacuously. Gazdar (1981), Chung & McCloskey (1983), Chomsky (1986: ch.9), Grimshaw (1997) and Haider (2004) argue in one way or the other that the wh-subject does not move. 39 Our analysis seems to imply that wh-subjects that originate as aboutness topics are obligatorily contrastive (ContrastT in our sense) since the operator chooses from a contextually determined set. One might therefore expect only D-linked wh-phrases to occur, but no simplex wh-items. For reasons of space, we have to leave a full discussion of this issue for another occasion. 40 One must be careful about one prediction that our analysis seems to make. We apparently predict that an XP that has undergone A’-movement to Spec, CP cannot be the topic in the root clause. The following two examples seem to contradict our analysis. (i) Speaking of Tom, Dick and Harry, … who (of them)1 do you believe that under the given circumstances t1 could have been the victim? [Subject extraction from a lower position] (ii) Speaking of Tom, Dick and Harry, … who (of them) do you believe that we should give the Nobel prize? [Object extraction]

324   

   Josef Bayer and Martin Salzmann

Two considerations militate against such a conclusion: First of all, the aboutness test which is used in (i) and (ii) (speaking of …, as for X … etc.) is not sensitive to the sentence topic but may rather pick up any discourse referent. Secondly, it would be a mistake to believe that topicality – essentially a discourse property – can be passed from one clause to the next like, say, grammatical function or Case. It is more likely that information structure is computed for each clause. The issue is too complex to address it in a footnote. For the moment it is only important to see that cases like (i) and (ii) are no challenge to the theory presented; cf. also fn. 54. 41 Cf. Steinbach (2007) and the discussion of further proposals in Viesel (2011). To be sure, the parenthesis analysis is by no means innocent. It creates questions about the V2-constraint because in this analysis the material immediately preceding the finite verb can actually not be its specifier. Furthermore, this material does not necessarily consist of a single constituent while this is is normally a solid column on which the grammar of V2 rests. 42 According to Cardinaletti (2004), (50a/b) should be ungrammatical. In her account this is so because they involve “weak subjects”. Weak subjects are universally restricted to a low nominative- and phi-feature-checking position. We admit that (50a/b) are degraded once the V1-construction is prosodically separated but since V1-parentheses are prosodically integrated, the problem does not arise. Notice also that there must be a phonological factor involved. Examples of this sort improve when the weak subject is immediately followed by the verb as in (i) Es hat, glaube ich, geregnet. (ii) Man sollte, finde ich, nicht so viel Lärm machen. Nevertheless, it is easy to find examples like (50) on the internet with nothing but the noncontrastable element before the parenthetical. (iii) Man finde ich sollte für das Buch der Richter den Titel „Wegbegradiger“ oder one find I should for the book (of) the judges the title “path-straightener” or was ähnliches wählen something similar choose ‘I think one should choose for the Book of Judges the title “straightener of the path” or something similar.’ http://wiki.volxbibel.com/Benutzer_Diskussion:Martin [04.05.2012] (iv) die Sachen zusammen gesucht, die man finde ich braucht the things together searched which one find I needs ‘collected the things which, I think, one needs’ http://www.das-hamsterforum.de/index.php?page=Thread&threadID=96704 [04.05.2012] (v) es finde ich hat auch etwas mit dem essen zu tun ob es schlimmer wird it find I has also something with the eating to do if it worse gets ‘I think it also has to do with eating whether it gets worse …’ http://www.rund-ums-baby.de/forenarchiv/schwanger-wer-noch/Neurodermitis_50219. htm [04.06.2012] 43 In spoken language, glaubst du and glaub ich tend to show enhanced cliticization and ultimately signs of grammaticalization as in 2nd person glaubsch and 1st person glaub. 44 In (53a), next week has been added to avoid the that-t effect and thus to guarantee that the example is excluded for only one reason, namely the deviant semantic selection of wonder. The issue is debated even outside linguistics as shown by an internet chat which can be found under: http://literalminded.wordpress.com/2009/01/27/who-may-i-ask-is-calling/ 45 Jane Grimshaw (p.c.) suggests that inversion (I-to-C movement) in the SAI-fragment is only possible where the scope of an “affective” operator is marked. Accordingly, What, do you wonder, is this all about? involves two instances of inversion; what must have taken scope

That-trace effects and resumption – How Improper Movement can be repaired   

   325

above as well as below wonder. The parenthesis solution offers a way out of this dilemma. There is only a simple wh-question (What is this all about?) into which the SAI-fragment is inserted. If the parenthesis solution turns out to be untenable, such examples as well as examples of subject questions like (53b) would have to be relegated to the realm of speech errors (whatever this would mean). More about this in the next footnote. 46 Examples which show apparent “double inversion” (see footnote 45) may not be standard but occur rather frequently. Here are some randomly collected examples: (i) Who do you think did he visit? http://twitter.com/#!/fashionkingkr [03.06.2012] (ii) Ideally – what do you believe did he need then? He asks softly. http://m.fanfiction.net/s/7756442/3/ [03.06.2012] (iii) How do you think did he make it? http://www.youtube.com/user/georgmeir [04.06.2012] (iv) How do you think did he simply disappear from history after 1945? http://in.answers.yahoo.com/question/index?qid=20110812214411AAhkWpw [04.06.2012] To us it would be surprising that these are all speech errors. The parenthesis hypothesis offers a serious alternative explanation: There is a matrix clause which – due to non-subject wh-movement – shows do-support. V1-subject-verb chunks are grammaticalized as integrated parentheticals; these parentheticals may enter simplex questions with inversion. Thus, it seems highly unlikely that we are facing some kind of unmotivated “double inversion” here. Interesting support for our conjecture comes from the following quote from an internet discussion about English grammar by non-linguists in which one discussant reveals his/her intuitions in a hesitating but nonetheless highly revealing way: “In vernacular, I suppose you could speak with the invisible commas to make the sentence grammatically correct (this is where I became so perplexed). I.e., „What, do you think, is the best way to learn English?“ I believe this would be my personally most natural way of asking this question. I tend to speak just a little differently than most people, however, so I‘m curious about what others think.” http://www.englishforums.com/English/Interrogatives/qxhlv/post.htm [03.06.2012] Although this is an unwarranted intuition by an untrained native speaker, we take it as constituting strong support for an analysis in terms of parenthetical insertion. 47 The only example from the internet we found is the following (i) What do you deny that it had, at the singing and shouting and … ? http://www.google.de/interstitial?url=http://hostflux.com/Wisconsin_Web_Hosting/ KOHLER_Web_Hosting/1613/botox-cluster-headaches.html 48 Here are some examples: (i) When you write only in English who do you assume will be your reader? (ii) So with that, who would you assume is now playing in the 2007 Stanley Cup? We wish to thank Benedikt Grimmler, who ran these searches in August 2007, and to Peter Culicover for stirring this discussion. Our Canadian informants David Bird and Amanda Pounder accept slifting in examples with assume as well as with suppose, imagine, guess, predict, all of which seem to be able to be used as integrated parentheticals: (iii) When will the guests arrive, do you {suppose/imagine}? (iv) When will the guests arrive, would you {guess/predict}? 49 For the transcription see Salzmann (2006: 320, fn. 259). For possessor relativization, cf. Salzmann (2011). Free relatives require wh-relative pronouns that leave gaps, cf. van Riemsdijk (1989). The syntax of dative relativization is more complex. With certain verbs (especially experiencer verbs), neither a gap nor a resumptive leads to a completely well-

326   

   Josef Bayer and Martin Salzmann

formed result, cf. Salzmann (2006b: 323–326). Additionally, there generally is a lot of interand intra-speaker variation: many speakers accept both gap and resumptive, cf. Salzmann (2008/2009b, 2012), Salzmann & Seiler (2010). 50 Wh-extraction does not improve with resumption. We will come back to the incompatibility of regular wh-movement with resumption in 6.7 below, cf. also Salzmann (2011). 51 See van Riemsdijk (1989) for an analysis where gap relatives are based on resumptive relatives. Salzmann (2009a: 41–42, 2009b: 143–145, 2009c: 66–67) argues extensively against this position, showing that gap relatives differ from resumptive relatives in a number of important properties so that a unification is undesirable. 52 A different (re-)analysis of long distance relativization is proposed in Salzmann (2006b), building on van Riemsdijk (2008), where it is argued that what looks like a long-distance dependency actually involves short extraction in the matrix clause and a somewhat different type of pronoun binding, i.e. a more abstract version of ex. (70) below; for reasons of space, we cannot provide a comparison of the two approaches; cf. also fn. 59. 53 Since relative operators always move on to a position above SpecTopP, the intermediate movement step can only be shown indirectly by means of floating quantifiers that can be stranded above discourse topics: (i) die Leute, die alle den Peter vermutlich nicht mögen the people who all the Peter probably not like (Standard German) Admittedly, the quantifier can also be stranded in lower positions. We will assume that it is essentially optional where a quantifier is stranded and that relative operators always move via SpecTopP even in cases where this cannot be inferred from the position of floating quantifiers. 54 Our analysis seems to predict that once a constituent has undergone A’-movement (to an intermediate position), it cannot become a (sentence) topic in the matrix clause (recall also the discussion in fn. 40). Long relativization would in principle seem to instantiate such a constellation (as does long-distance scrambling). In German and its varieties it is ruled out because of uContrast on the intermediate C (amount relatives constitute an exception, cf. 6.6.1 below, but they arguably do not involve topicality in the main clause). While Swiss German can resort to resumption in this case, long relativization in Standard German is strongly degraded if not ungrammatical, cf. section 6.6.2 below. Long relativization in English, however, is unproblematic, (as long as what is extracted does not correspond to the aboutness subject). This may be the case because topicality is not checked in the matrix clause syntactically or because information structural properties are not inherited from one clause to the next. We have to leave this intricate issue for future research. 55 Following Rizzi (1997) we assume that Relative operators eventually move to ForceP, the same position into which we took wh-operators to move. This may not be completely obvious since restrictive relatives (unlike appositives) do not have independent illocutionary force. Given that there is little positional evidence in German for an articulated left periphery, we will, however, not posit more positions; rather, we will just posit two, one that is related to clausetyping (“ForceP”) and one that is neutral (“FinP”); consequently, the label “Force” should not be taken too literally. Given that wo is a relative complementizer, it may be more desirable to have the uOp feature on wo rather than on Force. One possibility to achieve this is to assume that a single head can combine the features corresponding to those of FinP and ForceP, so that several constituents can be attracted to its specifiers. Alternatively, one can take wo to lexicalize Force°. This seems to create problems for amalgamation, which however, may not be relevant in relativization, cf. fn. 56

That-trace effects and resumption – How Improper Movement can be repaired   

   327

56  Operator and restriction remain unpronounced/are PF-deleted in Zurich German relatives, like in many other languages; in fact, the restriction is usually deleted even in languages with overt relative operators, at least under a traditional head-external analysis (or a Matching Analysis, cf. Salzmann 2006, where it is deleted under identity with material contained in the exernal head). This shows that there must be an independent trigger for movement of the restriction (viz. an EPP-feature). If, however, a head-raising analysis is adopted, the restriction would move on to become the head of the relative clause; this is not directly compatible with our assumptions about amalgamation. 57 We assume here that uOp on Force can be checked by means of external Merge. Alternatively, one would have to assume that Force comes in two variants, one with uOp (movement) and one with an EPP-feature only (base-generation), cf. Alexopoulou (2006: 88) for such an analysis of Greek pu. 58 It is interesting to look at Dutch in this context. Even though long-distance relativization is generally taken to be possible, it nevertheless seems restricted as well. Many speakers prefer alternative constructions similar to the Standard German one in (70). The situation in dialects is particularly interesting. The various patterns described in Boef (2008) show that many dialects resort to different means, which may be indicative of extraction difficulties. Barbiers et al. (2005) explicitly suggest that long-distance relativization is impossible in many dialects. To what extent the present analysis can be extended to Dutch is a question we intend to pursue in future work. 59 For reasons that are not clear to us, long-distance relativization appears to have been more acceptable in earlier stages of German (cf. Lühr 1988: 78–79); whether there has been a shift from long-distance relativization to the alternative construction in (70) is not quite clear since they seem to have co-existed for quite some time; cf. Lühr (1988: 79, fn. 24) for discussion. 60 It seems that the restriction on long relativization (and long topic movement more generally) is tied to the finite complementizer dass while non-restructuring infinitives as in (71), which are normally classified as CPs, do not have such a blocking effect. We attribute this to the absence of finiteness, a phenomenon that is widely attested since Chomsky (1973). 61 As indicated in the text, it seems to us that resumptives are often acceptable in non-finite clauses. Hodler (1969: 247) gives such an example: (i) Usdrück, wo si e rächte Möntsch schämti sen i ds Muu expressions C self a decent human.being would.be.ashamed them in the mouth z’ näh. to take ‘expressions that a decent human being would be ashamed to use’ If, as proposed in Salzmann (2006b), long-distance relativization in ZG can also be analyzed as an abstract version of the proleptic construction in (70), the possibility of resumption in (i) would be less surprising. It would be parallel to the fact that resumptive prolepsis is also possible with non-finite complements in standard German, cf. Salzmann (2006b: 205). The fact that long-distance movement and the proleptic construction do not block each other implies that they do not compete and thus do not belong to the same reference set. 62 There are no explicit statements about resumption in wh-movement and topicalization in the traditional dialect literature. One does, however, find examples in Weber (1964: 305) and Suter (1976: 186, §319). Crucially, all examples with resumption involve direct relations so that they can all be reanalyzed as A’-splits. 63 Since resumptives merely realize the restriction at PF, they are not part of the numeration. The only problematic case is the movement of the restriction which we have taken to be

328   

   Josef Bayer and Martin Salzmann

triggered by an EPP-feature on Fin° in (63). Such a feature will be absent under (local) basegeneration. To make the numerations perfectly identical, one would have to assume that movement of the restriction is not feature-driven (perhaps simply repair-driven to allow for amalgamation with Op). For reasons of space we have to leave a full discussion of this issue for a later occasion. 64 There is an optionality problem with R-pronouns, though, that we cannot account for: Next to the splitting construction as in (77), one also finds a pied-piping-like structure. At least in transparent contexts (unlike (77)), both derivations are possible: (i) Dadevoo1 verschtaan i t1 nüüt. there.there.of understand.1sg I nothing ‘I don’t understand anything of it.’ (ii) Da1 verschtaan i nüüt de1voo. there understand.1sg I nothing there.of We leave this for future research. See also Bader & Bayer (2007) for a treatment of prepositions as exponents of oblique Case; thereby one is no longer forced to say that Case-assignment by prepositions can be suppressed in the context of R-pronouns. 65 An alternative explanation for the opacity of V2-complements can be found in the following observation that embedded wh-sentences are incompatible with V2: (i) Ich möchte wissen, *wohin ist Susi gefahren / wohin Susi gefahren ist. I would.like know where is Susi driven where Susi driven is ‘I would like to know where Susi drove to.’ A possibility to unify the island nature of V2-complements with the data in (i) is to follow Heck (2010) in assuming that V2-complements are adjoined to VP. According to Heck, the impossibility of embedded wh-V2 as in (i) could then follow from the fact that the matrix predicate does not c-command the embedded C (since normal V2-complements satisfy a thetarole of the matrix verb, special provisions have to be made for theta-role assignment, e.g. as in Fanselow 2001). For a very different explanation of the long extraction restriction from V2complements, cf. Müller (2010).

9 References Ackema, Peter. 2010. Restrictions on subject extraction: a PF interface account. In: Raffaella Folli and Christiane Ullbrich (eds.), Interfaces in Linguistics. New Research Perspectives, 225–241. Oxford: Oxford University Press. Ackema, Peter and Ad Neeleman. 2004. Beyond Morphology. Oxford: Oxford University Press. Alexopoulou, Theodora. 2006. Resumption in Relative Clauses. Natural Language and Linguistic Theory 24: 57–111. Andersson, Sven-Gunnar and Sigmund Kvam. 1984. Satzverschränkung im heutigen Deutsch. Eine syntaktische und funktionale Studie unter Berücksichtigung alternativer Konstruktionen. Tübingen: Gunter Narr (= Studien zur deutschen Grammatik 24). Aoun, Josef, Lina Choueiri and Norbert Hornstein. 2001. Resumption, Movement, and Derivational Economy. Linguistic Inquiry 32/3: 371–403. Barbiers, Sjef, Hans Bennis, Magda Devos, Gunther de Vogelaer, and Margreet van der Ham. 2005. Syntactic Atlas of the Dutch Dialects, Volume 1. Amsterdam: Amsterdam University Press.

That-trace effects and resumption – How Improper Movement can be repaired   

   329

Bayer, Josef. 1984. Comp in Bavarian syntax. The Linguistic Review 3: 209–274. Bayer, Josef. 2001. Asymmetry in emphatic topicalization. In: Caroline Féry and Wolfgang Sternefeld (eds.), Audiatur Vox Sapientiae, 15–47. Berlin: Akademie-Verlag. Bayer, Josef. 2005. Was beschränkt die Extraktion? Subjekt – Objekt vs. Topic – fokus. In: Franz Josef D‘Avis (eds.), Deutsche Syntax: Empirie und Theorie, 233–257. Göteborg: Acta Universitatis Gothoburgensis. Bayer, Josef. 2012. From modal particle to interrogative marker: a study of German denn. In: Laura Brugè, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro and Cecilia Poletto (eds.), Functional Heads. The Cartography of Syntactic Structures, vol. 7, 13–28. Oxford: Oxford University Press. Bayer, Josef and Peter Suchsland. 1998. Some remarks on expletives in German. Paper for Noam Chomsky on his 70th birthday. Electronically available via the MIT Press websidehttp://mitpress.mit.edu/celebration. Bayer, Josef, Markus Bader and Michael Meng. 2001. Morphological underspecification meets oblique Case: syntactic and processing effects in German. Lingua 111: 465–514. Bayer, Josef and Markus Bader. 2007. On the syntax of prepositional phrases. In: Andreas Späth (ed.), Interface and Interface Conditions, 157–179. Berlin: Mouton de Gruyter. Behaghel, Otto. 1932. Deutsche Syntax. Eine geschichtliche Darstellung, Vol. 4. Heidelberg: Winter. Bennis, Hans. 1986. Gaps and Dummies. Dordrecht: Foris. Bhatt, Rakesh Mohan. 1999. Verb Movement and the Syntax of Kashmiri. Dordrecht: Kluwer. Bianchi, Valentina. 2004. Resumptive relatives and LF chains. In: Luigi Rizzi (ed.), The Cartography of Syntactic Structures, 76–114. Oxford: Oxford University Press. Boeckx, Cedric. 2003. Islands and Chains. Resumption as Stranding. Amsterdam u.a.: John Benjamins. Boef, Eefje. 2008. Long-distance relativization in varieties of Dutch. MA Thesis, University of Utrecht. Bresnan, Joan. 1977. Variables in the theory of transformations. In: Peter Culicover, Tom Wasow, and Adrian Akmajian (eds.), Formal Syntax, 157–196. New York: Academic Press. Bresnan, Joan. 1994. Locative inversion and the architecture of Universal Grammar. Language 70: 72–131. Bresnan, Joan and Sam Mchombo. 1987. Topic, pronoun, and agreement in Chichewa. Language 63: 741–782. Bruening, Benjamin. 2010. Language-particular syntactic rules and constraints: English locative inversion and do-support. Language 86: 43–84. Cardinaletti Anna. 2004. Toward a cartography of subject positions. In: Luigi Rizzi (ed.), The Structure of CP and IP. The Cartography of Syntactic Structures, Volume 2, 115–165. New York: Oxford University Press. Chao, Wynn and Peter Sells. 1983. On the interpretation of resumptive pronouns. Proceedings of the North East Linguistic Society 13: 47–61. Carlson, Gregory. 1977. Reference to Kinds in English. Ph.D. dissertation, University of Massachusetts, Amherst. Chomsky, Noam. 1973. Conditions on transformations. In: Stephen R. Anderson and Paul Kiparsky (eds.), A Festschrift for Morris Halle, 232–286. New York: Holt, Rinehart and Winston. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Barriers. Cambridge, Massachusetts: MIT-Press.

330   

   Josef Bayer and Martin Salzmann

Chomsky, Noam and Howard Lasnik. 1977. Filters and control. Linguistic Inquiry 8: 425–504. Chung, Sandra and Jim McCloskey. 1983. On the interpretation of certain island effects in GPSG. Linguistic Inquiry 14: 704–713. Cowart, Wayne. 1997. Experimental Syntax. Applying Objective Methods to Sentence Judgements. Thousand Oaks u.a.: SAGE. Culicover, Peter W. 1993. Evidence against ECP accounts of the that-trace effect. Linguistic Inquiry 14: 704–713. Demirdache, Hamida. 1991. Resumptive Chains in Restrictive Relatives, Appositives and Dislocation Structures. Doctoral Dissertation, Massachusetts Institut of Technology. Dikken, Marcel den. 2007. Questionnaire study on Dutch that-trace effects: Stimuli and results. Ms. City University of New York. Endriss, Cornelia and Stefan Hinterwimmer. 2008. Direct and indirect aboutness topics. Acta Linguistica Hungarica 55/3–4: 297–307. Erteschik-Shir, Nomi. 1997. The Dynamics of Focus Structure. Cambridge: Cambridge University Press. Erteschik-Shir, Nomi. 1999. Focus structure and scope. In: Georges Rebuschi and Laurie Tuller (eds.), The Grammar of Focus, 119–150. Amsterdam: Benjamins. Erteschik-Shir, Nomi. 2009. Canonical information structures. Lecture at Harvard University, 8 May, 2009. http://www.bgu.ac.il/~shir/harvard%208%20may%202009.pdf. Fanselow, Gisbert. 1987. Konfigurationalität: Untersuchungen zur Universalgrammatik am Beispiel des Deutschen. Tübingen: Narr. Fanselow, Gisbert. 2001. Features, theta-Roles, and free Constituent Order. Linguistic Inquiry 32: 405–437. Fanselow, Gisbert. 2002. Quirky „subjects“ and other specifiers. In: Ingrid Kaufmann and Barbara Stiebels (eds.), More than Words: a Festschrift for Dieter Wunderlich, 227–250. Berlin: Akademie Verlag. Fanselow, Gisbert and Denisa Lenertová. 2011. Left peripheral focus: Mismatches between syntax and information structure. Natural Language and Linguistic Theory 29: 169–209. Featherston, Sam. 2005. That-trace in German. Lingua 1091: 1–26. Frey, Werner. 2004. A medial topic position for German. Linguistische Berichte 198: 153–190. Frey, Werner. 2005. Zur Syntax der linken Peripherie im Deutschen. In: Franz-Josef d’Avis (ed.), Deutsche Syntax: Empirie und Theorie, 147–171. Göteburg: Göteborger Germanistische Forschungen 46. Frey, Werner. 2006. How to get an object-es into the German prefield. In: Patrick Brandt and Eric Fuß (eds.), Form, Structure, and Grammar. A Festschrift Presented to Günther Grewendorf on Occasion of his 60th Birthday, 337–352. Berlin: Akademie. Gazdar, Gerald. 1981. Unbounded dependencies and coordinate structure. Linguistic Inquiry 12: 155–184. Grewendorf, Günther. 1988. Aspekte der deutschen Syntax. Eine Rektions-Bindungs-Analyse. Tübingen: Narr. Grimshaw, Jane. 1997. Projection, heads, and optimality. Linguistic Inquiry 28: 373–422. Guilliot, Nicolas and Nouman Malkawi. 2006. When resumption determines reconstruction. Proceedings of the West Coast Conference on Formal Linguistics 25: 168–176. Gundel, Jeanette K. 1974. The Role of Topic and Comment in Linguistic Theory. Ph.D. dissertation, University of Texas. Gundel, Jeanette K. 1999. Topic, focus and the grammar pragmatics interface. In: Jim Alexander, Na-Rae Han and Michelle Minnick (eds.), Proceedings of the 23rd Annual Penn Linguistics

That-trace effects and resumption – How Improper Movement can be repaired   

   331

Colloquium. Penn Working Papers in Linguistics, 185–200. Philadelphia: University of Pennsylvania Press. Haider, Hubert. 1983. Connectedness effects in German. Groninger Arbeiten zur Germanistischen Linguistik 23: 83–119. Haider, Hubert. 2000. OV is more basic than VO. In: Peter Svenonius (ed.), The Derivation of VO and OV, 45–67. Amsterdam: Benjamins. Haider, Hubert. 2004. The superiority conspiracy: four constraints and a processing effect. In: Anton Stepanov, Gisbert Fanselow and Ralf Vogel (eds.), Minimality Effects in Syntax, 147–175. Berlin: Mouton de Gruyter. Hartmann, Jutta M. 2008. Expletives in Existentials: English there and German da. Utrecht: LOT. Heck, Fabian. 2010. A Constraint on recursion. Handout, DGfS 2010, Berlin. Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD dissertation, University of Massachusetts. Hodler, Werner. 1969. Berndeutsche Syntax. Bern: Francke Verlag. Kandybowicz, Jason. 2006. Comp-trace effects explained away. Proceedings of the West Coast Conference on Formal Linguistics 25: 220–228. Keenan, Edward L. and Bernard Comrie. 1977. Noun phrase accessibility and Universal Grammar. Linguistic Inquiry 8: 63–99. Kiss, Katalin E. 2002. The EPP in a topic-prominent language. In: Peter Svenonius (eds.). Subjects, Expletives, and the EPP, 107–124. Oxford: Oxford University Press. Kiziak Tanja. 2007. Long extraction or parenthetical insertion? Evidence from judgement studies. In: Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals, 121–144. Amsterdam: John Benjamins. Kiziak, Tanja. 2010. Extraction Asymmetries: Experimental Evidence from German. Amsterdam: John Benjamins. Koster, Jan. 2003. All languages are tense-second. In: Jan Koster and Henk van Riemsdijk (eds.), Germania et Alia: A Linguistic Webschrift for Hans den Besten, http://www.let.rug.nl/ koster/DenBesten/contents.htm. Krifka, Manfred. 2008a. What do Contrastive Topics and Frame Setters have in Common? The Role of Addressing and Delimitation in Information Structure. Presentation given at the conference on ‘Contrastive Information Structure Analysis’, University of Wuppertal, March 18–19, 2008. Krifka, Manfred. 2008b. Basic notions of information structure. Acta Linguistica Hungarica 55: 243–276. Lambrecht, Knud 1994 Information Structure and Sentence Form: Topic, Focus, and the Mental Representation of Discourse Referents. New York: Cambridge University Press. Lühr, Rosemarie. 1988. Zur Satzverschränkung im heutigen Deutsch. Groninger Arbeiten zur Germanistischen Linguistik 29: 74–87. Lutz, Ulrich. 1997. Parasitic gaps und Vorfeldstruktur. In: Franz-Josef d‘Avis and Ulrich Lutz (eds.), Zur Satzstruktur des Deutschen, 55–80. Tübingen: Arbeitspapiere des Sonderforschungsbereichs 340. Mayr, Clemens. 2010. On the necessity of phi-features: The case of Bavarian subject extraction. In: Phoevos Panagiotidis (ed.), The Complementizer Phase: Subjects and Wh-Dependencies, 117–142. Oxford: Oxford University Press. Meinunger, André. 2000. Syntactic aspects of Topic and Comment. Amsterdam: John Benjamins.

332   

   Josef Bayer and Martin Salzmann

Meinunger, Andre. 2006. Interface Restrictions on Verb Second. The Linguistic Review 23: 127–160. Müller, Gereon. 1995. A-bar Syntax: a Study in Movement Types. Berlin: Mouton de Gruyter. Müller, Gereon. 2010. Movement from Verb-Second Clauses Revisited. In: Thomas Hanneforth and Gisbert Fanselow (eds.), Language and Logos. A Festschrift for Peter Staudacher, 97–128. Berlin: Akademieverlag. Müller, Gereon. 2012. A Local Reformulation of the Williams Cycle. Ms. University of Leipzig. Müller, Gereon and Wolfgang Sternefeld. 1993. Improper Movement and Unambiguous Binding. Linguistic Inquiry 24: 461–507. Penka, Doris. 2007. Negative Indefinites. Doctoral dissertation, University of Tübingen. Pesetsky, David. 1998. Some Optimality Principles of Sentence Pronunciation. In: Pilar Barbosa, Danny Fox, Martha McGinnis and David Pesetsky (eds.), Is the Best Good Enough? Optimality and Competition in Syntax, 337-383. Cambridge: MIT Press. Pesetsky, David Michael and Esther Torrego. 2001. T-to-C Movement: Causes and Consequences. In: Michael Kenstowicz (ed), Ken Hale: A Life in Language, 355–426. Cambridge: MIT Press. Reinhart, Tanya. 1981. Pragmatics and linguistics: an analysis of sentence topics. Philosophica 27: 53–94. Reis, Marga. 1995. Extractions from Verb-Second clauses in German? In: Ulrich Lutz and Jürgen Pafel (eds.), On extraction and extraposition in German, 45–88. Amsterdam: John Benjamins. Rezac, Milan. 2004. Elements of Cyclic Syntax: Agree and Merge. Doctoral Dissertation, Department of Linguistics, University of Toronto. Riemsdijk, Henk van. 1989. Swiss Relatives. In: Dany Jaspers, Wim Klooster, Yvan Putseys & Pieter Seuren (eds.), Sentential Complementation and the Lexicon, 343–354. Studies in Honour of Wim de Geest. Foris: Berlin. Riemsdijk, Henk van. 2008. Identity Avoidance: OCP effects in Swiss relatives. In: Robert Freidin, Carlos P. Otero and Maria Luisa Zubizarreta (eds.), Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud, 227–250. Cambridge, Massechusetts: MIT Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Massachusetts: MIT Press. Rizzi, Luigi. 1997. The Fine Structure of the Left Periphery. In: Liliane Haegeman (ed.), Elements of Grammar, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2001. Reconstrution, Weak Island Sensitivity, and Agreement. In: Carlo Cechetto. Gennaro Chierchia and Maria Teresa Guasti (eds.), Semantic Interfaces, 145–176. Stanford: CSLI. Rizzi, Luigi and Ur Shlonsky. 2006. Satisfying the Subject Criterion by a non subject: English Locative Inversion and Heavy NP Shift. In: Mara Frascarelli Mara (ed.), Phases of interpretation, 341–361. Berlin: de Gruyter. Rizzi, Luigi and Ur Shlonsky. 2007. Strategies of subject extraction. In: Hans-Martin Gärtner and Uli Sauerland (eds.), Interfaces + Recursion = Language?, 115–160. Berlin: Mouton de Gruyter. Ross, John Robert. 1973. Slifting. In: Maurice Gross and Marcel Schützenberger (eds.), Formal analysis of natural languages, 133–172. Mouton and Company: The Hague. Salzmann, Martin. 2006a. Resumptive Pronouns and Matching Effects in Zurich German Relative Clauses as Distributed Deletion. In: Noureddine Elouazizi, Frank Landsbergen, Maika Poss & Martin Salzmann (eds.): Leiden Papers in Linguistics 3.1, 17–50.

That-trace effects and resumption – How Improper Movement can be repaired   

   333

Salzmann, Martin. 2006b. Resumptive Prolepsis. A Study in indirect A‘-dependencies. LOT: Utrecht. www.lotpublications.nl/publish/articles/001910/bookpart.pdf. Salzmann, Martin. 2008. Variation in resumption requires violable constraints – A case study in Alemannic relativization. In: Hans Broekhuis and Ralf Vogel (eds.), Optimality Theory and Minimalism: Interface Theories, 99–132. www.ling.uni-potsdam.de/lip. Salzmann, Martin. 2009a. When movement and base-generation compete – on the definition of the reference set, the typology of resumption, and ranked economy constraints. Groninger Arbeiten zur Germanistischen Linguistik 48, 27–63. http://gagl.eldoc.ub.rug.nl. Salzmann, Martin. 2009b. Different notions of variation and their reflexes in Swiss German relativization. In: Andreas Dufter, Jürg Fleischer and Guido Seiler (eds.), Describing and Modeling Variation in Grammar, 135–161. Berlin/New York: Mouton de Gruyter. Salzmann, Martin. 2009c. When movement and base-generation compete. The definition of the reference set and parameterized preferences for elementary operations. Linguistics in the Netherlands 2009: 64–77. Salzmann, Martin. 2011. Silent resumptives in Zurich German possessor relativization. In: Peter Gallmann & Melanie Wratil (eds.), Null Pronouns, 141–221. Berlin: Mouton de Gruyter. Salzmann, Martin. 2013. On three types of variation in resumption: evidence in favor of violable and ranked constraints. In: Hans Broekhuis & Ralf Vogel (eds.), Linguistic Derivations and Filtering, 76–108. Sheffield: Equinox. Salzmann, Martin. This volume. Repair-driven verb movement in English locative inversion. Salzmann, Martin and Guido Seiler. 2010. Variation as the exception or the rule? Swiss relatives, revisited. Sprachwissenschaft 35: 79–117. Salzmann, Martin, Jana Häussler, Markus Bader and Josef Bayer. (to appear). That-trace effects without traces. An experimental investigation. In: Stefan Keine and Shayne Sloggett (eds.), Proceedings of the 42nd Annual Meeting of the North East Linguistic Society. GLSA, University of Massachusetts, Amherst. Stalder, Franz Joseph. 1819. Die Landessprachen der Schweiz oder schweizerische Dialektologie. Sauerländer: Aarau. Steinbach, Markus. 2007. Integrated parentheticals and assertional complements. In: Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals, 53–87. Amsterdam: John Benjamins. Strawson, Peter F. 1964. Identifying reference and truth values. Theoria 3: 96–118. Suter, Rudolf. 1992. Baseldeutsch-Grammatik. Basel: Merian. Viesel, Yvonne. 2011. Glaubt er, glaub ich, glaub. Integrierte V1-Parenthesen, Extraktion aus V2.Komplementen, Grammatikalisierung. Linguistische Berichte 226: 130–169. Weber, Albert. 1964. Zürichdeutsche Grammatik: Ein Wegweiser zur Guten Mundart. Schweizer. Zürich: Spiegel Verlag.

Florian Schäfer

Passives of reflexive verbs: The repair of a Principle A violation* Abstract: The paper discusses Passives of Reflexive verbs in German and Icelandic. The construction is problematic for standard theories of case because accusative case on the reflexive pronoun in object position survives the passivization process. Furthermore, the construction challenges Binding Principle A as it can be shown that the implicit argument of the passive, although semantically the antecedent of the reflexive pronoun, is not the formal, i.e., syntactic antecedent of the anaphor. The solution to these two theoretical challenges is provided within a theoretical setup which uses syntactic Agree-relations to calculate the effects of both Binding as well as Case Theory. Such a system is enriched with the concept of Default Agreement which acts as a last resort syntactic repair process. It will be argued that only in the context of Passives of Reflexive verbs Default Agreement can formally antecede the reflexive pronoun and, thereby, triggers dependent accusative case on it. Such a formally repaired syntactic structure shows, however, a specific conceptual restriction: only inherently reflexive and naturally reflexive verbs can enter the construction because only such verbs allow to compute a reflexive interpretation in the absence of a DP acting as the syntactic antecedent of the reflexive pronoun.

1 Introduction In most languages, reflexive and reciprocal verbs do not passivize. German (1) and Icelandic (2) do, however, allow the formation of Passives of Reflexive/Reciprocal Verbs (henceforth PoRs).¹ (1) Zuerst wird sich geküsst, später dann geheiratet. first becomes refl kissed, later then married ‘First people kiss each other, then they marry.’ (2) Það var baðað sig á laugardögum. expl was bathed refl.acc on saturdays ‘People took a bath on Saturdays.’

336   

   Florian Schäfer

In order to identify the basic properties of PoRs and the theoretical questions that follow, consider the German active-passive pair in (3a, b) involving the reflexive use of the verb waschen (‘to wash’). As in ordinary passives, the lexical verb in (3b) appears as passive participle and the passive auxiliary werden (‘to become’) is used.² The external argument disappears (but it can reappear in a by-phrase as will be shown later). The PoRs in (1), (2) and (3b) lack a (overt) nominative DP, i.e. PoRs are (typically) instances of impersonal passives.³, ⁴ Finally, the reflexive pronoun does not change its shape under passivization: it does not shift to nominative as referential objects would but it keeps its accusative. While this is not easy to see in German, Icelandic, which has a case-inflected paradigm of reflexive pronouns, shows this clearly (cf. (2); see section 3 for a more detailed discussion).⁵ (3) a. Hier haben die Römer sich gewaschen. Here have the.nom Romans refl.acc washed b. Hier wurde sich (von den Römern) gewaschen. Here was refl (by the Romans) washed ‘Here, the Romans washed.’ PoRs immediately pose two questions concerning Binding Theory and Case Theory: (i) Binding Theory: How is Principle A of the Binding Theory satisfied in PoRs, i.e. what is the antecedent of the reflexive pronoun? (ii) Case Theory: How is accusative case on the reflexive pronoun licensed in PoRs? It stands in contradiction to Burzio’s Generalization, which predicts that passives, due to the absorption of the external argument, cannot license structural accusative case. In the present paper, I will propose an answer to these questions. But before I turn to a more detailed theoretical investigation of PoRs, I am first going to show that the formation of PoRs is a productive phenomenon in German (on Icelandic, see section 3). In fact, PoRs were often judged as ungrammatical in the theoretical literature on German (e.g. Reis 1982: 20–21; Haider 1985; Kiss 2003: fn. 15; Bierwisch 2006). However, other authors recognized that this view is not generally correct, and, more concretely, acknowledged the formation of PoRs to be a productive option provided by the grammar of German. First examples and preliminary discussions of PoRs can be found in Wunderlich (1985:222), Abraham (1986), Fanselow (1987,

Passives of reflexive verbs: The repair of a Principle A violation   

   337

1991), Sells, Zaenen and Zec (1987), Grewendorf (1988), Frey (1993) or Müller and Sternefeld (1993). Plank (1993) and Vater (1995) investigate PoRs in more depth and show that PoRs are, in principle, accepted among speakers of German (see also Ágel 1997, Müller 1999 and Hundt 2002 for more detailed discussions of PoRs). Three further observations suggest that PoRs are a common phenomenon in German. First, PoRs are clearly not substandard. The examples in (4) and (5) are from the evening news of the German public television (Tagesschau, ARD). Second, while PoRs are very frequent in spoken language, they are also found in written texts. The example in (6) is from the online version of a weekly German magazine (Spiegel online). Lastly, PoRs are not a recent invention. Behagel (1924, II:214) provides some Middle High German examples. (4) Während sich heute über Sonne gefreut werden konnte, while refl today about sun rejoiced become could, muss morgen mit Regen gerechnet werden. must tomorrow with rain calculated become ‘While today the sun could be enjoyed, rain is expected for tomorrow.’ (5) Bei der Kieler Koalitionskrise wird sich schon gar nicht at the Kieler.adj coalition-crisis becomes refl already really no mehr bemüht, die taktischen Mätzchen zu verstecken. longer bestirred the tactical tricks to hide ‘Politicians involved in the coalition crisis in the regional capital Kiel do not even try any longer to hide the tactical tricks.’ (6) Bei der ARD wurde sich eiligst für den Fauxpas entschuldigt. at the ARD was refl hastily for the faux-pas apologized ‘The people responsible at the ARD hastily apologized for the faux pas.’

2 A semantic/conceptual restriction on the formation of PoRs The last section showed that German allows the formation of PoRs. However, a closer examination shows that PoRs are restricted by a semantic or conceptual parameter. To illustrate this parameter, it is useful to make a short detour to the Dutch reflexive system. Dutch is a language with a so-called two-form reflexive system, that is, it has two reflexive pronouns: the simple reflexive pronoun zich

338   

   Florian Schäfer

and the complex reflexive pronoun zichzelf. Furthermore, Dutch makes a morphosyntactic distinction between three classes of reflexive verbs, i.e. verbs where the subject binds a reflexive element in object position. Crucially, the morphological distinction reflects a semantic or conceptual difference (see Kemmer 1993 and references there). The following verb-classes can be identified: Inherently reflexive (or inherently reciprocal) verbs: The reflexive pronoun cannot be replaced by a referential DP. Only the simple reflexive is allowed: (7) Jan schaamt zich/*zichzelf/*Marie. John shames refl/refl-self/Mary ‘John is ashamed.’ Naturally reflexive (or naturally reciprocal) verbs: The reflexive pronoun can be replaced by a referential DP. In out-of-the-blue contexts, the simple reflexive is strongly preferred. (The complex reflexive becomes acceptable under strong focus.) (8) a. Jan waste zich/??zichzelf/Marie. John washed refl/refl.self/Mary ‘John washed (Mary).’ b. Jan scheerde zich/??zichzelf/Peter. John shaved refl/refl.self/Peter ‘John shaved (Peter).’ Naturally reflexive verbs come from a number of semantic subclasses which all represent events that carry “[…] inherent in their meaning […] the lack of expectation that the two semantic roles they make reference to will refer to distinct entities […]” (Kemmer 1993: 58). So-called “grooming verbs” such as ‘shave’, ‘wash’ or ‘dress’ form one main subgroup of naturally reflexive verbs. Naturally reciprocal verbs involve, for example, verbs of social (‘meet’) or affectionate (‘kiss’) events but also verbs of antagonistic events (‘fight’). Naturally disjoint verbs (called other directed verbs in König and Vezzosi 2004): The reflexive pronoun can be replaced by a referential DP. The complex reflexive is strongly preferred. In opposition to naturally reflexive verbs, these verbs express events which carry the expectation that the two semantic roles they make reference to will refer to distinct entities (e.g. ‘hate’, ‘accuse’, ‘kill’, …).

Passives of reflexive verbs: The repair of a Principle A violation   

   339

(9) Zij haat ??zich/zichzelf/Peter. She hates refl/refl-self/Peter ‘John hates himself/Peter.’ German differs from Dutch in that the simple reflexive pronoun sich can be used with all three of the verb classes above (10a–c). The addition of the intensifier selbst (‘self’), while often possible, is hardly ever obligatory in German. Therefore, German does not make a (obligatory) morphological distinction between inherently/naturally reflexive verbs and naturally disjoint verbs. (10) a. Hans hasst sich/Maria. John hates refl/Maria ‘John hates himself/Mary.’

(naturally disjoint)

b. Hans wäscht sich/Maria. John washes refl/Maria ‘John washes himself/Mary.’

(naturally reflexive)

c. Hans schämt sich/*Maria. John shames refl/Maria ‘John is ashamed/ashames Mary.’

(inherently reflexive)

While German differs from Dutch in not making a morphological distinction between inherently/naturally reflexive verbs and naturally disjoint verbs, it turns out that the difference between these verb classes is, nevertheless, relevant in the grammar of German. The same lexical-semantic or conceptual aspects that determine the distribution of simple and complex reflexives in Dutch determine the formation of PoRs in German: PoRs are overwhelmingly formed with inherently and naturally reflexive verbs; naturally disjoint verbs are very rarely found in PoRs. Such a semantic restriction on the formation of PoRs has already been suggested in Abraham (1986), Sells, Zaenen and Zec (1987), Fanselow (1991), Kaufmann (2001) and especially Ágel (1997), but was never empirically corroborated. Here, I briefly present the results of two empirical studies that I undertook. First, I searched in Google for PoRs involving verbs of the three different verb classes. Inherently reflexive verbs and naturally reflexive verbs led to a number of hits involving PoRs. Naturally disjoint verbs hardly ever led to such hits.⁶, ⁷ In order to backup these findings (which might be misleading due to a general difference in frequency between the three reflexive verb classes), I also ran a questionnaire study with students at the University of Stuttgart. This study contained 48 target sentences all of which had the reflexive pronoun sich in the direct object position. 24 of the target sentences were reflexive active sentences, the other 24 sentences

340   

   Florian Schäfer

were the impersonal passive counterparts of these active sentences. The 24 activepassive pairs included four inherently reflexive verbs, four inherently reciprocal verbs, four naturally reflexive verbs and four naturally reciprocal verbs as well as eight naturally disjoint verbs, four of them used in a reflexive construal and four of them used in a reciprocal construal. The 48 target sentences were arranged in two questionnaires with 24 sentences each so that no verb occurred twice in one questionnaire (i.e., each verb occurred either in the active or in the passive in one questionnaire). Both questionnaires were filled with the same 48 distractors and were presented in two randomized orders. Each sentence was read by 24 speakers who judged the sentences on a scale from 1 (totally acceptable) to 7 (totally unacceptable). Table 1 below shows the results.⁸, ⁹

6 5 4 3 2 1

i-ref

i-rec

n-ref

n-rec

nd-ref

active

1,56

1,87

1,68

1,66

2,26

nd-rec 1,79

passive

2,67

3,21

3,1

3,19

5,21

4,66

Table 1: Results of the Questionnaire Study

The results in the above table lead to three conclusions:¹⁰ (i) In the active, the different verb classes were judged equally acceptable. That is, inherently reflexive/reciprocal verbs receive the same acceptance as naturally reflexive/reciprocal verbs and naturally disjoint verbs under a reflexive/reciprocal use. (ii) PoRs are generally less acceptable than the corresponding active counterparts. I take this result to be independent from reflexivization; it is well known that passive sentences are more marked than active sentences in an out-of-the-blue context. (iii) The most important observation for our purposes is that PoRs of naturally disjoint verbs are much less acceptable than PoRs involving inherently and naturally reflexive/reciprocal verbs. PoRs of the latter two verb classes, on the other hand, receive approximately the same acceptance. This confirms the results of the Google search presented above.

Passives of reflexive verbs: The repair of a Principle A violation   

   341

3 PoRs in Icelandic Icelandic PoRs were first discussed by Sigurðsson (1989: 355, fn. 60), who provides the active-passive pairs in (11) and (12). While Sigurðsson marks (11b) with one and (12b) with two question marks, Eythórsson (2008) judges these examples as totally acceptable (no question mark) and adds that most speakers he consulted agreed with him. Note that the verb in (11a, b) is inherently reflexive, and the verb in (12a, b) is naturally reflexive. Note furthermore, that Icelandic reflexive pronouns have a case-inflected paradigm (sig-acc, sér-dat, sín-gen), which proves that the reflexive pronoun in PoRs keeps the case that it has in the active. In (11b) this is a lexical dative, but in (12b) this is a structural accusative.¹¹ (11) a.

Börnin leika sér allan daginn. the.children play refl.dat all the.day ‘The children are playing all day.’

b. (?) Það var leikið sér allan daginn. expl. was played refl.dat all the.day (12) a.

Fólkið baðaði sig á laugardögum. the.people bathed refl.acc on Saturdays ‘The people took a bath on saturdays.’

b. (??) Það var baðað sig á laugardögum. expl was bathed refl.acc on Saturdays There is some indication that Icelandic PoRs are restricted by the same semantic/ conceptual parameter that was identified for German PoRs. They are accepted best in inherently and naturally reflexive contexts but are typically rejected in naturally disjoint contexts. The relevant data that show this are taken from a questionnaire study by Maling and Sigurjónsdóttir (2002). Before we can take a closer look at this data, some comments are necessary. Maling and Sigurjónsdóttir are only indirectly concerned with PoRs but investigate a broader phenomenon of Icelandic syntax, the so-called “New Passive” (also called “New Construction”), which is related to PoRs in a way that I cannot discuss here for reasons of space.¹² Since the acceptance of the ‘New Passive’ is subject to some speaker variation (concerning age and geographical origin), their questionnaire study differentiates between three classes of speakers. This is the reason why all the Icelandic examples that I will discuss below are followed by three acceptance rates for the three groups of speakers. I will not discuss any differences between these three groups of speakers (which clearly exist) but I will concentrate more on the

342   

   Florian Schäfer

question of whether or not the examples below show differences across all three groups of speakers. The passive sentences in (13a–c) all involve inherently reflexive verbs with the simple reflexive pronoun sig in object position. The three groups of speakers show quite the same acceptance rate for each of the three sentences. (Sentence (13c) is, for some reason, slightly less acceptable than the other two examples.) Since all three examples are PoRs with inherently reflexive verbs, I take their acceptance rates as a baseline. The question is then how acceptable PoRs involving naturally reflexive and naturally disjoint verbs are. (13) a. Svo var bara drifið sig á ball. Then was just hurried refl to the.dance ‘People hurried to the dance.’ (Elsewhere 78% | Inner Rvík 67% | Adults 40%)¹³ b. Það var haldið sig innan dyra út af óveðrinu. it was kept refl in doors due to bad.weather ‘People remained in the house due to the bad weather.’ (Elsewhere 82% | Inner Rvík 65% | Adults 37%) c. Það var skoðað sig um á svæðinu. it was looked refl around in the.area ‘People took a look-around in the area.’ (Elsewhere 72% | Inner Rvík 43% | Adults 31%) The examples in (14a, b) both involve a complex reflexive pronoun embedded in a prepositional phrase. Although both examples are syntactically identical, the first example receives much more acceptance than the second one in all three groups of speakers. I want to suggest that the reason for this difference is that the predicate ‘X looks at Y in the mirror’ is naturally reflexive (i.e., there is a strong expectation that X = Y), while the predicate ‘X points at Y in the picture’ is not naturally reflexive. Typically, people look at (a representation of) themselves when they look into a mirror. But there is no expectation that someone points at (a representation of) herself/himself when s/he points at a person in a picture. This conceptual difference is reflected by the acceptance rates of the two examples. (14) a. Það var horft á sjálfan sig í speglinum. it was looked at self refl in the.mirror ‘People looked at themselves in the mirror.’ (Elsewhere 58% | Inner Rvík 48% | Adults 34%)

Passives of reflexive verbs: The repair of a Principle A violation   

   343

b. Það var bent á sjálfan sig á myndinni. it was pointed to self refl in the.picture ‘People pointed at themselves in the picture.’ (Elsewhere 19% | Inner Rvík 11% | Adults 13%) The same conceptual parameter seems to be at play in the examples in (15a-c) involving possessive reflexive pronouns. These three examples show some syntactic differences. In (15a), the possessed noun phrase is embedded in a prepositional phrase selected by the verb, in (15b) the possessed noun phrase is the internal argument of the verb and in (15c), the possessed NP is embedded in a prepositional phrase modifying the internal argument of the verb.¹⁴ (15a) is accepted by many more speakers in all three groups than (15b, c). Once again, I suggest that the reason is that ‘X supports Y’s team’ is naturally reflexive (people typically support their own team) while this is not the case in the other two examples. For example, it is not more natural to push one’s own sister off the bike than it is to push someone else’s sister off the bike. (15) a. Það var haldið með sínu liði. It was held with self’s team ‘People supported their own team.’ (Elsewhere 63% | Inner Rvík 49% | Adults 36%) b. Í morgun var hrint systur sinni af hjólinu. this morning was pushed sister self’s off the.bike ‘People pushed their own sister from the bike.’ (Elsewhere 13% | Inner Rvík 7% | Adults 2%) c. Það var klippt hárið á dúkkunni sinni. it was cut the.hair on doll self’s ‘People cut the hair of their own doll.’ (Elsewhere 5% | Inner Rvík 2% | Adults 2%) The examples in (15) make an important point about the nature of the concept ‘natural reflexivity’. I proposed that (15a) is acceptable because there holds a naturally reflexive relation between an argument of the verb (its implicit external argument) and the possessor of an NP embedded in a PP selected by the verb. If this explanation is correct, it follows that ‘natural reflexivity’ cannot always be determined within a verb’s co-argument domain. More concretely, since possessors are not part of a verb’s lexical representation, ‘natural reflexivity’ cannot (always) be established in the lexicon but must be computed at a conceptual level following syntactic computation.

344   

   Florian Schäfer

To summarize this section, there is some indication that Icelandic PoRs are constrained by the same semantic/conceptual parameter as German PoRs. Inherently reflexive and naturally reflexive expressions allow the formation of PoRs while naturally disjoint expressions are largely rejected in PoRs. Furthermore, Icelandic shows that ‘natural reflexivity’ is not necessarily a relation between coarguments of a verb. Finally, Icelandic provides clear evidence that the reflexive pronoun in PoRs indeed keeps the case it has in the corresponding active counterpart. That is, structural accusative can survive in PoRs.

4 PoRs in other languages: Dutch and Norwegian As mentioned in the introduction, PoRs seem to be a rare phenomenon. In fact, German, Icelandic (and Lithuanian, see note 1) are the only languages with PoRs I know of. In this section, I exemplify the lack of PoRs in two other Germanic languages, Dutch and Norwegian. As the Dutch example in (16) and the Norwegian examples in (17a, b) show, PoRs are judged as ungrammatical in both languages. Note that these examples are ungrammatical even though (16) involves a naturally reflexive verb and (17a) involves an inherently reflexive verb.¹⁵ (16) * Er werd zich gewassen. There was refl washed ‘People washed (themselves).’ (17) a. * Det ble hygget seg. It was amused refl ‘People amused (themselves).’

(Reinhart and Siloni 2004)

(Maling and Sigurjónsdóttir 2002)

b. * Det ble låst seg (selv) inne i fabrikken. It was locked refl (self) inside in the.factory ‘People locked themselves in the factory.’ Note that Dutch and Norwegian lack PoRs although they share two properties with German and Icelandic that might be preconditions for the availability of PoRs: First, Dutch and Norwegian have impersonal passives (recall that (most) PoRs in German and Icelandic are impersonal passives; see note 4). Second, the reflexive system in Dutch and Norwegian is relatively similar to the system in Icelandic and German. For example, both Dutch and Norwegian have a light reflexive pronoun like German sich and Icelandic sig (a SE-anaphor in the terminology

Passives of reflexive verbs: The repair of a Principle A violation   

   345

of Reinhart and Reuland 1993) which is typically used in the context of inherently and naturally reflexive verbs.

5 Theoretical evaluation A theoretical account of PoRs should be able to answer the following questions: question (i) and (ii) were already formulated in the introduction, questions (iii) and (iv) are the result of our investigation in the last three sections: (i) Binding Theory: How is Principle A of the Binding Theory satisfied in PoRs, i.e. what is the antecedent of the reflexive pronoun/anaphor?¹⁶ (ii) Case Theory: How is the accusative case on the reflexive pronoun licensed in PoRs? (iii) Verb Class Restriction: Why are PoRs acceptable only with inherently/ naturally reflexive verbs but not with naturally disjoint verbs? (Recall that in the active, the three verb classes behave the same.) (iv) Language Restriction: Why don’t we find PoRs in other (Germanic) languages, especially, if they have impersonal passives and a similar anaphoric system involving SE-anaphors? Before I propose answers to these four questions, I want to discuss a number of theoretical proposals from the literature ― either specifically on PoRs or, more generally, on reflexivity ― that turn out to be insufficient to answer these questions. As an answer to question (i) (Binding Theory), one could suggest that the implicit argument of passives can act as the antecedent of the anaphor (e.g. Fanselow 1987, Barðdal and Molnár 2003, Sternefeld 2006, Sigurðsson 2011; see also the so-called “smuggling” approach to passives in Collins 2005). However, this proposal leaves questions (ii)–(iv) unanswered: it does not explain why the reflexive pronoun can have accusative case, it cannot explain why only inherently and naturally reflexive verbs can form PoRs and it cannot explain why only German and Icelandic form PoRs. It is worth noting in connection to the last point that the implicit argument of passives seems to have quite the same properties across languages otherwise (e.g. licensing of control, licensing of agentive adverbs, …). In order to answer both, question (i) about Binding Theory and question (ii) about Case Theory, one could also suggest that PoRs are hidden transitives

346   

   Florian Schäfer

with a covert external argument (for example pro, cf. Maling and Sigurjónsdóttir 2002 for Icelandic). If PoRs have a syntactically projected external argument they should behave like active clauses concerning Binding Theory and Case Theory. The covert external argument would act as the antecedent of the anaphor and Burzio’s Generalization would predict that accusative is available. However, this proposal cannot answer question (iii) either: Why should the covert external argument in PoRs be able to antecede an anaphor only if the verb is inherently or naturally reflexive but not if the verb is naturally disjoint? This is unexpected because in the active this difference does not exist (cf. Table 1). Furthermore, it remains unclear why only German and Icelandic can have this type of passive (question (iv)). German provides a further counter-argument against the idea that the anaphor in PoRs has an antecedent in the syntax. German impersonal passives combine with agentive by-phrases which can even introduce 1st and 2nd person pronouns (18a, b). (18) a. Von mir wurde nicht gelacht. by me was not laughed ‘I did not laugh.’ b. weil von dir zu viel gelabert wird because by you too much babbled becomes ‘because you babble too much’ 1st and 2nd person by-phrases are possible with PoRs, too. However, their effect on a reflexive pronoun differs from the effect that a 1st or 2nd person subject in active clauses has. In the active, the bound element agrees with the subject-antecedent in person and number (19a). Crucially and unequivocally, no agreement between the by-phrase and the bound element is possible in PoRs (19b); instead, the 3rd person reflexive pronoun is obligatory (Plank 1993). I conclude from these data that the anaphor in PoRs does not have any argument as its syntactic antecedent at all. Instead, the anaphor gets default realization (as a 3rd person reflexive pronoun).¹⁷ (19) a. Nur wir waschen uns / *sich hier täglich only we wash us.acc / refl here daily b. Nur von uns wird sich / *uns hier täglich gewaschen only by us is refl / us.acc here daily washed ‘Only we wash ourselves here everyday’

Passives of reflexive verbs: The repair of a Principle A violation   

   347

The existence of PoRs cannot be explained by approaches that treat reflexive verbs as being detransitivized, be it in the lexicon or in the syntax.¹⁸ A case in point is the so-called bundling-approach developed by Reinhart (2000) and Reinhart and Siloni (2005). These authors assume that reflexive verbs are derived by a process that bundles a verb’s internal theta-role () with its external thetarole (). As a result, only one argument with a complex theta-role () is merged in the external argument position. Crucially then, the reflexive element is not an argument of the verb. It is either merged as a lexical sign that indicates that the bundling-operation has taken place or it acts as a case-reducer which absorbs the accusative case of the basically transitive verb. Such an account could give the following answer to question (i) (Binding Theory): If the reflexive pronoun is not an argument of the verb, then it is not an anaphor either and therefore not subject to Binding Principle A.¹⁹ However, the bundling-approach fails to answer all other questions identified above. It cannot explain why the reflexive element can get accusative case. In fact, if the reflexive element is actually a case absorber (as proposed by Reinhart and Siloni (2005) for German sich), its presence in PoRs is totally unexpected. Since the passivization process should eliminate accusative case, there should be no need to insert a case absorber in PoRs. Furthermore, since the account does not make any difference between types of reflexive verbs (inherently and naturally reflexive vs. naturally disjoint) and since bundling is claimed to hold for all languages involving SE-anaphors, the questions (iii) and (iv) remain open, too. A further problem for the bundling approach is that the process of bundling, by definition, can only apply to co-arguments of a verb. However, we have seen that Icelandic PoRs can involve possessors. The relevant example is repeated in (20). (20) Það var haldið með sínu liði. it was held with self’s team ‘People supported their own team.’ Doron and Rappaport Hovav (2007) (see also Labelle 2008) provide a general argument against the bundling approach that can be transferred to PoRs. These authors observe that the bundling approach predicts that it should be impossible to focus only one of the two θ-roles that are involved in the bundling process. Recall that the bundling approach proposes that a verb’s internal θ-role () is not assigned to the object position, but is bundled together with the verb’s external θ-role () to a complex role (). This complex role is then assigned to the verb’s external argument position. As a consequence, it should be impossible to focus only one of the two θ-roles independently of the

348   

   Florian Schäfer

other. But as the authors observe this is empirically not correct. Both, the agent and the theme can be focused independently in a reflexive context as their French example in (21) shows. The German example in (22) illustrates the same point. (21) Jean-Pierre s’est dénoncé lui-même. Jean-Pierre refl is denounced himself (i) ‘Jean-Pierre denounced himself, it was not others who denounced him.’ (ii) ‘Jean-Pierre denounced himself, he did not denounce others.’ (22) Morgens wäscht sie sich immer/erst mal selber. at.morning washes she refl always/first-of-all self (i) agent focus: ‘She washes herself, no-one else washes her.’ (possible context: She is a disabled patient.) (ii) theme focus: ‘She washes herself, she washes no-one else.’ (possible context: She is a nurse.) Doron and Rappaport-Hovav (2007) conclude that reflexive verbs are transitive and that the reflexive pronoun bears a θ-role, that is, it is not just a marker of a bundling process or a case absorber but it is an anaphor in argument position. Turning to PoRs, we observe that, once again, the agent role and the theme role can be focused independently.²⁰ This shows that the reflexive pronoun bears a θ-role in PoRs and, therefore, is an anaphor in argument position. (23) Morgens wird sich immer/erst mal selber gewaschen. at.morning becomes refl always/first-of-all self washed (√ agent focus, √ theme focus) In the next section, I will propose a theory of the licensing of reflexive pronouns and accusative case that tries to answer the four questions mentioned above. Before I go into the technical details of my proposal, I will first describe the general idea.

6 Towards an analysis of PoRs I assume that reflexive elements like German sich, Icelandic sig or Dutch zich are always anaphors (SE-anaphors in the terminology of Reinhart and Reuland 1993). SE-anaphors can be either the (in-)direct object of a verb, the object of a preposition or the possessor of a DP.²¹ Being anaphors, they are subject to (some

Passives of reflexive verbs: The repair of a Principle A violation   

   349

version of) Binding Principle A (Chomsky 1981), no matter whether they occur in an inherently reflexive, naturally reflexive or naturally disjoint context.²² Inherent and natural reflexivity and their counterpart, natural disjointness, are not lexically specified but are post-syntactic phenomena, which are determined at the Conceptual-Intentional interface. They are computed on the basis of conceptual information about the material inside vP/VoiceP, that is, the verbal event and its arguments (cf. note 7; see also note 29). Above I have claimed that the anaphor in PoRs has no syntactic antecedent. If the implicit external argument of the passive could act as syntactic antecedent, we could not explain why naturally disjoint verbs are excluded form PoRs. Furthermore, the agreement facts in (19b) would be problematic. As a consequence, this means that the traditional version of Principle A of the Binding Theory is not fulfilled in PoRs. To explain why PoRs are, nevertheless, available in some languages under some contexts, I make the following proposal: (24) Some languages can formally repair a violation of Principle A. However, the formally repaired structure is acceptable (i.e., easily interpretable) only if the underlying predicate/event is conceptualized as inherently or naturally reflexive.

PoRs in all languages (all types of verbs)

Dutch

PoRs with inherently and naturally reflexive verbs

Norwegian

(formally) ungrammatical

PoRs in German/Icelandic (all types of verbs)

(formally) grammatical

(semantically) acceptable

Figure 1: The distribution of PoRs (≈ The availability of Repair for Principle A violations)

Figure 1 illustrates this idea. In most languages, PoRs are formally ungrammatical because they violate Principle A of the Binding Theory (dark area). Some languages have a mechanism that allows repairing the Principle A violation that goes along with PoRs (grey area). However, only a subset of these formally rescued PoRs is judged as acceptable (white area). This is so because the repair mechanism is purely formal and does not, by itself, lead to a successful inter-

350   

   Florian Schäfer

pretation of the reflexive pronoun. A successful interpretation is only possible if conceptual knowledge supports it. This is the case in inherently and naturally reflexive contexts.

6.1 Basic assumptions on anaphoric binding and structural case In this section, I provide some background assumptions concerning Binding Theory and Case Theory (see also Schäfer 2008, to appear). I will only discuss local binding of SE-anaphors and their pronominal 1st and 2nd person counterparts. I will not be concerned with non-local binding of pronouns, long-distance anaphors, or the licensing of self-anaphors. I follow earlier proposals in assuming that anaphoric binding is grounded in a syntactic Agree-relation between a DP-antecedent and a variable (e.g. Fanselow 1991; Burzio 1991; 1998; Reuland 2005; Fischer 2004, 2006; Heinat 2006; Chomsky 2008; Kratzer 2009). I assume that a variable is totally underspecified for φ-features: it is a set of a categorial D-feature and unvalued φ-features {D, uφ}. Therefore, it is referentially defective. Furthermore, a variable needs an antecedent to value its φ-features under syntactic Agree. Since the variable has unvalued φ-features it is active, thereby qualifying as a probe (Chomsky 2001). The antecedent is a full DP which has valued φ-features and, therefore, can act as a goal. Specifically, I propose that the variable probes the tree upwards to get is features valued by a c-commanding antecedent (cf. Baker 2008, Wurmbrand 2010, or Zeijlstra 2010 for the option of upward-probing). If Agree between the variable and a c-commanding antecedent has taken place, the structure is sent to the interfaces for interpretation. If Agree does not take place, the features of the variable remain unvalued and, as a consequence, the derivation will crash. Since syntactic Agree is local, this setup has an effect similar to Principle A of the standard Binding Theory (Chomsky 1981): A variable needs a local, c-commanding antecedent. However, rephrasing Principle A as an Agree-operation will allow me to implement a repair strategy for Principle A violations taking place in PoRs (see below). At LF, the syntactic Agree-relation between the variable and the antecedent is evaluated semantically as a binding relation (as expressed via coindexation in earlier stages of the theory; see Reuland 2001, 2005 for discussion). At PF, the Agree relation is evaluated morpho-phonologically and the Spell Out of the variable is determined. The specific Spell Out, either as a SE-reflexive or as a (locally bound) pronoun, depends on the φ-features of the antecedent and languagedependent morpho-phonological economy conditions (see Halle and Marantz

Passives of reflexive verbs: The repair of a Principle A violation   

   351

1993, Burzio 1998, Heinat 2006 for discussion).²³ In most Germanic languages the situation is as follows: If the variable is valued by a third-person antecedent, it gets spelled out as a SE-anaphor. If it is valued by a 1st or 2nd person antecedent, it is spelled out as a 1st or 2nd person object pronoun. However, some languages use the SE-anaphor for other than just 3rd person antecedents (e.g. Polish uses it for 1st, 2nd, and 3rd person antecedents) while other languages lack a SE-anaphor and always use a pronoun overtly matching the antecedent in φ-features (e.g. Frisian). Full-fledged referential pronouns, on the other hand, are the combination of a D-feature and a set of valued φ-features {D, φ}; that is, they always spell out their inherent φ-features and, therefore, refer independently. Note that I must assume that the PF-interface can differentiate between an element that started the derivation with all features valued and an element that started the derivation with unvalued features and gets these features valued only during the derivation. Only in the latter case the element is a variable that can get spelled out as a SEanaphor. I follow recent proposals that morphological case is mainly a PF phenomenon (Marantz 2000; McFadden 2004; Sigurðsson 2000, 2003, 2005, 2006, 2009). In accordance with these authors I assume that PF determines morphological case on the basis of syntactic information. Specifically, I propose that PF evaluates the syntactic Agree-relation involving T(ense) to determine structural case. T(ense) is equipped with unvalued φ-features and probes its c-command domain for the closest valued DP to agree with. At PF, this Agree-relation is evaluated to determine dependent case (accusative) and default case (nominative). I propose the following three principles to be at work: (25) a. Dependent case (acc): A DP is realized at PF with dependent case if a different DP has valued local T via Agree.²⁴ b. Default case (nom): A DP which is not realized with dependent case appears with default case. c. Inherent/lexical case takes precedence over default and dependent case. To illustrate the above assumptions, consider the derivation of the example in (26) involving a transitive verb with a subject-bound anaphor/variable in object position.²⁵

352   

   Florian Schäfer

(26) a. dass Hans sich mag that John.nom refl.acc likes ‘that John likes himself’ b.

TP T{uP, uN, uG}

vP v’

Hans {P, N, G} v

VP sich{uP, uN, uG}

V

T enters the derivation with unvalued φ-features. The variable in object position enters the derivation with unvalued ϕ-features. The external argument is merged with a full set of valued φ-features. Two Agree-processes take place. One is started by T which probes its c-command domain. The closest element with valued φ-features is the external argument in Spec,vP, which therefore agrees with and values T. (I assume that German does not have obligatory movement to Spec,TP, but nothing hinges on this). The second Agree-process is started by the variable. It probes the tree upwards and also agrees with and gets valued by the subject. These two Agree-relations will be evaluated at the interfaces. At PF, the internal argument is marked with dependent acc because there is a different DP (the subject) within the same clause which has valued the features on T via Agree (cf. 25a). The subject itself, on the other hand, gets default case as nothing more specific is said about its case (cf. 25b). Finally, the variable is spelled out according to the φ-features of its antecedent. Since the antecedent is 3rd person, the variable gets realized as a SE-anaphor. At LF, finally, the Agree-relation between the subject and the variable is interpreted as semantic binding. With this background, we can finally turn to a more technical discussion of PoRs.

6.2 Deriving PoRs Consider the German PoR in (27). The tree in (28) illustrates how the system sketched in the previous subsection handles this and similar examples. The following derivational steps take place: Both, the variable and T enter the derivation

Passives of reflexive verbs: The repair of a Principle A violation   

   353

with unvalued φ-features. T searches its c-command domain and the variable searches the tree upwards. Since PoRs are impersonal constructions,²⁶ there is no DP available that could value either T or the variable. But note that “Agree” and “valuation” are different processes if we assume with Frampton and Gutmann (2000) or Pesetsky and Torrego (2007) that Agree involves just feature sharing. Therefore, T and the variable agree with each other in (28), but since none of the two has valued features, no valuation can take place. (27) als sich gewaschen wurde when refl washed became TP

(28) T

vPpassive

{uP, uN, uG}

v’ v

VP DP{uP, uN, uG}

V

(28) involves an unvalued two-member agreement chain < T – variable >. Therefore, the derivation should crash at the interfaces. This looks like the correct result for Dutch or Norwegian, but it isn’t the correct prediction for German and Icelandic. Note, however, that German and Icelandic (but also Dutch and Norwegian) have a construction different from PoRs that provides a related problem, namely ordinary impersonal passives (or other impersonal constructions such as quirky intransitives in Icelandic) illustrated in (29a, b). As these examples show, German and Icelandic do not allow the insertion of an expletive in the context of impersonal passives. (29) a. weil (*es) hier gestern lange getanzt wurde because expl. here yesterday long danced was ‘because yesterday people danced for a long time’

(German)

354   

   Florian Schäfer

b. ĺ dag hefur (*það) verið dansað. today has expl. been danced ‘Today, people danced.’

(Icelandic)

The derivation of simple impersonal passives is sketched in (30). This time, only one element with unvalued ϕ-features is present, namely T. T probes its c-command domain but cannot find any other element that could value it. The derivation involves an unvalued one-member chain < T > which should lead to a crash at the PF-interface, contrary to fact. (30)

TP T{uP, uN, uG}

vP passive v’ v

VP V

?

Languages deal with impersonal passives in different ways. Some languages insert a nominal expletive with φ-features that can value T and check the EPP on T. In such languages, the problem depicted in (30) does not occur. Holmberg (2002) shows this to be the case for some Norwegian dialects where the expletive has a pronominal origin. But other languages lack expletives with nominal features and the problem depicted in (30) holds. The languages discussed in this paper (German Icelandic, Dutch, Norwegian) are of this latter type.²⁷ I assume with Ruys (2010) that in the latter type of languages default agreement (DA) values T and saves the derivation of impersonal passives in (30). In a minimalist framework which involves the syntactic operation Agree and the concept of Full Interpretation (i.e. the idea that derivations crash at the interfaces if unvalued features remain unchecked), default agreement must be a process taking place in core syntax, i.e. before Transfer to the interfaces. I assume the following conception of Default Agreement (see also Ruys (2010)): Default agreement (DA): In the absence of any appropriate nominal category, the ϕ-features on an unvalued probe undergo default valuation [3rd person, singular].

Passives of reflexive verbs: The repair of a Principle A violation   

   355

DA on T is a way to rescue impersonal constructions such as impersonal passives.²⁸ It is a last resort operation, i.e. a repair strategy that avoids that a derivation crashes for purely formal reasons at the interfaces. Furthermore, I assume that DA is a costly operation and, therefore, is not available in all languages. Languages lacking DA on T (as well as a nominal expletive) do not allow the formation of impersonal constructions. English is a case in point (cf. Ruys 2010). With this conception of DA at hand, I return to PoRs and provide answers to the questions (i)-(iv) formulated at the beginning of section 5. Recall that I concluded that the anaphor in PoRs does not have an antecedent. Question (i) asks, therefore, why PoRs are not filtered out as a Principle A violation. I proposed that variables (often spelled out as SE-anaphors) are similar to T in that both enter the derivation with unvalued ϕ-features. In the case of ordinary binding, the unvalued features on the variable get valued by a DP-antecedent. In PoRs, T and the variable form an agreement chain , which, however, remains unvalued (see the derivation in (28)). I argue that PoRs do not crash at the interfaces if this agreement chain can be valued by DA. That is, in PoRs, a violation of Principle A is formally precluded by DA: the variable gets its unvalued features valued although no c-commanding DP-antecedent is available. DA can, however, not provide any help concerning the interpretation of the variable in the absence of a DP-antecedent. The question of how to interpret the variable in PoRs will become relevant below. The idea that DA is a costly operation can provide an answer to question (iv). Recall that PoRs are available only in few languages and, more specifically, only in a subset of the languages with impersonal passives. Following Ruys (2010), I proposed that impersonal passives are rescued if the one member agreement chain is valued via DA. In order to rescue PoRs I proposed that the two-member agreement chain needs to be valued via DA. Arguably then, the DA-operation needed in simple impersonal passives is less complex than the DA-operation needed in PoRs. I propose, therefore, that languages differ in the complexitiy of the DA-operations they make available. Some languages lack DA. Other languages like Dutch or Norwegian make DA available only for heads of the extended verbal domain such as T. German and Icelandic, in addition, make DA available for non-homogeneous agreement-chains involving verbal heads and D-elements like variables/anaphors. Next, I turn to question (iii) about the verb class restrictions. Above I argued that DA formally avoids that PoRs crash at the interfaces. The variable (as well as T) gets its features valued as if it had a DP-antecedent. But crucially, no DP-antecedent is available that could semantically bind the variable at LF, i.e. the derivation provides no clue how to interpret the variable. At this point, I argue that conceptual knowledge about the verbal event expressed by the PoR enters the

356   

   Florian Schäfer

picture. The output of the syntactic derivation involving DA on a variable is comprehensible at the CI-interface only if conceptual knowledge provides some information about how to interpret the variable. With inherently reflexive verbs, it is conceptually clear that no further referent besides the external argument can be involved. Naturally reflexive events carry the strong conceptual expectation that the argument position realized by the semantically unbound reflexive pronoun denotes the same entity that acts as the external argument of the same event (see the discussion around example (8)). That is, in inherently and naturally reflexive contexts, the semantic coindexation that is usually computed from syntactic binding/Agree between a DP-antecedent and a variable can be reconstructed on conceptual grounds beyond any doubt. Naturally disjoint verbs or contexts do not provide any such conceptual clue about how to interpret the variable, and this is the reason, I argue, why they typically do not enter the formation of PoRs.²⁹ This leaves question (ii) about accusative case to be accounted for. Note that there is no reason to assume that the accusative case on the reflexive pronoun in PoRs is in any way different than ordinary structural accusative case. Furthermore, there is no reason to assume that the case system active in PoRs is different from the general case system of the languages under consideration. That is, it does not make sense to stipulate that only PoRs but not other passives can assign structural accusative case. The availability of structural accusative in PoRs should, therefore, follow from an interplay between the general theory of structural accusative case and some property present only in PoRs but not in ordinary passives. Arguably, this property must somehow be related to the presence of the reflexive pronoun. I argue, therefore, that the availability of accusative case in PoRs depends on one formal property of reflexive pronouns, namely their φ-features deficiency. In the previous section, I formulated the post-syntactic algorithm for object case (acc) repeated in (31) that builds crucially on a syntactic Agree-relation between T and a DP. In order to derive accusative on the reflexive pronoun in PoRs, we need to update the mechanism that determines dependent case (31) slightly as in (32) so that any kind of Agree-relation, either Agree with a DP or DA, is relevant for the determination of dependent case:³⁰ (31) Dependent case (acc) (old version): A DP is realized at PF with dependent case if a different DP has valued T via Agree. (32) Dependent case (acc) (updated version): A DP is realized at PF with dependent case if something else (either a different DP or DA) has valued T via (default) Agree.

Passives of reflexive verbs: The repair of a Principle A violation   

   357

7 Conclusions This paper discussed Passives of Reflexive Verbs that exist in a subset of the languages with impersonal passive and SE-reflexive, namely German and Icelandic. I showed that even in these languages the formation of PoRs is conceptually restricted to inherently and naturally reflexive contexts. I showed, furthermore, that lexical theories of reflexivity cannot account for the availability of PoRs because the reflexive pronoun in PoRs behaves like a syntactic and semantic argument and the reflexive relation found in PoRs is not confined to the verbal co-argument domain. However, a standard version of Principle A of the Binding Theory also cannot account for PoRs either because a closer inspection of PoRs showed that the implicit argument of the passive is not the formal, i.e. syntactic, antecedent of the reflexive pronoun. I proposed that the reflexive element in PoRs is an ordinary anaphor/a variable and should, therefore, be subject to some version of Principle A of the Binding Theory. However, PoRs argue in favor of a reformulation of the traditional Principle A in terms of a syntactic Agree-relation between the anaphor/variable and an antecedent. Since in PoRs the anaphor cannot find a c-commanding DP-antecedent, PoRs lead to a Principle A violation in most languages. In German and Icelandic, however, a formal violation of Principle A can be avoided because Default Agreement (DA), a last resort repair operation, can formally value the φ-features of the variable. DA only avoids a formal crash of PoRs at the interfaces but does not lead to a semantic interpretation of the variable. Therefore, a successful interpretation of PoRs depends on an inherently or naturally reflexive context. Finally, I proposed an Agree-based version of a dependent case approach which allows the application of Default Agreement to trigger structural accusative on the reflexive pronoun in PoRs.

Notes * I would like to thank the editors of the present volume and especially Andreas Pankau for his very helpful comments on an earlier version of this paper. All errors are mine. This work has been supported by a DFG (German Research Foundation) grant to the project B6, Underspecification in Voice systems and the syntax-morphology interface, as part of the Collaborative Research Center 732, Incremental specification in context, at the University of Stuttgart. 1 One further language with PoRs not discussed in this paper is Lithuanian (see Geniušienė 1987, Wiemer 2006).

358   

   Florian Schäfer

2 Crucially, the reflexive element is not part of the passivization process itself. PoRs are, therefore, not “reflexive passives” of the Romance or Slavic type as discussed, for example, in Cinque (1988). 3 When no by-phrase is present, I paraphrase PoRs as active sentences with the subject ‘people’. This is not the only possible interpretation for the implicit external argument of PoRs in that PoRs do not necessarily have a generic flavor. Furthermore, the possible interpretations of the implicit external argument of PoRs do not seem to differ from the interpretations of the implicit external argument in ordinary impersonal passives. 4 German, Icelandic, and Lithuanian have PoRs which are not impersonal (cf. (i, ii)). Such examples are unproblematic for Case Theory but not for Binding Theory. Since an integration of PoRs with nominative DP into the analysis proposed in section 6.2 is beyond the scope of the present paper, I will leave such examples aside and concentrate on the majority of PoRs, which are impersonal. (i) Da wurde wieder großer Unsinn von sich gegeben. there was again big nonsense.nom from oneself given ‘People uttered huge nonsense.’ (ii) Anschließend wurde sich ein Hamburger gekauft. Afterwards was refl.dat a.nom hamburger bought ‘People bought a hamburger.’ 5 I concentrate on passives involving reflexive pronouns such as German sich, which can have a reflexive and a reciprocal reading. PoRs can also be found with reciprocal pronouns such as German einander (‘each other’). I have not investigated the latter cases in any detail. 6 To identify naturally reflexive verbs, I used the classification in Kemmer (1993). She lists 13 semantically different subclasses of naturally reflexive verbs (verbs of body care, verbs of interlocution, verbs of antagonistic events, …). For each class, I randomly selected one verb and tested it on Google. For each of these verbs I found a number of hits (ranging from less than 10 to more than 50) where the verb was used in a PoR. A possible search query for the verb waschen (‘to wash’) would be sich gewaschen wird (‘refl washed becomes’) which brings about 10 hits (4.10.2010). Note that changes in the search string concerning word order or tense lead to many more hits. Note, furthermore, that although I used different word orders and time specifications when I tested naturally disjoint verbs, I hardly ever found a PoR with these verbs. 7 Hundt (2002) provides a list of 95 modern and 26 historical examples of PoRs. 7 of his examples involve verbs which might be best classified as naturally disjoint. Interestingly, such counterexamples to the above generalization very often contain the adverb gegenseitig (‘mutually’) as in (i), which necessarily triggers a reciprocal interpretation. Other counterexamples turn out to be from specialized communities or to involve irony (as if the event were naturally reflexive). (i) Hier wird sich nicht gegenseitig umgebracht. Here becomes refl not mutually killed ‘People do not kill each other here.’ 8 The following abbreviations are used: i-ref = inherently reflexive, i-rec = inherently reciprocal, n-refl = naturally reflexive, n-rec = naturally reciprocal, nd-ref = naturally disjoint under a reflexive construal, nd-rec = naturally disjoint under a reciprocal construal. 9 The individual subjects showed variable behaviour. Some rejected most PoRs; some accepted most PoRs. Note that half of the distractor sentences were strongly ungrammatical (wrong agreement, case or word order). This explains why the PoRs of naturally disjoint verbs were not judged as totally deviant.

Passives of reflexive verbs: The repair of a Principle A violation   

   359

10 A further observation is that it does not matter for the acceptability of PoRs whether the reflexive pronoun gets a reflexive or a reciprocal interpretation. 11 One might suspect that the accusative form of the Icelandic reflexive pronoun is actually ambiguous between a nominative and an accusative from. It has, however, been shown that – independently of Principle A (i.e. locality) considerations – Icelandic reflexive pronouns are incompatible with nominative positions (Everaert 1990; see Woolford 1999 on this so-called “anaphor agreement effect” in other languages). 12 Most of the recent literature argues that the New Passive is a real passive in that the external argument is absorbed although the (even definite) internal argument remains in situ and keeps accusative (Eythórsson 2008; Jónsson 2009; Sigurðsson 2011), not an active impersonal (cf. Maling and Sigurjónsdóttir 2002). While there is arguably a relation between the New Passive and PoRs (in that the latter might have grounded the way for the development of the former), the recent literature further concluded that PoRs are not genuine instances of the New Passive. Eythórsson (2008:189) mentions that the PoR “seems to be an innovation of Modern Icelandic which is increasingly gaining ground and is accepted by many speakers who do not accept the NC [New Construction] with non-reflexive verbs”. This conclusion can already be drawn from the questionnaire study by Maling and Sigurjónsdóttir (2002) which shows that PoRs receive much better acceptance than New Passives with a referential internal argument. 13 “Inner Rvík” stands for speakers that live in the center of Reykjavík. 14 A further difference is that the reflexive possessive in (15a) precedes the head noun while it follows the head noun in (15b, c). 15 The ungrammaticality of (16) was confirmed by Martin Everaert, Hans Kamp, and JanWouter Zwart (p.c.). The ungrammaticality of (17a, b) was confirmed by Terje Lohndal, Torgrim Solstad, and Kirsti Koch Christensen (p.c.). More empirical work is necessary on the availability of PoRs in these languages. For example, there is some preliminary indication that not all Norwegian speakers reject all PoRs to the same extend (p.c. Tor Afarli and Arild Hestvik). A further important question which is beyond the scope of this paper is whether impersonal and personal PoRs (see note 4. for the latter) are both ungrammatical in Dutch and Norwegian. There is some very preliminary indication that PoRs with a nominative DP might be more acceptable than impersonal PoRs in Dutch and Norwegian. 16 Below, I will use the term “reflexive (pronoun)” to refer to an element’s reflexive morphology and the term “anaphor” to refer to an element which is subject to Principle A of the Binding Theory. 17 An alternative, not implausible idea is that the existentially bound implicit argument of passives is always3rd person (or impersonal) and that by-phrases do not have to match the implicit argument in φ-features. But this would leave unexplained why naturally disjoint verbs do not form PoRs. 18 I concentrate here on a version of the intransitivity-account to reflexive verbs that assumes that these verbs are unergative. Some researchers proposed that reflexive verbs are unaccusative (e.g. Grimshaw 1981). PoRs are problematic for such an account simply because unaccusatives are not expected to passivize. McGinnis (1998, 2000, 2004) and Embick (2004) propose that reflexive verbs are transitive but involve an unaccusative derivation. The reflexive pronoun is located in the external argument position and the internal argument moves across the external argument position to bind the reflexive pronoun. Such an account is incompatible with PoRs because the external argument (i.e. the reflexive pronoun) should be absorbed by the process of passivization.

360   

   Florian Schäfer

19 Some authors explicitly argued that such an intransitivity approach to reflexive verbs (i.e., that reflexive verbs are unergative) would account for the binding problem that we identified for PoRs (e.g. Sells, Zaenen and Zec 1987; Ágel 1997; Hundt 2002; Eythórsson 2008). The arguments that I present below are problematic for all of these proposals. 20 See Hole (2006, 2008) on focusing the implicit argument of passives via an intensifier like German selber (‘self’). 21 I leave aside here the role of SE-anaphors in reflexive anticausatives and generic middles. In Schäfer (2008, to appear), I argued that the reflexive pronoun is located in the external argument position in these contexts. 22 Cf. Doron and Rappaport Hovav (2007), who propose that SE-anaphors are ambiguous between real anaphors and markers of reflexivization. In their conception, inherently reflexive verbs involve a marker of reflexivization, reflexive uses of naturally disjoint verbs involve an anaphor, and the reflexive element in reflexive uses of naturally reflexive verbs is ambiguous between the two options. 23 In the end, more has to be said about the technical aspects of this proposal. To make a binding relation and the Spell Out as a SE-anaphor possible, the information concerning which element has valued the variable must be available at both interfaces. Against standard minimalist assumptions, I have to assume, therefore, that the features of the variable, which get valued in the syntax are not deleted but remain available at both interfaces. 24 In Schäfer (2008, to appear) I argue that this combination of an Agree-based with a dependent case approach is necessary to derive the ergative case-pattern of reflexive anticausatives and middles. I argue that reflexive anticausatives/middles involve a nominative theme in object position and an accusative reflexive pronoun in Spec,vP. Since the reflexive has unvalued φ-features, the theme values T (and, indirectly, the reflexive). Baker and Vinokurova (2010) also propose the combination of an Agree-based with a dependent case approach. 25 I leave aspects of Phase Theory aside here (Chomsky 2001 et seq.). If vP/VoiceP is a phase, as standardly assumed, the system has to be reformulated in a way similar to Legate (2005) or Marantz (to appear). See Schäfer (2008) for a more detailed discussion of the present system that integrates Phase Theory and allows, thereby, also the integration of ECM-contexts. 26 But see note 4. 27 Holmberg (2002) shows that standard Norwegian uses an expletive in impersonal passives which has a locative origin and lacks nominal φ-features. The Dutch expletive er used in impersonal passives is of the same kind (see Ruys 2010). Such expletives check the EPP on T but cannot value T. I assume that German impersonal passives lack an expletive because German does not have an obligatory EPP-feature on T. 28 As said, I leave Phase Theory aside. If passive vPs are phases (e.g. Legate 2005), impersonal passives would involve DA on the phase head v. 29 Note that I assume that PoRs involving naturally disjoint verbs are formally grammatical in German and Icelandic because DA values the variable. They are however, unacceptable, due to the interpretative problems described in the text. I predict, however, that PoRs involving naturally disjoint verbs become more acceptable, if the bigger context (i.e. the context beyond the bare passive vP) provides some expectation about how to interpret the variable (see. fn. 7 for some exemplification). 30 Both versions of the dependent case approach free Burzio’s Generalization from any relation to theta-roles (see e.g. Sigurðsson 2006 and already Abraham 1986 for motivation). Both versions predict, furthermore, well-restricted, but slightly different, sets of counterexamples to the original version of Burzio’s Generalization.

Passives of reflexive verbs: The repair of a Principle A violation   

   361

8 References Abraham, Werner. 1986. Unaccusatives in German. Groninger Arbeiten zur germanistischen Linguistik (GAGL) 28: 1–72. Ágel, Vilmos. 1997. Reflexiv-Passiv, das (im Deutschen) keines ist. Überlegungen zu Reflexivität, Medialität, Passiv und Subjekt. In: Ch. Dürscheid, K. H. Ramers and M. Schwarz (eds.), Sprache im Fokus. Festschrift für Heinz Vater zum 65. Geburtstag, 147–187. Tübingen: Niemeyer. Baker, Mark. 2008. The Syntax of Agreement and Concord. Cambridge: Cambridge University Press. Baker, Mark and Nadya Vinokurova. 2010. Two modalities of case assignment: case in Sakha. Natural Language and Linguistic Theory 28: 593–642. Barðdal, Jóhanna, and Molnár, Valeria. 2003. Passive in Icelandic – compared to Mainland Scandinavian. In: J. Hetland and V. Molnár (eds.), Structures of Focus and Grammatical Relations, 231–260. Tübingen: Niemeyer. Behagel, Otto. 1924. Deutsche Syntax. Eine geschichtliche Darstellung. Volume II. Heidelberg: Winter. Bierwisch, Manfred. 2006. German Reflexives as Proper and Improper Arguments. In: P. Brandt and E. Fuß (eds.), Form, Structure, and Grammar: A Festschrift Presented to Günther Grewendorf on Occasion of his 60th Birthday, 15–36. Berlin: Akademie-Verlag. Burzio, Luigi. 1991. The morphological basis of anaphora. Journal of Linguistics 27: 81–105. Burzio, Luigi. 1998. Anaphora and Soft Constraints. In: P. Barbosa, D. Fox, P. Hagstrom, M. McGinnis and D. Pesetsky (eds.), Is the Best Good Enough? Optimality and Competition in Syntax, 93–114. Cambridge MA: MIT Press. Cinque, Guglielmo. 1988. On Si constructions and the theory of Arb. Linguistic Inquiry 19: 521–581. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 2001. Derivation by phase. Michael Kenstowicz (ed.), Ken Hale: A Life in Language, 1–52. Cambridge MA: MIT Press. Chomsky, Noam. 2008. On phases. In: R. Freidin, C. P. Otero and M. L. Zubizarreta (eds.), Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud, 133–166. Cambridge MA: MIT Press. Collins, Chris. 2005. A smuggling approach to the passive in English. Syntax 8: 81–120. Doron, Edith and Malka Rappaport Hovav. 2007. Towards a uniform theory of valence-changing operations. In: Y. N. Falk (ed.), Online-proceedings of IATL 23 (http://linguistics.huji.ac.il/ IATL/23/). Embick, David. 2004. Unaccusative syntax and verbal alternations. In: A. Alexiadou, E. Anagnostopoulou and M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface, 137–158. Oxford: Oxford University Press. Everaert, Martin. 1990. Nominative Anaphors in Icelandic: Morphology or Syntax? In: W. Abraham and E. Reuland (eds.), Germanic Syntax Workshop, 277–307. Amsterdam: John Benjamins. Eythórsson, Thórhallur. 2008. The New Passive in Icelandic really is a passive. In: T. Eythórsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers, 173–219. Amsterdam and Philadelphia: Benjamins. Fanselow, Gisbert. 1987. Konfigurationalität. Tübingen: Narr.

362   

   Florian Schäfer

Fanselow, Gisbert. 1991. Minimale Syntax. Habilitationsschrift, University of Passau. Groninger Arbeiten zur germanistischen Linguistik 32. Fischer, Silke. 2004. Towards an optimal theory of reflexivization. PhD thesis, University of Tübingen. Fischer, Silke. 2006. Matrix unloaded: Binding in a local derivational approach. Linguistics 44: 913–935. Frampton, John and Sam Gutmann. 2000. Agreement is feature sharing. Ms., Northeastern University. Frey, Werner. 1993. Syntaktische Bedingungen für die semantische Interpretation. Über Bindung, implizite Argumente und Skopus. Berlin: Akademie-Verlag. Geniušienė, Emma. 1987. The Typology of Reflexives. Berlin: Mouton de Gruyter. Grewendorf, Günther. 1988. Aspekte der deutschen Syntax. Tübingen: Narr. Grimshaw, Jane. 1981. On the lexical representation of Romance reflexivie clitics. In: J. Bresnan (ed.), The Mental Representation of Grammatical Relations, 87–148. Cambridge MA: MIT Press. Haider, Hubert. 1985. Von sein oder nicht sein: Zur Grammatik des Pronomens „sich“. In: W. Abraham (ed.), Erklärende Syntax des Deutschen, 223–254. Tübingen: Narr. Halle, Morris and Alec Marantz. 1993. Distributed Morphology and the pieces of inflection. In: K. Hale and S. J. Keyser (eds.), A View From Building 20, 111–176. Cambridge MA: The MIT Press. Heinat, Fredrik. 2006. Probes, pronouns, and binding in the Minimalist Program. PhD thesis, Lund University. Hole, Daniel. 2006. Agentive selbst and other instantiations of the identity function in German. Ms., University of Munich. Hole, Daniel. 2008. Focus on identity – The dark side of zìjĭ’. The Linguistic Review 25: 267–295. Holmberg, Anders. 2002. Expletives and agreement in Scandinavian passives. Journal of Comparative Germanic Linguistics 4: 85–128. Hundt, Markus. 2002. Formen und Funktionen des Reflexivpassivs im Deutschen. Deutsche Sprache 30: 124–166. Jónsson, Jóhannes G. 2009. The new impersonal as a true passive. In: A. Alexiadou, J. Hankamer, T. McFadden, J. Nuger and F. Schäfer (eds.), Advances in Comparative Germanic Syntax, 281–306. Amsterdam: John Benjamins. Kaufmann, Ingrid. 2001. Medium: Eine Studie zur Verbsemantik. Habilitationsschrift, University of Düsseldorf. Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam: John Benjamins. Kiss, Tibor. 2003. Die Genese der Ausnahmeanapher. In: L. Gunkel, G. Müller, and G. Zifonun (eds.), Arbeiten zur Reflexivierung, 157–188. Tübingen: Narr. Koenig, Ekkehard and Letizia Vezzosi. 2004. The role of predicate meaning in the developement of reflexivity. In: B. Wiemer, W. Bisang, and N. Himmelmann (eds.), What makes Grammaticalization? A Look from its Fringes and its Components, 213–244. Berlin: Mouton de Gruyter. Kratzer, Angelika. 2009. Making a pronoun: Fake indexicals as windows into the properties of pronouns. Linguistic Inquiry 40: 187–237. Labelle, Marie. 2008. The French reflexive and reciprocal se. Natural Language and Linguistic Theory 26: 833–876.

Passives of reflexive verbs: The repair of a Principle A violation   

   363

Legate, Julie A. 2005. Phases and cyclic agreement. In: M. McGinnis and N. Richards (eds.), MIT Working Papers in Linguistics 49: Perspectives on Phases, 147–156. Cambridge, MA: Department of Linguistics and Philosophy, MIT. Maling, Joan and Sigrídur Sigurjónsdóttir. 2002. The new impersonal construction in Icelandic. The Journal of Comparative Germanic Linguistics 5: 97–142. Marantz, Alec. 2000. Case and licensing. In: E. Reuland (ed.), Arguments and Case: Explaining Burzio’s Generalization, 11–30. Amsterdam: John Benjamins. Marantz Alec. To appear. Phases and words. MS, NYU. McFadden, Thomas. 2004. The position of morphological case in the derivation: a study on the syntax-morphology interface. PhD thesis, University of Pennsylvania. McGinnis, Martha. 1998. Reflexive external arguments and lethal ambiguity. In: E. Curtis, J. Lyle and G. Webster (eds.), Proceedings of WCCFL 16, 303–317. Stanford: CSLI. McGinnis, Martha. 2000. Reflexive clitics and the specifiers of vP. In: L. Pylkkänen, H. Harley and A. van Hout (eds.), MIT Working Papers in Linguistics 35: Papers from the UPenn/ MIT Round Table on the Lexicon, 137–160. Cambridge, MA: Department of Linguistics and Philosophy, MIT. McGinnis, Martha. 2004. Lethal ambiguity. Linguistic Inquiry 35: 47–95. Müller, Stefan. 1999. Deutsche Syntax deklarativ. Head Driven Phrase Structure Grammar für das Deutsche. Tübingen: Niemeyer. Müller Gereon and Wolfgang Sternefeld. 1993. Scrambling as A-bar movement. In: N. Corver and H. van Riemsdijk (eds.), Studies on Scrambling. Movement and Non-Movement Approaches to Free Word Order Phenomena, 331–385. Berlin: Mouton de Gruyter. Pestsky, David and Esther Torrego. 2007. The syntax of valuation and the interpretability of features. In: S. Karimi, V. Samiian and W.Wilkins (eds.), Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation, 262–294. Amsterdam: John Benjamins. Plank, Frans. 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 TL-00.002, Utrecht University. Reinhart, Tanya and Eric Reuland. 1993. Reflexivity. Linguistic Inquiry 24: 657–720. Reinhart, Tanya and Tal Siloni. 2004. Against the unaccusative analysis of reflexives. In: A. Alexiadou, E. Anagnostopoulou and M. Everaert (eds.), The Unaccusativity Puzzle, 288–331. Oxford: Oxford University Press. Reinhart, Tanya and Tal Siloni. 2005. The lexicon-syntax parameter: Reflexivization and other arity operations. Linguistic Inquiry 36: 389–436. Reis, Marga. 1982. Reflexivierung im Deutschen. In: E. Faucher (ed.), Actes du Colloque du Centre de Recherches Germaniques de l’Université de Nancy & Journée Annuelle des Linguistes de l’Association des Germanistes de l’Enseignement Supérieur 12, 1–40. Nancy. Reuland, Eeric. 2001. Primitives of binding. Linguistic Inquiry 32: 439–492. Reuland Eric. 2005. Agreeing to bind. In: H. Broekhuis, N. Corver, R. Huybregts, U. Kleinhenz and J. Koster (eds.), Organizing Grammar: Linguistic Studies in Honor of Henk van Riemsdijk, 505–513. Berlin: Mouton de Gruyter. Ruys, Eddy. 2010. Expletive selection and CP arguments in Dutch. Journal of Comparative Germanic Linguistics 13: 141–178. Schäfer, Florian. 2008. The Syntax of (Anti-)Causatives. External arguments in Change-of-State Contexts. Amsterdam: John Benjamins.

364   

   Florian Schäfer

Schäfer, Florian. To appear. Local Case, cyclic Agree and the syntax of truly ergative verbs. In: A. Alexiadou, T. Kiss and G. Müller (eds.), Local Modelling of Non-Local Dependencies in Syntax. Tübingen: Niemeyer. Sells, Peter, Annie Zaenen and Draga Zec. 1987. Reflexivization variation: Relations between syntax, semantics, and lexical structure. In: M. Iida, S. Wechsler and D. Zec (eds.), Working Papers in Grammatical Theory and Discourse Structure: Interactions of Morphology, Syntax, and Discourse, 169–238. Stanford: CSLI Publications. Sigurðsson, Halldór A. 1989. Verbal syntax and case in Icelandic. Doctoral dissertation, University of Lund. Sigurðsson, Halldór A. 2000. The locus of case and agreement. Working Papers in Scandinavian Syntax 65: 65–108. Sigurðsson, Halldór A. 2003. Case: abstract vs. morphological. In: E. Brandner and H. Zinsmeister (eds.), New Perspectives on Case Theory, 223–268. Stanford: CSLI. Sigurðsson, Halldór A. 2005. Accusative and the nom/acc alternation in Germanic. Working Papers in Scandinavian Syntax 76: 93–133. Sigurðsson, Halldór A. 2006. The nominative puzzle and the low nominative hypothesis. Linguistic Inquiry 37: 289–308. Sigurðsson, Halldór A. 2009. The No Case generalization. In: A. Alexiadou, J. Hankamer, T. McFadden, J. Nuger and F. Schäfer (eds.), Advances in Comparative Germanic Syntax, 249–280. Amsterdam & Philadelphia: John Benjamins. Sigurðsson, Halldór A. 2011. On the new passive. Syntax 14: 148–178. Sternefeld, Wolfgang. 2006. Syntax. Eine morphologisch motivierte generative Beschreibung des Deutschen. Tübingen: Stauffenberg. Vater, Heinz. 1995. Zum Reflexiv-Passiv im Deutschen. In: H. Popp (ed.), Deutsch als Fremdsprache: an den Quellen eines Faches. Festschrift für Gerhard Helbig zum 65. Geburtstag, 185-192. München: Iudicium. Wiemer, Björn. 2006. Relations between actor-demoting devices in Lithuanian. In: W. Abraham and L. Leisiö (eds.), Passivization and Typology (Form and Function), 274–309. Amsterdam, Philadelphia: Benjamins. Woolford, Ellen. 1999. More on the Anaphor-Agreement Effect. Linguistic Inquiry 30: 257–287. Wunderlich, Dieter. 1985. Über die Argumente des Verbs. Linguistische Berichte 97: 183–227. Wurmbrand, Susi. 2010. Parasitic morphology in Germanic: Consequences for the theory of feature checking. Talk at the guest lecture series of the SFB 732 ‘Incremental Specification in Context’, Universität Stuttgart. Zeijlstra, Hedde. 2010. There is only one way to agree. Talk at GLOW 33, Wroclaw, Poland.

Subject index A/A’-position 139–141, 146, 164, 171–2, 189, 199 (FN 17), 278, 287, 294, 297 aber 56–58, 70–74, 84, 86, 88, 92 (FN 40) ability 19, 31–33, 44–47 aboutness 275–276, 286, 290–292, 302, 312, 319 (FN 13), 320 (FN 15), 321 (FN 25) accent 9–10, 68, 200 (FN 28) accommodation 48, 65 accomplishment 32, 44–45, 101–105 accusative 13, 24, 119, 133, 137, 150–151 (FN 1), 214, 216, 224, 270 (FN 6), 304, 335–336, 344–347, 356–357, 359 (FN11, FN 12) achievement 32, 96, 101–104 activity 32–33, 39–44, 47–49, 100–101, 109, 126 (FN 12), 162, 193 actual world 32–33, 49 adjacency 187–188, 252, 277–279, 283, 291–294, 317, 321 (FN 18) adjective 96, 209, 247–248, 261–262, 270 (FN 11) adjunct 143, 155, 161, 163–165, 170–171, 186, 195, 198 (FN 12, FN 13), 233, 277, 304 adverb 18–19, 37, 56–57, 113, 165, 176–177, 185, 188, 195, 201 (FN 36), 209–211, 217, 232 (FN 7), 275, 282–284 adverbial 18–19, 44–47, 105, 184, 188, 209–211, 214–215, 279, 303 adversative 109, 111 agent 104, 113, 345–348 agreement 4, 24, 132, 156, 161, 169, 174–175, 179, 181–183, 199 (FN 24), 200 (FN 25), 243–245, 253–256, 258–259, 261–262, 267, 281, 293, 305, 335, 346, 350–357 AgroP 8, 179 AspP 136, 137 amalgamation 201 (FN 33), 275, 293–294, 296–297, 303, 307–309, 317–318, 326 (FN 55) amend 8, 10, 18, 21 amount 284–285, 294, 309–310, 315, 317

anaphor 24, 335, 344–348, 350–352, 355, 357, 359 (FN 16), 360 (FN 21, FN 22) anchoring 32–35, 36–38, 125 (FN 6) anticausative 98–100, 102, 125 (FN 2, FN 5), 360 (FN 21, FN 24) antisymmetry 151 (FN 1) argument alternation 95, 97–100, 102, 122–124 argument structure 2, 21, 95–96, 101–102, 120, 123, 160, 162–165 articulation 1, 11–12, 22–23, 223 (FN 4) aspect 17–19, 31–33, 36–39, 43, 47, 49–50, 175 asymmetry 14, 207, 278 assertion 4, 20, 33, 47–49, 186, 290–291, 297 atelicity 33–35, 39, 165 attribution 85, 115–117, 120, 123, 237, 239–241, 245, 248, 258, 265 autosegmental phonology 214, 217, 220–221, 233 (FN 18) auxiliary 6, 155, 158, 165, 175, 178–180, 190, 201 (FN 33), 224, 259, 269, 320 (FN 18), 336 background 4, 107–108 base-generation 8, 10, 163, 168, 171, 174–175, 180, 188, 191, 197 (FN 7), 243, 275, 292, 296, 303, 305–308, 313–314, 318, 323 (FN 35) big DP 242–245 binding 24, 107, 117, 148, 196 (FN 6), 305, 311, 326 (FN 52), 335–336, 345–347, 349–350, 357, 358 (FN 4), 360 (FN 19) blocking 32, 44, 56, 68, 71–73, 90 (FN 17), 101, 111, 116–117, 123, 159, 170, 178, 181, 184, 196 (FN 5), 199 (FN 19), 279, 281, 306, 313, 315–316, 327 (FN 60) Burzio’s generalization 336, 346 case assignment 2, 4, 12–13, 24, 135, 161, 179, 181, 184, 200 (FN 25), 214–217, 223–224, 226–227, 234 (FN 25, FN 27), 241–245, 253, 255, 258, 265, 270

366   

   Subject index

(FN 6), 305–306, 312, 316, 322 (FN 32), 328 (FN 64), 335–336, 345, 347–348, 350–352, 356, 360 (FN 24) category 1, 4, 14, 24, 106, 139, 142, 192, 216–217, 228, 243, 270 (FN 10), 350 causative 96, 98–100, 125 (FN 2), 225 c-command 12, 187, 203 (FN 45), 207, 223, 350–353, 357 chain 14, 23, 24 (FN 4), 138, 353–355 change of state 18–19, 193 clitic 12, 242–243, 289, 303, 324 (FN 43) co-argument 343, 347, 357 coda (syllable) 230–231 coercion 16–19, 25 (FN 9, FN 12), 45–46, 97, 111, 117, 122 comparative 116, 155, 172, 190, 309 competition 16–17, 31, 35, 43, 49–50, 184, 197 (FN 10), 254, 257, 264, 279, 290, 327 (FN 61) complement 21, 108, 117, 125 (FN 9), 136, 138–9, 142, 152 (FN 5), 165, 197 (FN 6), 201 (FN 37), 302, 311, 314–316, 328 (FN 65) complementizer 23, 159, 188–189, 201 (FN 37), 275–277, 289–290, 298, 307 compositionality 15, 31, 47, 50, 52 (FN 24), 111 conflict 12, 19, 23, 33, 50, 71, 134–135, 190, 192, 230, 238, 250, 258, 270 (FN 14), 300–301, 318 conjunction 17, 20, 57, 68, 73, 89 consonant 221, 225, 230–231, 234 (FN 1) constraint 16, 22, 76, 97, 117, 124, 144, 149, 155–156, 179–184, 189, 191–192, 195–196, 200 (FN 25, FN 26, FN 27, FN 30), 209, 244, 247, 277, 286, 291–292, 296–297, 305–306, 320 (FN 18), 322 (FN 31) construct state 226–227 content 14, 21, 32, 42, 51 (FN 20), 56, 72, 83, 108, 148 context 17, 20, 33, 40, 43, 61, 73–74, 77, 81–82, 84–86, 95–100, 103–106, 112–115, 121–124, 194, 237, 250–251, 271 (FN 15), 323 (FN 39), 338, 348–50

contradiction 23, 32, 52 (FN 24), 55, 71–72, 90 (FN 18), 109, 116, 125 (FN 7), 267, 270 (FN 12), 336 contrast 23, 106–107, 115–116, 122–124, 125 (FN 7), 151 (FN 4), 190, 200 (FN 28), 251, 270 (FN 11), 275–276, 286–287, 289–297, 302, 307–311, 317–318, 321 (FN 24), 322 (FN 27, FN 28), 323 (FN 39) convention 15, 25 (FN 13), 50 (FN 7), 56, 58–59, 88 convergence 135–136, 258, 296, 306, 319 (FN 10) copula 96, 131, 138–139, 143, 145–148, 165 copy 243–244, 246–247, 250, 261, 270 (FN 9) CP 2, 13, 21, 23, 134, 137, 177, 183, 188–190, 199 (FN 20), 201 (FN 37), 202 (FN 37), 210, 215–217, 233, (FN 15), 247, 250, 255–258, 260, 276–277, 280, 287, 289–290, 292–295, 297, 300, 301–303, 307–309, 311, 315, 317, 319 (FN 4), 321 (FN 24), 322 (FN 34), 323 (FN 40), 327 (FN 60) crash 7, 12, 21, 134–135, 192, 292–294, 317, 319 (FN 10), 350, 353–355 cycle 3, 5, 13, 133–134, 180, 183, 187, 188, 244, 256–257, 292, 298, 307 dative 99, 121, 125 (FN 3), 281, 303–304, 312, 321 (FN 22), 325 (FN 49), 341 default 3, 23, 24, 67, 70–74, 81–83, 85, 96, 103, 108, 114–115, 155, 179, 252–254, 271 (FN 15), 275, 296, 303, 305, 312, 335, 346, 351, 354, 356–357 defect 16–17, 243, 258, 261, 267, 350 definite 105, 218, 282, 297, 319 (FN 13), 359 (FN 12) deletion 6–8, 21, 23, 95, 97–98, 102, 111, 122, 146, 250–251, 301, 327 (FN 56), 360 (FN 23) denotation 19, 25 (FN 10), 33, 35, 39, 49, 51 (FN 20), 99, 106–108, 111, 122, 275, 287, 322 (FN 27), 356 derivation 5, 7, 16, 21, 124, 190, 208, 221, 279, 352 determiner 218, 226, 265 deverbal noun 214, 223–228, 234 (FN 25)

Subject index   

DP 13, 14, 156, 161, 193, 199 (FN 20), 215–217, 226–228, 242–245, 247, 257, 279, 282, 291, 295–297, 319 (FN 12), 320 (FN 14, FN 15), 321 (FN 25), 322 (FN 30, FN 32), 335, 336, 338, 348, 350–353, 355–357, 358 (FN 4), 359 (FN 15) displacement 4–5, 132, 174, 179, 200 (FN 27), (FN 31) cf. movement, extraction discourse 2, 20–21, 71, 75, 79, 82–85, 91 (FN 28), 193, 271 (FN 15), 281, 289–90, 319 (FN 13, 322 FN 29), cf. context Distributed Morphology 233 (FN 19), 238, 250–251, 255  do-support 6, 16, 158, 167, 177, 181, 189–190, 295–296, 325 (FN 46) economy 5, 9–10, 20, 24, 49, 135, 183, 190, 192, 200 (FN 27), 246–247, 257, 279, 316, 350 ellipsis 7, 11, 21–22, 110, 131–138, 143–147, 149–150, 151 (FN 1, FN 4), 178, 190, 305 emphasis 289, 295 enrichment 20, 42, 55–60, 62–63, 66–69, 72–79, 86, 88–89, 89 (FN 2), 90 (FN 15), 100 entailment 15, 40, 46, 48, 55, 84, 98, 103, 113, 122 error 65, 324 (FN 45), 325 (FN 46) evaluative 113–114, 117, 123 event 18–9, 32–3, 38–40, 44–47, 101–102, 111, 115–116, 165, 283, 338, 356 exclusivity 207–208 see nontangling, nontampering existential 33, 104, 107, 117, 122, 297, 359 (FN 17) experiencer 117–118, 125 (FN 11), 199 (FN 19), 281, 325 (FN 49) expletive 6, 155, 166–168, 278–279, 292, 353–355, 360 (FN 27) external argument 101, 117–118, 125 (FN 5), 336, 343, 346–347, 352, 356, 358 (FN 3), 359 (FN 18), 360 (FN 21) external merge 306, 327 (FN 57) externalisation 192, 220, 222, 233 (FN 22) extraction 7–8, 23–24, 148, 157–159, 187–189, 196 (FN 3, FN 5), 202 (FN 39, FN 40), 275–283, 286, 292, 295–302,

   367

314–318, 319 (FN 12), 320 (FN 14), 323 (FN 40), 328 (FN 65) extrametrical 225, 229–230 extraposition 155, 166, 168–170, 176, 194–195, 196 (FN 6), 201 (FN 36) Extended Projection Principle (EPP) 5,8, 12–14, 132–133, 135, 137, 144, 152 (FN 6), 169, 171–172, 179–180, 183–184, 186–188, 190–192, 194–195, 199 (FN 24), 200 (FN 25), 202 (FN 41), 269 (FN 3), 283, 292–294, 296–297, 307–309, 315, 317, 322 (FN 33), 327 (FN 56), 354, 360 (FN 27) feature checking/valuation 8, 11–12, 21, 132–133, 151 (FN 3), 174, 179, 183, 199 (FN 24), 201 (FN 37), 227, 231, 244, 276, 279, 290–293, 307–308, 322 (FN 33), 350, 352–355 feature inheritance 131, 135–137 felicity 9, 14, 36, 47, 120–121, 199 (FN 21) filter 12, 76, 80–82, 277, 283, 289, 320 (FN 14), 355 finiteness 139, 155–156, 159, 163, 171–172, 186–188, 196 (FN 6), 201 (FN 37), 202 (FN 40), 225–226, 258, 281, 311, 327 (FN 60, FN 61) FinP 287, 295–298, 308, 309, 322 (FN 34), 326 (FN 55) flexibility 2, 6, 155, 171, 199 (FN 20), 317 FocP 321 (FN 24) focus 9, 68, 145–148, 150 (FN 1), 155, 172, 180–185, 190, 192–196, 200 (FN 28), 201 (FN 36), 241, 250–251, 270 (FN 14), 287, 295, 297, 302, 321 (FN 24, FN 25), 338, 348 ForceP 287, 292–297, 308–309, 315–317, 326 (FN 55) free relative 12, 270 (FN 6), 325 (FN 49) freezing 148, 169, 278–279, 319 (FN 12) full interpretation 5, 24 (FN 4), 181, 354 functional category 136, 139 gap 23, 238–239, 245, 303–306, 312–313, 317, 325 (FN 49), 326 (FN 51) gender 261–261, 270 (FN 11)

368   

   Subject index

generalization 3–5, 25 (FN 11, FN 14), 173, 175, 294, 320 (FN 14), 321 (FN 22), 336, 346, 358 (FN 7), 360 (FN 30) genericity 6, 15, 21, 24 (FN 2), 25 (FN 14), 32, 46, 95–97, 102–107, 110–116, 121–123, 124 (FN 7), 287, 358 (FN 3), 360 (FN 21) genitive 226–227, 234 (FN 25) goal (theta role) 118–119 goal (feature theory) 136, 138, 183, 187, 197 (FN 8), 199 (FN 24), 242–244, 350 Government and Binding (GB) 2, 247, 277 grammatical function 156, 164, 283, 323 (FN 40) grammaticalization 16, 324 (FN 43), 325 (FN 46) greed 132, 143 habituality 19, 32, 37–38, 44–47, 51 (FN 13), 103 head movement 7, 12, 131–133, 138, 140–141, 143, 150, 182, 198 (FN 12), 200 (FN 31), 228, 231–232, 269 hiatus 32–33, 40, 47–49, 50 (FN 4) homogeneity 33, 39–40 imperfection 8, 12 imperfective 18–19, 35, 36–39, 43, 45, 49, 50 (FN 8) impersonal passive 336, 340, 344–345, 353–355, 357, 358 (FN 3), 359 (FN 12, FN 15), 360 (FN 27) implicature 14–15, 19, 50 (FN 7), 125 (FN 2) implicit argument 104–106, 118, 123, 335, 343, 345, 349, 357, 358 (FN 3), 359 (FN 17), 360 (FN 20) improper movement 23, 275–276, 286, 291–292, 302, 322 (FN 30) inclusion 39, 45, 50 (FN 8) indefinite 108, 122, 158, 281, 291, 320 (FN 15) indirect speech 25 (FN 13), 81, 85 individual 17–18, 21, 108, 315 individual-level predicate 111, 116–117, 121, 125 (FN 9), 285–286 inference 1, 33, 40–42, 51 (FN 17), 56, 70–71, 80–81, 84–85, 103–105, 109, 114–115

inflection 12, 174, 181, 223, 269, 270 (FN 11), 281, 336, 341 information structure 9, 22–24, 50 (FN 10), 70–72, 80–82, 85, 145–148, 150, 150 (FN 1), 155, 178, 183, 191–194, 199 (FN 19), 200 (FN 28, FN 29), 201 (FN 36), 250–251, 283, 287, 289–291, 318, 321 (FN 25), 323 (FN 40), 326 (FN 54) inherent reflexive 335, 338–342, 344–347, 348–351, 356–357, 360 (FN 22) input correspondence 228–229, 234 (FN 29) intention 37–38, 99, 113 interface 5–7, 11–12, 14–16, 19, 22–24, 24 (FN 4), 97, 112, 123, 180, 200 (FN 27, FN 30), 218, 221, 239, 258, 262–263, 265–266, 349–351, 352–357, 360 (FN 23) internal argument 96–97, 100–101, 103–105, 107, 109–112, 122, 126 (FN 12), 343, 347, 352, 359 (FN 12, FN 18) interpretability 6–7, 11–13, 21, 31, 45, 76, 96, 132, 136, 143, 151 (FN 3), 266, 269 (FN 2), 290, 292, 349, 360 (FN 29) interrogative 212–214, 216–217, 233 (FN 10, FN 15), 300 intervention 24 (FN 3), 186, 197 (FN 8, FN 10), 202 (FN 39), 241, 247, 277, 279 intonation 70, 79, 86, 212–213, 231, 250–251 intransitivity 35–37, 104–105, 107, 110–112, 122, 125 (FN 7), 126 (FN 12), 226, 353, 359 (FN 18), 360 (FN 18) inversion 12, 17, 22, 138–139, 150, 155–156, 160–161, 163–164, 166, 168, 170–172, 176, 180, 183–184, 188, 190–192, 195–196, 196 (FN 6), 198 (FN 11, FN 13), 199 (FN 17), 201 (FN 36, FN 37), 202 (FN 38), 203 (FN 44), 277, 279, 282, 294, 318 (FN 3), 320 (FN 18), 324 (FN 45), 325 (FN 46) IP 8, 9 irony 25 (FN 13), 91 (FN 28), 358 (FN 7) island 7–8, 24 (FN 3), 146, 149, 159, 186, 188, 196 (FN 5), 199 (FN 19), 202 (FN 39), 304–305, 315, 318, 319 (FN 12), 320 (FN 14), 328 (FN 65) isomorphism 2 iterativity 111, 123

Subject index   

kind 103, 106–107, 114–116, 118, 121, 123, 281–282 left periphery 242, 246, 249, 266, 287, 326 (FN 55) level 1–2, 4, 11, 16–17, 20, 75, 102, 123, 134, 141, 143, 214, 229, 233 (FN 13), 238, 244, 258, 266, 343 lexicon 11, 23, 25 (FN 7), 33, 39, 43, 51 (FN 20), 95–98, 100, 112, 123–124, 136, 143, 151 (FN 2), 198 (FN 13), 209, 217–218, 224, 232 (FN 5), 237–239, 243, 245, 251–252, 260–269, 270 (FN 5, FN 12), 339, 343, 347, 357 linearisation 3, 12, 22, 133–135, 200 (FN 26), 207–208, 210–214, 218–221, 223, 231–232, 233 (FN 23), 247, 296 literal meaning 14–15, 20, 55–56, 58–60, 62–63, 65, 67–68, 70–71, 74–79, 86, 88, 89 (FN 1), 90 (FN 4), 92 (FN 39) locality 5, 21, 24, 139, 151 (FN 4), 198 (FN 12), 247, 249, 252, 255, 270 (FN 5), 279, 291, 295, 297, 303–308, 313–314, 316–317, 319 (FN 6, FN 12), 350–351, 359 (FN 11) locative 17, 22, 118, 155, 156–163, 164, 168–170, 186–188, 191, 195, 196 (FN 3, FN 5), 197 (FN 8), 198 (FN 12), 203 (FN 48), 277, 279, 318 (FN 3), 360 (FN 27) Logical Form (LF) 6, 24 (FN 4), 31, 49–50, 246, 265, 293, 305, 322 (FN 31), 350, 352, 355 look-ahead 132, 257 markedness 9, 51, 80, 96, 100, 124 (FN 1), 148, 155–156, 161, 164, 183–184, 200 (FN 28, FN 33), 212–213, 238, 254, 257, 312, 316, 340–341 maxim 20, 25 (FN 13), 58–59, 61, 90 (FN 8), 284 merge 5, 13–14, 44–45, 136, 139–141, 178–179, 184, 199 (FN 24), 200 (FN 25), 207, 222–223, 226, 231–232, 238, 257, 263–264, 270 (FN 4), 293–294, 306, 317, 322 (FN 33), 347, 352 metaphor 25 (FN 13), 84

   369

middle 6, 15, 24 (FN 2), 25 (FN 14), 33, 46, 52 (FN 24), 95–97, 112–123, 125 (FN 7), 126 (FN 15, FN 17), 360 (FN 21, FN 24) Minimalism 5–7, 9, 11–13, 140, 175, 238, 241, 244, 264, 268, 292, 354, 360 (FN 23) mismatch 6, 12–13, 17, 33, 246, 262, 265, 267 modality 6, 15, 19–20, 24 (FN 2), 25 (FN 14), 33, 51 (FN 14), 52 (FN 24, FN 26), 116, 178, 209–210 modalization 20, 33, 55–58, 81 modularity 1–2, 11, 16, 19, 84, 210, 223, 231 monotonicity condition 21, 95, 97–98, 100–102, 107, 112, 122–123 morpheme 39, 43, 214, 225–226, 228–230 movement 5, 7–8, 12–14, 22–23, 131–143, 147–148, 150, 150 (FN 1), 151 (FN 2), 155–158, 161, 171, 174–185, 187–192, 195–196, 197 (FN 10), 198 (FN 12), 200 (FN 27, FN 31, FN 32, FN 33), 227–228, 231–232, 242–245, 247, 248–251, 255, 270 (FN 9), 275–276, 278–279, 281–283, 286–297, 302–303, 305–308, 311–318, 319 (FN 12), 321 (FN 24), 322 (FN 30), 323 (FN 36), 327 (FN 63), 352, 359 (FN 18) natural class 103, 115–116 naturally disjoint verb 338–344, 349 naturally reflexive verb 335, 338–346, 356–357, 358 (FN 6) negation 40, 42, 83, 109, 125 (FN 7), 171, 173, 177–178, 181, 199 (FN 22), 211–217, 221–222, 232, 233 (FN 10), 291 nominalization 216, 224–228, 231, 234 (FN 27) nominative 13, 175, 184, 214, 234 (FN 25), 240–241, 253, 255, 270 (FN 6), 279, 281, 304, 312, 316, 324 (FN 42), 336, 351, 358 (FN 4), 359 (FN 11, FN 15), 360 (FN 24) NP 7, 8, 17–18, 25 (FN 10), 99, 103, 114, 118, 120–122, 125 (FN 3, FN 6), 201 (FN 36), 217, 226–228, 305–307, 308, 322 (FN 32), 343 nucleus 230 number 35, 47, 259–261, 346

370   

   Subject index

object 9–10, 44, 108–109, 112, 117, 119–123, 131, 133–135, 145–146, 150 (FN 1), 157, 183, 193–194, 200 (FN 25), 211, 224, 226, 234 (FN 27), 276–279, 283, 294, 303, 304–306, 312, 317, 319 (FN 5), 321 (FN 22), 323 (FN 35), 335, 338, 339, 342, 347, 348, 351–352, 356, 360 (FN 24) operator 6, 15, 19, 20–21, 23, 40, 45, 47, 50, 52 (FN 23, FN 25, FN 26), 55–56, 66–67, 70, 75, 77, 88–89, 92 (FN 39), 109, 147, 173, 180, 189–190, 202 (FN 41), 248–250, 255–256, 258, 263, 270 (FN 8), 271 (FN 15), 275–276, 285, 291–197, 302, 307–310, 312, 316–318, 321 (FN 21), 322 (FN 32, FN 34), 326 (FN 53, FN 55), 327 (FN 56) Optimality Theory (OT) 16, 22, 179, 181 optionality 6, 176–177, 192, 201 (FN 37), 210, 261, 293, 313–314, 322 (FN 29), 323 (FN 35), 328 (FN 64) paradigm 88, 269, 336, 341 parallelism 145–147, 149 parenthesis 172, 189, 210, 276, 298–302, 314–316, 324 (FN 41, FN 42, FN 45), 325 (FN 46) participle 120–121, 199 (FN 23), 238, 258–259, 261, 267, 269, 270 (FN 11), 336 particle 150 (FN 1), 194–195, 233 (FN 10), 281, 319 (FN 11, FN 12) passive 8, 24, 99–100, 106, 115, 117, 160, 260, 335–337, 340–342, 344–347, 349, 353–357, 358 (FN 2, FN 3, FN 5), 359 (FN 12, FN 18), 360 (FN 27, FN 28, FN 29) patch 7–8, 11–12, 15, 21, 131 penthouse principle 21 perfective 17, 19, 32–35, 39, 40–43, 48–49, 50 (FN 6, FN 8), 51 (FN 18), 52 (FN 22) person 23, 237–242, 244–246, 250–254, 255–256, 258, 263, 265–269, 269 (FN 2), 270 (FN 8), 271 (FN 15), 294, 296–297, 299, 324 (FN 43), 346, 350–352, 354, 359 (FN 17) phase 5, 13, 134–135, 136, 138–141, 143, 151 (FN 4), 191, 197 (FN 10), 221–222, 231–232, 244, 257–258, 266, 322 (FN 34), 360 (FN 25, FN 28)

phonological composition 217–223, 227, 229, 231–232 Phonological Form (PF) 6–8, 12, 21–22, 24 (FN 3), 131–132, 136, 138, 146–147, 149–150, 150 (FN 1), 188, 202 (FN 40), 223–224, 238–239, 246, 251–252, 256–258, 261, 264–268, 275, 291, 305, 317, 321 (FN 20), 322 (FN 31), 350–352, 356 pied-piping 142, 278, 328 (FN 64) plural 46–7, 49 politeness 14, 25 (FN 13) polyphony 56, 73, 81–82, 84–85, 91 (FN 24) possible world 6, 15, 19–20, 31–33, 47, 49–50 PP 194, 197 (FN 7), 227, 228, 250, 258, 276, 279, 281, 282, 313, 343 predication 2–4, 164–165, 193, 261–262, 270 (FN 11) preemption 11–15, 24, 131, 143–144 preposition 100, 234 (FN 27), 248–249, 303, 312–313, 328 (FN 64), 342–343, 348 presupposition 18–20, 32–33, 35, 40–44, 47–49, 51 (FN 16, FN 17, FN 20), 56, 60, 77, 108, 114, 200 (FN 31), 290, 296–297, 301, 320 (FN 14), 321 (FN 26) Principle A 24, 335–336, 345, 347, 349–350, 355, 357, 359 (FN 11, FN 16) pro 67, 197 (FN 7), 207 (FN 10), 346 PRO 96 probe 132, 136, 138–139, 183, 189 (FN 24), 242–243, 319 (FN 9), 350–352, 354 pro-drop 284 progressive 37–38, 45, 50 (FN 11) projection 2, 14, 40–42, 51 (FN 17), 131, 133, 136, 139–143, 161, 175, 197 (FN 8), 201 (FN 37), 210, 215, 217, 222, 227, 229, 244, 285, 346 pronoun 12–13, 23, 24, 83, 108, 126 (FN 19), 151 (FN 2), 175, 183, 237, 239–242, 244, 251–256, 265, 270 (FN 6, FN 8), 271 (FN 15), 275, 278, 284, 287, 295, 303–304, 306–308, 311, 314, 319 (FN 13), 323 (FN 35), 326 (FN 52), 328 (FN 64), 335–339, 341–348, 350–351, 356–357, 358 (FN 5), 359 (FN 10, FN 11, FN 18)

Subject index   

property 18, 50 (FN 7), 52 (FN 26), 95–97, 103–109, 112, 114–116, 121–124, 125 (FN 6, FN 7), 164, 284, 286 proposition 2, 4, 18–19, 21, 75–76, 83, 110, 125 (FN 7), 180, 295, 307 prosody 12, 150, 183, 195, 201 (FN 36), 229, 231, 263, 270 (FN 14), 322 (FN 31), 324 (FN 42) pseudogapping 7, 131, 133, 137, 144, 150 (FN 1) psych-verb 111–112, 116–118, 123 quantifier 3–4, 15, 18, 25 (FN 10, FN 11), 174, 282, 291, 294, 319 (FN 10, FN 11), 320 (FN 15), 326 (FN 53) question 9, 40–41, 56–57, 72–74, 78, 85, 135, 155, 158, 167, 172, 177, 189, 196 (FN 5), 201 (FN 37), 276, 292, 295–296, 298, 301–302, 309, 324 (FN 45, FN 46) quirk 25 (FN 7), 240–242, 245–246, 266, 268, 281, 283, 321 (FN 22), 353 quotation 20, 73, 80–82, 84–86, 155, 175, 190–191, 199 (FN 21) raising 21–22, 131, 134–135, 137, 139–142, 144, 150, 151 (FN 2), 157, 167, 169, 187, 202 (FN 38), 242, 248–249, 255, 269 (FN 3), 327 (FN 56) reasoning 1, 20, 48, 55, 74, 76–79, 86–89 reciprocal 335, 338, 340, 358 (FN 5, FN 7), 359 (FN 10) reconstruction 197 (FN 7), 305 reflexive 24, 46, 99–100, 125 (FN 3), 225, 335–350, 356–357, 358 (FN 2, FN 5, FN 6), 359 relativization 7, 12, 23, 142, 147–148, 237–242, 244–246, 248–251, 253–255, 257–260, 262–263, 265–267, 268–269, 269 (FN 2), 270 (FN 6, FN 8), 271 (FN 15), 275–276, 302–304, 306–311, 314–315, 317, 325 (FN 49), 326 (FN 51, FN 54), 327 FN (58, FN 59) relevance 20, 73–74, 89, 91 (FN 19), 106 reordering 12, 176 resumption 23–24, 237–238, 240–242, 244–245, 249–251, 253–255, 258–259, 265–266, 268, 269 (FN 2), 271 (FN 15),

   371

275–276, 293, 302–307, 308, 310–318, 323 (FN 35), 325 (FN 49), 326 (FN 50, FN 54), 327 (FN 61, FN 62) rhetorical structure 20, 55–56, 79–80, 86–89 root 135–137, 151 (FN 2), 186, 200 (FN 33), 225–226, 229–231, 233 (FN 12), 234 (FN 26), 260, 270 (FN 10), 286, 299, 301, 315, 323 (FN 40) salience 83, 111, 115, 122 scope 178, 180, 222, 275, 278, 281, 293, 295, 301, 315, 317, 324 (FN 45) scrambling 10, 23, 183, 200 (FN 31), 281–283, 294, 319 (FN 12), 322 (FN 29, FN 30), 323 (FN 37), 326 (FN 54) selection 12, 14, 17, 138, 156, 166, 201 (FN 37), 226, 227–229, 300, 324 (FN 44), 343 semantic role 2, 117–118, 198 (FN 13), 226, 240–243, 258, 265, 267, 276, 281, 328 (FN 65), 338, 347–348, 360 (FN 30) semantic structure 2, 15, 17, 33, 51 (FN 14), 75–76, 97–98, 106–107, 116, 123, 246, 337 separation hypothesis 217–218, 221, 224, 228 simultaneity 3, 22, 145–146, 207, 209–211, 213–215, 217, 219–223, 225, 231–232, 232 (FN 4), 233 (FN 13, FN 22), 279 sluicing 7, 131, 135–136, 138, 147, 149, 151 (FN 2, FN 3), 182 small clause 139–143, 148, 161, 165–166, 191, 197 (FN 9, FN 10) spell-out 5, 11, 23, 25 (FN 7), 134, 143–144, 147, 151 (FN 2), 222, 229, 237–238, 241–242, 246–252, 255–260, 266, 268–269, 270 (FN 7, FN 9), 271 (FN 15), 292, 322 (FN 32), 350–352, 355, 360 (FN 23) split CP 287, 292–298, 308, 309, 315–317, 321 (24), 326 (FN 55) splitting 246–248, 276, 281–282, 291, 292, 312–314, 316, 319 (FN 12), 327 (FN 62) stage-level predicate 111, 284–285 stage topic 193–194, 199 (FN 20), 283–286, 292, 294, 310

372   

   Subject index

state 18–19, 39, 99, 101, 106, 110–111, 122, 125 (FN 11), 285 strengthening 14–15, 48 stress 9–10, 269 (FN 2), 270 (FN 11), 286–287, 289, 303 strong minimalist thesis (SMT) 5–6, 241–243, 267–268 subject 4–5, 8, 13–14, 22–23, 103, 107, 110–114, 116–118, 123–124, 125 (FN 6), 126 (FN 17), 135–136, 139–142, 145–148, 150, 155–158, 160–161, 166–168, 168–170, 171, 179, 181–182, 187–188, 190, 192, 195–196, 196 (FN 2), 197 (FN 10), 198 (FN 12), 201 (FN 36), 203 (FN 45), 209, 226, 234 (FN 25), 240–241, 246, 255-257, 269 (FN 3), 275-279, 280-283, 286–287, 294, 301, 306, 317, 318 (FN 3), 321 (FN 22), 324 (FN 4), 338, 346, 352, 358 (FN 3) suppression 96–97, 103, 107, 122, 126, 328 (FN 64) suspension 7–8, 152, 277 syllable 225, 229–231, 233 (FN 16, FN 18) symbol 2, 4–6, 15 symmetry 12–14, 258 syncretism 12–13, 192, 200, 202 syntactic structure 1–2, 5–6, 13, 22, 207–208, 211, 217, 221, 231–232, 250, 265, 335 tampering 21, 243, 256 tangling 207–208 telicity 32–35, 39–40, 47, 50, 95, 103–104, 106, 165 temporal 2, 32–33, 36–37, 44–45, 81, 91 (FN 22), 193, 208, 211, 219, 286, 310 tense 2, 4, 21, 37, 51 (FN 18), 117, 135, 175, 178–179, 199 (FN 21), 224, 226, 259–260, 286 that-trace effect 23, 157, 167, 169, 187–189, 202 (FN 39), 275–277, 279–281, 283–284, 291, 302, 317, 323 (FN 35) theme 4, 39, 100, 111–112, 117, 123, 161, 168, 196 (FN 5), 197 (FN 8), 347–348, 360 (FN 24) tone 213–214, 216, 221, 233

topic 4, 23, 71, 145–148, 151 (FN 1), 156, 158–161, 167–170, 179–180, 184, 186–188, 193–195, 199 (FN 17, FN 20), 201 (FN 37), 202 (FN 39), 212, 215–216, 233 (FN 15), 247, 275–276, 283–286, 289–294, 296–298, 307–309, 312–314, 317–318, 319 (FN 13), 320 (FN 14, FN 15), 321 (FN 25), 322 (FN 29), 326 (FN 54) TP 2, 8, 14, 21, 135–139, 144, 147, 149, 152 (FN 4), 155–157, 161, 164, 171–177, 179–181, 184, 186–190, 192, 193, 195, 197 (FN 6, FN 8), 198 (FN 12), 200 (FN 24, FN 26), 201 (FN 36), 202 (FN 37, FN 39, FN 40, FN 41), 203 (FN 43), 216, 217, 249, 257, 260, 269 (FN 3), 287, 295, 296, 320 (FN 16), 352–354 transformation 1, 186 transitivity 95, 101–104, 110, 137, 160, 173, 193–194, 203 FN 47, 226, 345, 347–348, 351, 359 (FN 18) translation 1–3, 5, 8, 11, 14, 15, 24, 90 (FN 14), 150 (FN 1), 223 transparency 6, 9, 15, 125, 328 (FN 64) trigger 8, 11–12, 14, 16, 20–21, 25 (FN 7), 51 (FN 16), 55, 68, 73–74, 88, 114, 116, 132, 135, 144, 156–157, 173, 181, 191–192, 200 (FN 30), 227–228, 251, 253, 261, 292–294, 307, 323 (FN 36), 327 (FN 56), 355, 357, 358 (FN 7) truth conditions 1–4, 14–16, 18, 21, 32–33, 56, 89, 95, 97–98, 105–106, 110–111, 122 type 17–18, 25 (FN 10, FN 11), 31, 45–47, 50 unaccusative 8, 98, 100, 125 (FN 2, FN 3), 156, 160, 165, 168, 170, 174, 178, 184–185, 189, 198 (FN 15), 200 (FN 33), 359 (FN 18) underspecification 12, 44, 78, 83, 109, 252, 262, 270 (FN 8), 271 (FN 15), 350, 357 (FN *) Universal Grammar (UG) 5, 22, 218, 233 (FN 18) update 71, 75–76, 356 V-2 260, 296, 298–299, 314–316, 324 (FN 41), 328 (FN 65)

Subject index   

variable 3, 6, 83–84, 104–106, 109–111, 116, 122, 147, 295, 350–353, 355–357, 360 (FN 23, FN 29) variation 3, 261, 268 verum-focus 68 vocabulary 1, 138, 232, 233 (FN 19), 238, 251–252, 254–255, 259, 262 Vorfeld 286 vowel 214, 221, 225, 227, 230–231, 233 (FN 18), 234 (FN 32), 303 vP 2, 13, 14, 21, 45, 51 (FN 21), 137, 156, 161, 163, 169, 174, 176, 177, 183, 187, 188, 190, 191, 197 (FN 10), 200 (FN 33), 201 (FN 34), 202 (FN 41), 203 (FN 45), 210, 217, 222, 226–229, 249, 255, 257, 260, 267, 281, 349, 352–354, 360 (FN 24, FN 28, FN 29)

   373

VP 2, 7, 8, 9, 21, 44, 114, 131, 133, 134, 136–138, 145–147, 151 (FN 1, FN 4), 152 (FN 4), 155, 161, 163, 170–171, 176–178, 185, 190, 197 (FN 10), 198 (FN 12), 200 (FN 33), 201 (FN 34), 204 (FN 45), 209–211, 215–217, 219, 222, 284, 301, 320 (FN 18), 321 (FN 24), 323 (FN 37), 328 (FN 65), 352–354 weakening 19, 25 (FN 13) wh-extraction 148, 158, 182–183, 188, 196 (FN 3), 292, 304, 306–308, 311–312 wh-operator 173, 276, 285, 296, 312, 326 (FN 55) wh-pronoun 12–13, 151 (FN 2), 270 (FN 6), 325 wh-question 9, 135, 189, 301, 309, 325