The Syntax of Topic, Focus, and Contrast: An Interface-based Approach 9781614511458, 9781614511564

Citation preview

The Syntax of Topic, Focus, and Contrast

Studies in Generative Grammar 113

Editors

Henk van Riemsdijk Harry van der Hulst Jan Koster

De Gruyter Mouton

The Syntax of Topic, Focus, and Contrast An Interface-based Approach Edited by

Ad Neeleman Reiko Vermeulen

De Gruyter Mouton

ISBN 978-1-61451-156-4 e-ISBN 978-1-61451-145-8 ISSN 0167-4331 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. ” 2012 Walter de Gruyter GmbH, Berlin/Boston Typesetting: Claudia Wild, Konstanz Printing: Hubert & Co. GmbH & Co. KG, Göttingen 앝 Printed on acid-free paper 앪 Printed in Germany www.degruyter.com

Preface

This book is part of the output of a project funded by the Arts and Humanities Research Council (A Flexible Theory of Topic and Focus Movement; grant number 119403), which ran from May 2006 to August 2009. We also benefitted from a British Academy grant to Reiko Vermeulen (for a threemonth visit to UCLA; grant number SG-50500) and from a grant from the Research Foundation Flanders (FWO) to Liliane Haegeman (grant number G091409) at Ghent University, which has supported Reiko’s research in recent years. The project had six members: Michael Brody, Ivona Kuþerová, Ad Neeleman, Kriszta SzendrĘi, Hans van de Koot and Reiko Vermeulen. In addition, there were three PhD projects that were carried out at UCL in the same period and that were closely associated with our work on the syntax-information structure interface. The students working on these projects were Axiotis Kechagias (who was funded by Hellenic State Scholarship Foundation (IKY) and the Leventis Foundation), Matthew Reeve (who was funded by the AHRC) and Elena Titov (who was funded by UCL). We are grateful to a number of individuals for important input at various stages of the project. We cannot mention all of them here, but would like to highlight what we think of as the Potsdam group (although not all of them are at Potsdam anymore): Caroline Féry, Gisbert Fanselow, Shinichiro Ishihara, and Malte Zimmermann. They gave us a unique and helpful forum for discussion. We received comments that led to major adjustments and clarifications of our proposals from Klaus Abels, Daniel Büring, Vieri Samek-Lodovici, Michael Wagner and Edwin Williams. We like to think that the influence of Tanya Reinhart is visible throughout. Thanks are also due to Joy Philip for her willingness to provide invaluable help with the editing of the manuscript. It is inevitable that the papers brought together in this book do not represent the complete output of the project. The works we selected address a limited set of related issues, with chapters supporting the same overall theoretical outlook. This meant that some work had to be left out, in particular that dealing with the syntactic marking of givenness by Kuþerová and by Neeleman and van de Koot. It is also inevitable that many questions remain unaddressed, have received only partial answers or have been answered in a way that may well be indicative of our igno-

vi

Preface

rance. We have tried to learn as much as we could about the mapping between syntax and information structure, effectively stretching the project to the birth of Reiko’s son Kai on the eighth of March 2012. But at some point it is time to stop. Ad Neeleman and Reiko Vermeulen

Table of Contents Chapter 1 The Syntactic Expression of Information Structure Ad Neeleman and Reiko Vermeulen

1

Chapter 2 Towards a Unified Encoding of Contrast and Scope Ad Neeleman and Hans van de Koot

39

Chapter 3 Word Order Variation and Information Structure in Japanese and Korean Reiko Vermeulen

77

Chapter 4 Encoding Focus and Contrast in Russian Elena Titov

119

Chapter 5 Against FP Analyses of Clefts Matthew Reeve

157

Chapter 6 Focus Movement Can be Destressing, but it Need not Be Kriszta SzendrĘi

189

Chapter 7 Types of Focus and their Interaction with Negation Ad Neeleman and Reiko Vermeulen

227

Chapter 8 Concluding Remarks Ad Neeleman and Reiko Vermeulen

265

References

275

Index

301

Chapter 1 The Syntactic Expression of Information Structure Ad Neeleman and Reiko Vermeulen 1.

Setting the scene

This book is about the syntax of discourse-related word order variation. We explore the idea that such word order variation is best explained in terms of effects at the interface between syntax and other components of grammar, namely information structure and semantics. This is in contrast to the cartographic approach, initiated by Rizzi (1997), where discourserelated information is argued to be directly encoded in the syntax in the form of designated, rigidly ordered functional projections and corresponding features. Work in the cartographic tradition has uncovered a wealth of data that will need to be captured by any alternative theory of discourse-related word order variation. For example, Rizzi (2004) observes that, in Italian, word order in the left periphery adheres to the following template (INT stands for a specific class of interrogative elements): (1)

Relative operator – Topic* – INT– Topic* – Focus – Modifier – Topic*

Rizzi’s explanation for this is phrase-structural. He proposes a sequence of functional projections in the higher regions of the clause. Each projection licenses an element with a particular interpretive function in its specifier. For example, focused constituents are licensed in FocusP. The order in which the various functional projections are merged then captures the linear order found in Italian: specifiers of functional heads merged later surface further to the left:

2 (2)

Ad Neeleman and Reiko Vermeulen ForceP Rel.O.

Force’

Force0

TopP

Topic

Top’

Top0

IntP INT

Int’ Int0

TopP Topic

Top’

Top0

FocP

Focus

Foc’

Foc0

ModP

Modifier

Mod’

Mod0

TopP

Topic Top0

Top’ …

However, relying on an elaborate structure like (2) is not the only way of deriving the discourse-related word order effects summarized in (1). As Abels (2012) shows in some detail, the Italian data can be derived by independently motivated considerations regarding the locality of movement. For example, relative operators can undergo long-distance movement across topics, but topics cannot undergo long distance movement across relative operators: (3)

a. [CP Rel.Op. … b. *[CP Topic …

[CP Topic … tRel.Op. …]] [CP Rel.Op. … tTopic …]]

Chapter 1 The Syntactic Expression of Information Structure

3

The above asymmetry can be explained as an effect of Relativized Minimality (given certain assumptions about the nature of Relativized Minimality (see Starke 2001) and the features that characterize relative operators and topics). Once we have this account in place for contrasts in long-distance movement, we can also use it to capture word order in the left periphery of a single clause. The ungrammatical order in (4b) is ruled out because it is generated by movement of a topic across a relative operator. (4)

a. [CP Rel.Op. … Topic … tRel.Op. …] b. *[CP Topic … Rel.Op. … tTopic …]

If Abels’ proposal is on the right track, it will allow us to eliminate the various functional projections in the left periphery without loss of empirical coverage. On this view, movement of topics and relative operators, for example, does not target specific pre-fabricated landing sites. As far as the syntax is concerned such movements can be to a variety of positions. Conditions like Relativized Minimality then filter out the unattested orders. Notice that the alternative proposal sketched above still requires a hierarchy of semantic and discourse-related features in order to derive the Italian data. This hierarchy may not determine the sequence of functional projections, but it is necessary to regulate movement. For example, the feature complex that identifies relative operators must be richer than that of topics. Similarly, the feature composition of foci must be richer than that of modifiers to capture their relative order. In other words, there is still a stipulated hierarchy underlying the ordering effects in (1). In view of this, making sense of the observed ordering effects requires a better understanding of the nature of the various semantic and discourse functions of the items in (1) and moreover an explicit theory of how these functions are mapped onto syntactic structures. This is of course a very large question that cannot be dealt with in a single book. For this reason we restrict our attention to three discourse notions – topic, focus and contrast – which we take to be primitives of information structure. The contributions in this book investigate the association of these notions to syntax in a variety of languages (including Dutch, Japanese, Korean, Russian and English). Our aim is to explain word order restrictions in terms of the mapping between syntax and information structure, without recourse to a stipulated order of functional projections or features in the syntax. One generalisation that will be central to our argumentation is that a focus cannot move across a topic. Thus, in languages in which both topics and foci move, the topic invariably lands in a position higher than the

4

Ad Neeleman and Reiko Vermeulen

focus, as in (5). If only one moves, a topic can cross an in-situ focus, but not vice versa, as in (6). If there is no movement, the relative ordering of topics and foci tends to be free, as in (7): (5)

a. [Topic [Focus [… tFocus … tTopic …]]] b. *[Focus [Topic [… tFocus … tTopic …]]]

(6)

a. [Topic [ … Focus … [ … tTopic …]]] b. *[Focus [ … Topic … [ … tFocus …]]]

(7)

a. [ … Topic … [ … Focus …]] b. [… Focus … [… Topic … ]]

We argue that this pattern of data results from the combination of two factors. Firstly, information-structural considerations require that topics be interpreted externally to foci. For now, we can represent this as in (8). Secondly, although mismatches between syntax and interpretation are tolerated, the mapping system is such that it does not allow movements that result in mismatches. This rules out (5b) and (6b), where focus movement turns a structure that allows transparent mapping into a structure that does not. (8)

[Topic [Comment Focus [Background … ]]]

We will discuss the above pattern in detail below, but first we elaborate on the notions of topic, focus and contrast. It is important to note, before we proceed, that there seems to be a clash between our claims about topic-focus order (as summarized in (5), (6) and (7)) and the claims Rizzi makes (as expressed in the order of functional projections in (2)). The tree in (2) allows topics to follow a moved focus, but we claim that this is not possible. This apparent contradiction is due to different usages of the term “topic”. Rizzi’s (1997 : 285) use of the term is rather broad: it covers elements “normally expressing old information somehow available and salient in previous discourse”. Our use of the term is considerably more restricted, as we explain below. We expect that, on this more restrictive use, the ordering effects summarized in (5) can also be observed in Italian. This will of course require that the properties of “high topics” and “low topics” are explicated in sufficient detail (see Benincà and Poletto 2004; Frascarelli and Hinterhölzl 2007; and SamekLodovici 2009).

Chapter 1 The Syntactic Expression of Information Structure

2.

5

Notions of information structure

There is overwhelming evidence from a wide range of languages that “topic”, “focus” and “contrast” are autonomous notions of information structure that interact in systematic ways with syntax (see, for example, Vallduví 1992; Rizzi 1997; Vallduví and Vilkuna 1998; Aboh 2004; and Frey 2004). We believe that these notions are part of a system that can be summarized as in the table below: (9)

Contrast

Topic

Focus

aboutness topic [topic]

new information focus [focus]

contrastive topic [topic, contrast]

contrastive focus [focus, contrast]

We will treat [topic], [focus] and [contrast] as discourse notions targeted by mapping rules operating between syntax and information structure (although none of our arguments are adversely affected if they are privative syntactic features). Since [topic], [focus] and [contrast] are notions relevant to mapping rules, they may have syntactic consequences. On the view that syntactic operations may be licensed by having an interpretive effect, movement can take place in order to feed a mapping rule associated with a particular discourse notion. The table in (9) expresses that topic and focus are basic notions in information structure that can be enriched to yield a contrastive interpretation. In other words, a contrastive topic is an aboutness topic interpreted contrastively. Similarly, a contrastive focus is a new information focus that receives a contrastive interpretation. We are not the first to make a suggestion along these lines; related ideas can be found in Vallduví and Vilkuna (1998), Molnár (2002), McCoy (2003), and Giusti (2006). The strongest evidence for the typology in (9) comes from cross-cutting generalizations, to be explored throughout this book, that jointly motivate a three-way typology. If contrast, topic and focus are privative features, we expect to find rules that refer to [topic] and therefore generalize over aboutness topics and contrastive topics, rules that refer to [focus] and therefore generalize over new information focus and contrastive focus, and rules that refer to [contrast] and therefore generalize over contrastive topic and contrastive focus. We do not expect to find rules that generalize over aboutness topics and new information foci, over contrastive topics and new information foci, or over aboutness topics and contrastive foci. None of these pairs share a feature.

6

Ad Neeleman and Reiko Vermeulen

The import of these predictions of course depends on what we mean by notions like “focus”, “topic” and “contrast”. We elaborate on this in the following two subsections.

2.1.

Focus and contrastive focus

Let us assume that a proposition P answers a wh-question Q. The focus of P is often taken to be that part of P that corresponds to the wh-expression in Q. By this criterion, The Selfish Gene is the focus in (10), while John is the focus in (11) (here and below, small capitals are used to indicate focus). (10)

A: What did John read? B: He read THE SELFISH GENE.

(11)

A: Who read The Selfish Gene? B: JOHN read The Selfish Gene.

Focus is clearly a grammatical notion as it affects linguistic phenomena like stress. In English and many other languages, a focused constituent receives the main stress of the sentence (see Selkirk 1984, 1995, among many others). Thus, the object carries sentence stress in (10), while the subject does so in (11): it is infelicitous to deviate from these stress patterns. Other languages have different or additional means of marking focus. In Thompson River Salish, for example, focused constituents are licensed at the edge of an intonational phrase, but do not need to carry stress (see Koch 2008). In Gùrùntùm, focus is marked by a designated particle a, which precedes focused constituents (see Hartmann and Zimmermann 2006). It is one thing to motivate the existence of focus, but it is another to explain why the relevant part of an answer to a wh-question must be marked as such. A widely accepted solution to this puzzle is proposed by Rooth (1985, 1992). The starting point of Rooth’s approach is that the semantics of questions is the set of potential answers, both true and false (see Hamblin 1973). So, the meaning of the question asked by A in (10) can informally be represented as the set in (12), while the meaning of the question in (11) corresponds to the set in (13). (12)

{[John read The Selfish Gene], [John read The Blind Watchmaker], [John read The Ancestor’s Tale], [John read The Extended Phenotype], …}

Chapter 1 The Syntactic Expression of Information Structure (13)

7

{[John read The Selfish Gene], [Johanna read The Selfish Gene], [Gerald read The Selfish Gene], [Jennifer read The Selfish Gene], …}

Rooth’s proposal, widely referred to as “alternative semantics”, is that focus itself evokes a set of alternative propositions. More specifically, it generates a set of alternative propositions that differ only in the focused position, and share all other material. This set of alternatives is called the focus value of a sentence, in contradistinction to its ordinary value (the proposition expressed by the sentence). Thus, the ordinary and focus values of the answers in (10) and (11) are as in (14) and (15), respectively: (14)

a. Ordinary value: [John read The Selfish Gene]. b. Focus value: {[John read The Selfish Gene], [John read The Blind Watchmaker], [John read The Ancestor’s Tale], [John read The Extended Phenotype], …}

(15)

a. Ordinary value: [John read The Selfish Gene]. b. Focus value: {[John read The Selfish Gene], [Johanna read The Selfish Gene], [Gerald read The Selfish Gene], [Jennifer read The Selfish Gene], …}

A congruent question-answer pair is one in which the set of potential answers that constitutes the meaning of the question matches the focus value of the answer. More precisely, a match is achieved if the former is identical to, or a subset of, the latter. (Why the subset relation must be allowed will not be discussed here.) Therefore, the focus marking in the answer in (10) fits the question asked, as the meaning of the question is the set in (12) and this set is identical to the focus value of the answer in (14). However, the question in (10) does not permit focus marking of the subject (as in (11)), because the latter has the focus value in (15), and (12) is not a subset of (15b). The above representations are widely used (see, for example, Büring 1997). We will use a slightly different notation, which explicitly represents the focus, as well as the set of alternatives to the focus found in the alternative propositions. Thus, we treat the information in (14) and (15) as triplets consisting of a function (corresponding to the background in (8)), the focus and a set of alternatives to the focus. The ordinary value is generated by applying the function to the focus, while the focus value is generated by applying it to members of the set:

8

Ad Neeleman and Reiko Vermeulen

(16)

a. b.

Since sentences must normally have a focus, the semantics of almost every sentence will contain this component of meaning. This is true also of allfocus sentences. Take an exchange like What happened? – John left. The answer can be represented as below, using an identity function as the first member of the triplet: (17)

It is important to note that the use of focus is not limited to answering a wh-question. There are other contexts in which focus is employed. For example, it is commonly assumed that in B’s contribution to the dialogue in (18) the subject is focused (it attracts main stress). (18)

A: John bought a Jaguar. B: (It’s a trend.) Even MARY bought an expensive car.

Of course, if Mary is focused, one must address the question of what principle of discourse makes the exchange in (18) congruent. One possibility, suggested by Krifka (2008) for comparable cases, is that the exchange in (18) includes a hidden question answered by the relevant part of B’s reply, something like: If it’s a trend, who (else) bought an expensive car? Although the presence of hidden questions is widely assumed, it is not readily testable, because there is no explicit theory about the process of contextual enrichment that introduces such questions, at least in examples like (18). We will therefore leave the matter open for the time being. We now turn to the notion of contrast. There seems to be an interpretive difference between examples like (19) and (20) on the one hand and examples like (10) and (11) on the other. In the former the focused constituent stands in opposition to an alternative explicitly mentioned in the discourse, while in the latter there is no explicit alternative and no sense of contrast. (Here and below we use boldface to mark constituents that we take to have contrastive reading.) (19)

A: What did John read, The Selfish Gene or The Extended Phenotype? B: He read THE SELFISH GENE.

(20)

A: John read The Extended Phenotype. B: (No, you’re wrong.) He read THE SELFISH GENE.

Chapter 1 The Syntactic Expression of Information Structure

9

However, there is nothing in the system described above that distinguishes a regular focus from a contrastive focus. It may well be that in certain cases the sense of contrast is only pragmatic (see also Krifka 2008). By general Gricean reasoning, the hearer infers that the answer to a wh-question he or she asked will be complete. It would therefore follow that alternative answers explicitly given in the context are taken to be false. So, in an example like (19), Gricean reasoning leads to the conclusion that John did not read The Extended Phenotype. Of course, this effect will be strongest when the set of alternative propositions is closed, as in the case at hand. However, a similar, but weaker effect can still be observed when the set is left open (as in “What did you read this summer? I read THE SELFISH GENE”). Although this could be the right analysis for (19), there are instances of contrast that have grammatical effects and that are therefore unlikely to be entirely pragmatic in nature. For example, correction contexts such as (20) allow movement of focused constituent, especially if the contrast is made explicit in the answer, as in (21). Wh-questions, however, do not normally provide a context compatible with focus movement in the answer, as (22) demonstrates (cf. É Kiss 1998). (21)

A: John read The Extended Phenotype. B: (No, you’re wrong.) THE SELFISH GENE he read. THE EXTENDED PHENOTYPE he only bought.

(22)

A: What did John read? B: #THE SELFISH GENE he read.

B’s answer is infelicitous in the context in (22) on a non-contrastive reading of the fronted constituent. (It is highly marked even when interpreted contrastively, because it presupposes a contrast that is not readily part of the common ground. Hence, B’s reply forces a non-trivial accommodation on the part of A, namely the assumption that there is a contrast between The Selfish Gene and some other reading material. B may have in mind what this reading material is, but this is not made accessible to A here. B’s answer is therefore likely to trigger a request for clarification, such as What do you mean? What did he not read? This effect can only be understood if the movement is linked to a contrastive reading.) The different behaviour of contrastive and regular focus is not limited to movement possibilities. Contrastive focus is often assumed to require a so-called A-accent in English (a plain high tone (H*), frequently followed by a default low tone; see Jackendoff 1972 and Pierrehumbert 1980),

10

Ad Neeleman and Reiko Vermeulen

whereas regular focus on objects is marked with nuclear stress. The contrast between an A-accent and a nuclear stress may not be obvious in all contexts, but there is clear evidence for the claim that contrast requires special prosodic marking. To begin with, Katz and Selkirk (2011) show that the phonetic prominence of contrastive focus is greater than that of discourse-new material in comparable syntactic positions (the latter of course includes regular foci) Moreover, there is some evidence for the phonetic relevance of contrast in the realm of second occurrence focus. A second occurrence focus is a contrastive focus that has already been introduced in the discourse. It is now widely recognized that it must bear some degree of prominence: Beaver et al. (2007) observe that it manifests itself with greater phonetic duration and intensity than material that is simply given. For example, wine is most prominent in the prosody assigned to the VP in B’s reply in (23), as indicated by the acute accent on wine. This is because a glass of wine, which is associated with the focus-sensitive particle only, is a contrastive focus (see chapter 7 for discussion). For similar examples, Beaver et al. show that this pattern carries over to the VP in C’s continuation, although the peak on wine is clearly significantly reduced in comparison to that in B’s utterance, presumably because it is now given (secondoccurrence focus is marked by dotted underlining here). (23)

A: What did John and Bill bring over to Mary? B: John only [VP brought A GLASS OF WÍNE over to Mary]. C: Oh, that’s funny. Bill only [VP brought A GLASS OF WÍNE over to Mary], too.

Although native speakers do not always perceive the greater phonetic prominence of the second occurrence focus, it is quite clear that in C’s reply in (23) it is at least possible for wine to bear a higher level of stress than Mary. The same effect is not found with a regular focus. A glass of wine in B’s reply below is intended to be a focus, but it is not intended to be contrastive (notice the absence of only). On this interpretation, the relative prosodic peak on wine cannot be maintained in C’s continuation, demonstrating that it is the feature [contrast] rather than [focus] that triggers the phonetic reflex in (23) (see also Selkirk 2008). (24)

A: What did John and Bill bring over to Mary? B: John brought A GLASS OF WÍNE over to Mary. C: Oh, that’s funny. Bill [brought a glass of wine over to Mary], too.

Chapter 1 The Syntactic Expression of Information Structure

11

Notice that a glass of wine in C’s utterance also corresponds to the whpart of A’s question, and by this criterion it must be classified as a focus. What (24) shows is that the destressing associated with givenness overrides the marking of regular focus. The special behaviour of contrastive focus is not a peculiarity of English, nor is it restricted to movement and prosody. In languages like Bole (West Chadic), the difference is expressed morphologically: there is a marker for contrastive focus that does not attach to regular focus (see Zimmermann 2008). Any theory of focus must account for these data. The most conservative approach is to leave the standard theory of focus intact, but add something to it that accounts for the distinct behaviour of contrastive focus. We propose that contrast, where it is linguistically encoded, is a quantifier. In general, quantifiers give information about the relationship between two sets in the universe of discourse. For example, the sentence most sheep eat grass expresses to what extent the set of sheep is contained in the set of grass eaters: it is asserted that most members of the set of sheep are also members of the set of grass eaters. In the same vein, contrast gives information about the relation between two sets. In an example like The Selfish Gene he read in (21), contrast expresses to what extent the set of (contextually relevant) books is contained in the set of things that John read. The sentence asserts that one member of the set of books is also a member of the set of things that John read. It also expresses that there is at least one other member of the set of books that is NOT contained in the set of things that John read. In the case at hand this other member is The Extended Phenotype. The positive statement and the notion that there are alternatives to The Selfish Gene derive from the normal interpretation of focus. The negative statement about an alternative, however, is part of the semantics of contrast. The negation is not just a pragmatic effect, as it is not cancellable. Consider the context in (25), which is organised around the implicit question “which books has John read?” We intend Dad’s reply to be understood as a contrastive focus. That is, it asserts that there is at least one other relevant book that John did not read. This interpretive effect is confirmed by the oddness of the continuation in (ii), which explicitly states that there is no such relevant book.1 1 Some speakers have indicated that they find it easier to pronounce the fronted constituent in Dad’s reply in (25) with a B-accent and to interpret it as a contrastive topic (see section 2.2). This is certainly a possibility in this context. However, the point that we are trying to make is related to the notion contrast:

12

Ad Neeleman and Reiko Vermeulen

(25)

(Mum and Dad know that John must read five books to prepare for the exam; they are discussing which books he has read so far.) Mum: John’s read The Selfish Gene. Dad: Yes, I know. THE SELFISH GENE he’s read. (i) But THE EXTENDED PHENOTYPE he hasn’t read. (ii) #In fact, he’s read all five books on the reading list.

Notice that since the negative statement is associated with the semantics of contrast rather than focus, we do not expect to find it with regular focus. Indeed, the continuation in (ii) is perfectly natural in the discourse below, where the object in Dad’s reply is a non-contrastive focus: (26)

(Mum and Dad know that John must read five books to prepare for the exam; they are discussing which books he has read so far.) Mum: John’s read The Selfish Gene. Dad: Yes, I know. He’s read THE SELFISH GENE. (i) But THE EXTENDED PHENOTYPE he hasn’t read. (ii) In fact, he’s read all five books on the reading list.

In sum, we propose that contrastive focus differs from regular focus in that it also encodes a negative statement. Thus, the interpretation of B’s initial answer in (21) differs from that in (10) (or (20), for that matter) in having a negative statement in addition to the normal focus semantics, as shown below: (27)

a. b. y [y  {The Blind Watchmaker, The Ancestor’s Tale, The Extended Phenotype, …} & ™[John read y]].

Having two semantic components as above is not particularly novel. This kind of analysis is standard for analyses of focus-sensitive particles like only. The semantics of a sentence like John read only The Selfish Gene is the implications of contrast are not cancellable. So it is orthogonal to at least some extent whether contrast is combined with focus or topic. Moreover, the possibility of interpreting the fronted constituent as a contrastive focus in this context can clearly be seen in languages like Japanese, which morphologically distinguishes contrastive foci and contrastive topics (the former bear regular case while the latter are marked by wa). There is no requirement to use wa here.

Chapter 1 The Syntactic Expression of Information Structure

13

generally analysed as consisting of two separate components (see Horn 1969; König 1991; Krifka 1999; among others). The first is essentially the semantics of the sentence without only (sometimes referred to as “the prejacent”), while the second is the quantificational statement “there is no alternative to The Selfish Gene such that John read that alternative”. In our notation, these components can be represented as below: (28)

a. b. ¬y, y  {The Blind Watchmaker, The Ancestor’s Tale, The Extended Phenotype, …}, [John read y]

One advantage of treating contrast as quantificational is that it implies that contrastive foci take scope, while regular foci do not. This can be tied in with the observation, already illustrated above, that contrastive foci may undergo A’-movement, which is known to mark scope. Notice that in view of the data in (29) and (30) it must be the element introducing the quantificational statement that licenses A’-movement. As is well-known, only can attach directly to the DP it modifies or be placed at a distance in an adverbial position. Movement is restricted to the case where only attaches directly to its associate, as in (29). One explanation for the ungrammaticality of the examples in (30)b and (30)b’ is that the fronted category does not contain the quantificational element that would license A’-movement. (29)

a. John read only THE SELFISH GENE. b. Only THE SELFISH GENE did John read. b’. %Only THE SELFISH GENE John read.

(30)

a. John only read THE SELFISH GENE. b. *THE SELFISH GENE did John only read. b.’ *THE SELFISH GENE John only read.

We will see that a very similar decompositional analysis can be given for contrastive topics.

14 2.2.

Ad Neeleman and Reiko Vermeulen

Topic and contrastive topic

We follow Reinhart (1981) in characterizing topics in terms of “aboutness”. Speakers generally have intuitions regarding what a given sentence is about. In fact, one could see the mere existence of expressions like “as for”, “about”, “regarding”, “concerning”, and so on, as evidence for the existence of aboutness. Yet, it is surprisingly difficult to pin down the exact content of the notion and how it is linguistically relevant. This is reflected in the variety of definitions of topic in the literature (compare Chafe 1976; Reinhart 1981; Givón 1983; Vallduví 1992; Lambrecht 1994). There is some consensus, however, that it is important to distinguish between the topic of a unit of discourse and the syntactic constituent used to introduce the referent that the sentence is about. This referent may then function as the topic of the subsequent discourse. We will refer to topics in this first sense as “discourse topic”, and to the second type of topic as “sentence topic” or simply “topic” when the distinction is clear. Since this book is concerned with the syntactic behaviour of constituents with a particular information-structural function, we will concentrate mainly on sentence topics, limiting the discussion of discourse topic to what is necessary to understand its opposition to sentence topic. Consider (31). Speakers generally have the intuition that Maxine is a sentence topic in the first sentence of this small monologue (here and below we use double underlining to mark sentence topics). The person that Maxine refers to continues to function as the topic of the subsequent discourse. The pronouns her and she that refer to this person are not sentence topics: they do not introduce the referent as the topic of discourse. Rather, they are discourse-anaphoric elements whose antecedent is the topic of discourse. It would be misleading to treat these pronouns as sentence topics just because they refer back to the discourse topic. After all, in other circumstances pronouns do not inherit the information-structural status of their antecedent either. For instance, a pronoun whose antecedent is a focus is not thereby itself a focus. Similarly, a pronoun whose antecedent is new is not thereby new itself. (31)

Well, Maxine was invited to a party by Claire on her first trip to New York. She was amazed by the strange crowd with their bell-bottom trousers and star-studded jackets.

Much of the confusion surrounding the notion of topic in the literature stems from complications involving discourse topics. A discourse always has an overarching topic, but in addition it may be divided into smaller

Chapter 1 The Syntactic Expression of Information Structure

15

units of discourse, each of which has their own discourse topic (see also Givón 1983). This structure is in principle recursive, so sub-topics of discourse can themselves have sub-topics. We can illustrate this by considering several contexts that may potentially precede the monologue in (31). The simplest context is (32)a. It is an explicit request to introduce Maxine as the topic of discourse, which is of course what happens in (31). Notice that Maxine in (32)a is not a topic. The sentence is not about Maxine. If it were, at least some (new) information about her would have to be provided. Moreover, in languages that have overt morphological marking of topics, such as Japanese and Korean, it is not marked as such. Therefore, even in the context of (32)a, Maxine in (31) is not simply discourse-anaphoric, but qualifies as a sentence topic. (32)

a. b. c. d.

Tell me about your friend Maxine. Tell me about one of your friends. Tell me about your friends’ experiences in New York. Do you know anything about parties in New York?

The context in (32)b is very similar. We have added it here to make it clear that a topic need not be old information (a point also made in Reinhart 1981). The contexts in (32c) and (32c) are more complex. The overarching topic about which information is requested in (32c) is more abstract and suggests that the subsequent discourse may be divided into smaller units that are each about one of the hearer’s friends. (31) is one such unit, with Maxine as its topic (a sentence topic and consequently a discourse topic). Maxine is therefore a sub-topic of the larger discourse.2 As for (32d), the overarching topic about which information is requested is parties in New York. Maxine is introduced as a sentence topic in (31), because her experiences are relevant to this overarching discourse topic.3 It is clear from these two examples that the relation between the overarching topic and the sub-topics can be quite diverse and complicated. 2 This “narrowing down” of the referent of the topic allows Maxine in (31) to be interpreted contrastively (see below for a discussion of contrastive topics). However, this is not obligatory in this context, and accordingly it is possible but not necessary to use the B-accent that contrastive topics. 3 The context in (32d) cannot be used as a test for topichood however, as it allows, but does not force, Maxine in (31) to be a sentence topic. This is apparent from languages like Japanese, where Maxine is only optionally marked with wa in the context of (32d).

16

Ad Neeleman and Reiko Vermeulen

Despite this complication, what remains constant across the contexts in (32) is that Maxine is a sentence topic in (31). Since we are mainly interested in sentence topics, we will put aside issues concerning the theory of discourse topics, in particular the division of topics into sub-topics in larger units of discourse. There are several grammatical effects associated with sentence topics. Our starting point is the observation that, all else being equal, unstressed pronouns have a strong tendency to refer to the topic of discourse (if they are interpreted through coreference as opposed to variable binding). There are various proposals that capture this fact as part of a larger theory of anaphora resolution, in particular Givón (1983) and Ariel (1987, 1990). An illustrative example, adapted from Reinhart (1995 : 80), is given in (33). The preceding context is about Max, who is therefore the topic of discourse. As a consequence, the unstressed pronoun he refers to Max, as opposed to the epithet the guy, which preferably refers to Felix.4 (33)

Max was on his way home from school, worrying about how things were going to turn out. After a while he ran into Felix, and … a. he proposed they go to a pub. (he refers to Max, not Felix) b. the guy proposed they go to a pub. (the guy refers to Felix, not Max)

In view of the above effects, we can use anaphora resolution to identify structures that mark a constituent as a sentence topic. Such a constituent is predicted to function as a preferred antecedent in anaphora resolution. The clearest case is the as for construction. This construction marks the DP-complement of as for as a topic, as shown by the fact that this DP is the preferred antecedent for following pronouns in examples like (34) and (35). (34)

As for Maxine, Claire invited her to a party in New York. She was amazed by the strange crowd with their bell-bottom trousers and star-studded jackets (#and wanted to share this experience with Maxine).

(35)

As for Claire, she invited Maxine to a party in New York. She was amazed by the strange crowd with their bell-bottom trousers and star-studded jackets (and wanted to share this experience with Maxine).

4 Judgments shift if the pronoun is stressed, in which case there is a preference for he to be construed as referring to Felix. This falls out from Accessibility Theory, as stressed pronouns are associated with less accessible antecedents.

Chapter 1 The Syntactic Expression of Information Structure

17

In the first example, the unstressed pronoun she is most naturally interpreted as Maxine. The preference for Maxine is further demonstrated by the garden path effect that the continuation between brackets creates. This effect is not surprising, because the content of the second sentence more readily applies to the person invited to a party than to the one inviting. However, in (35) the second occurrence of she preferably refers to Claire, despite Claire being the one inviting. Thus, there must be an overruling grammatical effect of Claire appearing as the complement of as for. This follows if as for marks its complement as the topic. A similar effect has been observed with subjects of passives (see Givón 1983 and Reinhart 1981, 1995). She in (36), when unstressed, has a strong tendency to refer to Maxine. This seems to be related to the fact that the initial sentence in (36) is passive. In (37), where the initial sentence is active, she can be associated with either subject or object (although there is a weak preference for a subject-oriented reading). It is much harder to construe she as referring to Claire in (36). (36)

Maxine was invited by Claire to a party in New York. She was amazed by the strange crowd with their bell-bottom trousers and star-studded jackets (#and she wanted to share this experience with Maxine).

(37)

Claire invited Maxine to a party in New York. She was amazed by the strange crowd with their bell-bottom trousers and star-studded jackets (and she wanted to share this experience with Maxine).

The traditional interpretation of these data is that the subject of a passive construction must, or has a very strong tendency to be, a topic. This description of the data is not optimal, given that the subject of a passive can be a focus, a wh-expression or a negative quantifier, elements that are incompatible with topichood. Perhaps, a better description would be to say that the demoted subject in the by-phrase cannot introduce, or refer to, a discourse topic, while there is a general preference to realise topics as subjects. This description receives some support from the contrast between the following examples. Coreference between John and him in (38)b is decidedly odd. (38)

a. As for John, he was seen by Mary. b. #As for John, Mary was seen by him.

If a passive construction is used at the beginning of a unit of discourse (such as a narrative like (36)), the pressure to establish a topic will force

18

Ad Neeleman and Reiko Vermeulen

the derived subject to function as such. This explains why she in (36) preferably refers to Maxine. There is a range of other grammatical effects of sentence topics as opposed to old information or constituents that refer back to the topic of discourse. Vallduví (1992), for instance, demonstrates that in Catalan the distinction is formerly marked by the direction of dislocation: sentence topics must be left-dislocated, while elements that refer back to the topic of discourse must be right-dislocated together with other backgrounded material. Frascarelli and Hinterhölzl (2007) show that in Italian and German sentence topics bear a different intonation from discourse-given items. Moreover, the former cannot be right-dislocated, while the latter can (see also Lambrecht 1994). Choi (1999) shows that in Korean the socalled topic marker nun typically marks sentence topics and not elements that refer back to the topic of discourse, which pattern with other discourse-given material. We take the above to be sufficient to establish the linguistic relevance of topichood. We would now like to sketch how the interpretive effects of topic might be captured. Like foci, we take sentences containing topics to be associated with a set of alternatives. However, in the case of foci the function introduced by the lambda operator generates propositions, whereas in the case of topics it generates utterances. These utterances vary only in the value for the position occupied by the topic. For example, the passive structure in (36) can be represented as a triplet, as in the case of focus. The triplet consists of a function (corresponding to the comment in (8)), the topic and a set of alternatives to the topic. The difference with the representation of focus is that the function contains an assertion operator, which means that its application derives an utterance, rather than a plain proposition (see also Tomioka 2010). Applying the function to the topic generates an assertion whose propositional content is the ordinary value of the sentence. Applying it to members of the set of alternatives generates a set of utterances that one might call the topic value of the sentence. The triplet thus represents the intuition that when a speaker utters the sentence in (36), he or she performs the following speech acts: (i) Consider Maxine (out of a set of possible topics); (ii) I assert that Maxine was invited by Claire to a party in New York (compare Jacobs 1984; Kuroda 1992; Krifka 2008; Portner 2007; Brunetti 2009; Endriss 2009; and Ebert and Hinterwimmer 2010). (39)

Chapter 1 The Syntactic Expression of Information Structure

19

The above representation of topichood captures the contrast between (40a) and (40b), which was explored first by Strawson (1964) and subsequently discussed by Reinhart (1981) (we have adjusted the examples somewhat). In the first sentence, the King of France is a sentence topic, which implies that this utterance consists of the following two speech acts: (i) Consider the King of France; (ii) I assert that he visited the exhibition yesterday. However, considering that there is no King of France, the first speech act cannot be performed: it presupposes that there exists such an individual. The utterance in (40b) can be paraphrased as follows: (i) Consider the exhibition; (ii) I assert that it was visited by the King of France yesterday. In this case, the first speech act will not lead to a presupposition failure, assuming that there is indeed an identifiable exhibition in the domain of discourse. Rather, the example is judged to be false. If we check the set of visitors to the exhibition, we will not find the King of France, as there is no such individual.5 (40)

a. As for the King of France, he visited the exhibition yesterday. b. As for the exhibition, it was visited by the King of France yesterday.

We are aware that topicality is not the only factor that gives rise to a difference in judgements of the sort in (40). However, the claim in Lasersohn (1993) and von Fintel (2004) that topicality is irrelevant seems to be too strong. Abrusán and SzendrĘi (2011) call the referent used to evaluate the truth value of a proposition a “pivot”. Based on experimental data, they show that pivots do not have to be topics, but that the default choice of a pivot is a topic. Thus, even though the original form of the argument in Strawson (1964) is too simplistic, a more sophisticated version of Strawson’s argument can be developed. Since topicality is relevant for the choice of a pivot and since the choice of a pivot is crucial in accounting for a contrast like (40), it follows that the notion of topic must be accessible to sentence grammar (including truth-value judgements). 5 The example at hand is intended to be a claim about the actual world. Of course, when discussing a fictional world, topics may refer to entities in that world that have no correspondent in the actual world: As for Emma Bovary, she is a beautiful woman. A truth value can be assigned to this sentence within the context of the fictional world described in the relevant novel. “Emma Bovary is a beautiful woman” is true if an only if, within the world depicted in the novel Madame Bovary by Gustave Flaubert, Emma Bovary is a beautiful woman. (See Reicher (2010) for further discussion on this topic.)

20

Ad Neeleman and Reiko Vermeulen

In the previous section we have argued that a focus can have an additional contrastive interpretation. The same is true for topic, as (41) demonstrates. In reply to the question by A, Maxine in B’s reply is interpreted contrastively. B is unable or unwilling to make about Bill the comment that he/she makes about Maxine. Notice that there is no sense of contrast in the earlier example in (31). (41)

A: Tell me about Bill. Was he invited to a party when he went to New York? B: Well, I don’t know about Bill, but Maxine was invited to a party on her first trip to New York by Claire.

We argued that contrast is quantificational and therefore licenses A’movement. Thus, contrastive focus can undergo A’-movement, but as we saw, new information focus must remain in situ. The same is true for topics. If a topic undergoes A’-movement, it must be interpreted contrastively. Consider B’s reply in (42), where the topic the female popstars is fronted and therefore requires a contrastive interpretation. Such an interpretation is felicitous because the female popstars denotes a subset of the topic introduced in the context question (the popstars) and therefore stands in opposition to the complement set (the male popstars). (42)

A: What about the popstars? Who showed them around? B: Well, the female popstars, Bill gave a tour.

The same utterance is awkward in contexts that do not require or suggest a similar contrastive interpretation. To the extent that B’s answer in (43) is acceptable, it requires accommodation of contrast, or it would trigger a further request for clarification such as What do you mean?, Was there a problem with the male popstars? (43)

A: Tell me about the female popstars. B: #Well, the female popstars, Bill gave a tour.

Like contrastive focus, a contrastive topic must be prosodically marked in English. Contrastive topics typically carry what Jackendoff (1972) calls a B-accent, maximally realised as L+H*, followed by a default low tone and a high boundary tone (L H%). Regular topics do not require any such marking. Finally, like the opposition between contrastive and non-contrastive focus, the opposition between contrastive and non-contrastive topics can

Chapter 1 The Syntactic Expression of Information Structure

21

be marked morphologically, as has been observed for Tsez by Polinsky and Potsdam (2001). The effects of a contrastive interpretation for topics are different from those for focus. In the case of focus, the alternatives are propositions, (at least) one of which is claimed to be false. In the case of topics, the alternatives are utterances, and the interpretive effect of contrast is therefore that the speaker is unwilling to utter (at least) one alternative. This could be for a variety of reasons. It could be that the speaker knows that the propositional content of the alternative is false, but is not a position to share this information or does not want to be held responsible for it. It is also possible that the speaker is uncertain of the truth value of the proposition expressed by the alternative (as is the case in (41)). This difference in the interpretation of contrast is unsurprising. It follows naturally from the fact that focus is a notion relevant to propositions, whereas topic is notion relevant to utterances (Tomioka 2010). Consequently, contrastive foci deny an alternative proposition, whereas contrastive topics indicate that the speaker is unwilling or unable to make an alternative utterance. We therefore propose the following representation for the interpretation of B’s answer in (41).6 (44)

a. b. y [y  {Bill, Susan, …} & Ox ™ASSERT [x was invited by Claire to a party in New York](y)].

Notice that the assertion regarding the female popstars in (42) indeed does not imply that Bill did not show the male popstars around, as is apparent from the fact that all the continuations below are felicitous. (45)

a. I don’t know about the male popstars. Maybe Bill gave them a tour, too. b. The male popstars, Gary took out for lunch. c. The male popstars didn’t turn up.

The felicitousness of (42) in the contexts in (45) demonstrates that the use of contrastive topic is licensed in a variety of situations, ranging from 6 The ASSERT operator can be read as “I assert that …”. In other words, it derives a proposition with the speaker as the subject. The fact that negation is usually taken to be a propositional operator is therefore compatible with its being prefixed to the ASSERT operator.

22

Ad Neeleman and Reiko Vermeulen

uncertainty of the speaker about the truth of an alternative statement (as in (45a)) to explicit knowledge about the relevant alternative (as in (45b); see Hara and Van Rooij 2007). The latter case may include cancellation of the presupposition of the question (as in (45c)). There is a link between Büring’s (1997, 2003) work on contrastive topics and the proposal made above. Büring defines contrastive topics as elements whose meaning requires a set of sets of alternative propositions, whereas foci simply require a set of alternative propositions, as standardly assumed on the Alternative Semantics approach (see Rooth 1985, 1992). On our proposal a sentence containing both a contrastive topic and a focus will indeed involve representation in which there is a set of alternative utterances, each of which contains a set of alternative propositions. This is very similar to the D-tree representation in Büring 2003, but notice that according to our analysis the order of embedding of the sets associated with topic and focus is a direct consequence of topic being an utterance-level notion and focus being propositional. There is also a difference, however. Büring’s view of topics implies that no sentence that contains a contrastive topic can lack a focus. As Büring himself points out, this may be problematic in view of the following example from O’Connor and Arnold (1973). Note that the interpretation of B’s answer is as expected if it is a contrastive topic: it implies that B is unwilling or unable to assert the same for an alternative to Bill (namely Jack). However, the sentence does not appear to contain an obvious focus.7 (46)

A: Can Jack and Bill come to tea? B: Bill can.

On our view, nothing in the meaning of a contrastive topic hinges on there being a focus in the same sentence. Therefore, if there are sentences without a focus, we could allow contrastive topics in the absence of focus. Even though topic is an utterance-level notion, our analysis assumes that it is encoded linguistically (see also Krifka 2001). This means that, as in the case of contrastive focus, the negative statement implied by a contrastive topic must be semantically encoded as well, which predicts that it is not cancellable. This point was demonstrated for Japanese by Oshima (2008), but his conclusion carries over to English (see also Constant 2006). The discourse in (47) is about John and Bill. A’s initial statement is about

7 Notice that the absence of stress on can in B’s utterance suggests that this context does not require polarity focus in B’s answer.

Chapter 1 The Syntactic Expression of Information Structure

23

John. The how about question guarantees that Bill in B’s reply will be a contrastive topic (that stands in opposition to John). The reply in (i) is infelicitous, as it implies that there is an alternative to Bill about whom the speaker could not make the same assertion. However, this is an odd thing to imply, given that the only alternative to Bill is John, and the speaker has met him. The reply in (ii) works better, because it implies that B could not assert that he didn’t meet John, which is consistent with the context. (47)

John and Bill (and nobody else) came to town. A: So, you met John. How about Bill? B: (i) #Bill, I met. (ii) Bill, I didn’t meet.

The different semantics proposed for contrastive topic and contrastive focus makes a further prediction concerning their respective effects in contexts of correction. A contrastive focus is used for correcting propositional content (and it therefore denies that the original statement is true). Thus, a Bentley in A’s continuation in (48) must be a contrastive focus, a prediction that is confirmed by the fact that it can carry an A-accent, but not a B-accent. (48)

A: John bought a car B: I know. It was a Rolls Royce A: No, you’re wrong. It was a BENTLEY. / #It was a Bentley.

On the other hand, a contrastive topic can be used for correcting an invalid conclusion. In such cases, it is not clear that the conclusion is correct given the evidence available to the participants in the discourse. The use of a contrastive topic then indicates that the assertion is not warranted (but its propositional content could still be true). An example is given below, where the female popstars in C’s statement must be a contrastive topic and must therefore be marked with a B-accent: (49)

B and C went to a music festival. They saw only the female acts: Madonna, Kate Bush and Joni Mitchell. Surprisingly, these performers were all wearing kaftans. They missed the other acts (U2, Prince and Bob Dylan), so they have no idea what the male stars were wearing. A: So, how was the festival? B: It was great. All the popstars were wearing kaftans! C: You can’t say that. The female popstars were wearing kaftans. / #THE FEMALE POPSTARS were wearing kaftans. (But we didn’t see the male popstars.)

24

Ad Neeleman and Reiko Vermeulen

The contrast between an A-accent and a B-accent in examples like the one above may be difficult to hear. However, there is strong confirmation in Japanese for the claim that the context in (49) indeed requires a contrastive topic. In Japanese, contrastive topics are morphologically marked with wa, while contrastive foci bear regular case (at least in the absence of focus-sensitive particles). The point is demonstrated by the following example. (50)

sore-wa i-e-nai yo. that-WA say-can-not SFP. kahutan-o kite-ita kaftan-ACC wearing-PAST minakatta desyoo.) see-not-PAST SFP

[zyosee-no kasyu]-wa/#ga female-GEN popstar-WA/NOM (kedo [dansee-no kasyu]-wa but male-GEN popstar-WA

So far, we have treated the B-accent that identifies contrastive topics as an autonomous tune. However, it has been proposed by Steedman (2000) and Büring (2003), among others, that the B-accent is in fact a composite of an A-accent plus a high boundary tone (which presumably marks the end of an intonational phrase that contains the contrastive topic; cf. Féry 2007). One piece of evidence supporting this claim comes from examples in which the contrastive topic is a larger constituent ending in given material. In such cases, the B-accent is realised as a high pitch accent on the contrastive material, indicated by H*, and a separate high boundary tone at the end of the phrase, H% (Büring 2003 : 537): (51)

A: Where will the guests at Ivan and Theona’s wedding be seated? (L+)H* (L+)H*LH% H*L-L% B: [Friends and relatives of the couple] will sit AT THE TABLE. (Reporters have to sit in the back.)

One interpretation of the above observation is to analyse contrastive topic as a topic containing a focus (Krifka 2006; Steedman 2000; Büring 2003; Tomioka 2010). On our definition of focus, this cannot be true, as focus is associated with alternative propositions, while topics are associated with alternative utterances. Note however that the notion of focus as used by some of these authors is rather different from ours. For Krifka (2006) and Tomioka (2010), focus merely signals the relevance of alternatives for interpretation, and the presence of focus inside a topic would therefore signal the existence of alternative aboutness topics.

Chapter 1 The Syntactic Expression of Information Structure

25

This comes close to our own interpretation of data like (51). As we have argued in the previous section, the A-accent marks the presence of the feature [contrast]. This feature does not by itself give rise to alternatives, but it can only be interpreted in the context of alternatives, which are made available by the features [topic] and [focus]. We can therefore understand that a contrastive topic, being composed of the features [topic] and [contrast], would have an A-accent as part of its tune.8 There is considerable similarity in the overall characterisation of contrastive topics on the view advanced by Krifka and others and the one put forward here. However, a difference emerges when one considers the total set of discourse notions that are under discussion here. On our view there is a four-way distinction, as in the table in (52), whereas the system proposed by Krifka and others gives rise to the three-way distinction given in (53). The difference stems from our system being built on three unitary features, while the system in (53) is built on only two.

(52)

Contrast

Topic

Focus

aboutness topic [topic]

new information focus [focus]

contrastive topic [topic, contrast]

contrastive focus [focus, contrast]

Topic

(53)

aboutness topic [topic] Focus

contrastive topic [topic, focus]

new information / contrastive focus [focus]

The main motivation for treating [contrast] as a grammatical feature is that it allows us to capture different behaviour of new information focus

8 The high boundary tone should presumably be attributed to the feature [topic]. This would predict that non-contrastive topics can also be followed by a high boundary tone, which appears to be correct for cases like “Speaking of John, he was going to Italy”. Of course, in the case of non-contrastive topics the high boundary tone is optional. On the other hand, the high boundary tone is obligatory with contrastive topics. We speculate that this has a functional explanation. Without the high boundary tone, the pronunciation of contrastive topic would be identical to that of contrastive focus.

26

Ad Neeleman and Reiko Vermeulen

and contrastive focus, for example in relation to A’-movement. As we have already seen above, contrastive foci and contrastive topics can undergo such movement in English, but new information foci and aboutness topics cannot. On the typology in (53), it is hard to capture this generalization, as there is no feature that is shared by the elements that can move and that is absent in the elements that cannot.

3.

Case study: Dutch A’-scrambling

The following section provides empirical motivation from Dutch for the distinct features [contrast], [topic] and [focus], as in the typology in (52). In particular, we will argue that there is a movement known as A’-scrambling that requires the presence of [contrast] and restrictions on this operation can only be explained with additional reference to [topic] and [focus]. There is general agreement that, in Germanic and beyond, there are two types of scrambling. A-scrambling feeds and bleeds binding and secondary predication, does not give rise to weak crossover effects, is clausebounded, and does not give rise to scope-reconstruction. We cannot illustrate all these properties here, but for relevant discussion, see Vanden Wyngaerd (1989), Mahajan (1990), Zwart (1993), Neeleman (1994), and Neeleman and Van de Koot (2008). In contrast, A’-scrambling does not affect binding or secondary predication, gives rise to weak crossover effects, is not clause-bounded, and reconstructs (obligatorily) for scope. Again, we will not demonstrate these properties here, but refer the reader to Neeleman (1994), Jacobs (1997), Haider and Rosengren (1998), and Neeleman and Van de Koot (2008) for discussion. In Dutch, the two types of scrambling can be easily told apart, because in this language only A’-scrambling can alter the basic order of arguments (subject – indirect object – direct object). A-scrambling is restricted to the reordering of arguments and adjuncts (see Zwart 1993 and references cited there).9 The two types of scrambling are also associated with different interpretive effects. A-scrambling operations typically mark the scrambled DP as discourse-anaphoric (that is, the DP refers back to an entity introduced earlier in the discourse); see Reinhart (1995) and Neele-

9 Some speakers of Dutch marginally allow A-scrambling of a direct object across an indirect object, a possibility more generally available in German. The judgments reported here are from speakers who reject such scrambling.

Chapter 1 The Syntactic Expression of Information Structure

27

man and Reinhart (1998), among others, for discussion. Abstracting away from pied-piping, A’-fronting operations typically require the moved DP to be interpreted as contrastive (either as a contrastive focus or a contrastive topic; see Neeleman 1994 and Frey 2001). We will not demonstrate the point here, but the data showing the necessity of a contrastive interpretation for movement are very similar to those given for English in (21) and (22). What distinguishes the Dutch data from English is that A’-scrambling can target a variety of positions. Irrespective of whether the moving phrase is a contrastive topic or a contrastive focus, it can land in a position between the complementizer and the subject, as in (54), in a position between the subject and the indirect object, as in (55), or in the first position in main clauses, as in (56). Further landing sites are available in structures containing adverbs, as these are (usually) freely ordered with respect to moved topics and foci. (54)

a. Ik geloof dat [DIT BOEK]1 Jan Marie t1 gegeven I believe that this book John Mary given heeft. has ‘I believe that John has given this book to Mary.’ b. Ik geloof dat [zo’n boek]1 alleen JAN Marie t1 I believe that such-a book only John Mary gegeven heeft. given has ‘I believe that only John has given such a book to Mary.’

(55)

a. Ik geloof dat Jan [DIT BOEK]1 Marie t1 I believe that John this book Mary heeft. has ‘I believe that John has given Mary this book.’

gegeven given

b. Ik geloof dat Jan [zo’n boek]1 alleen MARIE I believe that John such-a book only Mary gegeven heeft. given has ‘I believe that John has given such a book only to Mary.’ (56)

a. [DIT BOEK]1 zou Jan Marie t1 geven. this book would John Mary give ‘John would give Mary this book.’

t1

28

Ad Neeleman and Reiko Vermeulen b. [Zo’n boek]1 zou alleen JAN Marie such-a book would only John Mary ‘Only John would give Mary such a book.’

t1 geven. give

One might conjecture that A’-scrambling is licensed by a mapping rule that licenses a contrastive interpretation for a moved constituent. A drawback to this suggestion is that constituents that remain in situ can also be interpreted contrastively, so it is difficult to argue that this interpretation arises as a result of movement. We propose an alternative account based on the idea that A’-movement of a contrastive topic or focus determines its scope, much like A’movement of other quantificational elements. What allows contrastive items to take scope is the negative operator in statements like (27)b and (44)b. The movement marks what material is included in the scope of this operator, which we call the domain of contrast (DoC). The proposal can be formalized as below (where the M-diacritic encodes a movement dependency, see Neeleman and Van de Koot 2002, 2010). (57)

DoC Marking In (58), N2 is interpreted as the domain of contrast for XP.

(58)

N1 [M # ] XP[contrast]

N2 [M ]

If a contrastive topic or focus remains in situ, the domain of contrast is not marked. In interpreting the sentence, the hearer must therefore construe an appropriate domain of contrast based on contextual clues. This can be the sister of the contrastive category, but it can also contain more material and as a result be discontinuous. In other words, the DoC for XP in (59a) can be its sister or YP minus XP itself. A’-movement of the contrastive category has a disambiguating effect, as it creates a match between the syntactic representation and the information structure associated with the sentence, as shown in (59b).

(59)

a.

YP

b. XP[contrast]

(Doc)

XPDoc

XP[contrast] (Doc)

tXP

Chapter 1 The Syntactic Expression of Information Structure

29

The proposal, then, is that the examples in (54a) and (55a) differ in that the DoC for the moved focus includes the subject in the former example, but not in the latter. A technical question arises at this point. Given that a DoC is always in the scope of negation, it must minimally be of the type that can be negated. If so, there must be a procedure of existential closure that transforms an incomplete DoC into an expression of the right type (compare Schwarzschild 1999). So, the DoC for (54a) is as in the underlined part of (60b), while the one for (55a) is as given in (61b) (for reasons of clarity we only represent the embedded clauses in the examples). (60)

a. < Ox[John has given x to Mary], this book, {flowers, chocolate, …}> b. y [y  {flowers, chocolate, …} & ™[John has given y to Mary]].

(61)

a. < Ox[John has given x to Mary], this book, {flowers, chocolate, …}> b. y [y  {flowers, chocolate, …} & z ™[z has given y to Mary]].

The interpretation of the existentially bound variable in (61) is then provided by the context (it must be as specific as the context allows). So far there appears to be no difference in the behaviour of contrastive topic and contrastive focus in Dutch: both can undergo A’-scrambling. This may be sufficient to motivate the existence of [contrast]. However, the theory generates several additional predictions regarding the order in which contrastive topics and foci can surface. We will show that these predictions are correct. If so, it supports the existence of the notions [topic] and [focus] and the Dutch data therefore confirm the three-way typology of information-structural functions in (52). The additional predictions come about through the interaction of two assumptions already introduced above. The first is that A’-scrambling is a scope-marking operation. As such, it gives rise to a scope freezing effect: material c-commanded by the moved category cannot outscope that category. Our second assumption is that topics have to be interpreted externally to foci. This is because topic is an utterance-level notion, while focus is a notion relevant to propositions. As a matter of course, propositions are contained inside utterances, rather than the other way around. Thus, in information structure (62a) is well-formed, while (62b) is not. (For relevant discussion, see Prince 1981; Reinhart 1981, 1995, 2006; Vallduví 1992; Lambrecht 1994; and Hajiþová, Partee, and Sgall 1998.) (62)

a. [utterance topic [proposition FOCUS … ]] b. #[proposition FOCUS [utterance topic … ]]

30

Ad Neeleman and Reiko Vermeulen

The predictions that follow from these two assumptions all pertain to sentences that contain both a contrastive topic and a contrastive focus. To begin with, if no A’-scrambling takes place, no domain of contrast is overtly marked, which implies that no instructions are given concerning the mapping between syntax and information structure. Consequently, the ordering between contrastive topic and contrastive focus is free (see (63a,b)). However, if a contrastive focus scrambles across a contrastive topic, the scope marking nature of the movement implies that the topic must be interpreted in the scope of the negative operator associated with the focus. As this negative operator is attached to the proposition, the topic must be part of the proposition as well, but this gives rise to the ill-formed structure in (62b). If a contrastive topic moves across a contrastive focus, however, the result is well formed. This is because the negative operator in this case attaches to an utterance, and an utterance can of course contain a focus (see (63c,d)). (63)

a. b. c. d.

[ … CT … CF … ] [ … CF … CT … ] #[CF [ … CT … tCF … ]] [CT [ … CF … tCT … ]]

The following data bear out the predictions in (63). By the criteria already introduced, ‘the beans’ in B’s answer is a focus: it corresponds to the whexpression in A’s question. Even though alternatives to ‘the beans’ are not mentioned explicitly, it is fairly easy to interpret it contrastively, as the food served at parties usually consists of more than beans. On a contrastive reading, ‘the beans’ carries an A-accent. Wim in B’s answer is a contrastive topic. A’s question is about Fred, but B is unable to say anything about Fred, instead offering information about Wim. In line with this, Wim is (or at least can be) pronounced with a B-accent.10 What (64) shows, then, is that a contrastive focus can surface in a base-generated position to

10 Judgments in this section are based on pronunciation of the examples such that the constituent marked as focus carries a plain high tone, and the constituent marked as topic carries a tune consisting of a high tone, a low tone and a high boundary tone (this intonation of Dutch topics is in line with the observations reported in Van Hoof 2003). As far as we can judge these matters, this pronunciation is very similar to what is found in English.

Chapter 1 The Syntactic Expression of Information Structure

31

the right of a contrastive topic (see (63a)), but cannot move across the contrastive topic (see (63c)).11 (64)

A: Hoe zit het met Fred? Wat heeft hij dit jaar op het feest gegeten? ‘What about Fred? What did he eat at this year’s party?’ B: Nou, dat weet ik niet, maar ik geloof ‘Well, I don’t know, but I believe’ (i) dat Wim van DE BONEN meer gegeten heeft dan that Bill from the beans more eaten has than vorig jaar. last year ‘that Bill has eaten more from the beans than last year.’ (ii) #dat [van DE BONEN]1 Wim t1 meer gegeten heeft dan that from the beans Bill more eaten has than vorig jaar. last year

There is a further test that can be used to corroborate the classification of contrastive topics and foci: negative quantifiers can function as foci, but not as topics (for obvious interpretive reasons: one cannot say something about nothing). This is corroborated by the fact that negative quantifiers cannot appear in the English as for construction, which marks topics:12

11 As was pointed out to us by Michael Wagner (p.c.), the context in (64) and comparable ones below do not force the interpretations indicated, but merely favour them. All the contexts we use are based on implicit multiple wh-questions. In (64), this question is “Who ate what at this year’s party?”, while in (67) below, it is “What was eaten by whom at this year’s party?” Answers to multiple wh-questions tend to be constructed in such a way as to line up the contrastive topic with the fronted wh-phrase and the focus with the in situ WHphrase. Thus, “Who ate what?” is most commonly answered by something like “John ate the BEANS, Mary ate the CHEESE”, etc. However, when there is reason to do so, it is also possible to swap the topic and focus functions, yielding answers like “JOHN ate the beans, MARY ate the cheese”, etc. (see Roberts 1996 and Büring 2003). Given that the contexts we use presuppose implicit multiple wh-questions, the possibility of a topic-focus swap also presents itself. Hence, in evaluating our empirical claims, one should not just rely on the effects of context, but also consider other indicators of topic- and focushood. 12 We treat this test as a test for topicality, which means that neither a contrastive topic nor a regular topic can be a negative quantifier. This certainly seems to be true. There is a question, however, as to whether a contrastive focus can be a negative quantifier. We believe that it can, contra Geurts and van der Sandt

32

Ad Neeleman and Reiko Vermeulen

(65)

#As for no boy, I like him.

Consequently, if in the relevant context a constituent can be replaced by a negative quantifier, it cannot be a topic. Indeed, when Wim in (64B-i) is replaced by niemand ‘nobody’, the result is decidedly odd, as shown in (66a).13 (The hearer is left to wonder which person is referred to as ‘nobody’.) However, the variant of (64B-i) in (66b), in which de bonen ‘the beans’ has been replaced by nergens ‘nothing’, is perfectly natural.

(2004), as negative quantifiers can move when focused, at least in some languages such as Dutch: (i) Ik geloof dat NIETS VAN WAT WE VOORSTELLEN Jan I believe that nothing from what we propose Jan zal accepteren. will accept ‘I believe that Jan will not accept anything that we will propose.’ There are however some questions surrounding the contrastive interpretation of negative quantifiers. In particular, it is not clear what the alternatives are to the negative quantifier. In most circumstances, the alternative to nothing will not be individuals but quantities like “some”. If this is the case, the meaning contributed by contrast will be equivalent to or weaker than the assertion. So, the meaning of the embedded clause of (i) can be represented as in (ii). Trivially, if John accepts none of what we propose, then he does not accept some or all of what we propose. As a result, contrast does not have detectable semantics when applied to negative quantifiers that have a quantity interpretation. (ii) a. < Oq[John will accept q(what we propose)], none, {some, all, ...}> b. q [q  {some, all, ...} & ™[John will accept q(what we propose)]]. There are special circumstances where the alternatives to a contrastive negative quantifier can be individuals, for instance, where a list is provided and the negative quantifier is one of the options. An example which is acceptable to some speakers we consulted is given in (iii). Here, the meaning of contrast is the usual rejection of an alternative proposition, namely “I would prefer strawberries” in this case. (iii) A: You can have strawberries or nothing. B: NOTHING I would prefer. 13 The answer in (66a) is felicitous if the B-accent on the subject is omitted. Doing so allows the example to be construed as providing indirect information about Fred, namely that he did not eat more from the beans than he did last year.

Chapter 1 The Syntactic Expression of Information Structure (66)

33

a. #dat niemand van DE BONEN meer gegeten heeft dan that nobody from the beans more eaten has than vorig jaar. last year ‘that nobody has eaten more from the beans than last year.’ b. dat Wim NERGENS van meer gegeten heeft dan vorig that Bill nothing of more eaten has than last jaar. year ‘that Bill has not eaten more from anything than last year.’

The data in (67) show that, by contrast, an in-situ contrastive topic may follow an in-situ focus or move across it (see (63b,d)). This observation is corroborated by the fact that in neither (67a) nor (67b) can ‘the beans’ be replaced by a negative quantifier (see (68)), whereas replacing Wim by ‘nobody’ is unproblematic in both of these examples (see (69)). The results of this test are consistent with a classification of ‘the beans’ as topic and of Wim as focus. (67)

A: Hoe zit het met de soep? Wie heeft die dit op het feest gegeten? ‘What about the soup? Who ate that at this year’s party?’ B: Nou, dat weet ik niet, maar ik geloof … ‘Well, I don’t know, but I believe …’ (i) dat WIM van de bonen meer gegeten heeft dan that Bill from the beans more eaten has than vorig jaar. last year (ii) dat [van de bonen]1 WIM t1 meer gegeten heeft that from the beans Bill more eaten has dan vorig jaar. than last year ‘that Bill has eaten more from the beans than last year.’

(68)

a. #dat WIM nergens van meer that Bill nothing of more jaar. year

gegeten heeft dan vorig eaten has than last

b. #dat [nergens van]1 WIM t1 meer gegeten heeft dan that nothing of Bill more eaten has than vorig jaar. last year ‘that Bill has not eaten more from anything than last year.’

34 (69)

Ad Neeleman and Reiko Vermeulen a. dat NIEMAND van de bonen that nobody from the beans vorig jaar. last year

meer gegeten heeft more eaten has

dan than

b. dat [van de bonen]1 NIEMAND t1 meer gegeten heeft that from the beans nobody more eaten has dan vorig jaar. than last year ‘that nobody has eaten more from the beans than last year.’

Let us summarize the results of this section. We have argued that A’scrambling in Dutch is associated with the notion contrast. In particular, this type of movement marks the scope of the negative operator associated with a contrastive meaning. As a consequence, not all topics and foci can undergo A’-movement, but only those that are interpreted contrastively. Despite the fact that contrastive topic and contrastive focus movement have an identical trigger, they behave differently in certain respects. This is because a topic cannot be contained inside the DoC of a focus (which is propositional). As a consequence, a focus cannot undergo A’scrambling to a position above a topic. But this account of the data makes crucial reference to notions of topic and focus, and is therefore compatible with the typology in (52), but not that in (53).

4.

Overview of the book

Chapter 2 (by Ad Neeleman and Hans van de Koot) considers the distribution of topic and focus in Dutch in much greater detail and embeds the proposal in a more general theory of scope marking. The data in the previous section can in principle be explained in terms of an intervention effect (e.g., foci cannot move across topics.) However, it can be shown that a contrastive focus cannot move out of a constituent containing a contrastive topic even if the base position of the focus c-commands the position of the topic. This implies that the ungrammaticality of focus movement under the relevant circumstances cannot be due to intervention. The chapter develops a theory of scope marking aiming to explain differences between scope marking by overt movement and scope marking at LF. For example, if a quantifier moves into a scope marking position in overt syntax, then quantifiers c-commanded by it can escape its scope by overt movement, but not by covert movement (for example, by QR). This is claimed to follow from a principle that blocks scope extension at LF

Chapter 1 The Syntactic Expression of Information Structure

35

across the landing site of scope marking movement. The extension to topic and focus is straightforward. The proposal is to treat the domain of contrast on a par with the scope of a quantifier. In other words, movement of a contrastive item is a scope-marking operation. Given that a topic must be interpreted externally to focus, problems arise if the landing site of a moved focus c-commands a topic. Due to the principle just introduced, the topic cannot escape the domain of contrast associated with the focus at LF, yielding an uninterpretable configuration in the information structural component. This proposal makes the crucial prediction that the distribution of topic and focus should interact with quantifier scope. As is well-known, such interactions exist. The chapter discusses generalisations from the literature as well as a new range of data confirming the proposal. Chapter 3 (by Reiko Vermeulen) explores in great detail the implications of the syntactic typology of information-structural notions given in (9) for Japanese and Korean. It systematically compares the distribution of topic and focus, both contrastive and regular types, in the two languages, and argues that the patterns observed can be explained straightforwardly if we treat contrastive topics and contrastive foci as composites of [contrast] and [topic] and [contrast] and [focus], respectively. It is demonstrated that the two languages have the same mapping rule for [topic] and the same mapping rule for [contrast], but for contrastive topics Japanese adopts the mapping rule for [topic], while Korean adopts the one for [contrast]. The rule pertaining to [topic] requires topics to be licensed in clause-initial position, marking the sister constituent as the comment, while the effects of the rule concerning [contrast] is the same as in Dutch, namely that the relevant movement marks the scope of contrast. The proposal captures similarities and differences between the two languages, which are often overlooked in the literature. Moreover, the predictions regarding the distribution of contrastive topic and contrastive focus with respect to each other, discussed above, are shown to be correct for Japanese and Korean too. The chapter also investigates the meaning of so called contrastive wa and nun in Japanese and Korean, respectively, which are suffixed to contrastive topics. There is much confusion in the literature as to whether phrases marked by these markers are topics or foci, as they sometimes exhibit properties unlike topics. Closer examination reveals that the meaning encoded by contrastive wa and nun is in fact compatible with both topicality and focus-hood. The context determines which discourse function a particular contrastive wa- or nun-marked phrase has. Evidence for this analysis includes the fact that contrastive wa- or nun-marked

36

Ad Neeleman and Reiko Vermeulen

phrases interpreted as foci show the distribution predicted in the previous section for foci, while those interpreted as topics display the distribution predicted for topics. Chapter 4 (by Elena Titov) continues the theme of a decompositional approach to contrastive focus in Russian: it provides empirical motivation for the distinct features [contrast] and [focus]. Russian data discussed in this chapter fall into place, if contrastive focus is a composite of features [contrast] and [focus], each of which is responsible for a specific syntactic effect. Evidence for [contrast] comes from the fact that only contrastive foci can undergo A’-movement, while evidence for [focus] comes from the observation that all types of focused constituents in Russian share an underlying clause-final position. Non-contrastive new information focus consistently surfaces in clause-final position, whereas contrastive focus, when it undergoes movement to positions further to the left, consistently reconstructs to clause-final position for interpretation. Thus, the launching site for the movement of contrastive foci must be the position in which new information foci surface. The chapter argues that the distribution of Russian focus is best analyzed as being regulated by mapping principles that favour a particular type of correspondence between representations in syntax and in information structure. Less than optimal results are achieved on an analysis which assumes that contrastive focus moves to a designated position in the left periphery and the movement is triggered by a syntactic feature. The latter approach fails to account for some scope facts, the optional nature of this type of movement and the fact that the movement can target a variety of positions. The proposal developed up to this point implies that there cannot be designated syntactic positions for discourse-related notions such as topic, focus and contrast. The following chapter (by Matthew Reeve) addresses this issue with respect to focus, arguing that clefts in English and Russian do not contain a FocusP. Rather, cleft constructions behave as we would expect if they are a subtype of specificational copular sentence with a semantically contentful pronominal subject. It is first argued that É. Kiss’s (1998) analysis of English clefts cannot easily capture the common properties between clefts and specificational copular sentences. The chapter then examines an apparently functionally equivalent, but syntactically distinct, cleft construction in Russian, exploring and rejecting the assumptions that a Focus Phrase analysis would have to make about its structure. An alternative unified analysis of English and Russian clefts is proposed, assimilating them to specificational copular sentences and treating the subject pronoun as non-expletive. It does depend on a functional head,

Chapter 1 The Syntactic Expression of Information Structure

37

but it encodes equative semantics, rather than information-structural content. The analysis improves on a Focus Phrase account by capturing parallels between clefts and other copular sentences and by accounting for the non-expletive nature of the cleft pronoun, the possibility and nature of wide focus clefts in Russian, restrictions on focus-movement of the clefted constituent, evidence that the “cleft clause” in English behaves like an extraposed relative clause, and an “adjacency condition” on the focus of Russian clefts. The chapter further contributes to the general theme of the book by providing evidence that the syntax-information structure interface need not make reference to syntactic features (e.g., focus features) encoding the relevant information-structural notions. Chapter 6 (by Kriszta SzendrĘi) investigates the possibility of focus movement inside the noun phrase and examines whether such movement marks the domain of contrast as in the clausal domain. The chapter considers two phenomena that are often argued to involve DP-internal focus movement, adjective reordering in English and the polydefinite construction in Greek. Adjective reordering in English is analysed as derived by focus movement, marking the domain of contrast. Some of the evidence for this analysis comes from the possibility of reconstruction in such cases. On the other hand, it is argued that the Greek polydefinite construction does not involve movement and does not mark the domain of contrast. Rather, the construction involves givenness marking (Lekakou and SzendrĘi 2012). This raises the theoretical question of whether contrastive focus marking and givenness marking are two sides of the same coin, as suggested by Schwarzschild (1999) and more recently by Wagner (2005, 2010). The chapter presents some new theoretical and empirical arguments against this position, supporting earlier proposals that focus marking and givenness marking are two separate operations (Neeleman and SzendrĘi 2004; Féry and Samek-Lodovici 2006; Krifka 2006; and Reinhart 2006). Chapter 7 (by Ad Neeleman and Reiko Vermeulen) addresses an empirical challenge to the decompositional approach to contrastive focus, which arises from the interaction between focus and negation. As already mentioned above, we hypothesise that contrast contributes a component of meaning that builds on focus semantics (see discussion around (27)), which is very much in line with the standard analysis for focus sensitive particles such as only, even, and also. Both contrast and focus-sensitive particles specify further the relation between the focus value and the ordinary value of a sentence. They do so by introducing a quantificational statement over the alternatives to the focus, which incidentally also licenses A’-movement.

38

Ad Neeleman and Reiko Vermeulen

The empirical challenge to this proposal is that sentences with some types of focus-sensitive particles yield a meaning that is unexpected and seems to falsify the general approach that the component of meaning introduced by the focus-sensitive particle is computed based on the focus semantics of the sentence. For instance, John did not invite only Pia does not imply that John did not invite Pia. Rather, it implies that John invited Pia in addition to someone else. In other words, although realised as sentential negation, not must be interpreted as constituent negation, on a par with John invited not only Pia. The chapter proposes that these kinds of effects are seen with focussensitive particles that contribute a negative component of meaning. Only belongs to this class, as is obvious from the presence of a negative operator in (28b). Constituents of the form not Y but X also belong to this class, but even and also do not. The recalcitrant data do not require adjustment of the standard semantic theory of focus-sensitive particles, but can be explained if elements contributing a negative component of meaning are treated as positive polarity items. This is important, because the decomposition of contrastive focus into [contrast] and [focus] relies exactly on this theory.

Chapter 2 Towards a Unified Encoding of Contrast and Scope* Ad Neeleman & Hans van de Koot 1.

Introduction

As argued in chapter 1, contrastive elements are associated with a particular syntactic domain, which we have called their domain of contrast. In particular, we have suggested that the semantics of contrastive topics and foci involves a negative operator, which makes it possible to analyze a domain of contrast as a kind of ‘contrastive scope’. This line of thinking suggests that we should investigate whether the theory of contrastive scope could be subsumed under the theory of quantifier scope. That such a move is desirable is not just a matter of theoretical parsimony but is also suggested by interactions between information structure and quantifier scope observed in the literature. To give two examples that we discuss below, Sæbø (1997) notices that scope shift of one quantifier over another is favoured if the lower quantifier is a topic, while Krifka (1998) observes that inverse scope in German requires a special intonational contour, otherwise associated with topic and focus readings. In this chapter we present a theory that does indeed unify the mechanisms by which quantifier scope and contrastive scope are encoded. This unification takes the following form. To begin with, we show that certain * This paper is an extended reply to a question raised by Edwin Williams regarding an earlier paper of ours, which left open the LF mechanism that allows interpretation of an in situ topic when c-commanded by an in situ focus. In addition to Edwin Williams, we would like to thank the following people for help and advice: Klaus Abels, David Adger, Greville Corbett, Danny Fox, Ray Jackendoff, Nathan Klinedinst, Terje Lohndal, Matthew Reeve, Uli Sauerland, Kriszta SzendrĘi, Shoichi Takahashi, Rob Truswell, and Reiko Vermeulen. Various parts of this paper were presented at the University of Potsdam in 2008, the University of Seoul (SICOL 2010), the University of Sussex, the University of Leeds (LAGB 2010), the University of Utrecht (Semantics in the Netherlands IX, 2011), and the University of Verona (Workshop on Anaphora and Information Structure, 2012). We thank the participants for valuable comments and discussion. Finally, we thank the very many students at UCL who were willing to help us with scope judgments of sentences containing three quantifiers.

40

Ad Neeleman & Hans van de Koot

positional restrictions on topics and foci follow from a general condition that holds at information structure, in conjunction with a condition on scope shift (the CSS, to be introduced below). We then need to show that the CSS also regulates quantifier scope. We do this in two steps. First, we argue that several interactions between scope and information structure are explained by this condition (in conjunction with independently motivated principles of interpretation). Second, we argue that the CSS gives an adequate account of scopal restrictions in sentences with three quantifiers. The overall theory is strikingly simple, but has a broad empirical basis. The proposal we develop also supports a specific view of quantifier scope, advocated in particular by Reinhart (see Reinhart 1983, 1995, 2006), according to which LF is a representation of relative scope only. That is, scope marking at LF is limited to cases where scope diverges from surface c-command. The CSS is incompatible with the perhaps more standard view that every quantifier undergoes covert movement to a position from which it c-commands its scope domain. Thus, to the extent that our proposals are on the right track, they not only allow a unification of the mechanisms of encoding quantifier scope and contrastive scope, but they also shed light on the nature of the LF interface.

2.

A new condition on scope shift

For reasons that will become apparent as we develop our proposals, it will be advantageous to adopt an encoding of covert scope extension first introduced in Williams 1994 (with some modifications). Williams argues that scope extension is represented at LF as percolation of an index originating in the quantified expression. On this view, the scope of a quantified argument coincides with the largest category that carries its index, minus the argument itself. Thus, the scope of QP1 in (1) is Ȗ. Throughout this chapter, we place inherited indices after a colon (to distinguish them from an index introduced by a quantifier, which precedes the colon; we omit the colon when there is no inherited index). α

(1)

γ:1

β

ε:1

δ ζ

QP1

Chapter 2 Towards a Unified Encoding of Contrast and Scope

41

Of course, scope can also be marked by overt movement, as is common in languages such as German and Japanese. We assume that scope-marking movement involves percolation of the scope index from the moved quantifier to the node that immediately dominates its landing site. This percolation has the same interpretation as before, so that in (2) the scope of QP is Į minus QP, which equals Ȗ. α:1

(2) QP1

γ δ

ε ζ

t

We take the index percolation in (2) to be an expression of the scopemarking nature of the movement. In the absence of index percolation, QP undergoes scope reconstruction, which implies that its scope coincides with the c-command domain of its trace or is extended by index percolation from its trace. Therefore, A’-movement not accompanied by index percolation is not a scope-marking operation. Since the work of Chomsky (1976) and May (1977), the standard view in transformational generative grammar has been that there is a syntactic level of Logical Form (LF) whose representations are transparent in that they are (largely) isomorphic to semantic structures (see McCawley 1970 for related ideas). LF representations are derived through the operation of Quantifier Raising (QR), which moves quantified expressions from their surface position and attaches them to a node of a suitable semantic type, typically , that constitutes their scope (see Barwise and Cooper 1981 and Heim and Kratzer 1998). The trace left behind by QR functions as the variable required for interpretation of the quantifier. The claim of restrictiveness that empirically supports the standard view is linked to the null hypothesis that Quantifier Raising obeys restrictions on movement and hence that an interpretation is only available for a surface string containing a quantifier if it can be derived through one or more well-formed applications of move Į. A translation of this view into the index-based representations of Williams 1994 yields the representation in (3a) for a structure containing two quantifiers interpreted as taking surface scope and that in (3b) for the same structure with inverse scope.

42 (3)

Ad Neeleman & Hans van de Koot a. [:1 … [:1,2 … QP1 [:2 … QP2 … ]]] b. [:2 … [:1,2 … QP1 [:2 … QP2 … ]]]

QP1 > QP2 QP2 > QP1

There is an alternative view of scope according to which LF merely represents deviations from surface scope. Thus, there is no requirement that quantifiers must take a sister of type in the syntactic representation. Rather, scope extension only takes place when a quantifier must take scope over a quantifier that c-commands it. In other words, scope extension is limited to structures in which it generates an interpretation that is otherwise unavailable.1 This is essentially the view of quantifier scope in Reinhart 1983, 1995 and 2006 (see also Lakoff 1972, Huang 1982, Reinhart 1983, and Hoji 1985). Reinhart treats scope extension as QR; a translation of her proposal in terms of indices expresses the readings in (3) with the slightly simpler structures in (4). (4)

a. [ … [ … QP1 [ … QP2 … ]]] b. [:2 … [:2 … QP1 [:2 … QP2 … ]]]

QP1 > QP2 QP2 > QP1

We summarize the main tenets of our proposal in (5).2,3 The Economy principle in (5b) is intended to block scope extension where it does not give rise to inverse scope, while (5c) is a mapping principle that, in the

1 The view that scope extension is subject to Economy and relativized to an interpretation has been argued for in Fox 1999, 2000. However, unlike Reinhart, Fox assumes that every quantifier must move. 2 Of course, this list of assumptions is incomplete. It omits conditions on quantifier lowering, economy conditions on scope shift (Fox 2000), the condition that QR is (often) clause-bounded, and so on. Where relevant, these will be introduced in what follows. 3 To clarify, we claim that LF represents scope relations between quantifiers but does not give a full representation of quantifier scope. This implies that while quantifiers can in principle be interpreted without scope extension through index percolation or movement, scope extension is necessary for inverse scope. There are several ways of integrating this proposal in a larger theory of the syntax-semantics interface. Quantifiers might belong to a limited set of semantic types that can express scope but cannot give rise to scope inversion (see Montague 1974). An alternative would be to assume that scope is represented in a separate module (see Jackendoff 1996 and Williams 2003 for different characterizations of this module). Our LF structures are mapped onto this module using a holistic notion of shape conservation rather than through a syntactic operation (such as index percolation).

Chapter 2 Towards a Unified Encoding of Contrast and Scope

43

absence of scope extension, regulates the association of syntactic structures at LF with their semantic representations. (5)

a. Scope Extension If a quantifier percolates its index to a dominating node Į, then its scope coincides with Į minus the quantifier itself. b. Economy Scope extension must give rise to an otherwise unavailable interpretation. c. Default Scope Rule If a quantifier does not percolate its index, it takes scope over its c-command domain.

The rule in (6) is the condition at the heart of this chapter that allows us to build the case that restrictions on placement of topics and foci fall out from the theory of scope. It states that index percolation is severely restricted: no node can belong to the extended scope of two quantifiers.4 (6)

Condition on Scope Shift (CSS) No node may inherit two scope indices.

Our argument for the view that LF merely represents scope shift is based on the condition in (6). If the CSS is correct, then the representations in (3) are inadmissible. Those in (4) are of course grammatical. An important fact about the CSS is that it creates an asymmetry between covert scope shift, analyzed here as index percolation, and scope marking by overt movement. Because a moved quantifier only percolates an index in its landing site, the movement is insensitive to scope extensions in its path, whether created by movement or by index percolation. By contrast, the index percolated by an in situ quantifier will violate the CSS whenever another quantifier attached to a node in the intended percolation path percolates an index as well. This contrast is illustrated in (7), where a quantifier QP1 (which may be in situ or may have moved) percolates its index to a node dominating a 4 There is a superficial similarity between the pattern captured by the CSS and the intervention effect discussed in Beck 1996, 2006. However, Beck’s data yield to a semantic explanation, which we believe is unlikely for CSS effects. Moreover, the CSS deals with the interaction between two scope shifts, whereas Beck’s work is concerned with a single scope shift across a c-commanding quantifier.

44

Ad Neeleman & Hans van de Koot

second quantifier, QP2. The CSS makes it impossible for QP2 to extend its scope beyond QP1 through index percolation (see (7a), where Ȗ inherits two scope indices), but it does not prohibit movement of QP2 across QP1 (see (7b)). Thus, overt scope marking is freer than covert scope shift, a prediction we will explore at various places in this chapter. As will be clear, this predicted difference is hard to reconcile with theories that treat covert scope shift as movement. (7)

a.

*

α:2

β

γ:1,2 QP1

3.

b.

α:2 QP2

QP2

γ:1 QP1

t2

Topic/focus interaction and the CSS

In chapter 1 an initial proposal was developed to account for informationstructural restrictions on A’-scrambling in Dutch. This fairly informal account relied on the notion of DoC-marking. (DoC stands for ‘domain of contrast’.) It was suggested there that DoC-marking should be considered a type of scope marking: if a contrastive phrase moves, then the movement marks the scope of the contrast. No characterization was given of the analysis of in situ topics and foci. We will now attempt to put the intuitive ideas of chapter 1 on a firmer footing by assimilating them to the general theory of scope sketched in section 2. In particular, we will argue that, like quantifiers, contrastive categories that remain in situ may percolate an index to extend their scope, whereas A’-scrambling of a contrastive category only requires percolation of the scope index to the node that immediately dominates its landing site (on the assumption that A’-scrambling is a scope-marking operation). The resulting theory explains the observations made in chapter 1 regarding the interaction between topics and foci in A’-scrambling structures. That is to say, the CSS, which we will argue is a condition that regulates quantifier scope, also explains the distribution of topic and focus. The theory that emerges will also capture a number of new observations about the interaction between A’-scrambling and quantifier scope. These will be explored in section 4. Recall that our theory of information structure requires a topic to be interpreted externally to the minimal information-structural unit containing all foci in a given utterance. In a nutshell, this is because ‘topic’ is an utterance-level notion, while ‘focus’ operates at the level of propositions.

Chapter 2 Towards a Unified Encoding of Contrast and Scope

45

Whereas utterances can contain propositions, the reverse is not true. Thus, (8a), but not (8b), is a well-formed information structure (as before, topics are doubly underlined, foci appear in small caps and contrastive categories are italicized). (8)

a. topic [ … FOCUS … ] b. *FOCUS [ … topic … ]

What are the consequences of this information-structural restriction for sentences that contain an in situ contrastive topic and an in situ contrastive focus? Recall that LF is not a full representation of scope, but only encodes the relative scope of quantificational categories. Therefore, if the topic c-commands the focus neither category needs to extend its scope, as in (9a). If the c-command relations are reversed, however, the topic will be forced to percolate its index to a node dominating the focus, as in (9b). (9)

a. [ … topic1 [ … FOCUS2 … ]] b. [:1 … FOCUS2 [:1 … topic1 … ]]

The situation is more complex in structures involving A’-scrambling. In line with the suggestions in chapter 1, we take A’-scrambling to be a scopemarking operation (it marks the domain of contrast of the moved category). As discussed earlier, a movement is scope-marking if the moved category percolates its index to the node that immediately dominates it. For A’-scrambling this implies that it creates a barrier for scope extension of categories in its c-command domain. This is unproblematic if a topic is scrambled across a focus, as in (10a), but it leads to a violation of the CSS if a focus is scrambled across a topic, as in (10b). In the latter case, the resulting syntactic representation can only be mapped onto a well-formed information structure if the topic percolates its index to a node dominating the moved focus. But this will give rise to overlapping paths of index percolation. (10)

a. [:1 … topic1 [ … FOCUS2 [ … ttop … ]]] b. *[:1 … [:1,2 … FOCUS2 [:1 … topic1 [ … tFOC … ]]]]

This effect of the CSS is one factor that governs topic-focus interactions. A second factor is an independent scope-freezing effect associated with A’movement. In particular, as already observed by Van Riemsdijk (1978), once an element moves to a scope position in overt syntax, further scope

46

Ad Neeleman & Hans van de Koot

extension at LF is unavailable. For example, the questions Who knows what John bought? and Who knows that John bought what? differ in that only the second can be construed as a multiple wh-question. This is because in the former what has been moved to a scope position, which makes it impossible to extend its scope at LF (as a result, wh-absorption is blocked). In the current framework, the scope-freezing effect associated with movement can be expressed as a restriction on index percolation: the index of a moved category cannot percolate higher than the mother node of the landing site. This restriction on index percolation can be understood as follows. We have assumed that the movement of topics and foci is licensed as long as these movements mark the scope of contrast. If the index were to percolate beyond the node that immediately dominates the moved category, then the movement can no longer be said to be scope-marking and therefore it lacks an appropriate trigger. The CSS and the restriction on index percolation after A’-movement give rise to the overall patterns in (11) and (12). In all the grammatical cases, the topic can be interpreted externally to the focus, either because this interpretation matches surface scope (as in (11a), (11b), (11c), (11e), (12d) and (12f)), or because index percolation from the topic to a node dominating the focus is possible (as in (12a)). In the remaining structures, interpretive requirements force the topic to percolate its index across the focus, but this is blocked by the CSS (as in (11d) and (12b)), by the restriction that overt movement of contrastive elements must mark their scope (as in (12c)), or by both restrictions (as in (11f) and (12e)). (11)

a. [… topic … FOCUS …] b. … topic1 … [ … t1 … FOCUS …] c. [… topic … FOCUS1 … t1 …] d. * … FOCUS1 … [ … topic … t1 …] e. … topic1 … FOCUS2 … [ … t1 … t2 …] f. * … FOCUS2 … topic1 … [ … t1 … t2 …]

(12)

a. [ … FOCUS … topic …] b. * … FOCUS1 … [ … t1 … topic …] c. * [ … FOCUS … topic1 … t1 …] d. … topic1 … [ … FOCUS … t1 …] e. * … FOCUS1 … topic2 … [ … t1 … t2 …] f. … topic2 … FOCUS1 … [ … t1 … t2 …]

In sum, the order topic-focus is well-formed, whether the topic has moved or not, but the order focus-topic is well-formed only if both the topic and the focus are in situ.

Chapter 2 Towards a Unified Encoding of Contrast and Scope

47

The empirical consequences of our proposal depend on the exact definitions of topic and focus. These were outlined in chapter 1, but in the present context it is worth repeating that we reserve the term topic for syntactic constituents that introduce a new discourse topic, narrow down the current discourse topic, or change it. We thus exclude constituents that are merely discourse-anaphoric. We return to this issue at the end of this section. A subset of the above predictions were already discussed in chapter 1, where we dealt with in situ structures and with structures in which a topic moves across a focus or vice versa (see (11a), (11d), (12a) and (12d)). We repeat the relevant data in (13) and (14). (13)

Hoe zit het met FRED? Wat heeft HIJ gegeten? ‘What about Fred? What did he eat?’ Nou, dat weet ik niet, maar ik geloof ‘Well, I don’t know, but I believe’ a. dat Wim van de BONEN meer gegeten heeft dan vorig jaar. that Bill from the beans more eaten has than last year ‘that Bill has eaten more from the beans than last year.’ b. # dat [ van de BONEN]1 Wim t1 meer gegeten heeft dan vorig jaar. that from the beans Bill more eaten has than last year

(14)

Hoe zit het met de SOEP? Wie heeft DIE gegeten? ‘What about the soup? Who ate that?’ Nou, dat weet ik niet, maar ik geloof … ‘Well, I don’t know, but I believe …’ a. dat WIM van de bonen meer gegeten heeft dan vorig jaar. that Bill from the beans more eaten has than last year b. dat [ van de bonen]1 WIM t1 meer gegeten heeft dan vorig jaar. that from the beans Bill more eaten has than last year ‘that Bill has eaten more from the beans than last year.’

The movement structures in the above examples involve configurations in which one contrastive category moves across another. This may be suggestive of an account in terms of intervention. However, the restriction that falls out from the CSS has nothing to do with intervention; the constituent that is the sister of a moved focus can simply not contain a topic, as that topic would have to percolate its index to a node dominating the moved focus. Consequently, even if the focus precedes and c-commands the topic in the base order, the movement of the focus should lead to a violation of the CSS (see (12b)). The prediction is confirmed by the data in (15), which involve a ditransitive verb. The context is set up in such a way as to favour a reading of the indirect object as focus (it answers the WH-question), and the direct object

48

Ad Neeleman & Hans van de Koot

as topic (it switches the topic of the discourse from the antique sideboard to the clock). While the answer in (15a) is felicitous, the answer in (15b), where the focus has moved, cannot be produced with a focus-topic intonation. (15)

Hoe zit het met het antieke DRESSOIR? Wie heeft grootvader DAT nagelaten? ‘How about the antique sideboard? To whom has granddad bequeathed that?’ Nou, dat weet ik niet, maar ik geloof ‘Well, I don’t know, but I believe …’ a. dat grootvader zijn BUREN de klok heeft nagelaten. that granddad his neighbours the clock has bequeathed b. #dat [ zijn BUREN]1 grootvader t1 de klok heeft nagelaten. that his neighbours granddad the klok has bequeathed ‘that granddad has bequeathed the clock to his neighbours.’

We can corroborate our classification of the indirect object and direct object as focus and topic by trying to replace them with a negative quantifier (see chapter 1). In the context of (16), such a replacement is only possible for the indirect object. (16)

a. #dat grootvader zijn BUREN niets heeft nagelaten. that granddad his neighbours nothing has bequeathed ‘that granddad did not bequeath anything to his neighbours.’ b. dat grootvader NIEMAND de klok heeft nagelaten. that granddad nobody the clock has bequeathed ‘that granddad did not bequeath the clock to anyone.’

As expected (see (11b)), it is possible to move a topic from a position preceding a focus, as shown in (17). Such structures do not require index percolation from the focus. (17)

Hoe zit het met tante JO? Wat heeft grootvader HAAR nagelaten? ‘How about auntie Jo? What has granddad bequeathed to her?’ Nou, dat weet ik niet, maar ik geloof ‘Well, I don’t know, but I believe …’ a. dat grootvader zijn buren de KLOK heeft nagelaten. that granddad his neighbours the clock has bequeathed b. dat [ zijn buren] grootvader t1 de KLOK heeft nagelaten. that his neighbours granddad the clock has bequeathed ‘that granddad has bequeathed the clock to his neighbours.’

Chapter 2 Towards a Unified Encoding of Contrast and Scope

49

Our classification of ‘neighbours’ as topic and of ‘clock’ as focus in the examples in (17) receives support from the by now familiar negativequantifier test (see (18) and (19)). (18)

a. # dat grootvader niemand de KLOK heeft nagelaten. that granddad nobody the clock has bequeathed b. # dat [ niemand]1 grootvader t1 de KLOK heeft nagelaten. that nobody granddad the clock has bequeathed ‘that granddad did not bequeath the clock to anyone.’

(19)

a.

dat grootvader zijn buren NIETS heeft nagelaten. that granddad his neighbours nothing has bequeathed b. dat [ zijn buren]1 grootvader t1 NIETS heeft nagelaten. that his neighbours granddad nothing has bequeathed ‘that granddad did not bequeath anything to his neighbours.’

Further evidence showing that an analysis based on intervention is insufficiently general comes from structures in which a contrastive element moves to a position below a topic or focus. Where the moving element is a topic (see (12c)), it would have to percolate its index to a node dominating a c-commanding focus, but since the topic has moved, such scope extension is impossible. This pattern is indeed what we find. A topic can move, but if it does so, it must land in a position c-commanding any focus: (20)

Hoe zit het met de NIETMACHINE? Wie heeft een collega DAAR om gevraagd? ‘What about the stapler? Who has asked a colleague for that?’ Nou, dat weet ik niet, maar … ‘Well, I don’t know, but …’ a. ik geloof dat PIET Jan om de liniaal heeft gevraagd. I believe that Peter John for the ruler has asked b. #ik geloof dat PIET [ om de liniaal]1 Jan t1 heeft gevraagd. I believe that Peter for the ruler John has asked c. ik geloof dat [ om de liniaal]1 PIET Jan t1 heeft gevraagd. I believe that for the ruler Peter John has asked ‘I believe that Peter has asked John for the ruler.’

In contrast, no difficulties are predicted if a focus moves to a position below a topic (see (11c)), as such structures do not require index percolation from the focus. Indeed, there is a sharp contrast between (21b), where the focus lands in a position below the topic, and (21c), where it lands in a position c-commanding the topic.

50 (21)

Ad Neeleman & Hans van de Koot Hoe zit het met MARIE? Waar heeft ZIJ een collega om gevraagd? ‘What about the stapler? Who has asked a colleague for that?’ Nou, dat weet ik niet, maar … ‘Well, I don’t know, but …’ a. ik geloof dat Piet Jan om de LINIAAL heeft gevraagd. I believe that Peter John for the ruler has asked b. ik geloof dat Piet [ om de LINIAAL]1 Jan t1 heeft gevraagd. I believe that Peter for the ruler John has asked c. #ik geloof dat [ om de LINIAAL]1 Piet Jan t1 heeft gevraagd. I believe that for the ruler Peter John has asked ‘I believe that Peter has asked John for the ruler.’

Finally, the logic of our proposal makes the further prediction that focus movement across a topic will not lead to ungrammaticality if the topic subsequently moves across the focus (see (11e) and (12f)). This is because scope-marking through movement does not create a path of index percolation to accompany the path of movement. Rather the index percolates from the moved category only. Consequently, in a structure like (22) no node needs to belong to the extended scope of two categories. (22)

[:1 … topic1 [:2 … FOCUS2 [ … ttop [ … tFOC … ]]]

That configurations like (22) are indeed grammatical is confirmed by the example in (23). (23)

Hoe zit het met Jan z’n OUDERS? Welk boek heeft hij HEN gegeven? ‘What about John’s parents? Which book has he given to them?’ Nou dat weet ik niet. Maar ik denk … ‘Well, I don’t know. But I think …’ dat [ z’n kinderen]1 [ DIT boek]2 Jan t1 zeker t2 niet zal geven. that his children this book John certainly not will give ‘that John will certainly not give his children this book.’

The opposite order of movement (see (11f) and (12e)) is correctly predicted to be ungrammatical, as in the resulting structure the topic would have to percolate its index to a node dominating the focus. This would violate the CSS as well as the restriction that such percolation is not possible from moved categories. (24)

#dat [ DIT boek]2 [ z’n kinderen]1 Jan t1 zeker t2 niet zal geven. that this book his children John certainly not will give ‘that John will certainly not give his children this book.’

Chapter 2 Towards a Unified Encoding of Contrast and Scope

51

For reasons of space, we do not demonstrate here that the contrast between (23) and (24) is independent of the base order of topic and focus. The above completes our survey of the structures in (11) and (12). The generalization that captures all patterns is that topics and foci may appear in either order, with the topic c-commanding the focus or vice versa, but as soon as one of these categories moves, a surface structure must be generated in which the topic c-commands the focus. This generalization must be qualified in two ways. First, as mentioned earlier, it requires a particular definition of the notion ‘topic’, namely one that excludes categories that are discourse-anaphoric and hence merely refer to the current topic of discourse. We are aware that this conception of topic is narrower than assumed in much of the literature. Note, however, that it is not without precedent (compare Lambrecht’s 1994 notion of ‘reference-oriented topic’ and Büring’s 1997 notion of ‘S-Topic’). To give an example, Maxine in (25) is a topic, but on our classification this is not true of her. Rather, the function of this pronoun is to indicate how the topic is related to the comment (we mark discourse anaphoricity by wavy underlining). (25)

As for Maxine, I really like her.

We believe that this analysis should be generalized to other discourseanaphoric constituents that index the current topic. For example, the continuations below of the utterance in (25) are all-comment sentences linked to the topic Maxine through the pronoun she and the definite DP the girl (see Lambrecht’s 1994 discussion of sentences containing only a ‘role-oriented topic expression’, and Vallduví 1992 and Vallduví and Engdahl 1996 for discussion of sentences without a ‘link’). (26)

a. She always wears such nice scarfs. b. The girl always wears such nice scarfs.

This proposal makes two predictions. If discourse-anaphoric material coreferential with the current topic is interpreted as part of the comment, it should not undergo topic movement. Moreover, it should be allowed to surface in a constituent that is the sister of a moved focus. Both predictions seem to be correct: (27)

a. As for Felix, he hit John. Nobody has talked to him ever since. b. As for Felix, he hit John. #Him nobody has talked to t ever since.

52 (28)

Ad Neeleman & Hans van de Koot a. As for Felix, he hit John. He stayed clear of BILL. b. As for Felix, he hit John. BILL he stayed clear of t.

Our second qualification is that our information-structural constraint in (8) holds within a domain that can be construed as an utterance, but not across utterances. In other words, when an utterance is embedded in a larger utterance, the well-formedness of the two utterances vis-à-vis information-structural constraints like (8) is evaluated independently. This is clearly relevant for verbs like tell or say, which may embed an utterance as their complement: (29)

[CP-1 … V [CP-2 …]], where CP-2 is interpreted as an utterance Domains of evaluation: (i) CP-2 and (ii) CP-1 excluding CP-2.

Although we have argued that a focus may not move to a position c-commanding a topic, (29) has the implication that this is not true across-theboard. A focus may move in the matrix clause headed by a verb like tell, even if the embedded CP contains a topic. This is indeed possible: (30)

Jan zei tegen Sofie [dat dit soort boeken zelfs PIET niet leest] ‘John said to Sophie that even Peter doesn’t read such books.’ Nee dat heb je fout. Ik weet wel zeker … ‘No, you’re wrong about that. I’m pretty sure …’ dat [ tegen MARIE]1 Jan t1 zei dat [ dit soort boeken]2 zelfs PIET t2 that to Mary John said that this kind books even Peter niet leest. not reads ‘that John said to Mary that even Peter does not read such books.’

By contrast, if a topic or focus moves out of an embedded CP to a matrix CP, then the entire sentence must be construed as a single utterance. The consequence is that the entire sentence constitutes a single domain for evaluation against (8). But since the topic must percolate its index to a node dominating the moved focus, (8) is violated: (31)

Jan zei [dat dit soort boeken zelfs PIET niet leest] ‘John said that even Peter doesn’t read such books.’ Nee dat heb je fout. Ik weet wel zeker … ‘No, you’re wrong about that. I’m pretty sure …’ #dat [ zelfs KAREL]1 Jan zei dat [ dit soort boeken]2 niet that even Charles John said that this kind books not leest. reads ‘that John said that even Charles does not read such books.’

Chapter 2 Towards a Unified Encoding of Contrast and Scope

4.

53

Interactions BETWEEN scope shift and A’-scrambling

In the previous section we have considered overt and covert operations that mark the scope of contrastive categories. If we are correct in claiming that quantifier scope and contrastive scope are part of the same system, then we would expect interaction between these phenomena (i) when a sentence contains both a contrastive element and a quantifier or (ii) when a quantifier receives a contrastive interpretation. These interactions come about through two conditions, the CSS and the Q-Scope Inclusion Corollary. The latter will be introduced shortly. Consider first sentences containing both a quantifier and a contrastive phrase. It is conceivable that the CSS is relativized with respect to type of scope marking, so that marking of a domain of contrast would not interfere with marking of quantificational scope. Alternatively, if the CSS generalizes across quantificational and contrastive scope, movement of a contrastive category should create a barrier for scope extension of quantifiers in its c-command domain. In particular, a focus movement that lands between two quantifiers, as in (32a), should have a scope freezing effect that is absent when the landing site c-commands both quantifiers, as in (32b). This is because only in (32a) the index percolated by the moved focus blocks percolation of the index of QP2 to a node dominating QP1. (32)

a.

*

α:2

QP1

b. γ:2,3

Foc3

α:3 Foc3

ε:2 QP2

QP1

ε:2 QP2

η tFoc

γ:2

ι

η tFoc

ι

The data in (33) provide strong support for the view that the CSS is not relativized with respect to the two types of scope marking. (33)

Het was een gekkenhuis in de foyer. Ik geloof … ‘It was a madhouse in the foyer. I believe …’ a. dat iemand iedere ober om champagne gevraagd heeft. that someone every waiter for champagne asked has (i) some > every; (ii) every > some

54

Ad Neeleman & Hans van de Koot b. dat iemand [PP om CHAMPAGNE] iedere that someone for champagne every (i) some > every; (ii) *every > some c. dat [PP om CHAMPAGNE] iemand iedere that for champagne someone every (i) some > every; (ii) every > some

ober tPP gevraagd heeft. waiter asked has ober tPP gevraagd heeft. waiter asked has

The sentence in (33a) is our baseline example, showing that the universal object can extend its scope across the indefinite subject. In (33b), the scope ambiguity is lost, presumably as a consequence of the A’-scrambling operation that places the PP-complement between the subject and the object. The scrambled PP percolates an index and thereby prevents the universal from percolating its scope index across the landing site of the scrambled PP. The example in (33c), which instantiates the structure in (32b), is again ambiguous. This is as expected, as A’-scrambling of the PPcomplement to a position above the subject does not prevent the universal from percolating its scope index to a node that dominates the indefinite but not the fronted PP. Some speakers find A’-scrambling of contrastive foci to a position below the subject, as in (33b), somewhat marginal. However, the pattern of scope judgments that obtains when the PP is interpreted as a contrastive topic runs parallel to that in (33): (34)

Zeg Jan, hoe zit het met PROSECCO? Is DAT populair in jouw café? ‘Hey John, what about Prosecco? Is that popular in your pub?’ Nou, over Prosecco heb ik de cijfers niet bij de hand, maar wel weet ik … ‘Well, I don’t have the figures for Prosecco, but I do know …’ a. dat iemand iedere ober om champagne telkens in het WEEKEND that someone every waiter for champagne always in the weekend gevraagd heeft. asked has (i) some > every; (ii) every > some b. dat iemand [PP om champagne] iedere ober tPP telkens in het that someone for champagne every waiter always in the WEEKEND gevraagd heeft. weekend asked has (i) some > every; (ii) *every > some c. dat [PP om champagne] iemand iedere ober tPP telkens in het that for champagne someone every waiter always in the WEEKEND gevraagd heeft. weekend asked has (i) some > every; (ii) every > some

Chapter 2 Towards a Unified Encoding of Contrast and Scope

55

In addition to the type of interaction illustrated in (33) and (34), there are also noticeable effects when a quantifier receives a contrastive interpretation. In particular, it can be demonstrated that the domain of contrast of a contrastive quantifier can be larger, but not smaller, than its quantificational scope. We will refer to this restriction as the Q-Scope Inclusion Corollary: (35)

Q-Scope Inclusion Corollary (QIC) The domain of contrast of any contrastive quantifier must include its quantificational scope.

The QIC can be derived from the assumption that the minimal domain of contrast for a contrastive category must contain all elements that depend on that category. This assumption is no more than what is implicitly assumed in most theories of contrastive interpretation: a contrastive element is interpreted with respect to a syntactic or semantic ‘frame’ against which alternatives are evaluated. This frame, which we have called the domain of contrast, must at least contain those elements with which the contrastive category entertains a semantic relation. For example, if the contrastive element is an object, then its domain of contrast must at least contain the selecting verb, but not necessarily the subject. But if the subject happens to be a quantifier that is scopally dependent on the object, then it, too, must be part of the domain of contrast. The QIC emerges from the interaction between the above assumption regarding the minimal domain of contrast and our earlier claim that scope extension is only licensed if it gives rise to scope shift. What the latter claim amounts to is that scope extension must create a new scopal dependency. Thus, extension of the scope of the universal in some boy admires every teacher is licensed because it gives rise to a reading in which some boy is dependent on every teacher. However, examples like (36) are unambiguous because the wide scope reading of the universal in the first sentence would require a parallel scope extension in the second sentence that fails to create a new scopal dependency (see Fox 2000 for extensive discussion). (36)

Some boy admires every teacher. Mary does admire every teacher, too. (i) some > every; (ii) *every > some

A derivation that violates the QIC would be one in which a contrastive quantifier takes quantificational scope over material that is not contained in its domain of contrast. This material cannot depend on the quantifier,

56

Ad Neeleman & Hans van de Koot

because if it did, it would have to be contained in its domain of contrast. But if it does not depend on the quantifier, then scope extension is not licensed, as this operation must give rise to a new scopal dependency. Therefore, the QIC must hold. There is ample empirical support for the QIC. To begin with, there are a number of structures in which the domain of contrast of a contrastive quantifier is indeed larger than its scope. Consider the examples in (37). The domain of contrast of the fronted universal is presumably everything contained in its sister node (the matrix IP). After all, fronting in English – like A’-scrambling in Dutch – requires a contrastive interpretation of the fronted category. However, whereas (37b) is ambiguous for all speakers, only a subset of speakers accept a wide scope reading of the universal in (37a). This suggests that for speakers in the complement of this subset, the maximal quantificational scope of the universal in (37a) is the embedded clause. Moreover, even the more permissive speakers allow narrow scope of the universal in this example.5 (37)

a. Every one of the motorbikes, a mechanic said that he will fix t in one day. (i) %every > a; (ii) a > every) b. Every one of the motorbikes, John claimed that a mechanic will fix t in one day. (i) every > a; (ii) a > every)

The same divergence between quantificational scope and domain of contrast arises with contrastive quantifiers that remain in situ. The focused indefinite in the answer in (38) can depend on the universal, indicating that it can be interpreted below the subject for quantifier scope, even though the context strongly suggests that the universal must be part of its domain of contrast. Thus, its contrastive scope must include every student, but its quantificational scope may exclude this constituent. (38)

Q: What did every student read after reading the morning paper? A: Every student read a BOOK. (i) every > a; (ii) a > every)

5 The example in (37a) is taken from Bianchi and Frascarelli 2010, who report that 13 out of 28 speakers can get the wide scope reading of the universal. We speculate that the availability of this reading correlates with the ability of speakers to perform non-local scope shift.

Chapter 2 Towards a Unified Encoding of Contrast and Scope

57

An even stronger case is provided by the example in (39), where the context forces the universal to scope under at least one student, even though this constituent must be contained in its domain of contrast. (39)

Q: What has at least one student read? A: At least one student has read every BOOK. (i) *every > at least one; (ii) at least one > every)

In conclusion, the domain of contrast of a contrastive quantifier can be larger than its quantificational scope. The QIC predicts that the reverse situation, in which the quantificational scope is larger than the domain of contrast, cannot exist. We can test this using overt A’-movements, such as topicalization in English and A’scrambling in Dutch. These operations mark a domain of contrast and should therefore have a freezing effect on the quantificational scope of the scrambled category.6 This is because – by the QIC – the constituent moved by these operations cannot extend its quantificational scope from the landing site. That topicalization in English has a scope freezing effect was already observed by Bianchi and Frascarelli (2010). None of the speakers in their sample accepted a wide scope reading of the universal in the following example: (40)

A mechanic said that every one of the motorbikes, he will fix t in one day. (i) *every > a; (ii) a > every)

The predicted freezing effect with A’-scrambling can be illustrated with the following data set. In (41a), the contrastive quantifier remains in situ and hence its scope can be extended across the indefinite subject, giving rise to scopal ambiguity. In (41b), however, the contrastive quantifier has scrambled to a position below the subject, marking VP as its domain of contrast. This implies that its quantificational scope cannot include the subject. Indeed, this example very strongly resists a wide-scope reading of

6 Note that this effect should not be subsumed under the general scope-freezing effect of A’-movement. As Van Riemsdijk (1978) observes, overt WH-movement blocks covert scope extension of the moved WH. We have suggested that this is because scope extension would destroy the motivation for the scopemarking movement. In the case at hand, however, the movement is motivated by that fact that it marks the scope of contrast, which implies that extension of quantifier scope is in principle possible. In order to block it, we need something like the QIC.

58

Ad Neeleman & Hans van de Koot

the universal.7 In (41c), A’-scrambling crosses the subject, thereby making it possible for the universal to take scope over the indefinite. (41)

a. Ik geloof dat een jongen Marie ieder BOEK gegeven heeft. I believe that a boy Mary every book given has (i) a > every; (ii) every > a b. Ik geloof dat een jongen [DP ieder BOEK] [DoC Marie tDP gegeven I believe that a boy every book Mary given heeft]. has (i) a > every; (ii) *every > a c. Ik geloof dat [DP ieder BOEK] [DoC een jongen Marie tDP gegeven I believe that every book a boy Mary given heeft]. has (i) a > every; (ii) every > a ‘I believe that a boy has given Mary every book.’

Striking confirmation of this account is provided by the dative construction in (42), where the surface position of the universal is its base position, rather than the landing site of A’-scrambling. The QIC therefore predicts the observed ambiguity of this example, despite its surface similarity to (41b). (42)

Ik geloof dat een jongen ieder BOEK aan Marie gegeven heeft]. I believe that a boy every book to Mary given has (i) a > every; (ii) every > a

Finally, the QIC explains an observation made by Sæbø (1997), who notices that a topical quantified object can take scope over a focused quantified subject, but a focused quantified object cannot take scope over a topical quantified subject. A topical quantifier is only felicitous in a discourse about quantities. For example, universal quantifiers normally resist topichood, but in a conversation about how many books various pupils have read, it is possible to say something like as for every book, only John 7 There are two complications regarding the judgment of (41b). First, some speakers disprefer scrambling of focused constituents to a position below the subject. Obviously, such speakers will classify the example as altogether ungrammatical. Second, another group of speakers marginally allow A-scrambling of a direct object across an indirect object. For such speakers, (41b) will be on a par with (41a) (and with (42)). The exact prediction made by the QIC, then, is that there will be no speakers who resist local A-scrambling in the double-object construction and for whom (41b) is ambiguous.

Chapter 2 Towards a Unified Encoding of Contrast and Scope

59

has read that much. The following examples should therefore be understood to be part of a discourse about how many members of various groups of attendants to a conference (students, staff, …) went to how many events (talks, meetings, wine receptions, …). By the tests introduced in the previous section, several (students) in (43) qualifies as a contrastive topic, while most (meetings) is a contrastive focus. In this example, the inverse scope reading is unavailable, in line with Sæbø’s generalization: the focused quantifier cannot take scope over the topical subject.8 (43)

Tell me about the STUDENTS. How many meetings did THEY attend? Well, I’m not sure about all students, but I think that several students attended MOST meetings. (i) several > most; (ii) *most > several

As opposed to (43), the example in (44) is ambiguous. The topical object may be interpreted with narrow scope, but alternatively it may take scope over the focused subject. (44)

Tell me about the MEETINGS. How many students attended THEM? Well, I’m not sure about all meetings, but I think that SEVERAL students attended most meetings. (i) several > most; (ii) most > several

The contrast between (43) and (44) can be understood as follows. In order for the focused object in (43) to take wide scope, it must percolate its scope index to a node dominating the topical subject. By the QIC, this implies that the domain of contrast of the object must include the subject. However, as explained in the previous section, this gives rise to an illicit information structure, in which the topic is in a position subordinate to the focus. Repair of this illicit configuration through covert scope extension by the topic is ruled out by the CSS, because the node dominating the topical subject is already the extended scope of the object. In (44), the object can take scope over the subject. This reading implies that the domain of contrast of the topic includes the focus, a configuration that satisfies information-structural well-formedness conditions. Of course, the QIC also allows surface scope for (44), as also on that reading 8 The examples in (43) and (44) are adapted from Sæbø’s paper. We have made modifications in the contexts in order to bring out more clearly which categories function as topics and which as foci.

60

Ad Neeleman & Hans van de Koot

the quantificational scope of most (meetings) is contained in its domain of contrast. Sæbø’s explanation of the pattern in (43) and (44) differs dramatically from ours. He denies the existence of a covert rule that extends quantifier scope and claims instead that scope shift is always parasitic on covert movement of a topic across a focus (which is triggered by informationstructural considerations). We believe that there are some serious objections to this analysis. As we have already seen in the examples in (37) and (38), the quantificational scope of a contrastive focus need not coincide with its domain of contrast. A similar picture emerges when we consider the quantificational scope of contrastive topics in more detail. Crucially, a topical quantifier can be scopally dependent on a focused quantifier, even though the latter must be contained in its domain of contrast (note that the QIC allows the quantifier scope of a category to be smaller than its contrastive scope). As far as we can tell the dependent reading of several (meetings) in (46) is as easy to get as it is in (45), even though the topic and focus functions in these examples are reversed. (45)

Tell me about the STUDENTS. How many meetings did THEY attend? Well, I’m not sure, but I think that most students attended SEVERAL meetings. (i) (*)several > most; (ii) most > several

(46)

Tell me about the MEETINGS. How many students attended THEM? Well, I’m not sure, but I think that MOST students attended several meetings. (i) several > most; (ii) most > several

The crucial reading of the examples in (45) and (46) for our argument is the one in which several is dependent on most. Notice, however, that an inverse scope reading also seems to be available for both examples. This is unsurprising for (46). But the apparent wide scope reading of several (meetings) in (45) cannot be the result of covert scope extension, as it would require a focus to percolate its index to a node dominating a topic. This would inevitably result in either an ill-formed information structure (if the topic does not extend its domain of contrast) or a violation of the CSS (if the topic does percolate its contrastive index). It must therefore arise as a consequence of either a specific reading of the indefinite or as a special subcase of the dependent reading (namely one in which the choice of meeting happens to be the same for every student). A second problem for Sæbø’s approach is the availability of scope extension for non-topical quantifiers. There are a number of contexts,

Chapter 2 Towards a Unified Encoding of Contrast and Scope

61

including relative clauses, that cannot contain topics (Maki et al. 1999, Haegeman 2002, Emonds 2004, Heycock 2006, and Bianchi and Frascarelli 2010), but that can contain quantifiers that undergo scope shift:9 (47)

The book that a boy has given to every girl in my class appears to be about spy planes. (i) a > every; (ii) every > a

The data presented above provide key evidence for the proposal developed in this chapter. We maintain that information-structural restrictions observed in A’-scrambling are scopal in nature and part of the same system that regulates the covert scope extension of quantifiers. This section has presented initial evidence supporting this claim. Interactions between quantifier scope and contrastive scope partly fall out from the CSS and partly from the QIC, a constraint that specifically mentions both contrastive scope and quantifier scope. Such interactions would be unexpected if quantifier scope and information structure were represented separately.

5.

Predictions of the CSS for quantifier scope

As mentioned earlier, our proposal does not restrict scope inversion in structures with two quantifiers (although there might be economy effects along the lines of Fox 2000). However, the consequences of the CSS are felt in sentences containing three quantifiers. Consider the structure in (48).

9 One might think that the example in (i) provides a counterexample to the claim that relative clauses cannot contain topics. (i) What about the book that Bill bought? How long did that burn for? Well, I don’t know about the book that Bill bought, but the book that John bought only burned for a few seconds. In this example, the constituents that are highlighted prosodically in the answer seem to be Bill and John. However, the actual contrastive topics are the book that Bill bought and the book that John bought. This is apparent from the leading question, which asks about books and not about people. Indeed, the answer in (ii) is odd. (ii) What about the book that Bill bought? How long did that burn for? #Well, I don’t know about Bill, but as for John, the book that he bought only burned for a few seconds.

62

Ad Neeleman & Hans van de Koot

(48)

α QP1

β QP2

γ QP3

δ

The surface scope of the quantifiers in this structure is QP1 > QP2 > QP3. The scope of QP2 and QP3 can be extended through index percolation, as explained above. If QP3 percolates its index to ȕ, the reading in (49b) is derived. If its index is percolated to Į, the resulting interpretation is that in (49d), while (49c) is derived if the index of QP2 targets this node. In the absence of further restrictions, we would expect two more readings in which both QP2 and QP3 percolate their index to a node dominating QP1. However, (49e) and (49f) are ruled out by the CSS, because they require that Į belongs to the extended scope of two quantifiers. (49)

a. [Į QP1 [ȕ QP2 [Ȗ QP3 į]]] QP1 > QP2 > QP3

d. [Į:3 QP1 [ȕ:3 QP2 [Ȗ:3 QP3 į]]] QP3 > QP1 > QP2

b. [Į QP1 [ȕ:3 QP2 [Ȗ:3 QP3 į]]] QP1 > QP3 > QP2

e. *[Į:2,3 QP1 [ȕ:2,3 QP2 [Ȗ:3 QP3 į]]] QP2 > QP3 > QP1

c. [Į:2 QP1 [ȕ:2 QP2 [Ȗ QP3 į]]] QP2 > QP1 > QP3

f.

*[Į:3,2 QP1 [ȕ:3,2 QP2 [Ȗ:3 QP3 į]]] QP3 > QP2 > QP1

Truth value judgment of sentences containing three quantifiers are notoriously difficult and we therefore intend to test our predictions experimentally in future work. However, we have been able to obtain initial native speaker intuitions and the discussion in the remainder of this section is based on these.

5.1.

The dative construction

Consider first the dative construction in (50). On its surface scope interpretation, given in (50a), the example fits a situation in which there is one teacher who makes a particular selection that contains most papers on the syllabus and gives this selection to every student. The reading in (50b) differs from the surface scope interpretation in that the teacher in question may make different selections consisting of most papers on the syllabus

Chapter 2 Towards a Unified Encoding of Contrast and Scope

63

and give every student one such selection. The reading in (50c) is like the surface scope reading in that there is a single selection of papers that gets given to every student, but each paper in that selection is handed out to every student by a (potentially) different teacher. The final available reading is given in (50d). It fits a situation in which for every student there is some teacher who gave that student a selection consisting of most papers on the syllabus. As in (50b), the set of papers can vary per student. (50)

Some teacher gave most papers on the syllabus to every student. a. some > most > every b. some > every > most c. most > some > every d. every > some > most e. * most > every > some f. * every > most > some

While the readings in (50a) through (50d) may differ in accessibility, we believe that they are genuine interpretations of the example. By contrast, the readings in (50e) and (50f) seem to be unavailable. To see this, let us construct situations under which these readings would be true. We assume that there are five papers, three students and five teachers. Consider first a situation compatible with the scope relations in (50e). One such situation involves a single selection of three papers, such that each paper in that selection is given to every student. Furthermore, each giving event may be initiated by a different teacher. As a result, we end up with a state of affairs in which teachers may vary per student and per paper:

paper-1

student-1 student-2 student-3

teacher-1 teacher-2 teacher-3

paper-2

student-1 student-2 student-3

teacher-4 teacher-5 teacher-1

paper-3

student-1 student-2 student-3

teacher-2 teacher-3 teacher-4

(51)

paper-4 paper-5

64

Ad Neeleman & Hans van de Koot

This situation can be described as “for most papers it is the case that every student was given the relevant paper by a possibly different teacher.” It is, however, not captured by any of the scope relations in (50a) through (50d). In (50a) through (50c), the choice of teacher cannot vary with the choice of student, while this is the case in (51). In (50d), the choice of teacher cannot vary with the choice of paper, but again this is what we find in (51). Given that the sentence in (50) cannot be used to describe the situation in (51), we must conclude that the reading in (50e) is unavailable. This is of course exactly what follows from the CSS, as (50e) can only be derived through percolation of multiple scope indices to the same node. We now turn to a situation compatible with the scope relations in (50f). In this case, every student receives a different selection of papers. Each paper in each selection is given to that student by a potentially different teacher. As before, we end up with a state of affairs in which teachers may vary per student and per paper. In addition, papers may vary per student:

student-1

paper-1 paper-2 paper-3

teacher-1 teacher-3 teacher-5

student-2

paper-4 paper-5 paper-1

teacher-2 teacher-4 teacher-5

student-3

paper-2 paper-3 paper-4

teacher-4 teacher-3 teacher-1

(52)

This situation can be described as “for every student it is the case that he or she was given most papers, but each paper by a possibly different teacher.” Again, it is incompatible with the scope relations in (50a) through (50d). As already explained above, in (50a) through (50c) the choice of teacher cannot vary with the choice of student, whereas in (50d) the choice of teacher cannot vary with the choice of paper. Given that the sentence in (50) cannot be used to describe the situation in (52), we must conclude that the reading in (50f) cannot be generated, as expected if the CSS holds.10

10 Some informants describe their experience with the readings in (50e,f) as follows. They can understand the situation or they can parse the sentence, but they are unable to connect the two.

Chapter 2 Towards a Unified Encoding of Contrast and Scope

5.2.

65

The double-object construction

Further corroboration of the hypothesis that scope extension is subject to the CSS comes from the double-object construction. There is a well-known observation that the scopes of the direct and indirect objects are frozen with respect to each other (see Bruening 2001 and others). This observation does not follow from the CSS and for this reason we will not discuss it in this chapter (but see footnotes 11 and 12 for some comments). However, when we consider scopal relations not affected by scope-freezing, we find exactly the same pattern of readings that we found with the dative construction: each object may take scope over the subject, but they may not do so simultaneously. The scope-freezing effect is responsible for (53) lacking any reading in which every paper on the syllabus takes scope over most students. This leaves us with the surface scope reading in (53a) and the scope-shifted readings in (53c) and (53e). (53)

Some teacher gave most students every paper on the syllabus. a. some > most > every b. * some > every > most (scope-freezing effect) c. most > some > every d. * every > some > most (scope-freezing effect) e. * most > every > some f. * every > most > some (scope-freezing effect)

The first two of these are indeed easily accessible. The scope relations in (53c) fit a situation in which for most students there is some teacher who gave them every paper on the syllabus. This reading should be distinguished from the unavailable reading in (53e), in which the choice of teacher does not only depend on the choice of student but also on the choice of paper. Assuming there are five students, three papers and five teachers, one situation described by (53e) can be depicted as below: (54)

student-1

paper-1 paper-2 paper-3

teacher-1 teacher-2 teacher-3

student-2

paper-1 paper-2 paper-3

teacher-4 teacher-5 teacher-1

student-3

paper-1 paper-2 paper-3

teacher-2 teacher-3 teacher-4

student-4 student-5

66

Ad Neeleman & Hans van de Koot

This state of affairs can be described as “For most students it is the case that every paper was given to him or her, but each paper by a possibly different teacher.” The example in (53) seems incompatible with the situation in (54), however, indicating that the reading in (53e) is not available. This cannot be due to some restriction on percolation of the scope index of the universal across the subject, given that such percolation is independently motivated by the ambiguity of (55). (55)

Some teacher gave Mary every paper on the syllabus. (i) some > every; (ii) every > some

In other words, both the direct and the indirect object may take scope over the subject, but it is not possible for both of these arguments to have the subject in their scope. This is of course exactly what the CSS predicts. The unavailability of the reading in (53e) has been observed previously by Shoichi Takahashi, who gives it a single question mark in his 2003 NELS paper but indicates in personal communication that it is worse than that and deserves at least two question marks. His suggestion is that it may be difficult for a quantifier to take narrow scope relative to two quantifiers it c-commands in the surface representation, something that would follow directly from the CSS.11,12,13

11 Bruening (2001) does not give examples in which a quantified subject takes scope below two quantified objects, even though his analysis predicts the existence of such a reading. Sauerland (2000) and Lechner (2009) suggest that such a reading exists for the example in (i). (i) Two boys gave every girl a flower. However, one of the readings independently available for this example has the scopal relations  > 2 > . This reading subsumes the one in which the subject would scope below the two objects, namely for the special case that the same flower is selected for each boy in the pair of boys associated with a girl. For this reason, we have used examples in which the subject can in principle be interpreted as dependent on the direct object. If we adjust the example in (i) accordingly, as in (ii), the low reading of the subject ( > most > 2) turns out to be unavailable, although there is an intermediate reading ( > 2 > most). (ii) Two boys gave every girl most flowers. The low reading can be paraphrased as follows: for every girl it is true of most of the flowers she received that they were given to her by a potentially different pair of boys. The intermediate reading is subtly different as the pair of boys cannot vary by flower.

Chapter 2 Towards a Unified Encoding of Contrast and Scope

5.3.

67

The CSS does not ban all multiple scope shifts

Of course, scope inversion in general is marked. One might therefore speculate that the unavailability of the readings in (50e), (50f) and (53e) is due the fact that they rely on two applications of a marked operation (two quantifiers must percolate their scope index across the subject)12. How12 One observation often construed as an argument for the hypothesis that all quantifiers undergo movement at LF is the scope-freezing effect found in the double object construction (see Bruening 2001 and Lechner 2009). The basic idea in Bruening’s work is that quantifier raising maintains the order of arguments and hence does not lead to scope shift. Instead, variation in the scope of objects with respect to the subject results from reconstruction of the subject below the LF positions of the objects. By assumption, the semantic type of quantifiers is incompatible with in situ interpretation of quantified objects, so that reconstruction of objects is ruled out. We are not persuaded by this line of argumentation because scope-freezing effects are not only found in double object constructions but also with certain monotransitive predicates. For example, a wide-scope reading of the universal quantifier seems unavailable in (i)-(iii) (comparable examples can be found in Gruber 1965). (i) A box contained every present. (ii) An Australian wine received every prize. (iii) At least one student possessed every book. We do not think that the type of analysis proposed by Bruening and Lechner can be extended to the scope-freezing effect observed in these examples. There is no reason why reconstruction of the subject should be blocked. We speculate that what may be at stake is an interaction between argument structure and quantifier raising, more or less along the lines of Williams (2006). Suppose that the unmarked realization of a ș-grid containing a theme and what one might call an anchor (goal, beneficiary, location, etc.) is such that the theme is projected higher than the anchor. It is possible to produce marked structures in which this order is reversed. The generalization over such marked orders, which include the examples in (i)-(iii) and the double object construction, is that they resist scope shift. Working out this proposal is beyond the scope of the present article. 13 The CSS implies that the interpretation of WH-expressions that remain in situ in English is not achieved through index percolation but rather through a mechanism akin to unselective binding, as argued by Reinhart (1998). This process requires the presence of a QP that takes scope over the clause and on which lower WH-phrases can be parasitic. As Reinhart notes, this analysis straightforwardly explains why in English WH-in-situ is not subject to conditions on QR (or index percolation, on our proposal). Following Ackema and

68

Ad Neeleman & Hans van de Koot

ever, this cannot be true. There are examples in which two scope reversals are possible without this giving rise to the interpretive difficulties associated with the multiple scope shifts in (50e,f) and (53e). Such multiple shifts reversals are attested exactly where a derivation is available that respects the CSS. There are three cases to consider. The first two involve covert scope shift and will be dealt with in section 5.3.1. The third involves scope marking by overt movement and will be discussed in section 5.3.2.

5.3.1. Multiple covert scope shifts Multiple covert scope shifts are predicted to be possible if the paths of index percolation do not overlap. To begin with, consider (56). Here the universal indirect object can take scope over the existential subject, while at the same time the scope between the direct object and a sentence-final adverbial may be inverted. Whatever one’s view of the structure of the English VP, one of the examples in (56) will require two non-overlapping scope shifts. As predicted, the relevant readings are very easily accessible. (56)

a. A nurse gave every patient an injection on every day of the week. b. A nurse gave every patient each of the ten required injections on a different day of the week.

Although the examples in (56) may seem trivial, it is important to note that the ease with which the relevant readings can be computed makes it implausible to attribute the inaccessibility of the readings in (50e,f) either to the number of quantifiers in the sentence or the number of scope shifts required to generate these readings. What is crucial is that the readings in (50e,f) and (53e) results from an interaction between three quantifiers that requires overlapping paths of index percolation. A second, more intricate case of multiple covert scope shift allowed by the CSS is provided by the total scope reversal found in certain cases of inverse linking. Inverse linking is the term used to describe the scope Neeleman (1998), we suggest that, in languages in which all WH-expressions remain in situ, one WH-expression extends its scope, thereby making it possible for the other WH-expressions to be unselectively bound. This explains Watanabe’s (1992) observation that subjacency effects are systematically observed in single questions, whereas, in multiple questions, WH-expressions can surface in XPs that are islands for QR, as long as there is at least one WH-expression external to the island.

Chapter 2 Towards a Unified Encoding of Contrast and Scope

69

reversal in examples like (57), where the indefinite depends on the universal. (57)

One apple in every basket was rotten.

In early work, May (1977) argued that inverse linking is achieved through movement of the universal out of the containing indefinite DP to a landing site at the edge of the clause. In later work, however, he and others rejected this approach in favour of an analysis in which the universal adjoins to the DP that contains it, essentially to account for the absence of the reading in (58c), as discussed below (see May 1985, Rooth 1985, Larsson 1985 and Barker 2001, among many others).14 (58)

Two politicians spy on someone from every city. a. 2 > every > someone b. every > someone > 2 c. * every > 2 > someone

On the index-based theory of scope shift, this latter analysis translates into a structure containing two index percolation paths that do not overlap. The universal quantifier every city (QP3 in (59)) percolates its index to the top node of the containing indefinite quantifier someone from every city (QP2). By the interpretive rule for scope percolation, this means that QP3’s scope consists of QP2 minus QP3 itself. If no other scope shift takes place, this generates the reading in (58a). The reading in (58b) results from the additional percolation of the index of QP2 to a node dominating two politicians (QP1 in (59)). (The reading in (58c) is excluded if a DP is an island for scope percolation.) α:2

(59) QP1

γ:2 δ

QP2:3 Q

QP3

14 There is work, in particular by Sauerland (2005), that challenges the idea that DP is a scope island. We turn to this issue at the end of the section.

70

Ad Neeleman & Hans van de Koot

Note that although QP2 has two indices, its own index and the index it inherits from QP3, it does not violate the CSS, which bans multiple index percolation but does not prevent a node from having more than one scope index. Put differently, the CSS makes it impossible for a node to be part of the extended scope of more than one quantifier, but this is not true of QP2. QP2 is part of the extended scope of QP3 but its other scope index is there to encode its own scope. Strictly speaking the scope rule in (5c), repeated here as (60a), only assigns a partial scope interpretation to (59). The DP every city has scope over the DP [someone from every city] (interpreted as someone from x). Similarly, the DP [someone from every city] takes scope over two politicians. However, the structure does not seem to encode a scope relation between every city and two politicians. We can fix this by subjecting the mapping between LF and semantics to the transitivity condition in (60b).15 (For related discussion, see Rullmann 1988; it is perhaps worthwhile noting that transitivity conditions are a familiar property of other mapping systems, for example the mapping from syntactic structures to strings.) (60)

a. Default Scope Rule If a quantifier does not percolate its index, it takes scope over its c-command domain. b. Scope Transitivity If Į takes scope over ȕ and ȕ takes scope over Ȗ, then Į takes scope over Ȗ.

The transitivity condition has the consequence that the two scope shifts encoded in (59) give rise to an interpretation in which the universal takes scope over two politicians. The mapping principles in (60) provide a natural solution to a longstanding problem in inverse-linking structures, namely the availability of variable binding in structures like (61), where the quantifier fails to c-command the pronoun at Spell-Out and also appears not to take scope over it at LF. In a theory based on QR, the universal adjoins to DP, where it fails to c-command the pronoun. In a theory based on index percolation there is a comparable problem, as the index of the universal fails to percolate to a node dominating the pronoun.

15 Greville Corbett (p.c.) suggest that alternatively one could adopt the convention that a percolating index may carry along an index that it itself has inherited.

Chapter 2 Towards a Unified Encoding of Contrast and Scope (61)

71

[IP [DP2 : 1 someone from [DP1 every city]] despises it]

However, it follows from the transitivity condition in (60b) that the universal does take scope over the pronoun, as long as it takes scope over the indefinite. This is so because the indefinite itself uncontroversially takes scope over the pronoun. Therefore, percolation of the index of the universal to the root node of the subject is sufficient for the bound variable reading to become available. It follows from this account that the bound variable reading will be lost if the universal does not take scope over the indefinite. It has indeed been observed by Larson that inverse linking is a prerequisite for variable binding in examples like (61). We can illustrate this using the data in (62). Whereas (62a) is ambiguous, the variable binding in (62b) has the consequence that the narrow scope of the universal with respect to the indefinite is no longer available. (62)

a. [At least one reader of [every novel in the library]] also likes classical music. (i) at least one > every; (ii) every > at least one b. [At least one reader of [every novel in the library]1] hates it1. (i) *at least one > every; (ii) every > at least one

For completeness sake, we give the data in (63), which show that total scope reversal is available not just with the quantifiers in (58), but also with those used in the earlier example in (50): (63)

Some politician spies on most inhabitants from every city. a. some > every > most b. every > most > some c. *every > some > most

The above implies that the contrast between the ease of total scope reversal in examples like (58) and (63) and its impossibility in (50) requires a structural account. Such an account is provided here: whereas the scope reversal in (58) and (63) can be generated without violating the CSS, the scope reversal in (50) would require overlapping paths of index percolation. The argumentation in this section is based on Larson’s claim that DP is a scope island. As already mentioned, this claim receives support from the unavailability of the reading in (58c). If it were possible for a universal quantifier to extend its scope across a DP node, this reading could easily

72

Ad Neeleman & Hans van de Koot

be generated. Its absence therefore suggests that such scope extension is not possible and hence that total scope reversal in (58) must rely on two separate scope shifts that interact through the transitivity condition in (60b). In recent work, Sauerland (2005) has cast doubt on Larson’s generalization concerning the absence of readings like (58c), arguing that intensional verbs and certain other elements can take scope between the universal and the indefinite DP in which the universal originates. However, Sauerland’s conclusions have been challenged, successfully in our view, by Charlow (2009). Our assessment of the existing literature, then, is that there is a strong case for the derivation required by the CSS. This case is further strengthened by the fact that the example in (64b), as opposed to that in (64a), is unambiguous (these examples are adapted from Koster 1991; see also Rullman 1988). (64)

a. Two Protestants spy on the father of every Catholic. (i) two > every; (ii) every > two b. Two Protestants spy on the Holy Father of every Catholic. (i) two > every; (ii) *every > two

In our terms, the wide scope reading of every Catholic in (64a) is the result of two operations of index percolation. First, the index of every Catholic percolates to the DP that contains it. Normally, DPs introduced by the do not bear a scope index, as they denote a unique individual. However, once in the scope of the universal every Catholic, the entire DP denotes a set of individuals and thereby acquires quantificational properties licensing a new scope index on this DP. This second index may percolate to a node dominating two Protestants, and if it does, scope transitivity gives rise to a reading in which for every Catholic it is true that the father of that Catholic is spied upon by a (potentially different) pair of Protestants. The logic of this analysis implies that if the first operation of index percolation is blocked, it will be impossible for every Catholic to take scope over two Protestants. This prediction is confirmed by the fact that the example in (64b) is unambiguous. The relevant operation of index percolation would give every Catholic scope over the Holy Father of x. However, this scope extension is ruled out by Economy, because it does not yield an interpretation that is unavailable if the universal takes surface scope. Without scope extension there can be only one Holy Father for all Catholics. With scope extension, a reading must be chosen in which every Catholic is associated with the same Holy Father (in view of our knowledge of the world). But these two readings are indistinguishable and

Chapter 2 Towards a Unified Encoding of Contrast and Scope

73

therefore scope extension is blocked by Economy. This means that the containing DP will not be able to function as a quantifier, so that it becomes impossible for the universal quantifier to piggy-back on scope extension of this DP.16 All this of course presupposes that DPs are scope islands. If this were not the case, every Catholic in (64b) could directly extend its scope beyond the DP it is contained in, so that it should be possible for two Protestants to depend on it, contrary to fact. One might hypothesize that the unavailability of the wide scope reading of the universal in (64b) has its source in the fact that the possessor of Holy Father must be a variable bound by the universal. Such a relation of variable binding might be unexpected, given that there is only one Holy Father. However, an explanation along these lines cannot be maintained in view of the grammaticality of (65), where there is a perfectly coherent reading on which his varies with the choice of Catholic priest but his Holy Father always denotes the same individual. (65)

That evening, every Catholic priest watched a video of his Holy Father.

5.3.2. Scope-marking by movement So far we have looked at covert scope shift, but many languages allow scope marking through overt movement. Recall that if a category is moved to mark its scope, it percolates a scope index to the node that immediately dominates the landing site. Since scope-marking movement does not involve index percolation from the launching site, overt scope-marking is predicted to be freer than covert scope-shift (compare the discussion around (7)). In particular, while an in situ quantifier cannot percolate its scope index past the landing site of a moved scope-taking element, it can be fronted across that element. As a consequence, readings can be overtly encoded that are not derivable through index percolation. Thus, the in situ structure in (48) does not permit the reading QP3 > QP2 > QP1, but the movement structure in (66) does. (The same is true, mutatis mutandis, for the reading QP2 > QP3 > QP1.) 16 The same line of argumentation explains the contrast between (i) and (ii) (for related discussion see Rullman 1988 and Koster 1991). (i) The father of every Catholic boy thinks that he is smart. (ii) * The Holy Father of every Catholic boy thinks that he is smart.

74

Ad Neeleman & Hans van de Koot

(66)

α:3 QP3

γ:2 QP2

ε QP1

η tQP2

ι tQP3

λ

This prediction is borne out. In German, Japanese and other languages that allow or require overt scope marking, it is possible to move two quantified objects across the subject giving rise to surface scope. Some German examples illustrating this are given in (67) and (68) (see Kiss 2001 for extensive discussion). (Notice that this observation parallels an observation made earlier, namely that movement of a contrastive topic across a moved contrastive focus is possible, but covert scope extension of the topic is not.) (67)

Ich glaube … ‘I believe …’ dass jedem Studenten die meisten Artikel irgendein Professor that every student the most papers some professor gegeben hat. given has every > most > some ‘that some professor has given most articles to every student.’

(68)

Ich glaube … ‘I believe …’ dass jeden Artikel den meisten Studenten irgendein Professor that every article the most students some professor gegeben hat. given has every > most > some ‘that some professor has given most students every article.’

To sum up, we have seen in this section that assignment of quantifier scope is subject to the CSS. In particular, multiple scope shift is ruled out where it requires a single node to inherit two or more scope indices, but allowed

Chapter 2 Towards a Unified Encoding of Contrast and Scope

75

where this is not the case. This demonstrates that the CSS is a condition on quantifier scope, as much as it is a condition on the scope of contrastive categories. In conjunction with the results of the previous section, this confirms our claim that restrictions on the placement of topics and foci and restrictions on quantifier scope belong to one and the same system.

6.

Conclusions

In this chapter we have investigated whether contrastive scope can be considered a kind of quantifier scope. This line of inquiry was suggested by the semantic analysis of contrast, as developed in the introduction, according to which contrast is associated with negative quantification. We have seen that a unification of the theories of quantifier scope and contrastive scope is indeed feasible. It was demonstrated not only that scope extension of either type obeys the CSS, but also that this condition is instrumental in explaining subtle interactions between quantifier scope and information structure. The CSS is incompatible with the standard view that every quantifier undergoes covert movement to a position from which it c-commands its scope domain. The proposal advanced here therefore also provides support for the view that LF is representation of relative scope only. This implies that the research reported here confirms the approach by Reinhart (2006) and others, according to which covert scope extension is a marked operation whose application is limited to scope inversion. The resulting theory is able to draw a distinction between overt scope marking by movement and covert scope extension through index percolation, with the latter more restricted in application than the former. The asymmetry between overt and covert scope extension has allowed us to resolve several issues that have remained elusive in the literature on scope. Most importantly in the context of this book, it has made it possible to provide an explanation of why a structure in which a topic is located in the c-command domain of a moved focus cannot be repaired by covert scope extension of the topic, whereas it can be repaired by topic movement.

Chapter 3 Word order variation and information structure in Japanese and Korean* Reiko Vermeulen 1.

Introduction

Much work has been devoted to the grammatical properties of discourserelated notions such as topic and focus in Japanese and Korean (Kuroda 1965, 2005, Kuno 1973, Saito 1985, Hoji 1985, Heycock 2008, Miyagawa 2006, Kishimoto 2009, Tomioka 2010, a. o. for Japanese; Choe 1995, Choi 1999, Han 1998, C. Lee 2003, 2006, M. Lee 2006, a. o. for Korean). However, although the two languages are strikingly similar in a number of respects, relatively little attention has been paid to the similarities and differences that exist between the two languages. This chapter systematically compares the syntactic distribution of topic and focus, both contrastive and non-contrastive types, in the two languages and offers a uniform account. The account is couched within the general interface-based approach proposed in Chapter 1 of this book. In particular, I argue that Japanese and Korean display an additional pattern of cross-linguistic variation that is expected on the typology in (1), which was introduced in Chapter 1.

* Earlier versions of parts of this paper were presented at GLOW 34, Japanese/ Korean Linguistics 20, Workshop on Formal Altaic Linguistics 7, and On Linguistic Interfaces II in 2011, Colloquium on Generative Grammar in 2010, and at the University of Potsdam in 2009. I thank the participants for their helpful comments. I would also like to thank Ad Neeleman, Kriszta SzendrĘi, and Michael Wagner for their insightful comments and discussions. Thanks are also due to my informants for their immeasurably patient help. Parts of this research have been funded by the FWO (No. G091409), the AHRC (No. 119403) and the British Academy (No. SG50500).

78 (1)

Reiko Vermeulen Syntactic typology of topic, focus and contrast

contrastive

Topic

Focus

non-contrastive topic [topic]

non-contrastive focus [focus]

contrastive topic [topic] [contrast]

contrastive focus [focus] [contrast]

Recall that the main motivation for the typology comes from the observation that languages show cross-cutting generalisations over grammatical properties of items sharing one of the notions [topic], [focus] and [contrast]. Some of these generalisations are syntactic.1 Contrastive topics and contrastive foci undergo A’-movement in English and Dutch. This observation can be explained straightforwardly if we assume that the three discourse notions can be targeted by mapping rules operating between syntax and information structure, and that contrastive topics and contrastive foci are composites of the features [contrast] and [topic], and [contrast] and [focus], respectively. In English and Dutch, [contrast] licences A’-movement. Notice that the typology does not, in principle, preclude a situation where a language has distinct mapping rules relevant for different notions. However, if it has a rule relevant for [topic] and another rule relevant for [contrast], for instance, a conflict potentially arises for contrastive topics. If the two rules cannot be satisfied simultaneously, it is not immediately obvious which rule should apply to contrastive topics. I propose in this chapter that there is parametric variation as to which rule is adopted in such a situation. This predicts that if a language has conflicting rules for [topic] and [contrast], contrastive topics in that language would systematically behave either like non-contrastive topics or contrastive foci with respect to the mapping rules. I demonstrate that Japanese and Korean bear out these situations precisely. The two languages have the same rule relevant for [topic] and the same rule relevant for [contrast], but the two rules cannot be satisfied simultaneously. Specifically, the rule for [topic] states that a topic must be licensed in clause-initial position, while the rule for [contrast] states that a contrastive item may undergo scrambling to a 1 Following the practice in Chapter 1, I will sometimes refer to these notions as features. This is simply for a matter of presentational convenience. It will not make a difference to the arguments presented here if they are discourse notions that are not visible to syntax, or are represented as syntactic features that syntax-internal operations may target.

Chapter 3 Word order variation and information structure

79

variety of positions. For contrastive topics, Japanese adopts the rule relevant for [topic] and Korean adopts the one relevant for [contrast]. Thus, Japanese contrastive topics are restricted to clause-initial position, while those in Korean can occupy a variety of positions. In considering the distribution of contrastive topics, I will also examine the meaning associated with so-called contrastive wa in Japanese and contrastive nun in Korean, which are particles that are suffixed to contrastive topics. I argue that the meaning is in fact compatible with both topic and focus, and whether a particular contrastive wa- or nun-marked phrase is a topic or a focus is determined by discourse. In Japanese, the distinction is also marked syntactically: one that functions as a topic must move to clause-initial position, while one that functions as a focus need not. Further syntactic evidence for this claim comes from examining certain predictions concerning the distribution of contrastive topics and contrastive foci with respect to each other. The predictions derive from the combination of the mapping rule I will propose for [contrast], and independent, well-formedness conditions in the information structural component. Contrastive wa- and nun-marked phrases that I argue function as contrastive topics show the predicted distribution for contrastive topics, but those that I argue are foci do not. I adopt the notions topic, focus and contrast defined in Chapter 1, which I will briefly review in Section 2, explaining how these notions are morphologically realised in Japanese and Korean. Section 3 examines the syntactic distribution of non-contrastive topics and shows that the two languages share the same mapping rule relevant for [topic]. In Section 4, I argue that contrastive focus in the two languages is subject to the same rule relevant for [contrast]. I show that the kind of scrambling that a contrastive item may undergo is not licensed by the feature [focus], but rather by [contrast]. Section 5 investigates the distribution of contrastive topics in the two languages. Section 6 shows that the predictions concerning variation in word order between contrastive topic and contrastive focus are borne out. In Section 7, I examine the meaning of contrastive wa in Japanese and contrastive nun in Korean, and provide evidence that contrastive wa- and nun-phrases that function as foci do not exhibit the syntactic behaviour predicted for contrastive topics discussed in Section 6. Section 8 concludes the chapter.

80

Reiko Vermeulen

2.

Discourse notions

2.1

Topic

I follow Reinhart (1981) in characterising topics in terms of aboutness. One can see the mere existence of expressions like as for, about, regarding, concerning, and so on, as evidence for the existence and the linguistic relevance of aboutness. Moreover, speakers generally have intuitions regarding what a given sentence is about (Reinhart 1981, 1995; Endriss 2009). Yet, it is notoriously difficult to pin down the exact content of the notion and how it is linguistically relevant.2 This state of affairs is reflected by the variety of definitions of topic offered in the literature (Chafe 1976, Reinhart 1981, Givón 1983, Vallduví 1992, Lambrecht 1994). There is some consensus, however, that it is important to distinguish between the topic of a unit of discourse and the syntactic constituent used to introduce the referent as what the sentence is about. This referent typically also functions as the topic of discourse and may continue to do so in the subsequent discourse (Givón 1983). Following Reinhart (1981), I will refer to a topic in the first sense as a ‘discourse topic’ and to a topic in the second sense as a ‘sentence topic’, or simply a ‘topic’ when the distinction is clear. As far as topics are concerned, this chapter deals mainly with the syntactic behaviour of sentence topics. The discussion of discourse topics will be limited to what is necessary to understand their opposition to a sentence topic. Sentence topics are variously known as ‘chain-initial topic’ (Givón 1983), ‘link’ (Vallduví 1992), ‘aboutness topic’ (Frascarelli and Hinterhölzl’s 2007). A sentence topic can be identified as the item X in the reply to requests such as tell me about X. Such a request explicitly instructs the hearer to give a response that encodes X as the sentence topic. Thus, John in Speaker B’s utterance in the exchange in (2) is a sentence topic. In an exchange such as (2), no obvious contrast is implied between John and any other individual. I will therefore refer to such sentence topics as ‘non-contrastive’ topics. (Sentence topics are indicated by double-underlining throughout this chapter.)

2 Portner and Yabushita (1998, 2001) propose a formal account of aboutness. Their account, however, does not distinguish sentence topics and those that refer back to discourse topics, discussed immediately below. The syntactic generalisations discussed in later sections are therefore difficult to capture on this definition.

Chapter 3 Word order variation and information structure (2)

81

A: Tell me about John. B: John is a student from Canada.

The fact that John in B’s utterance in (2) is indeed a sentence topic introducing its referent as the discourse topic, rather than John in A’s request, is suggested by three facts. First, the sentence uttered by A is not about John: it does not provide any new information about John. Second, B’s utterance is felicitous even if the request is less specific about what is to be the topic of discourse, such as tell me about someone in your class. Finally, a well-known property of a sentence topic is that it must be referential (Reinhart 1981). The fact that a request such as tell me about someone in your class is felicitous, where the target of the request is not specific, shows that the constituent X in tell me about X is not a sentence topic. It is also important to distinguish a sentence topic from an item that refers back to it (and hence to the discourse topic). The point is illustrated by the following exchange. (3)

A: Who did Max see yesterday? B: He saw Rosa yesterday.

Uttered discourse-initially, Max in A’s question in (3) is most typically interpreted as a sentence topic, introducing Max as the topic of this discourse. (A definite, human subject is very likely to be construed as a topic (Givón 1976).) The pronoun he in B’s reply in (3), on the other hand, is not a sentence topic. Its referent is indeed what the rest of the sentence is about, but the referent is not newly introduced by it. Rather, he is simply a discourse anaphoric item that refers back to the referent of the sentence topic Max and hence also the discourse topic Max. Thus, B’s sentence is interpreted as being about the referent of he, because he refers back to the referent that the discourse is about, not because he functions as a sentence topic. (See also Vallduví 1992, Lambrecht 1994, and Vallduví & Engdahl 1996 for the importance of this distinction.) As proposed in Chapter 1, a topic can be represented as a triplet consisting of a function, the topic and a set of alternatives to the topic. Thus, B’s utterance in (2) has the representation in (4). (4)

Topic is an utterance-level notion and the function contains an assertion operator. The complement of the assertion operator represents the comment of the topic. Applying the function to the topic generates an asser-

82

Reiko Vermeulen

tion whose propositional content is the ordinary value of the sentence. Applying it to members of the set of alternatives to the topic generates a set of alternative utterances. The triplet thus represents the intuition that when the speaker B in (2) utters the sentence, he or she performs the following speech acts: (i) Consider John (out of a set of possible topics); (ii) I assert that John is a student from Canada. The intuition that John is selected out of a set of possible topics might be more clearly felt, if the request is something like tell me about someone in your class. However, I maintain that the set of alternatives is part of the meaning of a topic even in instances that a particular topic has been suggested before uttering the sentence, as in the case of (2). In Japanese, non-contrastive topics are marked by the particle wa. In Vermeulen (2011), I argue that wa marks items that refer to discourse topics, that is, non-contrastive sentence topics, as well as items that refer back to discourse topics, such as he in (3). I show that non-contrastive topics and items that refer back to discourse topics exhibit different syntactic behaviour. I will not discuss this distinction in this chapter. Korean non-contrastive topics are typically marked by the particle nun, but can be marked by a case marker. Choi (1999) claims that nun on a clause-initial phrase specifically marks items with the function of ‘link’ in Vallduví’s (1992) terminology, (which is sufficiently similar to the notion of topic adopted in this chapter for the current purpose,) while case markers are neutral with respect to the discourse function of the host item. Thus, the latter can appear on non-contrastive topics too. Moreover, unlike wa in Japanese, nun cannot mark items that refer back to discourse topics. (See Shimojo & Choi 2001 for further comparison between wa and nun). As we will see below, contrastive topics are also marked by wa and nun in Japanese and Korean, respectively. However, for Japanese, it is generally assumed that wa that attaches to non-contrastive topics and wa that appears on contrastive topics are distinct lexical items, as they have different sets of syntactic, prosodic and interpretive properties. For example, a phrase marked by the latter, but not the former, is tonally prominent, shows movement properties and can appear on predicates (Kuno 1973, Kuroda 1979, 2005, Hara 2006b, Oshima 2008, pace Kuroda 1965, 1992, Shibatani 1990, Tomioka 2010). I adopt the general assumption that they are distinct lexical items. For Korean, the standard assumption is that nun which attaches to non-contrastive topics and nun that attaches to contrastive topics are one and the same lexical item (Choi 1997, 1999, Han 1998, Gill & Tsoulas 2004, Y. Lee 2005, M. Lee 2006, Oh 2007). However, there is overwhelming evidence that nun should also be treated as

Chapter 3 Word order variation and information structure

83

two distinct lexical items. Nun that appears on a contrastive topic, like its Japanese counterpart, bears tone prominence, can appear on a variety of categories, including predicates, and its host shows syntactic properties associated with movement, such as sensitivity to island and weak crossover effects. Nun that appears on non-contrastive topics, on the other hand, does not display these properties (Choe 1995, Choi 1999, C. Lee 2006). I assume therefore that nun that appears on non-contrastive topics and nun that attaches to contrastive topics are also two distinct lexical items. I will call the type that appears on non-contrastive topics ‘noncontrastive wa/nun’ and the type that attaches to contrastive topics ‘contrastive wa/nun’. Contrastive wa and nun will be examined in detail in Section 7.

2.2

Focus

I adopt the widely held view that a focus provides a highlighted piece of information with respect to the rest of the sentence. As such, it can be identified as the item that corresponds to the wh-part of a preceding question. Thus, Rosa in B’s utterance in (3) is a focus, and the remaining items in this particular instance constitute its background. In an exchange such as (3), no obvious contrast is implied between Rosa and any other individual. I will therefore refer to this kind of focus as ‘non-contrastive’ focus. Following Rooth (1985, 1992), I assume furthermore that focus is associated with a set of alternative propositions. Like topic, focus can be represented as a triplet consisting of a function, the focus and a set of alternatives to the focus. The focus interpretation of Rosa in B’s utterance in (3) can be represented as in (5). The difference with topic is that focus is a notion relevant to a proposition. Thus, the function does not contain an utterance-level operator and application of the function to the focus derives a plain proposition, the ordinary value of the sentence. Its application to the alternatives to the focus will generate a set of alternative propositions to the ordinary value. This set of alternative propositions and the ordinary value together correspond to what is generally referred to as the focus value of the sentence. (5)

In Japanese and Korean, non-contrastive foci typically bear case markers.

84 2.3

Reiko Vermeulen

Contrast

Contrast is a notion that combines with topic or focus. Contrast implies the negation of (at least) one of the alternative utterances or propositions generated by application of the function in the triplet to members of the set of alternatives to the topic or focus, respectively. The interpretive effects of contrast are different for topic and focus. A contrastive topic implies that the speaker is unwilling to make the same assertion regarding an alternative to the topic. Topic being an utterance-level notion, the reason for the unwillingness could be pragmatic. Perhaps, the speaker is unsure about the truth of the proposition asserted by the utterance, or the speaker wishes not to share the relevant information regarding the alternative topic. On the other hand, since focus is a propositional notion, a contrastive focus implies that the truth value of an alternative proposition is false (Tomioka 2010). The interpretations are illustrated with examples below. (6) is a typical discourse context which requires a contrastive topic. A’s question is about Bill, but B’s response provides information about John instead. It implies that he or she is unwilling to make the same assertion about Bill. Thus, John in B’s utterance in (6) is a topic because it newly introduces its referent as what the sentence is about and it is contrastive because it implies the negation of (at least) one alternative utterance ‘the assertion that Bill bought a hat’ being the most salient candidate in this case. A contrastive topic usually bears the so-called B-accent in English (Jackendoff 1972, Büring 2003). (Contrast is indicated by bold here and below). (6)

A: Did Bill buy a hat? B: Well, John bought a hat.

The interpretation of John as a contrastive topic can be represented as in (7). (7a) is the meaning of John as the topic of the sentence and (7b) is the additional contrastive interpretation, where the negation of (at least) one alternative utterance is stated. (See Hara 2006a on how negation, a propostional level operator, may operate on a speech act.) (7)

a. b. y [y  {Bill, Mary, …} & ™(Oy ASSERT [y bought a hat])]

A typical context that requires a contrastive focus is an instance of correction, such as (8) (Repp 2009). Rosa in B’s utterance in (8) is a focus, as it provides a highlighted piece of information with respect to the back-

Chapter 3 Word order variation and information structure

85

grounded material, expressed by the rest of the sentence, ‘John saw x yesterday’. Its contrastive interpretation is made clear by the explicit denial of a proposition containing the alternative, Claire, to the focus. (Focus is indicated by SMALL CAPS here and below.) (8)

A: John saw Claire yesterday. B: That’s not true. He saw ROSA yesterday.

Another context which requires a contrastive focus is provided by a disjunctive question. In the following exchange, Rosa in B’s utterance is a contrastive focus. The explicit mention of the alternative Claire in the question makes the contrastive reading of Rosa in the reply clear: the reply implies that the relevant alternative proposition, ‘John saw Claire yesterday’, is false. The contrastive reading can be further strengthened by the use of the focus-sensitive particle only (see Chapter 7 for discussion on differences between contrastive focus and only.) (9)

A: Did John see Claire or Rosa yesterday? B: He saw (only) ROSA yesterday.

The interpretation of Rosa as a contrastive focus in B’s utterances in (8) and (9) can be represented as in (10). (10a) is the meaning of Rosa as the focus, and (10b) is the additional contrastive interpretation, which includes negation of (at least) one alternative proposition, ‘John saw Claire yesterday’ being the most salient candidate in the two contexts. (10)

a. b. y [y  {Claire, Jennifer, Jason, …} & ™[John saw y]]

As mentioned above, contrastive topics in Japanese and Korean are marked by contrastive wa and contrastive nun, respectively. Contrastive topics are examined in Section 5. However, as I will argue in Section 7 contrastive wa and nun do not only mark contrastive topics and therefore are not markers of contrastive topic. Contrastive foci identified in contexts such as (8) and (9) are marked by case markers, as we will see in Section 4. In the following sections, I will examine the syntactic distribution of non-contrastive and contrastive types of topics and foci in Japanese and Korean as identified by the reelvant discourse contexts discussed above.

86 3.

Reiko Vermeulen

[Topic] and non-contrastive topic

In this section, I argue that the following rule is operative in Japanese and Korean. (See also Vermeulen (2009, 2011)). (11)

[Topic] is licensed in clause-initial position.

(11) is very much in line with what is generally assumed in the literature for both languages. Below, I demonstrate with the appropriate discourse context that the general assumption is correct. In most examples provided in this chapter, the object, rather than the subject, is either topic or focus. This is to demonstrate most clearly the non-canonical placement of the relevant item. The same observations obtain for subjects. The following exchange illustrates the point for Japanese. In replying to the request in (12), the object, marked with the particle wa, ano boosiwa ‘that hat-wa’, must appear in clause-initial position, as in (13a). A reply in which the wa-phrase occupies a non-clause-initial position, as in (13b), is infelicitous. (12)

ano boosi-nituite nanika osiete-kudasai. that hat-about something tell-give ‘Tell me something about that hat.’

(13)

a. ano boosi-wai John-ga kinoo ei that hat-WA John-NOM yesterday b. #John-ga ano boosi-wa kinoo John-NOM that hat-WA yesterday ‘John bought that hat yesterday.’

(J)

katta. bought katta.3 bought

3 For reasons not entirely clear to me, a wa- or nun-marked phrase sometimes prefers not to surface adjacent to the verb. This is perhaps due to the phonological properties of wa/nun-phrases and verbs in these languages: a non-contrastive wa/nun-phrase forms a separate prosodic domain, but the verb typically forms a domain with an immediately preceding phrase (Nagahara 1994, Nakanishi 2001, 2007, Jun 1993). An adverbial is inserted to circumvent this issue. Following Neeleman & Reinhart (1998), I assume that a structure in which an object precedes an adverbial can be base-generated, hence the absence of a trace or an empty position below the adverbial in (12b) and (15b). This does not affect the claims made in the main text.

87

Chapter 3 Word order variation and information structure

In Korean, a non-contrastive topic must also occupy clause-initial position:4 (14)

ku moca-eytayhayse this hat-about ‘Tell me about this hat.’

mal-hay-po-a. tell-do-try-IMPERATIVE

(15)

a. ku moca-nun/luli John-i this hat-NUN/ACC John-NOM b. #John-i ku moca-nun/lul John-nom this hat-nun/acc ‘John bought this hat yesterday.’

ecey yesterday ecey yesterday

(K)

ei

sasse. bought sasse. bought

I propose that a non-contrastive topic surfaces in an adjoined position to the highest maximal projection in the clause in Japanese and Korean. Thus, in a simple declarative clause, the topic is adjoined to TP. Following the standard literature on Japanese and Choe (1995) for Korean, I assume that a non-contrastive, argument topic is base-generated in the adjoined position, binding an empty pronominal internally to the clause: (16)

[TP XP-wa/nuni

[TP

… proi … ]]

The above analysis is motivated by the fact that non-contrastive topics in the two languages do not show properties associated with movement. For instance, they are insensitive to islands. Thus, in the Japanese example in (17), ano boosi-wa ‘that hat-WA’ appears in the initial position of the main clause, but is interpreted as the object inside the relative clause. The sentence can be uttered as a response to the request tell me about that hat. The presence of the empty pronominal pro can be seen from the fact that it is possible to overtly realise it (Perlmutter 1972, Kuno 1973, Saito 1985, also Hoji 1985).5

4 In Korean, the nominative case marker is realised as ka after a vowel and as i elsewhere. Similarly, the accusative marker is realised as lul if following a vowel and as ul elsewhere, and the particle nun as nun after a vowel and as un elsewhere. 5 Kuroda (1988), Sakai (1994) and Ishizuka (to app.) argue that topicalisation always involves movement. However, the possibility of linking to a position inside a relative clause is still considered to be a characteristic of (a construction that can feed into) topicalisation on their accounts.

88 (17)

Reiko Vermeulen ano boosii-wa John-ga [NP [TP ej proi / sorei-o kabutteita] hitoj]-o (J) that hat-WA John-NOM it-ACC wearing-PAST person-ACC yoku sitteiru. well know ‘Speaking of that hat, John knows well the person who was wearing it.’ (slightly modified from Hoji 1985 : 152)

The same construction is possible in Korean (Choe 1995 : 312). The following example is a felicitous response to the request tell me about this hat. (18)

ku mocai-nun/lul Chelswu-ka [NP [TP ej proi/kukesi-ul sacwu-n] this hat-NUN Chelswu-NOM it-ACC bought-ADN salamj]-ul cal arayo. (K) person-ACC well know ‘This hat, I know the person who bought it.’

I propose that the derivation in (16) is motivated by its effect at the interface (Neeleman & van de Koot 2008; see also Reinhart 2006). Specifically, I argue that the placement of a topic in clause-initial position marks the rest of the sentence as its comment, allowing for a transparent mapping between syntax and information structure, as illustrated below. I will show in Section 5 that contrastive topics in Japanese are also subject to this rule, but those in Korean are not. As mentioned in Section 2, there is evidence that contrastive topics in Japanese undergo movement to this position. A trace is therefore provided as an alternative to pro below.6 (19)

Mapping rule for [topic]: Syntax: [YP Information Structure:

4.

XPi | Topic

[YP … proi/ti …

]]

Comment

[Contrast] and contrastive focus

In this section, I argue that Japanese and Korean have the same mapping rule relevant for the notion [contrast]. An item interpreted contrastively can undergo scrambling in the two languages to a variety of positions and 6 For concreteness, I assume that so-called ‘scene-setting topics’ (Lambrecht 1994) or ‘stage topics’ (Erteschik-Shir 1997), are base-generated in the adjoined position (Tateishi 1994), while other adverbials such as manner adverbials, have undergone movement to this position. I concentrate on argument topics in this chapter.

Chapter 3 Word order variation and information structure

89

this operation is also motivated by its effects at the interface. The idea is formulated as in (20). I will discuss the syntactic distribution and its effects at the interface in turn. (20)

4.1

[Contrast] licenses scrambling.

The syntactic distribution of contrastive focus

The following examples demonstrate that a contrastive focus in a correction context can remain in situ, or undergo scrambling to a clause-medial or clause-initial position in both Japanese, (21)/(22), and Korean, (23)/(24). (21)

John-wa Sue-ni John-WA Sue-to ‘John gave a CD to Sue.’

CD-o CD-ACC

(22)

a. Ie, John-wa Sue-ni b. Ie, John-wa ANO HON-Oi c. Ie, ANO HON-Oi John-wa no, that book-ACC John-WA ‘No, John gave that book to Sue.’

ageta. gave

(J)

ANO HON-O

ageta. ti ageta. ti ageta. gave

Sue-ni Sue-ni Sue-to

(23)

John-i Sue-eykey CD-ul John-NOM Sue-to CD-ACC ‘John gave a CD to Sue.’

cwuess-e. gave-DECL

(24)

a. Ani, John-i Sue-eykey KU CHAYK-UL b. Ani, John-i KU CHAYK-ULi Sue-eykey c. Ani, KU CHAYK-ULi John-i Sue-eykey no, that book-ACC John-NOM Sue-to ‘No, John gave that book to Sue.’

(K)

ti ti

cwuess-e. cwuess-e. cwuess-e. gave-DECL

A contrastive focus can also undergo long-distance scrambling in both languages: (25)

Bill-wa [John-ga Sue-ni CD-o Bill-WA John-NOM Sue-to CD-ACC ‘Bill thinks that John gave Sue a CD.’

ageta to] omotteiru. gave COMP thinking

(26)

Ie, ANO HON-Oi Bill-wa [John-ga Sue-ni ti ageta to] no, that book-ACC Bill-WA John-NOM Sue-to gave COMP omotteiru (no desu yo). thinking (NMZ COP PRT) ‘No, Bill thinks that John gave a book to Sue.’

(J)

90 (27)

(28)

Reiko Vermeulen Swuni-ka [CP Yenghi-ka ku kwutwu-lul sasse-ta-ko] Swuni-NOM Yenghi-NOM that shoes-ACC bought-DECL-COMP sayngkakha-n-ta. think-DECL ‘Swuni thinks that Yenghi bought those shoes.’

(K)

Ani, KU MOCA-LULi Swuni-ka [CP Yenghi-ka ti sasse-ta-ko] no, that hat-ACC Swuni-NOM Yenghi-NOM bought-DECL-COMP sayngkakhan-ta. thought-DECL ‘No, Swuni thinks that Yenghi bough that hat.’

The same observation obtains in a context where the preceding question is a disjunctive question. The preceding question in (29) in Japanese and (31) in Korean makes hon/chayk ‘book’ and CD the two possible alternatives in this context. The use of dake/man ‘only’ in the answers strengthens and ensures the contrastive implicature that an alternative proposition regarding the CD is false. (29)

John-wa Sue-ni hon-to CD-to dotti-o ageta John-WA Sue-to book-and CD-and which-ACC gave ‘Which of the book and the CD did John give to Sue?’

(30)

a. John-wa Sue-ni HON(-DAKE)-O b. John-wa HON(-DAKE)-Oi Sue-ni ti c. HON(-DAKE)-Oi John-wa Sue-ni ti book-only-ACC John-WA Sue-to ‘John gave (only) the book to Sue.’

(31)

John-un Sue-eykey chayk-kwa CD-cwung enu kes-ul John-NUN Sue-DAT book-and CD-among which thing-ACC cwuess-ni? gave-Q ‘Which of the book and the CD did John give to Sue?’

(32)

a. John-i Sue-eykey CHAYK-MAN/UL b. John-i CHAYK-MAN/ULi Sue-eykey ti c. CHAYK-MAN/ULi John-i Sue-eykey ti book-only/ACC John-NOM Sue-to ‘John gave (only) the book to Sue.’

no? Q

(J)

ageta. ageta. ageta. gave

(K)

cwuess-e. cwuess-e. cwuess-e. gave-DECL

Again, a contrastive focus in this kind of context can undergo long-distance scrambling:

Chapter 3 Word order variation and information structure

91

(33)

John-wa [CP Bill-ga Sue-ni hon-to CD-to dotti-o John-WA Bill-NOM Sue-to book-and CD-and which-ACC ageta to] omotteiru no? (J) gave COMP thinking Q ‘Which of the book and the CD does John think that Bill give to Sue?’

(34)

HON(-DAKE)-Oi

(35)

John-un [CP Bill-i Sue-eykey chayk-kwa CD-cwung John-NUN Bill-NOM Sue-Dat book-and CD-among enu kes-ul cwuess-ta-ko] sayngkakha-ni? (K) which thing-ACC gave-DECL-COMP think-Q ‘Which of the book and the CD does John think that Bill give to Sue?’

(36)

CHAYK-MAN/ULi

John-wa [CP Bill-ga Sue-ni ti ageta to] omotteiru. book-only-ACC John-WA Bill-NOM Sue-to gave COMP thinking ‘John thinks that Bill gave (only) the book to Sue.’

John-i [CP Bill-i Sue-eykey book-only/ACC John-NOM Bill-NOM Sue-to sayngkakhan-ta. think-DECL ‘John thinks Bill gave (only) the book to Sue.’

ti cwuess-ta-ko] gave-DECL-COMP

Thus, a contrastive focus can undergo scrambling to a variety of positions. However, it has sometimes been proposed in the literature that ‘focus’, among other factors, licenses scrambling in the two languages (Kitagawa 1990, Miyagawa 1997, 2006, Aoyagi & Kato 2008, a. o. for Japanese; Choe 1995, M. Lee 2006, a. o. for Korean). One may wonder, therefore, whether the relevant notion licensing the optional scrambling in the above examples is in fact [focus], rather than [contrast]. Some speakers indeed allow the scrambled (b)/(c)-version in (22) and (24) as answers to a simple object-wh-question such as What did John give to Sue?. However, there are reasons to believe that the relevant notion is [contrast] rather than [focus]. First, most of my Japanese and Korean informants report that a contrast between the book and another item is implied if scrambling takes place (see also Choi 1997, 1999). In other words, a contrastive interpretation must be accommodated if focus is scrambled. This is in line with the cross-linguistic observation that in languages that have ‘focus movement’, it is only the contrastive type of focus that undergoes movement (Vallduví & Vilkuna 1997 for Finnish, É Kiss 1998 for Hungarian, Rizzi 1997, Samek-Lodovici 2005 for Italian, Neeleman & van de Koot 2008 for Dutch, Neeleman & Titov 2009 for Russian, Zimmermann 2008 for Hausa, among many others).

92

Reiko Vermeulen

Secondly, there is consensus in the literature on Japanese that longdistance scrambling requires a contrastive focus reading (Saito 1985, Kitagawa 1990, Miyagawa 2006) and most of my informants allow longdistance scrambling only in cases of correction. Thus, all native speakers of Japanese I have consulted found long-distance scrambling in cases of correction perfectly natural, as in the exchange in (25)/(26), but most found the same sentence infelicitous in answering a simple wh-question, as in the exchange in (37)/(38), and those who found it acceptable reported that they accommodated some sort of a contrastive interpretation on the scrambled object. (37)

(38)

Bill-wa [John-ga dare-ni ano hon-o ageta to] Bill-WA John-NOM who-to that book-ACC gave COMP omottteiru no? thinking Q ‘Who does Bill think John gave the book to?’

(J)

??SUE-NI

ti ano hon-o ageta to] omottteiru. i Bill-wa [John-ga Sue-to Bill-WA John-NOM that book-ACC gave COMP thinking ‘Bill thinks that John gave that book to Sue.’

For Korean too, it is reported in the literature that long-distance scrambling requires a contrastive interpretation (Choe 1995, Tsoulas 1999 and Hwang, et al. 2010 for evidence from parsing). Indeed, my Korean informants confirm that long-distance scrambling is perfectly natural in a context of correction as in (27)/(28), and requires accommodation of contrast in answering a simple wh-question context as in (39)/(40).7 (39)

(40)

Swuni-ka [CP Yenghi-ka mwuess-ul sasse-ta-ko] Swuni-NOM Yenghi-NOM what-ACC bought-DECL- COMP sayngkakhan-ni? think-Q ‘What does Swuni think Yenghi bought?’

(K)

??KU MOCA-LUL i

Swuni-ka [CPYenghi-ka ti sasse-ta-ko] this hat-ACC Swuni-NOM Yenghi-NOM bought-DECL- COMP sayngkakhan-ta. think-DECL ‘Swuni thinks that Yenghi bough this hat.’

7 The reported judgement is confirmed by all but one of my informants. For the one informant, a contrastive interpretation need not be accommodated in (39)/(40). I do not have an explanation for this variation.

Chapter 3 Word order variation and information structure

93

Finally, as we will see in the next section, a contrastive topic in Korean displays the same scrambling possibilities as observed above for a contrastive focus, and it does not have the syntactic behaviour of a non-contrastive topic we saw in the previous section. Such an observation can be captured in a most straightforward way if we assume that contrastive focus and contrastive topic in Korean are both subject to the same operation sensitive to [contrast]. Thus, I propose that it is indeed the feature [contrast] that licenses scrambling in the above examples, rather than [focus].

4.2

Scope of Contrast

Scrambling clearly does not make an item contrastive, since the item can remain in situ and still be interpreted contrastively. Rather, I argue that contrast is quantificational and takes scope, and scrambling explicitly marks this scope. The scope contains material that is required for calculating the contrast. Recall that the meaning of contrastive focus is a composite of [focus] and [contrast]. Thus, the contrastive focus Rosa in B’s utterance in (8) has the representations in (10a) for the [focus] component and (10b) for the [contrast] component, repeated below. What allows a contrastive item to take scope is the negative operator in the statement in (10b). (8)

A: John saw Claire yesterday. B: That’s not true. He saw ROSA yesterday.

(10)

a. b. y [y  {Claire, Jennifer, Jason, …} & ™[John saw y]]

Scrambling of a contrastive focus in Japanese and Korean marks what material is to be included in the scope of this operator, creating a transparent mapping between syntax and information structure. The idea can be understood in the form of a mapping rule for [contrast], illustrated below. I assume that scrambled items are adjoined to a maximal projection.8 8 An issue that potentially arises is whether clause-internal scrambling licensed by [contrast] is an instance of A-scrambling or A’-scrambling. As is wellknown, scrambling in Japanese and Korean shows properties associated with A-movement as well as A’-movement (Hoji 1985, Saito 1985, 1989, 1992, 2003, Tada 1993, Miyagawa 1997, 2005, Ueyama 1998, a. o. for Japanese, Cho 1994, Choe 1995, Tsoulas 1999, McGinnis 2008 for Korean, among others). However,

94

Reiko Vermeulen

(41)

Mapping rule for [contrast]: Syntax: [YP Information Structure:

XPi | CF

[YP

…

ti …

]]

Scope of Contrast

Let us see how the above mapping rule applies to the examples in (21)/ (22) and (23)/(24). Only the Japanese examples in (21)/(22) are repeated below, as the Korean examples in (23)/(24) are identical in the relevant respects. (21)

John-wa Sue-ni CD-o John-WA Sue-to CD-ACC ‘John gave a CD to Sue.’

(22)

a. Ie, John-wa Sue-ni b. Ie, John-wa ANO HON-Oi c. Ie, ANO HON-Oi John-wa No, that book-ACC John-WA ‘No, John gave that book to Sue.’

ageta. gave

ANO HON-O

Sue-ni Sue-ni Sue-to

(J)

ageta. ti ageta. ti ageta. gave

The [focus] component of the meaning for ano hon ‘that book’ is represented below: (42)

In (22b), scrambling of the contrastive focus is to below the subject, and therefore, the subject is not included in the scope of contrast, the scope of negation. However, a VP excluding the subject is not of a semantic type that can be negated. I assume that there is a procedure of existential closure that transforms an incomplete scope of contrast into an expression of the right type that can be negated (compare Schwarzschild 1999). After application of existential closure, the negated proposition is ‘someone gave y to Sue’, where y is an alternative to ‘that book’, ‘the CD’ in this context. This is represented in (43a) and this is what is implied by the sentences in (22b). In (22c), on the other hand, scrambling is to above the subject, and the subject is therefore included in the scope of negation. initial investigation shows that [contrast] is not a distinguishing factor for the two kinds of scrambling at least for Japanese. Thus, for instance, an object contrastive focus scrambled to above the subject can bind an anaphor in the subject, but it can also reconstruct to obviate a Condition C violation. I will not discuss this issue here.

Chapter 3 Word order variation and information structure

95

Here, the negated proposition is ‘John gave y to Sue’, and this is represented in (43b). In (22a), the scope of contrast is not overtly marked, so the hearer must construct it from the context, which may be either as in (43a) or (43b), or just the verb. (43)

a. y [y  { flowers, CD, chocolate, …} & z ™[z gave Sue y]] b. y [y  { flowers, CD, chocolate, …} & ™[John gave Sue y]]

The proposal also extends straightforwardly to cases involving long-distance scrambling such as (26) and (27). I argue that marking of the scope of contrast is obligatory when a contrastive item undergoes scrambling. Thus, although scrambling of a contrastive item in itself is optional, if it takes place, the scope of contrast is always marked. I provide evidence for this in Section 6. I propose in Section 5 that essentially the same procedure applies to the contrastive interpretation of contrastive topics in Korean, which have the same syntactic distribution as contrastive foci. One difference is that what is in the scope of negation in the case of contrastive topics is an assertion, not a proposition. The effects of the difference between (43a) and (43b) is difficult to show with examples such (22) and (24), where the contrastive focus is the only contrastive element in the sentence. Like other scope-taking items, the effects of overtly marking the scope of contrast can be seen most clearly when it interacts with other contrastive items. I will examine sentences containing both a contrastive focus and a contrastive topic in Section 6.

5.

[Contrast] and [topic]: Contrastive Topics

Consider the two mapping rules proposed so far, which are repeated below: (11)

[Topic] is licensed in clause-initial position.

(20)

[Contrast] licenses scrambling.

The placement of a topic in clause-initial position has the effect of marking its sister constiutent as the comment, and scrambling of a contrastive item marks the scope of contrast. The two rules cannot be satisfied simultaneously. The comment of a topic is not the same as the scope of contrast of a contrastive topic. If a contrastive topic undergoes scrambling to below the

96

Reiko Vermeulen

subject and therefore its scope of contrast excludes the subject, it is not the same as the comment, which necessarily contains all the arguments in the rest of the sentence. I argue that where there is a conflict such as this, the language in question simply chooses one operation over the other. Thus, if a language adopts the rule relevant for [topic], a contrastive topic in this language would systematically show the same behaviour as a non-contrastive topic with respect to this particular rule. I show that this is the case in Japanese. On the other hand, if a language chooses the rule pertaining to [contrast] for contrastive topics and contrastive foci are also subject to this rule in this language, then contrastive topics and contrastive foci behave alike with respect to this rule. I argue that this is the case in Korean. Let us first consider Japanese. The following exchange shows that contrastive topics in this language must move to clause-initial position. The question in (44) is about ano CD ‘that CD’, but the hearer of this question appears not to know the relevant information with respect to the CD and offers information regarding ano hon ‘that book’. In doing so, she has shifted the topic of discourse from the CD to the book, making the latter a contrastive topic. As the answers in (45) shows, ano hon ‘that book’ must be displaced to clause-initial position.9 9 Unlike the Japanese examples we considered in the previous section, the subject in this example is marked with the nominative case marker, as opposed to non-contrastive wa. Nominative subjects are typically interpreted as a focus or part of a focus (Kuno 1973), especially when uttered discourse-initially (Hinds, et al. 1987). Thus, it is more natural to use non-contrastive wa, which allows the subject to function as a non-contrastive topic. However, having a wa-marked subject seems to affect the distribution of the object contrastive wa-phrase in parallel examples to (45) for some speakers. Two of my five informants disallow movement of an object contrastive wa-phrase over a subject non-contrastive wa-phrase here. This is unlikely to be due to a violation of Relativized Minimality (Rizzi 1990), as the same speakers allow such movement, if the predicate is mono-transitive, as in (i), or if there is a contrastive focus later in the sentence, as in (73). I have no explanation for this discrepancy. As the informants found the question and the answer with a nominative subject less than perfect, but still acceptable, I have used nominative subjects in the exchange in (44)/(45) (See also footnote 14). (i) A: ano inu-wa John-o kanda no? ‘Did that dog bite John?’ B: (Well, I don’t know about John, but...) Bill-wai ano inu-wa moo sudeni kyonen Bill-WA that dog-WA already last.year ‘that dog has bitten Bill already last year.’

ti

kandeiru. bite.PERF.

Chapter 3 Word order variation and information structure ageta gave

no? Q

97

(44)

John-ga Sue-ni ano CD-o John-NOM Sue-to that CD-ACC ‘Did John give that CD to Sue?’

(J)

(45)

Hmm, ano CD-wa doo-da-ka siranai-kedo, Well, that CD-WA how-COP-whether not.know-but ‘Well, I don’t know about that CD, but …’ a. # John-ga Sue-ni ano hon-wa kinoo ageteita (yo). b. ??John-ga ano hon-wai Sue-ni kinoo ti ageteita (yo). c. ano hon-wai John-ga Sue-ni kinoo ti ageteita (yo). that book-WA John-NOM Sue-to yesterday gave PRT ‘as for that book, John gave (it) to Sue yesterday.’

Thus, contrastive topics in Japanese, like non-contrastive topics, are subject to the rule for [topic] and must appear in clause-initial position. As noted above in Section 2, there is abundant evidence reported in the literature that contrastive topics undergo movement to its surface position, as opposed to being base-generated there. Thus, it is possible for a contrastive topic to undergo long-distance movement, as shown in (46), but it is sensitive to islands, as illustrated by (47) (slightly modified from Hoji 1985 : 161). (46)

A: Does Bill think that John gave a CD to Sue? B: Well, I don’t know about the CD, but … ano hon-wai Bill-wa [John-ga Sue-ni ti ageta to] omotteiru that book-WA Bill-WA John-NOM Sue-to gave COMP thinking (yo).

(J)

PRT

‘as for that book, Bill thinks that John gave (it) to Sue.’ (47)

?*(Susan-zyanakute) Maryj-wa John-ga [NP[TP ei (kanozyoj-o) butta] (Susan-not.but) Mary-WA John-NOM she-ACC hit hitoi-o sagasiteiru. person-ACC looking.for Lit.: ‘(Not Susan, but) Mary, John is looking for a person who hit (her).’

The fact that different structures are involved for contrastive and noncontrastive topics is not important here. What is important is that they are both in clause-initial position, and that has a specific effect at the interface. The effect is that it creates a transparent mapping between the syntactic structure and a topic-comment structure. The rule for [topic] is repeated below from (19). I assume that a subject, whose canonical posi-

98

Reiko Vermeulen

tion is clause-initial, also undergoes string-vacuous movement, when it is interpreted as a contrastive topic. (19)

Mapping rule for [topic]: Syntax: [YP Information Structure:

XPi | Topic

[YP … proi/ti …

]]

Comment

Thus, unlike contrastive focus in this language, movement of a contrastive topic does not mark the scope of contrast, but the comment. The interpretation of the contrastive topic in (45c) can be represented as below, in the form of a triplet as before. The comment corresponds to what is in the scope of the assertion operator: (48)

The scope of contrast is not overtly marked. It must be calculated from the context, just as in the case of in situ contrastive foci (see the discussion around (43)). Unlike Japanese contrastive topics, Korean contrastive topics can remain in situ, or scramble to a clause-medial or clause-initial position. This is illustrated by the following parallel exchange to the Japanese example in (44)/(45):10 (49)

John-i Sue-hantey ku CD-lul ecey John-NOM Sue-to this CD-ACC yesterday ‘Did John give this CD to Sue yesterday?’

cwuesse? gave

(50)

Hmm, ku CD-nun molu-keyss-ko, Well, this CD-NUN not-know-but ‘Well, I don’t know about this CD, but …’ a. John-i Sue-hantey i chayk-un (ecey) b. John-i i chayk-uni Sue-hantey (ecey) ti c. i chayk-uni John-i Sue-hantey (ecey) ti this book-NUN John-NOM Sue-to yesterday ‘as for this book, John gave it to Sue yesterday.’

(K)

cwuesse. cwuesse. cwuesse. gave

10 My informants report that it is also possible to mark contrastive topics with a case marker, as was the case for non-contrastive topic (see Section 3), but they also expressed strong preference for marking with nun in cases of contrastive topics. I will therefore consider only nun-marked contrastive topics here.

Chapter 3 Word order variation and information structure

99

A contrastive topic may also undergo long-distance scrambling: (51)

A: Does Chelswu think that John bought this coat? B: Well, I don’t know about this coat, but … ku mocai-nun Chelswu-ka [John-i ei sasse-ta-ko] this hat-NUN Chelswu-NOM John-NOM bought-COMP sayngkakhan-ta. think-DECL ‘as for this hat, Chelswu thinks that John bought (it).’

(K)

The above distribution is identical to that of contrastive focus in Korean and Japanese, examined in Section 4. Thus, I argue that contrastive topics in Korean are subject to the rule relevant for [contrast]. The rule for [contrast] is repeated below from (41): (41)

Mapping rule for [contrast]: Syntax: [YP Information Structure:

XPi | CF

[YP

…

ti …

]]

Scope of Contrast

The moved contrastive topics in (50b) and (50c) mark their scope of contrast. Recall that contrastive topic is a composite of the features [contrast] and [topic]. The [topic] component of the meaning can be represented as in (52) for both examples. For (50b), after application of existential closure, what is implied is the negation of ‘the assertion that someone gave y to Sue’, as shown in (53a). For (50c), what is implied is the negation of ‘the assertion that John gave y to Sue’, as represented in (53b). (52)

(53)

a. y [y  {that book, CD, chocolate,…} & z ™[z Oy ASSERT [z gave yWR Sue]]] b. y [y  {that book, CD, chocolate,…} & ™[y Oy ASSERT [John gave y to Sue]]]

Thus, a contrastive topic in Japanese must surface in clause-initial position, like its non-contrastive counterpart, while a contrastive topic in Korean can appear in a variety of positions like a contrastive focus. The proposed idea based on the typology in (1) that a language adopts one rule over the other in cases of conflict provides a uniform account of the differences in the syntactic distribution of contrastive topics in Japanese and Korean.

100

Reiko Vermeulen

The above generalisations regarding the distribution differ from the standard views in the literature, however. For Japanese, it is widely assumed that any phrase marked by contrastive wa is a contrastive topic and it only ‘optionally’ undergoes movement to clause-initial position (Heycock 2007, Tomioka 2010, and references in those works). Indeed, the examples in (45a) and (45b) are not ungrammatical, only infelicitous in the given context. For Korean, there is less clear consensus on the treatment contrastive topics. The predominant view is that contrastive nunphrases are interpreted as contrastive topics in clause-initial position, but they are contrastive ‘foci’ elsewhere (Choi 1997, Han 1998, Gill & Tsoulas 2004). Some authors argue alternatively that contrastive nun-phrases are contrastive foci everywhere and do not discuss contrastive topics at all (Choe 1995, Choi 1999, M. Lee 2006), while C. Lee (2003, 2006) argues that any contrastive nun-marked phrase is a contrastive topic. I will provide interpretive and syntactic arguments in the following two sections that these characterisations are inadequate and offer an alternative account of the data discussed by these authors.

6.

Interaction between contrastive topic and contrastive focus

So far, we have discussed the distribution of contrastive topics and contrastive foci separately in both Japanese and Korean. I have proposed that the two discourse functions are subject to two different rules in Japanese, one relevant for [topic] and the other for [contrast], while in Korean, one rule pertaining to [contrast] is relevant for both discourse functions. Moreover, I claimed that marking of the scope of contrast by scrambling licensed by [contrast] is obligatory. This section provides evidence for this claim by investigating the scope interaction of contrastive focus and contrastive topic in sentences that contain both. In combination with wellformedness conditions in the information structural component, the proposed mapping rule for [contrast] gives rise to several predictions concerning the syntactic distribution of contrastive topic and focus with respect to each other. The empirical predictions are the same as those we discussed for Dutch in Chapter 1, although they are derived slightly differently for Japanese. Below, I demonstrate that these predictions are also borne out for Japanese and Korean. The specific predictions come about through the interaction of two assumptions already introduced above. The first is that scrambling licensed by [contrast] is a scope-marking operation. As such, it gives rise to a scope freezing effect: material c-commanded by the moved category

Chapter 3 Word order variation and information structure

101

cannot outscope that category. Our second assumption is that topics have to be interpreted externally to foci. This is because topic is an utterancelevel notion, while focus is a notion relevant to propositions. As a matter of course, propositions are contained inside utterances, rather than the other way around. Thus, in the information structural component, (54a) is well-formed, while (54b) is not. (For relevant discussion, see Reinhart 1981, 1995, 2006, Vallduví 1992, Lambrecht 1994, and Hajiþová et al. 1998.) (54)

a. [utterance topic [proposition FOCUS … ]] b. #[proposition FOCUS [utterance topic … ]]

The predictions that follow from these two assumptions are as follows. First, if no scrambling by a contrastive item takes place, its scope is not overtly marked, which implies that no instructions are given concerning the mapping between syntax and information structure. Consequently, the ordering between contrastive topic and contrastive focus is free (see (55a,b)). However, if a contrastive focus scrambles across a contrastive topic in either Japanese or Korean, the scope marking nature of the movement implies that the topic must be interpreted in the scope of the negative operator associated with the contrastive focus (see (43)). As this negative operator must attach to a proposition, the topic must be interpreted as part of the proposition as well, but this gives rise to the ill-formed structure in (54b). Thus, such a structure, shown in (55c), should be disallowed. On the other hand, if a contrastive topic undergoes scrambling across a contrastive focus in Korean, the result should be well formed. This is because the negative operator in this case attaches to an utterance (see (53)), and an utterance can of course contain a focus. Similarly, movement of a contrastive topic to clause-initial position in Japanese marks the comment. A comment is a proposition that is predicated of a topic via an assertion operator and can therefore contain a focus. Thus, it is predicted that a contrastive topic in Korean and Japanese can move across a contrastive focus, as in (55d). (55)

a. b. c. d.

[ … CT … CF … ] (CF = contrastive focus; CT = contrastive topic) [ … CF … CT … ] #[CF [ … CT … tCF … ]] [CT [ … CF … tCT … ]]

I will discuss these predictions for each language in turn, starting with Korean. The following exchange illustrates that (55a) and (55c) are cor-

102

Reiko Vermeulen

rect. (56) is a question about John, but the reply is about Bill, making the latter a contrastive topic. The object khong ‘beans’ is a contrastive focus, answering the wh-part of the disjunctive question. In (57a), the contrastive topic and contrastive focus appear in their base-generated positions in that order, while in the infelicitous (57b), the contrastive focus has scrambled across the contrastive topic. The fact that (57b) is infelicitous also shows that marking of the scope of contrast is obligatorily when a contrastive item undergoes scrambling. If a scrambled contrastive focus can undergo reconstruction for its scope of contrast, (57b) should be acceptable. (56)

John-i pathi-eyse khong-kwa pasutha-cwung enu kes-lul John- NOM party-at beans and pasta-among which thing-ACC mekesse? (K) ate ‘Which of the beans and the pasta did John eat at the party?’

(57)

Hmm, John-un molu-keyss-ko, Well, John-NUN not-know-but ‘Well, I don’t know about John, but …’ a. Bill-un 8-si-ey KHONG-UL b. # HONG-ULi Bill-un 8-si-ey ti beans-ACC Bill-NUN 8 o’clock-at

mekesse. mekesse. ate

(CT CF) (#CFi CT ti)

The following example shows that the predictions in (55b) and (55d) are also correct: (58)

ecey party-eyse John-kwa Bill-cwung nwuka pasutha-lul yesterday party-at John-and Bill-among who pasta-ACC mekesse? (K) ate ‘Which of John and Bill ate the pasta at the party yesterday?’

(59)

Hmm, pasutha-nun molu-keyss-ko, Well, pasta-NUN not-know-but ‘Well, I don’t know about the pasta, but …’ a. BILL-I khong-un 8-si-ey mekesse. b. khong-uni BILL-I 8-si-ey ti mekesse. beans-NUN Bill-NOM 8 o’clock-at ate ‘as for the beans, Bill ate them at 8 o’clock.’

(CF CT) (CTi CF ti)

We saw above that contrastive focus cannot move across a contrastive topic, (55c). However, what deems the resulting structure ill-formed is the fact that the scope of contrast for the contrastive focus contains a contras-

Chapter 3 Word order variation and information structure

103

tive topic, rather than the fact that it has moved across a contrastive topic. Thus, scrambling of a contrastive focus to a position above a contrastive topic should generally be disallowed even if the launching site is above the contrastive topic, as in (60a). On the other hand, a contrastive topic should be able to scramble to a position above a contrastive focus from a launching site above the contrastive focus, as shown in (60b) without comparable problems. (60)

a. #[CF [ … tCF … CT … ]] b. [CT [ … tCT … CF … ]]

The examples in (61)/(62b) and (63)/(64b) show that these further predictions are borne out, respectively. These data also show that the ill-formedness of the example in (57b), cannot be explained in terms of a Relativized Minimality violation (Rizzi 1990), by assuming that the feature composition of a contrastive topic may be richer than that of a contrastive focus, for instance. In addition, again, the fact that (62b) is unacceptable shows that marking of the scope of contrast is obligatory. (61)

(62)

John-i Sue-kwa Mary-cwung nwuku-hantey ku CD-lul John-NOM Sue-and Mary-among who-to this CD-ACC cwuesse? gave ‘To which of Sue and Mary did John give this CD?’ Hmm, ku CD-nun molu-keyss-ko, Well, this CD-NUN not-know-but ‘Well, I don’t know about this CD, but …’ a. John-i SUE-HANTEY i chayk-un b. #SUE-HANTEYj John-i Sue-to John-NOM

tj i chayk-un this book-NUN

‘as for this book, John gave it to Sue.’ (63)

(K)

ecey

cwuesse. (CF CT) ecey cwuesse. yesterday gave (#CFi ti CT)

John-i Mary-hantey chayk-kwa CD-cwung enu kes-ul John-NOM Mary-to book-and CD-among which thing-ACC cwuesse? (K) gave ‘Which of the book and the CD did John give to Mary?’

104 (64)

Reiko Vermeulen Hmm, Mary-nun molu-keyss-ko, Well, Mary-NUN not-know-but ‘Well, I don’t know about Mary, but …’ a. John-i Sue-hantey-nun I CHAYK-UL

ecey cwuesse. (CT CF) b. Sue-hantey-nunj John-i tj I CHAYK-UL ecey cwuesse. (CTi ti CF) Sue-to-NUN John-NOM this book-ACC yesterday gave ‘as for Sue, John gave her this book yesterday.’

The above data provide support for the proposal that scrambling of a contrastive focus and a contrastive topic in Korean indeed has the effect of marking the scope of contrast. Moreover, the data provide further evidence against the predominant view in the literature that the interpretation of a contrastive nun-phrase as a contrastive topic is limited to clauseinitial position. I now turn to Japanese. The predictions are repeated below for convenience. (55)

a. b. c. d.

[ … CF … CT … ] [ … CT … CF … ] #[CF [ … CT … tCF … ]] [CT [ … CF … tCT … ]]

(60)

a. #[CF [ … tCF … CT … ]] b. [CT [ … tCT … CF … ]]

Due to the fact that contrastive topics in Japanese must independently appear in clause-initial position, some of the predictions above cannot be tested. In particular, it is not possible to test (55a), (55b) and (60a), where the contrastive topic has not moved to clause-initial position. At first sight, it may seem that (55c) can also not be tested. However, this prediction can be shown to be correct by examining data involving an embedded clause. First, it is possible for a contrastive topic to appear in an embedded clause, as shown in (66), uttered in a context such as (65). The context makes kono CD ‘this CD’ a contrastive topic, as it shifts the topic of discourse from the book. The presence of kare ‘his’, that is coreferential with the matrix subject Bill, ensures that the embedded clause is indeed embedded and not a direct quotation (Fukui 1995).

Chapter 3 Word order variation and information structure

105

(65)

Context: John finds a book on Sue’s desk and he asks Bill to tell him something about the book. Bill does not know anything about the book, but he knew how Sue obtained a CD that was also on the desk. So, he decides to tell John about the CD. In describing this situation, you utter (66).

(66)

Billj-wa [CP kono CD-wai Mary-ga karej-no mise-de Sue-ni ti Bill-WA this CD-WA Mary-NOM he-GEN shop-at Sue-to ageta to] itta. gave that said ‘Billj said that as for this CD, Mary gave it to Sue in hisj shop.’

(J)

Independently, we saw in Section 4 that a contrastive focus can undergo long-distance scrambling in cases of correction (Saito 1989, Miyagawa 2006). Thus, correcting the statement in (67), one could say (68), where the indirect object of the embedded verb provides correct information and is fronted to sentence-initial position: (67)

(68)

Billj-wa [CP Mary-ga Jane-ni kono CD-o karej-no mise-de Bill-WA Mary-NOM Jane-to this CD-ACC he-GEN shop-at ageta to] itta. gave that said ‘Billj said that Mary gave this CD to Jane in hisj shop.’

(J)

Tigaimasu yo. SUE-NIi Billj-wa [CP Mary-ga ti kono CD-o incorrect PRT Sue-to Bill-WA Mary-NOM this CD-ACC karej-no mise-de ageta to] itta no desu yo. he-GEN shop-at gave that said NMZ COP PRT Lit.: ‘That’s not true. It’s to Sue that Billj said that Mary gave this CD in hisj shop.’

The precise prediction is that it should be impossible to combine the above two operations, as this will result in the ill-formed structure in (55c). The prediction is borne out: the example in (70), uttered in correcting the statement in (69), is infelicitous. The contrastive topic kono-CD-wa ‘this CD-WA’ is moved to initial position in the embedded clause, while the contrastive focus Sue-ni has undergone scrambling across it to initial position of the embedding clause.11

11 Slight unnaturalness arises here due to repeated mention of Bill-wa in (69), but this does not affect the argument here, as the same informants found (71) acceptable.

106 (69)

(70)

Reiko Vermeulen Billj-wa [CP Mary-ga Jenny-ni kono hon-o karej-no mise-de Bill-WA Mary-NOM Jenny-to this book-ACC he-GEN shop-at ageta to] itta. gave that said ‘Billj said that Mary gave this book to Jenny in hisj shop.’

(J)

Tigaimasu yo. Bill-wa ano hon-nituite-wa sir-anakat-ta-kedo, incorrect PRT Bill-WA that book-about-WA know-not-PAST-but ‘That’s not true. Bill didn’t know anything about the book, but …’ # SUE-NIi Billk-wa [CP kono CD-waj Mary-ga karek-no mise-de ti tj Sue-to Bill-WA this CD-WA Mary-NOM he-GEN shop-at ageta to] itta no desu yo. gave that said NMZ COP PRT Lit.: ‘it’s to Sue that Billk said that as for this CD, Mary gave it to her in hisk shop.’ (CFi CT ti)

Crucially, the sentence is acceptable if the focus remains in situ, which is possible in the same context, in accordance with the prediction in (55d): (71)

… Billk-wa [CP kono CD-waj Mary-ga karek-no mise-de SUE-NI tj … Bill-WA this CD-WA Mary-NOM he-GEN shop-at Sue-to ageta to ] itta. gave that said ‘… Billk said that as for this CD, Mary gave it to Sue in hisk shop.’

(CT CF) Finally, the following exchange shows that (60b) is also correct: (72)

John-wa Mary-ni nani-o John-WA Mary-to what-ACC ‘What did John give to Mary?’

ageta gave

no? Q

(73)

Hmm, Mary-wa doo-da-ka siranai-kedo, Well, Mary-WA how-COP-whether not-know-but ‘Well, I don’t know about Mary, but …’ a. ??John-wa Sue-ni-wa ANO HON-O kinoo John-WA Sue-to-WA that book-ACC yesterday ageteita (yo). gave PRT b. Sue-ni -wai John-wa ti ANO HON-O kinoo Sue-to-WA John-WA that book-ACC yesterday ageteita (yo). gave PRT ‘as for Sue, John gave her that book yesterday.’

(J)

(CT CF)

(CTi ti CF)

Chapter 3 Word order variation and information structure

107

In sum, the predictions in (55) and (60) are borne out in both Korean and Japanese, where testable. This observation lends strong support for the idea that displacement of topics and contrastive items in these languages has direct consequences for the mapping between syntax and information structure. In particular, marking of the scope of contrast is obligatory when scrambling takes place and scrambling cannot create a structure in which the topic must be interpreted within a proposition.

7.

Contrastive wa- and nun-phrases as foci

Recall that the generalisations I have proposed for the distribution of contrastive topics in Japanese and Korean, discussed in Section 5, are different from the standard analyses of contrastive topics in the literature. In this section, I examine the meaning of contrastive wa and nun and propose that contrastive wa- and nun-phrases that are not identified as contrastive topics on the definition adopted in this chapter are in fact foci whose interpretation is enriched by the meaning encoded by the particles. In this sense, they are similar to foci whose interpretation is modified by focus-sensitive particles such as only, even and also. I will call such non-topical contrastive wa- and nun-phrases ‘focal contrastive wa- and nun-phrases’. There has recently been much work on the precise interpretation of contrastive wa-phrases in Japanese (Kuroda 2005, Hara 2006b, Oshima 2008, Tomioka 2010). Adapting Büring’s (1997, 2003) analysis of contrastive topics in German, Hara (2006b) argues that a contrastive wa-phrase induces the presupposition that a scalar alternative stronger than the assertion of the sentence exists and also has the implicature that the speaker thinks the stronger alternative could be false. Let us consider the following example. (Here and below, focal contrastive wa- and nun-phrases are marked by single underlining in adddition to bold font and small caps.) (74)

NANNINKA-WA kita. some people-WA came ‘Some people came.’ (Implicature: ‘Not everyone came’)

(J)

(74) has the meaning in (75a). It presupposes that there is a stronger scalar alternative, such as (75b), and implicates that the speaker thinks this alternative could be false, resulting in the implicature indicated above. (75)

a. (x) [[person(x)] [came (x)]] b. stronger scalar alternative:

(x) [[person(x)] [came (x)]]

108

Reiko Vermeulen

The analysis explains the infelicity of (76), where the subject is a universally quantified item: there is no stronger alternative and thus the presupposition is not satisfied.12 (76)

*MINNA-WA everyone-WA

kita. came

(J) (Hara 2006b: 32)

Hara extends the analysis to non-quantified DPs. A contrastive wa-phrase can answer the wh-part of a preceding question with the implicature that the speaker is unsure about the alternatives. In a context where there are only two relevant individuals, Mary and John, the example in (77b) has the implicature that John probably did not pass the exam. (77)

A: dare-ga siken-ni ukatta who-NOM exam-DAT passed ‘Who passed the exam?’

no? Q

(J)

B: MARY-WA ukatta. Mary-WA passed ‘Mary passed’ (Implicature: ‘John probably didn’t pass’)

The stronger alternative that the utterance in (77b) induces is that both Mary and John passed. However, the speaker just asserted that Mary passed. The hearer can therefore infer that the intended implicature is that John did not pass. The data considered in the literature involve predominantly cases where the subject bears contrastive wa.13 The same contrastive interpretation obtains with contrastive object wa-phrases in situ in similar contexts, and Hara’s analysis can be extended straightforwardly to these cases (Yurie Hara, p.c.). The sentence in (78) gives rise to the implicature ‘John did not help everyone’. A universal quantifier minna ‘everyone’ cannot be 12 Note that wa can mark minna ‘everyone’ if the sentence is negative, as in (i), as this allows for a stronger scalar alternative. See Hara (2006b) for further discussion. (i) minna-wa ko-nakatta. everyone-WA come-not.PAST ‘Not everyone came.’ 13 Fiengo & McClure (2002) argue that the contrastive interpretation depends on the wa-phrase occupying a non-clause-initial position. However, as many examples in this article and other works cited in the main text show, the contrastive reading is not limited to clause-medial positions.

Chapter 3 Word order variation and information structure

109

an object marked with contrastive wa, as in (79), similarly to (76). The sentence in (80) has the implicature ‘John did not help Bill’ in a context where Bill and Mary are the only relevant individuals in the discourse, in the same way as in (77b). It is also a felicitous answer to the question who did John help?. (78)

John-ga NANNINKA-WA tasuketa. (J) John-NOM some.people-WA helped ‘John helped some people.’ (Implicature: ‘John didn’t help everyone.’)

(79)

* John-ga MINNA-WA John-NOM everyone-WA ‘John helped everyone.’

(80)

John-ga MARY-WA tasuketa. John-NOM Mary-WA helped ‘John helped Mary.’ (Implicature: ‘John didn’t help Bill.’)

tasuketa. helped

(J)

(J)

Several authors have proposed that contrastive nun in Korean has a meaning very much like that encoded by contrastive wa in Japanese (Choi 1999, C. Lee 1999, 2003, 2006, M. Lee 2006, Hetland 2007). For instance, Choi (1999), provides the following description with respect to the example in (81), which contains a contrastive nun-phrase as the object: “[(81)] implies that ‘Swuni met Inho, but she probably did not meet other people’, or ‘Swuni met Inho at least, but we do not know whether she met other people as well.’” (Choi 1999 : 168, emphasis mine). The implicature expressed by such paraphrases seems identical to the implicature induced by contrastive wa-phrases in Japanese. Thus, (81) can be used to answer the whquestion who did Swuni meet? and in a context where only Swuni and Yenghi are the relevant individuals, the sentence has the implicature that Mary did not meet Yenghi. (81)

Swuni-ka INHO-NUN manna-ss-e. (K) Swuni-NOM Inho-NUN meet-PAST-DECL ‘Swuni met Inho (but maybe not others)’ (Choi 1999 : 168, gloss and notation modified)

Similarly, C. Lee (2006) proposes that contrastive nun evokes a scalar implicature very much in line with Hara’s proposal for contrastive wa in Japanese. The sentence in (82) gives rise to the scalar implicature that no more than one person came, because a stronger scalar alternative is that

110

Reiko Vermeulen

‘two people came’ and it could be false, and (83) shows that a universal cannot be marked by contrastive nun. (82)

HAN SARAM-UN

(83)

?*MOTWU-NUN w-ass-e. all-NUN came ‘At least all came.’ (C. Lee 2006, ex. (69), gloss and notation modified)

w-ass-ta. one person-NUN come-past-DEC ‘One person came.’ (C. Lee 2006, ex. (112b), gloss and notation modified)

There are obviously differences amongst the proposals offered in the literature. However, one common feature is that a contrastive wa- or nunphrase generates a set of alternatives and has a particular implicature akin to uncertainty regarding the alternatives. I believe that this line of analysis provides a correct characterisation of the interpretation of contrastive waand nun-phrases in general. However, nothing inherent in the kind of interpretation makes a contrastive wa- or nun-phrase a topic, i.e., what the rest of the sentence is about, introducing the referent as the topic of discourse. I propose that contrastive wa- and nun-phrases that do not function as contrastive topics are foci whose interpretation is further modified by contrastive wa and nun, in a similar fashion to foci whose interpretation is modified by focus-sensitive markers such as dake/man, ‘only’, sae/ to, ‘even’ and mo/to ‘also’ (Kuroda 1965, 2005, Oshima 2008; see also Chapter 7, this volume). There are several reasons to believe that ontrastive wa- and nunphrases that are not topics are indeed foci. First, contrastive wa- and nunphrases have identical prosodic properties to regular foci in the two languages: they carry the same type of tone prominence / pitch boost, trigger suppression of pitch movement on the following material and can be the sole focal accent of the sentence (Tomioka 2010, Kim 2009, and references therein). Secondly, recall that topics must be referential (Reinhart 1981, Endriss 2009). The fact that quantifiers such as nanninka ‘some people’ in (74) and han saram ‘one person’ in (78) can be marked by contrastive wa and nun and receive a non-specific reading suggests that phrases marked by these particles are not inherently topics. Thirdly, a contrastive wa- or nun-phrase can correspond to a wh-expression in the preceding question. We have already observed this in (77) and noted this fact for (80) and (81). If we take this diagnostic test for focus-hood seriously, then one must conclude that such contrastive wa- and nun-phrases are foci. Moreover, focal

Chapter 3 Word order variation and information structure

111

contrastive wa- and nun-phrases have the same distribution as foci marked with focus-sensitive particles. We saw in Section 4.1 that a focus modified by -dake/man can remain in situ or undergo scrambling to a variety of positions and long-distance. The following examples show that focal contrastive wa- and nun-phrases may do so too: (84)

(85)

John-ga Sue-ni nani-o John-NOM Sue-to what-ACC ‘What did John give to Sue?’

ageta gave

no

desu

NMZ

COP

ka?14 Q

(J)

a. John-ga Sue-ni ANO HON-WA ageta. b. John-ga ANO HON-WAi Sue-ni ti ageta. c. ?ANO HON-WAi John-ga Sue-ni ti ageta. that book-WA John-NOM Sue-to gave ‘John gave that book to Sue.’ (Implicature: ‘John did not give the CD to Sue.’)

14 Some Japanese speakers find wh-questions where a nominative subject precedes a wh-phrase, such as (84), less than perfect. A nominative subject in the matrix clause is typically interpreted as either focus or part of focus (Kuno 1973), especially when uttered discourse-initially (Hinds, et al. 1987). Tomioka (2007) analyses the marginality as an instance of intervention effects. There are two ways to circumvent the effects. One is to front the wh-phrase, yielding nani-o Mary-ga Sue-ni ageta no desu ka?. Due to the pitch suppression triggered by the wh-phrase, the subject is interpreted as backgrounded. However, not all of my informants found the question in (84) marginal. Moreover, all speakers found sentences in (85a-b) infelicitous as answers to the question with the alternative word order. This is presumably due to the fact that the subject is explicitly marked as backgrounded in the question, but not in the answer. Another way is to mark the subject with non-contrastive wa, as in John-wa Sue-ni nani-o ageta no desu ka? However, some speakers disallow movement of a contrastive wa-phrase over a non-contrastive wa-phrase in parallel examples to (85c), resulting in ANO HON-WA, John-wa Sue-ni ti agreta, for reasons I do not understand. This is unlikely to be due to Relativized Minimality, however, for reasons discussed in footnote 9. In Vermeulen (To app. q), I report the observation that a non-topical object contrastive wa-phrase cannot move to clause-initial position, using examples in which the subject is marked by noncontrastive wa, and provide a functional account for this observation. In light of the relative acceptability of (85c), however, that analysis may not be on the right track. I leave this issue for future research. To avoid this confound, I provide the question with a nominative subject and with the canonical word order, as in (84). The crucial point here is that what I am claiming to be a focal contrastive wa-phrase may appear in positions other than clause-initial position.

112

Reiko Vermeulen

(86)

John-i Sue-hantey mwu-lul John-NOM Sue-to what-ACC ‘What did John give to Sue?’

(87)

a. John-i Sue-hantey I CHAYK-UN cwuesse. b. John-i I CHAYK-UNi Sue-hantey ti cwuesse. c. I CHAYK-UNi John-i Sue-hantey ti cwuesse. this book-NUN John-NOM Sue-to gave ‘John gave the book to Sue.’ Implicature: ‘John did not give a CD to Sue.’)

(88)

‘What does Bill think that John gave to Sue?’15 a. ANO HON-WAi Bill-wa [CP John-ga Sue-ni that book-WA Bill-WA John-NOM Sue-to omotteiru. thinking b.

cwuesse? gave

(K)

ti

ageta to] gave COMP (J)

I CHAYK-UNi

Bill-i [CP John-i Sue-hantey ti cwuessta-ko] this book-NUN Bill-NOM John-NOM Sue-to gave-COMP sayngkakhan-ta. (K) think-DECL ‘John gave that book to Sue.’ (Implicature: ‘John did not give the CD to Sue.’)

Finally, we saw in the previous section that there are restrictions on the distribution of contrastive topics and foci with respect to each other. These predictions allow us to provide further support for the claim that contrastive wa- and nun-phrases can function as foci, if identified as such by the discourse. Focal contrastive wa- and nun-phrases should not be subject to

15 Foci modified by sae/to, ‘even’ and mo/to ‘also’ can also undergo long-distance scrambling. This fact and the observation in (88) suggest that long-distance scrambling can be licensed by features other than [contrast], contrary to the discussion in Section 4.1. The definition of [contrast] adopted in this chapter implies the negation of at least one salient alternative. However, the meaning encoded by contrastive wa/nun is weaker than that and sae/to, ‘even’ and mo/ to ‘also’ do not imply negation of any alternative. Thus, the definition of [contrast] obviously needs to be modified. One feature of the meaning of these particles that is shared by the definition of [contrast] proposed in Section 2.3 is that there is a specific implicature with respect to the alternatives that are not selected. Perhaps, such a feature suffices to characterise [contrast], and negation of at least one alternative can be viewed as a default interpretation of [contrast]. I will leave this issue for future research.

Chapter 3 Word order variation and information structure

113

the restrictions in (55) and (60), in particular, (55c) and (60a), repeated below: (55)

c. #[CF [ … CT … tCF … ]].

(60)

a. #[CF [ … tCF … CT … ]]

The prediction is correct for both languages. The following exchange illustrates the point for (55c) for Japanese. The utterance in (89) contains a non-specific contrastive wa-phrase in situ in the embedded clause. In correcting this statement, the contrastive focus Sue-o can undergo long-distance scrambling to sentence-initial position, as illustrated in (90). The contrast between (90b) and (70) is unexpected if all contrastive wa-phrases were contrastive topics, as assumed in the standard literature. Moreover, the acceptability of (90) shows that the unacceptability of (70), where a contrastive focus undergoes scrambling across a contrastive topic, cannot be due to Relativized Minimality (Rizzi 1990). In the acceptable (90b) too, the focus moves across a contrastive wa-phrase. (89)

Billj-wa [CPMary-ga [sukunakutomo 3-NIN-NI]-WA Jane-o (J) Bill-WA Mary-NOM at.lesat 3-CL.-to-WA Jane-ACC karej-no mise-de syookai-sita to] itta. he-GEN shop-at introduced that said ‘Bill said that Mary introduced Jane to at least three people in his shop.’

(90)

a. Ie, Billj-wa [CPMary-ga [sukunakutomo 3-NIN-NI]-WA SUE-O no, Bill-WA Mary-NOM at.least 3-CL.-to-WA Sue-ACC karej-no mise-de syookai-sita to] itta no desu yo. (FOC-WA CF) he-GEN shop-at introduced that said NMZ COP PRT b.

SUE-Oi Billj-wa [CPMary-ga [sukunakutomo 3-NIN-NI]-WA ti no, Sue-ACC Bill-WA Mary-NOM at.least 3-CL.-to-WA karej-no mise-de syookai-sita to] itta no desu yo. (CFi FOC-WA ti) he-GEN shop-at introduced that said NMZ COP PRT ‘No, it is Sue that Bill said that Mary introduced to at least three people in his shop.’

?Ie,

Similarly, the following example illustrates the point for (60a). In (91), the embedded direct object is a contrastive wa-phrase in situ. In correcting this statement, the embedded indirect object can undergo long-distance scrambling, moving to a position above the focal contrastive wa-phrase.

114

Reiko Vermeulen

(91)

Billj-wa [CP Mary-ga Jane-ni sukunakutomo Bill-WA Mary-NOM Jane-to at.least [3-NIN-NO HITO]-WA karej-no mise-de syookaisita to] omotteiru. (J) 3-CL-GEN person-WA he-GEN shop-at introduced COMP thinking ‘Billj thinks that Mary introduced at least three people to Jane in hisj shop.’

(92)

a.

Billj-wa [CP Mary-ga SUE-NI sukunakutomo no, Bill-WA Mary-NOM Sue-to at.least [3-NIN-NO HITO]-WA karej-no mise-de syookaisita to] omotteiru 3-CL-GEN person-WA he-GEN shop-at introduced that thinking (ndayo). PRT (CF FOC-WA) b. ?Ie, SUE-NI Billj-wa [CP Mary-ga ti sukunakutomo no, Sue-to Bill-WA Mary-NOM at.least [3-NIN-NO HITO]-WA karej-no mise-de syookaisita to] omotteiru 3-CL-GEN person-WA he-GEN shop-at introduced COMP thinking (ndayo). PRT (CFi ti FOC-WA) Lit.: ‘No, Suei Billj said that Mary introduced at least three people ti in hisj shop.’) ?Ie,

For Korean, I have argued that contrastive nun-phrases are topics only if identified as such by the context and that they are not linked to clauseinitial position, contrary to the predominant view in the literature. It is predicted then that a focal contrastive nun-phrase in clause-initial position should not be subject to the predicted distribution for a contrastive topic in (55c). The prediction is correct. In the following exchange, the contrastive nun-phrase, ceketo sey-myeng-uy haksayng ‘at least three students’, is non-specific and therefore cannot be a contrastive topic. As the example in (94b) shows, a scrambled contrastive focus Yenghi-lul ‘YenghiACC’ may precede it. (93)

Ceketo [SEY-MYENG-UY HAKSAYNG(TUL)]-NUN at least 3-CL-GEN student-PL-NUN mannass-e. saw-DECL. ‘At least three students saw Chelswu.’

Chelswu-lul Chelswu-ACC (K)

Chapter 3 Word order variation and information structure (94)

a. Aniyo. Ceketo [SEY-MYENG-UY HAKSAYNG(TUL)]-NUN No. at least 3-CL-GEN student-PL-NUN (ecey) mannass-e. yesterday saw-DECL

115

YENGHI-LUL Yenghi-ACC (FOC-NUN CF)

b. Aniyo. YENGHI-LULi ceketo [SEY-MYENG-UY HAKSAYNG(TUL)]-NUN ti No. Yenghi-ACC at least 3-CL-GEN student-PL-NUN (ecey) mannass-e. (CFi FOC-NUN ti) yesterday saw-DECL ‘No. At least three students saw Yenghi.’

Moreover, the following example shows that the prediction regarding the restriction in (60a) is also borne out, which demonstrates that not all contrastive nun-phrases are topics, contra C. Lee (2003, 2006): (95)

Chelswu-ka Sue-eykey ceketo [SEY-MYENG-UY HAKSAYNG(TUL)]-NUN Chelswu-NOM Sue-to at least 3-CL-GEN student-PL-NUN (ecey) sokayhayssta. (K) (yesterday) introduced ‘Chelswu introduced at least three people to Sue (yesterday).’

(96)

a. Aniyo. Chelswu-ka No Chelswu-NOM HAKSAYNG(TUL)]-NUN student-PL-NUN

YENGHI-EYKEY ceketo [SEY-MYENG-UY Yenghi-to at least 3-CL-GEN (ecey) sokayhayssta. (CF FOC-NUN) (yesterday) introduced

b. Aniyo. YENGHI-EYEKEYi Chelswu-ka ti ceketo [SEY-MYENG-UY No Yenghi-to Chelswu-NOM at least 3-CL-GEN HAKSAYNG(TUL)]-NUN (ecey) sokayhayssta. (CFi ti FOC-NUN) student-PL-NUN (yesterday) introduced ‘No, Chelswu introduced at least three people to Yenghi (yesterday).’

Thus, the proposal is that contrastive wa- and nun-phrases in general have the particular contrastive interpretation proposed by Hara (2006b), and this interpretation is compatible with both topic and focus. The context determines whether the phrase marked by these particles is interpreted as a topic or a focus. In the case of Japanese, the distinct discourse function is marked syntactically: only topical contrastive wa-phrases obligatorily move to clause-initial position, as we saw in Section 5. Thus, Japanese contrastive topics illustrate clearly that topicality and contrast are two independent attributes of a contrastive topic, as proposed in Section 2. (See also Kuroda 2005 and Tomioka 2010 for similar decompositional approach to contrastive topics in Japanese.)

116

Reiko Vermeulen

So, why are contrastive topics always marked by wa and nun in Japanese and Korean, respectively? I argue that this is because there is a significant overlap in the implicature induced by contrastive wa and nun and the interpretation of a contrastive topic. The implicature induced by the particles, as discussed above, is that the speaker thinks that a stronger scalar alternative to the proposition expressed by the sentence could be false and the interpretation of a contrastive topic is that the speaker is not making an alternative assertion for some reason or other, as discussed in Section 2.3. If the particles are not markers of contrastive topics, one might expect that contrastive topics need not always be marked by these particles. This is indeed true at least for Japanese: they may be marked by nara, as illustrated below (Munakata 2006):16 (97)

Dare-ga pasuta-o tabeta ‘Who ate the pasta?’

no?

(J)

(98)

Hmm, pasuta-wa doo-ka siranai-kedo, suupu-nara MARY-GA well, pasta-WA how-whether know.not-but soup-NARA Mary-NOM tabete ita yo. eating.past PRT ‘Well, I don’t know about the pasta, but as for the soup, Mary was eating (it).’

I am not the first to claim that contrastive wa-/nun-phrases are not always contrastive ‘topics’. For Japanese, some authors refrain from using the term ‘contrastive topic’ and refer to them as ‘contrastive wa-phrases’ or talk in terms of the ‘function’ of contrastive wa (Hara 2006b, Oshima 2008, among others), precisely because they do not appear to be what the rest 16 Note that in English too, the so-called B-accent can mark items that are not topics in the sense of Reinhart (1981). The following examples show that the accent can be used to mark contrast on verbs or quantifiers. It is difficult to see in what sense these non-specific items are what the sentences are about. (i) A: How’s your revision going? B: Well, I [bought]B the book, but I haven’t [READ]A it yet. (ii) A: How many people expressed interest in your house? B: Well, [lots]B of people called, and [three]B looked at it, but [nobody] B made an offer. (modified from McNally 1998: 152) Thus, in English too, it seems plausible that the B-accent only indicates contrast of the type that is compatible with contrastive topics and the topic status of a contrastive topic is identified by discourse. I will not pursue this issue here, but see Molnár (2002), Hetland (2007) and Wagner (2008) for related ideas.

Chapter 3 Word order variation and information structure

117

of the sentence is about. Some even describe them as focus (Endo 2007). However, like other authors who claim that all contrastive wa-phrases are contrastive topics, these authors do not distinguish contrastive wa-phrases displaced to clause-initial position from those appearing elsewhere. Their accounts therefore cannot easily capture the observation that contrastive wa-phrases must occupy clause-initial position if they function as contrastive topics. As mentioned above, contrastive nun-phrases in Korean are sometimes described as contrastive foci. However, this description differs from the standard understanding in the discourse literature of the term ‘contrastive focus’. This is clear from Choi’s (1999) paraphrase provided for the example in (81) above. Contrastive foci identified in a correction context are not loaded with an implicature expressing the speaker’s uncertainty (see Sections 2.3 and 4.1). Moreover, C. Lee (2003, 2008) shows that in a correction context the lexical item that provides the correct information must be case-marked and cannot bear the contrastive particle nun (see also (23)/(24)), as illustrated below: (99)

Yumi-ka atul-ul nah-ass-e. Yumi-NOM son-ACC give.birth.to-PAST-DEC ‘Yumi gave birth to a son’

(K)

(100)

Ani.ya, TTAL-UL/*UN nah-ass-e. no daughter-ACC/NUN give.birth.to-PAST-DEC ‘No, (she) gave birth to a daughter.’ (modified from C. Lee 2003, ex. (46)/(47))

Thus, contrastive nun is not a marker of contrastive focus on the definition adopted in this chapter. In sum, I argue that contrastive wa and nun encode a meaning that is compatible with both topic and focus. Contrastive topics are typically marked by these particles, because the meaning of the particles is very close to and is compatible with that of a contrastive topic. In Japanese, whether a given contrastive wa-phrase is a topic or focus is also marked syntactically: a contrastive topic must move to clause-initial position, but a focal contrastive wa-phrase need not. The observed correlation between the syntactic distribution and the interpretation is difficult to capture on the standard analysis of contrastive wa-phrases in the literature, where they are treated uniformly as contrastive topics. In Korean, where contrastive topics are subject to the mapping rule pertaining to [contrast], the distinction between the two discourse functions is not marked syntacti-

118

Reiko Vermeulen

cally, contrary to the standard view that a contrastive nun-phrase in clauseinitial position must be interpreted as a contrastive topic.

8.

Conclusion

In this chapter, I have examined the syntactic distribution of various discourse-related items in Japanese and Korean. I argued that the differences and similarities between the two languages as well as between different discourse-related items within one language can be explained in a principled manner if we assume the syntactic typology of information structural notions given in (1). Specifically, the two languages have the same rule relevant for the notion [topic], requiring a topic to occupy clause-initial position, and the same rule pertaining to the notion [contrast], allowing a contrastive item to undergo scrambling. The syntactic operations mentioned by the two rules have direct effects at the interface. Placement of a topic in clause-initial position marks its sister constituent as the comment, as a result of either base-generation in or movement to that position, thus creating a transparent mapping between the syntactic structure and the topic-comment structure. Scrambling of a contrastive item overtly marks the scope of contrast for the contrastive item. The two rules cannot be satisfied simultaneously and this gives rise to a conflict for contrastive topics. I argued that where such a conflict arises, a language chooses one rule over the other. Based on comparison with the behaviour of non-contrastive topics and contrastive foci in these languages, I have argued that in Japanese, the rule for [topic] is adopted, while in Korean, the rule for [contrast] is adopted. These patterns are predicted to exist by the typology in (1). Crucially, neither of the languages exhibits patterns that are predicted not to exist, namely patterns where there is a generalisation over non-contrastive topic and regular focus, or non-contrastive topic and contrastive focus, or contrastive topic and regular focus. These patterns are predicted not to exist, as there is no common discourse-related feature in each pair. In addition, I considered several predictions that derive from the current proposal regarding the word order between a contrastive topic and a contrastive focus in sentences that contain both and showed that they are correct. I have also examined the meaning encoded by the particles contrastive wa and contrastive nun, and have presented interpretive and syntactic arguments that the particles are compatible with both topic and focus.

Chapter 4 Encoding focus and contrast in Russian* Elena Titov This chapter investigates the distribution of focused constituents in Russian and provides empirical motivation for the distinct features [contrast] and [focus]. I demonstrate that Russian data support the conclusion that contrastive focus is a composite of features [contrast] and [focus], each of which is responsible for a specific syntactic effect. This conclusion is based on the observation that different types of focused constituents in Russian share an underlying clause-final position. Non-contrastive new information focus consistently surfaces in this position, whereas contrastive focus is allowed to undergo A’-movement to positions further to the left, but reconstructs interpretatively to clause-final position. This suggests that the launching site for the movement of contrastive foci is the position in which new information foci must surface. The behaviour of Russian foci follows if contrastive foci are a composite of the features [focus] and [contrast], while new information foci are characterized by [focus] only. Movement of contrastive foci to the left periphery is then licensed by the feature [contrast], but the launching site of that movement is dictated by the feature [focus]. I argue that the distribution of Russian focus is better analysed as being regulated by mapping principles that favour a particular type of correspondence between representations in syntax and in information structure, rather than as a result of movement to a designated syntactic position in the left periphery triggered by a syntactic feature. The latter approach is not supported by the scopal readings and fails to account for the fact that movement of contrastive focus is optional and can target a variety of positions.

* Many thanks to Klaus Abels, John Fred Bailyn, Gisbert Fanselow, Ivona Kuþerová, Ad Neeleman, Matthew Reeve, Rob Truswell, Hans van de Koot and Reiko Vermeulen for their valuable comments and discussion.

120

Elena Titov

1.

Introduction

This chapter is concerned with the distribution of focused constituents in Russian. Russian has received considerable attention in the linguistic literature due to its relatively free word order. However, the word-order freedom found in this language is not unrestricted. On the contrary, each structure that deviates from the canonical SVO order is used to convey a discourse interpretation that an alternative order fails to express. As a result, a given word order can be acceptable in one context but impossible in another. It is a widely accepted assumption that word-order alternations in Russian result from the need for syntactic structures to reflect the linear ordering of Focus and Background (Adamec 1966, Daneš 1974, Fibras 1964a, 1964b, Hajiþova 1974, Isaþenko 1976a, 1976b, Krylova and Khavronina1988, Mathesius, 1964, Sgall 1972).1 However, an attempt to account for the existing orders on the basis of the respective ordering of only these two discourse interpretations inevitably fails, as the same scrambled orders can be used for quite distinct distributions of Background and Focus (compare (1b) with (1f), and (1c) with (1e)). (1)

a. Ivan þitaet Ivan reads ‘Ivan reads a/the book.’

KNIGU book.ACC

SVO

b. Ivan KNIGU Ivan book.ACC ‘Ivan reads a/the book.’

þitaet reads

SOV

c. KNIGU Ivan book.ACC Ivan ‘Ivan reads a/the book.’

þitaet reads

OSV

d. Knigu þitaet book.ACC reads ‘Ivan reads a/the book.’

IVAN Ivan

OVS

e. Knigu IVAN book.ACC Ivan ‘Ivan reads a/the book.’

þitaet reads

OSV

1 In the Russian linguistic literature the terms background and focus are traditionally replaced with Theme (also occasionally Topic) and Rheme, respectively, which stand for the same interpretations.

Chapter 4 Encoding focus and contrast in Russian f.

IVAN knigu Ivan book.ACC ‘Ivan reads a/the book.’

þitaet reads

121 SOV

Only the sentence in (1a) corresponds to the canonical SVO, with the rest of the sentences representing scrambled orders. As can be seen from the translations, all the sentences in (1) are truth-conditionally identical, suggesting that the reordering of constituents is related to information packaging. At the same time, focus shows a chaotic distribution in (1), surfacing in sentence final (see (1a) and (1d)), intermediate (see (1b) and (1e)), and sentence-initial positions (see (1c) and (1f)). Moreover, in the sentences in (1a)–(1c), the interpretation of focus and background is assigned to the same constituents, as is the case with (1d)–(1f), but the position of focus with respect to the background varies. However, the seeming chaos in the distribution of focus and background in (1) can be resolved if an additional discourse interpretation is taken into consideration that is not captured by the notions focus and background per se. The hypothesis defended in the present chapter is that this additional interpretation is Contrast. As argued below, the syntactic processes involved in creating a scrambled structure with a sentence final focus, as in (1d), are quite distinct from those used for structures with non-final focus, as in (1b) and (1c). In particular, a scrambled structure of the former type exhibits properties of A-scrambling, whereas the latter constructions must be analysed as involving A’-scrambling. Plausibly the interpretation assigned to sentence final focus is distinct from that conveyed by non-sentence final foci, as they license distinct syntactic processes. Indeed, as discussed in section 3, the interpretation of non-final foci in Russian is ‘richer’ than that of clause final focus in that it also includes contrastive reading. The present chapter argues that a canonical structure and an A-scrambled structure with clause final focus, as in (1a) and (1d), respectively, can serve as input to constructions with A’-scrambled non-final focus. Assuming that each order corresponds to a specific information structural interpretation, an A-scrambled OVS structure, as in (1d), can be analysed as resulting from the requirement for background to linearly precede focus (see also Adamec 1966, Hajiþova 1974, Isaþenko 1976a, 1976b, Krylova and Khavronina1988, Mathesius, 1964, Sgall 1972 for a similar observation), whereas supplementing this focus with a contrastive interpretation results in optional A’-fronting (see (1e) and (1f)). Similarly, enriching the focus in the canonical structure in (1a) with a contrastive interpretation also results in optional A’-scrambling (see (1b) and (1c)).

122

Elena Titov

The proposed hypothesis, if feasible, has the following consequences. First, contrast must be analysed as a separate information structural notion that can independently license syntactic operations such as A’movement. That is, to account for the distribution of focused constituents in Russian, the notion of Contrastive Focus (henceforth CF) must be decomposed into the notions of focus and contrast, each of which is responsible for a specific syntactic effect. Second, no designated syntactic position for focus in Russian is conceivable for the following reasons. If it were assumed that the default clausefinal position of Russian focus results from the movement of an argument to the left-peripheral SpecFocusP followed by remnant movement of IP to a higher SpecTopicP (see Wallenberg 2009, to appear, for a corresponding analysis of Heavy NP Shift), one would have to hypothesise an additional KONTRAST-projection above TopP* but below ForceP to account for the distribution of Russian focus. Molnár (2001) postulates such a projection and argues that movement of contrastive categories to the Specifier-position of the KONTRAST-projection is triggered by a K-feature that guarantees “discourse connection” and therefore requires “the absolutely leftmost position”. However, as argued below, such an analysis fails to account for the fact that contrastive interpretation is available for clause-final and intermediate foci. Moreover, the remnant movement account of clausefinal focus is not supported by the scope facts discussed in sections 2 and 3.2 as well as by the Russian data presented in section 4 that show that even non-contrastive focus can surface in a non-default position, whenever a structure with a clause-final focus fails to be generated for a given truthconditional interpretation, strongly suggesting that no designated position for Russian focus is plausible even on an account that assumes a separate KONTRAST-projection in the left-periphery. The chapter is organized as follows: Section 2 examines the distribution of non-contrastive focus in Russian; Section 3 establishes the definition of contrast that is slightly different from that proposed in the introduction to this volume but accounts more accurately for the distribution of contrastive focus in Russian; Section 4 discusses scrambled orders licensed by considerations other than focus encoding; Section 5 proposes an analysis that captures the optional status of A’-scrambling; Section 6 concludes the chapter.

Chapter 4 Encoding focus and contrast in Russian

2.

123

The Distribution of Non-Contrastive Focus in Russian

I will assume two interpretative features, [focus] and [contrast], to characterize the information-structural categories I discuss in this chapter. A key hypothesis on which my analysis will rely is that contrastive and non-contrastive focus share the feature [focus]. While the above features are not intended as syntactic, but merely specify interpretative properties of the categories that carry them, they do have impact on the syntactic distribution of these categories. However, I will argue that these distributive effects come about as a result of mapping principles that relate syntactic structures to information structural representations. In this section, I consider the distribution of focused material and propose a mapping principle that captures the observation that non-contrastive focus in Russian consistently surfaces in clause final position. As a null hypothesis, I propose that Russian is subject to the generalization introduced by Neeleman and Titov (2009): (2)

[Focus] is licensed in clause final position

As already pointed out, the generalization in (2) does not hold on the surface, as contrastive categories typically occupy positions further to the left (see Krylova and Khavronina 1988, King 1995, and Brun 2001). Noncontrastive focus, however, consistently shows up clause-finally:2 (3)

a. [ýto þitaet Anja?]CONTEXT What does Anna read? Anja þitaet KNIGU Anna reads book.ACC ‘Anna reads the/a book.’

Russian SV[O]F

b. [Kto þitaet knigu?]CONTEXT ‘Who reads the/a book?’ Knigu þitaet ANJA book.ACC reads Anna ‘Anna reads the/a book.’

OV[S]F

2 The focused constituents in (3) must surface in clause final position unless they are enriched with emphatic interpretation (Krylova and Khavronina 1988). Emphatic focus is analyzed in the present chapter as contrastive, which accounts for the fact that it is allowed to undergo A’-fronting in Russian (see the discussion around the examples in (17) in the main text).

124

Elena Titov c. [Komu Anja dala knigu?]CONTEXT ‘Who did Anna give a book to?’ Anja dala knigu Anna gave book.ACC ‘Anna gave a book to Catherine.’

KATE Catherine.DAT

SVO[IO]F

d. [ýto Anja dala Kate?]CONTEXT ‘What did Anna give to Catherine?’ Anja dala Kate KNIGU Anna gave Catherine.DAT book.ACC ‘Anna gave a book to Catherine.’

SVIO[O]F

The sentences in (3) are divided into background and focus, with the former consisting of discourse-given material and the latter representing discourse-new information. It must be noted that a description of the notions background and focus in terms of discourse-anaphoricity is not always adequate. Although background arguably cannot include any discourse-new material, focus can quite easily contain discourse-anaphoric material. In fact, the entire focused constituent can be discourse-anaphoric, as shown in (4). (4)

[Kto poceloval ženu Ivana?]CONTEXT Who kissed Ivan’s wife? Ženu Ivana poceloval Ivan’s wife.ACC kissed ‘Ivan kissed Ivan’s wife.’

IVAN Ivan

Russian OV[S]NIF

The sentence in (4) does not contain any discourse-new material. At the same time, it must have some sort of focus, as no focusless sentences are possible. Indeed, the reply in (4) contains information that is not pragmatically presupposed prior to the sentence being uttered, namely that the person whom Ivan’s wife kissed was in fact Ivan.3 3 The term presupposition has traditionally been used in two different fields of linguistics to describe two separate phenomena (see Kratzer 2004 for a discussion of why the two phenomena cannot be collapsed): In semantics, it is used to refer to a condition that has to be fulfilled for a sentence to be either true or false; in works on Information Structure, it denotes the background of a sentence. In the present chapter I use the term presupposition on the latter definition. Hence, a focus of a sentence is always pragmatically non-presupposed, whereas a background is presupposed.

Chapter 4 Encoding focus and contrast in Russian

125

Notably, the requirement for the discourse-anaphoric focus to linearly follow the background licenses an A-scrambled OVS order in (4), strongly suggesting that the notions background and focus are independent from that of discourse-anaphoricity. In (3a-d), the background also consists of pragmatically presupposed material, while the focus offers non-presupposed information. For instance, in (3a), the question in the context presupposes that Anna reads something but it is not known exactly what, whereas the answer contains a non-presupposed part that fulfils the background and turns it into a true proposition (Büring 1997). In other words, the focused constituent in (3a) provides a value for x in ‘Anna reads x’. The examples in (3) and (4) contain a non-contrastive focus or the socalled New Information Focus (henceforth NIF). This type of focus is either not linked to anything in the preceding discourse, or, when used in a question-answer context, is linked to a wh-phrase. Prosodically, NIF is marked with a falling intonational contour across a variety of languages. In Russian, it also bears a falling contour that has been referred to in the Russian linguistic literature as IK1. This intonational contour is similar to IK2 carried by CF but the latter is slightly higher-pitched and more intense (Bryzgunova 1971, 1981). The variation in word order found in Russian sentences with non-contrastive focus exhibits properties of A-scrambling (Titov 2007). That is, it feeds anaphoric binding (see (5a) vs. (5b))4, does not give rise to weak

Rochemont and Culicover 1990 reject the term ‘presupposed’ on the basis of its ambiguity and replace it with ‘c(ontext)-construable’. However, the latter notion fails to distinguish pragmatic presupposition from discourse-anaphoricity. 4 Ionin (2001) argues on the basis of the examples like (i) below that anaphoric binding is impossible in Russian scrambled sentences. (i) a. b.

* Roditeli parents.NOM

drug druga1 videli each other.GEN saw

* Detej1 videli children.ACC saw

roditeli parents.NOM

DETEJ1 children.ACC

SVO

DRUG DRUGA1 OVS each other.GEN (Ionin 2001 : 44)

However, careful examination of the Russian data reveals that the Russian reciprocal resists being embedded in an animate argument carrying the most prominent ș-role in the predicate’s argument structure. This suffices to explain the ungrammaticality of (i). It is beyond the scope of the present chapter to investigate this selective behavior of the Russian reciprocal. What matters is

126

Elena Titov

crossover effects (see (6)), is clause-bound (not demonstrated here), and does not give rise to scope-reconstruction (see (7)).5,6,7 (5)

a. *Vystrely shots.NOM

drug druga1 ubili each other.GEN killed

MILICIONEROV1 milicia-men.ACC

b. Milicionerov1 ubili vystrely DRUG DRUGA1 militia-men.ACC killed shots.NOM each other.GEN ‘Militia men were killed by each others shots.’ Každuju devoþku ljubit every girl.ACC loves ‘Every girl is loved by her mum.’

(6)

(7)

a. Každuju otkrytku podpisali every postcard.ACC signed ‘Two students signed every postcard.’

eë her

MAMA mum

SVO OVS

OVS

[DVA STUDENTA]NIF two students > ; ?>

b. Dve otkrytki podpisal [KAŽDYJ two postcards.ACC signed every ‘Every student signed two postcards.’

STUDENT]NIF student >; *> 

that embedding the reciprocal in an inanimate argument, as in (5b), results in a grammatical sentence, strongly suggesting that anaphoric binding is possible in Russian neutrally scrambled sentences. The hypothesis that the Russian reciprocal resists being embedded in an animate argument carrying the most prominent ș-role in the predicate’s argument structure is further supported by the data in (ii) below (I am grateful to Klaus Abels for suggesting this example to me). Here, the sentence differs from the one in (5b) only in the interpretation of the subject embedding the reciprocal with respect to animacy. As can be seen from (ii), embedding the reciprocal in an animate argument renders a sentence of the type given in (5b) ungrammatical: (ii) *

Milicionerov1 ubili kollegi drug drUga1 OVS militia-men.ACC killed colleagues.NOM each other.GEN 5 In (7a), the apparent wide scope reading of the existential quantifier is accessible due to the availability of a specific interpretation for the indefinite. 6 Nominative case marking is indicated in the present chapter only in possessive constructions. In all other cases, an NP whose case is not indicated should be understood as carrying nominative case. 7 The OVS constructions in (5b) and (6) are translated as passives because this is the only way to capture the binding properties of these sentences in English.

Chapter 4 Encoding focus and contrast in Russian

127

It has been claimed that scope reconstruction and WCO effects are unreliable tests for an A-position in Russian because it has so-called ‘frozen’ scope and obviates WCO effects in general (King 1995, Ionin 2001, Bailyn p.c., 2004a). However, the examples in (8) below demonstrate that WCO violations and scope reconstruction obtain whenever an A’-moved quantifier undeniably crosses an argument, suggesting that the scrambled sentences that are taken to have ‘frozen’ scope or to lack WCO violations involve reconstruction of an A’-moved object to an A-position above the sentence-final focused subject, as in (9) below (Titov 2007).8 (8)

devoþku]l, girl.ACC

eë1 her

þtoby tl that

poceloval kissed

IVAN Ivan

[Každuju every

devoþku]l, girl.ACC

dva two

a. * [Každuju every

b.

mama mum

xoþet, wants

mal’þika boys

xotjat, want

þtoby tl poceloval IVAN that kissed Ivan ’Two boys want every girl to be kissed by Ivan.’ (9)

a.

[Každuju every

devoþku]l, girl.ACC

ty you

>; *>

xoþeš’, want

þtoby tl pocelovala EË1 MAMA that kissed her mum ’You want every girl to be kissed by her mum.’ b.

[Každuju

devoþku]l,

Ivan

xoþet,

every

girl.ACC

Ivan

wants

þtoby tl pocelovali DVA MAL’ýIKA that kissed two boys ‘Ivan wants every girl to be kissed by two boys.’

> ; ?>

Russian A-scrambling can be analyzed either as resulting from A-movement (Bailyn 2004a, King 1995) or from variation in the base component (Titov 2007).9 However, to sustain the former analysis, it must be stipu-

8 The fronted objects in (8) and (9) are interpreted as Contrastive Topics. 9 Richards’ (2008) ‘tucking in’ analysis of A-scrambling is not discussed here due to space limitations.

128

Elena Titov

lated that there is no scope reconstruction in A-chains.10 Regardless of the approach taken, base-generation or A-movement with no scope reconstruction, A-scrambled structures are marked with respect to those exhibiting neutral order (cf. the idea of movement as a Last Resort in Chomsky’s version of Minimalism, Chomsky 1995). Since both an unmarked structure and a marked one exist in the language, economy considerations demand that the latter is generated only to achieve an interpretation that the former fails to convey. In other words, a marked structure must receive an interpretative license (cf. Neeleman and van de Koot 2008). The interpretative license for Russian scrambled constructions can be provided by a mapping rule that maps a structurally flat Information Structure (henceforth IS) onto a syntactic structure in such a way that an argument in a scrambled position is interpreted as more prominent with respect to an argument in the position across which scrambling takes place: (10)

Argument prominence mapping rule Provided that no syntactic restrictions are placed on generation of an unmarked structure with a given truth-conditional interpretation, interpret an A-scrambled structure as reflecting the relative prominence of two arguments, where an argument in a scrambled position is construed as discourse-prominent and an argument in the position across which scrambling takes place as not prominent.

I would like to argue that in scrambling languages such as Russian the relative prominence of arguments can be established on the basis of not only theta prominence but also discourse prominence; with the former predicting the unmarked order of arguments and the latter licensing scrambled orders. To be precise, an argument can be construed as either prominent or not prominent on the basis of the discourse interpretation it is linked to. For instance, an argument that conveys information that is already part of a Common Ground can be understood as having been made prominent through context, whereas an argument representing information that is not yet known or not yet taken for granted cannot. If so, the generalization in (2) can be understood as following from the mapping rule in (10), because the relative prominence of arguments must be

10 It must also be stipulated that this type of A-movement is not subject to locality restrictions, as it allows for A-movement of NPs/DPs across c-commanding NPs/DPs.

Chapter 4 Encoding focus and contrast in Russian

129

reflected by the linear precedence of the prominent, pragmatically presupposed argument with respect to the non-prominent, non-presupposed argument.11 The resulting information structure is mapped onto syntax, with the outcome that the non-discourse-prominent NIF consistently follows the prominent background. Furthermore, given the right-branching structure of Russian clauses, the mapping rule also determines that the prominent argument outscopes the non-prominent one (see (7)). So far we have only considered the distribution of NIF with respect to background. The next section looks in more detail at the distribution of contrastive focus with the aim to provide a coherent definition of contrast that captures the differences in the syntactic behaviour between contrastive and non-contrastive foci in Russian.

3.

Contrast

3.1

What does it mean to be contrastive?

In the introduction to this chapter it was proposed that Russian nonclause-final foci are enriched with a contrastive reading. Before this can be demonstrated, it is necessary to establish what it means for a focus to be contrastive. The standard assumption in the literature on information structure is that for a constituent to be interpreted as contrastive, it must be construed as belonging to a contextually salient set of alternatives (Chafe 1976, Jackendoff 1972, Halliday 1967, Rooth 1985, and Rooth 1992). I will refer to a set of alternatives that is contextually salient as a pragmatic set of alternatives, as opposed to a semantic set of alternatives, which is usually taken to form the basis for the interpretation of foci generally (Krifka 2008). Unlike a semantic set, a pragmatic set of alternatives is relevant specifically for the discourse under consideration (see Wagner 2007 for a similar idea). The hypothesis put forward here is that, although the semantic interpretation of focus might involve selection from a set, for a focus to be contrastive, the set of alternatives must become active in the discourse at the point the sentence containing the contrastive element is uttered, no sooner and no later. That is, it must be indicated either through a link to the context or within the utterance itself that the set to which the 11 The discourse interpretation of the verb is not discussed here because it does not seem to have any effect on the structural position of the verb. In monotransitive constructions the verb is consistently sandwiched between the two arguments: S[VO]FOC, SV[O]FOC, O[VS]FOC, OV[S]FOC.

130

Elena Titov

focused constituent belongs indeed contains alternative members that are relevant for the discourse at hand. For example, the focused object NPs in (11) are not contrastive because no contextually salient set of alternatives is available for them. (11)

[Kogo Ivan pokormil?]CONTEXT Who did Ivan feed? a. Ivan pokormil Ivan fed ‘Ivan fed the cat.’

KOTA cat.ACC

(a nasþët Ivana ne znaju) b. [Boris]TOP pokormil KOTA Boris fed cat.ACC (but I don’t know about Ivan) ‘Boris fed the cat (but I don’t know about Ivan).’

The NIF ‘cat’ provides a value for a variable introduced by the ‘wh-phrase’ in the context, but it is not indicated either through a link to the context or within the utterance itself that there are alternative members of the set to which ‘cat’ belongs that are relevant for the discourse at hand. That is, it is not made explicit by the utterances in (11) that for the proposition ‘y fed x’, more than one entity is competing for x. Importantly, the interpretation of a non-contrastive focus is not necessarily exhaustive and further members of the set to which ‘cat’ in (11) belongs can be added in the following discourse. What is crucial for the non-contrastive reading is merely that the utterance containing a non-presupposed element does not pragmatically ‘activate’ a set that contains the non-presupposed element along with at least one alternative member. Similarly, in (12), the focused object is not contrastive because the set to which it belongs and which contains at least two members is made salient before the relevant sentence is uttered. (12)

[Ivan pokormil kota ili sobaku?]CONTEXT Did Ivan feed the cat or the dog? Ivan pokormil Ivan fed ‘Ivan fed the cat.’

KOTA cat.ACC

In (12), the reply does not activate the interpretation that the object belongs to a pragmatic set along with alternative members. This interpretation is already activated by the contextual question, which presupposes that alternative properties hold of ‘cat’ and ‘dog’ (i.e. one has the property of being fed by Ivan and one doesn’t). Therefore, the object is interpreted as non-presupposed, discourse-anaphoric and non-contrastive.

Chapter 4 Encoding focus and contrast in Russian

131

Notably, if the contextual question in (12) introduced other members of the set of alternatives but not ‘cat’, whereas the reply conveyed that ‘cat’ also belongs to this set, the object would be construed as contrastive because in that case, the interpretation that the object belongs to a set of alternatives would be activated by the reply despite the set being introduced by the context: (13)

[Ivan pokormil xomjaka ili sobaku?]CONTEXT Did Ivan feed the hamster or the dog? Ivan [KOTA]CF 1 pokormil t1 (a ne xomjaka i ne sobaku) Ivan cat.ACC fed (and not hamster and not dog) ‘Ivan fed the cat (not a hamster or a dog).’

As expected, unlike the non-contrastive foci in (11) and (12), a contrastive focus in (13) is allowed to A’-scramble. Here it moves to a position immediately before the verb. It is, however, possible for it to move further to the left, suggesting that this type of movement does not target a particular structural position but rather involves adjunction to any maximal projection (see section 5 for a discussion of the length of A’-chains). Similarly, in (14) and (15a), the non-presupposed objects must be construed as contrastive: In (14), the proposition ‘Ivan fed x’ has two contextually salient members of a set that compete for x, ‘dog’ and ‘cat’. This set is not activated until the mention of ‘cat’ in the answer.12 (14)

[Ivan pokormil sobaku?]CONTEXT Did Ivan feed the dog? a. [Kota]CT1 Ivan POKORMIL t1 (a nasþët sobaki ne znaju) cat.ACC Ivan fed (and about dog not know) ‘Ivan fed the cat (but I don’t know about the dog).’ b. (Net,) Ivan [KOTA]CF1 (no) Ivan cat.ACC ‘Ivan fed the cat (not the dog).’

pokormil t1 fed

(a ne sobaku) (and not dog)

12 The object in (14a) has the interpretation of a contrastive topic. Contrastive topics are also characterized by the feature [contrast] and therefore A’-scramble in Russian. The structures that host contrastive topics are, however, interpretatively distinct from those containing a contrastive focus. Since the present chapter is concerned with the distribution of Russian focus, contrastive topics are mentioned here only to demonstrate the properties of Russian contrastive categories.

132

Elena Titov

In (15), the question in the context contains a plural noun. The reply in (15a) treats this noun as generalizing over a set of animals to which ‘cat’ belongs along with alternative members. Importantly, the construal of the noun ‘animals’ in the context as generalizing over a set of alternatives is activated by the sentence that contains ‘cat’ and not by the contextual question itself. As demonstrated in (15b) and (15c), the following discourse can treat this noun as a member of a set of alternatives (see (15b)), or refer back to it without invoking contrast, as in (15c).13 (15)

[Ivan pokormil životnyx?]CONTEXT Did Ivan feed the animals? a. [Kota]CT1 Ivan

POKORMIL t1 (a nasþët ostal’nyx životnyx ne znaju) cat.ACC Ivan fed (but about remaining animals not know) ‘Ivan fed the cat (but I don’t know about the rest of the animals).’

b. Net, Ivan [LJUDEJ]CF1 no Ivan humans.ACC ‘No, Ivan fed the humans.’

pokormil t1 fed

c. Da, Ivan POKORMIL yes Ivan fed ‘Yes, Ivan fed the animals.’

životnyx animals.ACC

Crucially, the set of alternatives to which the object ‘cat’ belongs in (14) and (15a) becomes active at the point the utterances containing it are produced, no sooner and no later.14 The interpretation of contrastive constitu13 The sentences in (14a), (15a) and (15c) contain non-contrastive verum focus, i.e. focus on the truth-value, which is prosodically realized on the inflection. 14 The reason the set of alternatives must be activated no later than the mention of the relevant focused category lies in the fact that a late activation of the set is possible for non-contrastive focus, as in (i) below. Here, the set for the noncontrastive new information focus ‘cat.ACC’ is activated at the time ‘dog. ACC’ is mentioned. This, however, does not have an effect on the interpretation of the focus as non-contrastive, which is further confirmed by the fact that such focus does not move and is marked with IK1. (i) [Kogo Ivan pokormil?]CONTEXT Who did Ivan feed? [Boris]TOP pokormil KOTA a Ivan sobaku Boris fed cat.ACC and Ivan dog.ACC ‘Boris fed the cat and Ivan fed the dog.’

Chapter 4 Encoding focus and contrast in Russian

133

ents in (14) and (15a) is not exhaustive. For instance, the CF in (14b) is not interpreted as the only member of the set of animals that has the property of having been fed by Ivan. That is, the construal of (14b) is such that the dog has not been fed by Ivan, whereas the cat has been fed by him and some other animals might have been fed by him as well. In fact, more members of the set the CF belongs to can potentially be present in the discourse, suggesting that the semantic set for a CF is not closed (contra Kiss 1998, Halliday 1967, Chafe 1976 and Rooth 1992). What distinguished the contrastive constituents in (14) and (15a) from NIF in (11) and (12) is not the nature of the semantic set they belong to, or whether the number of members is limited in this set, but the fact that the utterance that hosts the former activates the interpretation that the set they belong to contains alternative members relevant for the discourse at hand, whereas the utterance hosting the latter either refers back to an already introduced set, as in (12), or treats the NIF as the only member of its pragmatic set (see (11)). As already mentioned, in Russian, contrastive constituents are allowed to undergo A’-fronting. Notably, the focused objects in (11), and (12) have to remain in situ (in their thematic postverbal positions), strongly suggesting that they are not contrastive. Conversely, in (13), (14) and (15a,b), the contrastive constituents are allowed to move (although not demonstrated here, they can also undergo long-distance movement). Membership of a pragmatic set of alternatives can be indicated not only through a link to an alternative member of a set in the context, as in (14) or a superset, as in (15a); it can also be specified with the help of a special marker attached to a constituent that signals that this constituent belongs to a set along with alternative members (see (16)). Here, attachment of either a prosodic marker (see (16a) where the subject is marked with a B-accent) or a morphological marker (see (16b)) to a discourseanaphoric constituent activates the interpretation that there is at least one more member of the set it belongs to that is significant for the exchange at hand. (16)

a. [What did the teachers drink at the party?]CONTEXT [The teachers]CT drank WATER, (but I am wondering what the students drank) b. [ýto uþitelja pili na veþerinke?]CONTEXT What did teachers drink at the party?

134

Elena Titov [Uþitelja-to]CT

pili

teachers – TO

drank

VODU,

(a vot interesno, þto studenty pili) water.ACC (but interesting what students drank). ‘The teachers drank water (but I wonder what the students drank).

In (16), membership of a pragmatic set of alternatives is signalled not through a link to the preceding context but by a property of the utterance itself. Similarly, sentences containing a so-called Emphatic Focus (henceforth EF) also activate the interpretation that this type of focus belongs to a pragmatic set of alternatives but this time it is done not with the help of a special marker on the focused constituent but via its marked structural position (see (17)). (17)

a. [Kogo ty tol’ko þto videl?]CONTEXT Who did you just see?

Russian

(Predstavljaeš’,) (Imagine) ja tol’ko þto [ýELOVEKA S RUŽ’ËM]EF1 videl I just man.ACC with gun saw ‘(Can you imagine?) I just saw a man with the a gun!’

t1 !

b. [ýto ty loviš’?]CONTEXT What are you fishing for? Ja [RYBU]EF1 lovlju t1 (þto že ešþë) ! I fish.ACC catch (what else) ‘I’m fishing for fish (what else can I be fishing for)!’

Here, a focused constituent cannot be interpreted as NIF because, unlike non-contrastive focus, it surfaces in a preverbal position. Yet, the contrastive interpretation is not achieved through a link to the context. Crucially, the focus in (17) must be construed as occupying a certain scalar position with respect to all other members of the set it belongs to, each of which could potentially fulfil the background. Logically, only two such positions can be indicated when the alternative members stay implicit, the lowest and the highest. According to the first reading, the focused constituent is interpreted as the weakest member of a set with respect to all other potential members (see (17a)); the second reading, in contrast, interprets the focused object as the strongest member of its set (see (17b)). The lowest scalar position of the non-presupposed constituent in a set of alternatives in (17a) conveys surprise as to the fact that out of a set of individuals the speaker expected to see, it was the least expected ‘a man

Chapter 4 Encoding focus and contrast in Russian

135

with the gun’ that was seen. The interpretation of the non-presupposed object in (17b), on the other hand, is the directly opposite one. This time the focused constituent is perceived as the strongest member as regards all other potential members of the set of alternatives. That is, out of the set of objects that one can be expected to be fishing for, ‘fish’ is the most obvious choice. It can therefore be said that in (17b) the interpretation is not that of surprise but rather of annoyance as to the fact that one is asked a question that has a rather obvious answer. Since EF often occurs out of the blue or in a context that does not force a contrastive interpretation (see (17)), it has traditionally not been grouped together with contrastive categories. However, EF must be interpreted as belonging to a pragmatic set of alternatives. This reference is achieved by appealing to the shared knowledge of interlocutors about the scalar position of the focused constituent with respect to other potential members of the relevant set.15 Plausibly, no item can be perceived as occupying either the highest or the lowest position in a set if this set does not contain alternative members. Since alternative members must become active in the discourse for such an interpretation to be available, our definition of contrast suggests that such foci must be analysed as contrastive. The analysis of contrast as membership of a pragmatic set of alternatives activated by the utterance containing the relevant non-presupposed element groups contrastive topic, contrastive focus and emphatic focus together as contrastive, whereas simple new information focus must be analysed as associated with a non-contrastive reading. At the same time, all these categories are interpreted as not pragmatically presupposed. In the majority of cases, they consist of discourse-new material, and can therefore not be construed as belonging to the background of a sentence. Moreover, in the rare instances where they are discourse-anaphoric, they still convey information that is not known or taken for granted. Thus, a discourse-anaphoric NIF consistently fulfils the background by providing a value for the variable introduced by a wh-phrase (see (4)), whereas contrastive categories are non-presupposed simply in virtue of being contrastive. That is, even when a contrastive interpretation is assigned to a discourse-anaphoric constituent, as in (16), it provides this constituent with the non-presupposed information that it must be construed as 15 The fact that the knowledge about the scalar position of EF must be shared by the interlocutors is confirmed by the observation that whenever the hearer is unaware of it, a sentence with a moved focus is perceived as odd in contexts that do not license contrast and requires clarification (i.e. the speaker is perceived as being either surprised or annoyed for no apparent reason).

136

Elena Titov

belonging to a pragmatic set of alternatives. As this information is not known or taken for granted before the utterance hosting a contrastive constituent is uttered, the contrastive interpretation must itself be nonpresupposed. The proposal advanced in this section contradicts the analysis advocated in Bolinger (1961) and Dretske (1972) that sees all foci as contrastive. Contrast is seen here as an extra interpretative property that can be added to a focused constituent and that licenses it to undergo A’-fronting. At the same time, it follows from the theory proposed here that only pragmatically non-presupposed constituents can be contrastive, as contrastive interpretation is itself non-presupposed.

3.2

The Distribution of Contrastive Foci16

One way of capturing the observation that NIF and CF are both pragmatically non-presupposed is to propose that they share the feature [focus]. If so, they are both subject to the generalization in (2) in Russian. However, as already mentioned, this generalization does not hold on the surface, as CFs (including EFs) are typically fronted.17 This misalignment can easily be captured if it is assumed that NIF and CF are both merged in clause-final position, but that a CF may undergo subsequent A’-scrambling in virtue of also carrying the feature [contrast]. This proposal is summarized in (18).18 (18)

a. [CP … [focus]] b. [CP [focus, contrast]1 … t1]

This hypothesis is indeed supported by the scopal properties of CFs, and in particular by the observation that they take scope in the same position

16 This subsection is partially based on Titov (2007) and Neeleman and Titov (2009). 17 The judgments in the main text presuppose that the constituents marked as CF bear IK2, while the rest of the sentence is destressed. For further discussion, see Bryzgunova 1971, 1981, Yokoyama 1986, Pereltsvaig 2004, and Krylova and Khavronina 1988. 18 Movement is assumed to be licensed rather than triggered by the feature [contrast], since the relevant type of A’-scrambling is optional in many languages including Russian. For further discussion of the issue see section 5 as well as Neeleman and Van de Koot 2008 and Neeleman et al. 2008.

Chapter 4 Encoding focus and contrast in Russian

137

as NIFs. The data fall out from (10), if A’-scrambling obligatorily reconstructs for scope.19 Recall that Russian exhibits surface scope in sentences with non-contrastive focus, as, in line with the rule in (10), quantifiers that constitute NIF scope under quantifiers that belong to the background (see (7)). The pattern of surface scope breaks down in the case of contrastive foci. Even though these are fronted, they systematically take lowest scope. That is, they reconstruct obligatorily to a position below backgrounded quantifiers. Thus, the fronted CF in (19b) takes scope in exactly the same position as the clause final new information focus in (7b).20 (19)

a. KAŽDUJU OTKRYTKU1, every postcard.ACC

ja I

xoþu, want

þtoby that

dva studenta podpisali t1, (a ne každuju knigu) two students signed (and not every book) ‘I want two students to sign every postcard (not every book).’ >; *> b. KAŽDYJ every

STUDENT1 student I

ja want

xoþu, that

þtoby

dve otkrytki podpisal t1, (a ne každyj docent) two postcards.ACC signed (and not every lecturer) ‘I want every student to sign two postcards (not every lecturer).’ >; *> 19 That A’-moved focused constituents obligatorily reconstruct is apparent from examples like (i) below, which are unambiguous: the indefinite cannot be dependent on the universal. For further discussion, see Neeleman & Van de Koot 2008. The same judgment holds for Russian (contra Bailyn 2001; see (ii) for A’-moved CF and (8b) in the main text for A’-moved CT). (i) [Every bOy]CF1 two girls said [that Mary kissed t1]

>; *>

(ii) [Každogo mAl’þika]CF1 dve devoþki xotjat, [þtoby Maša every boy.ACC two girls want that Masha pocelovala t1] kissed ‘Two girls want every boy to be kissed by Masha.’ >; *> 20 The object in (19b) is in a scrambled preverbal position due to its interpretation of belonging to the background. Keeping the object in a postverbal position gives the sentence a somewhat degrading status, presumably due to distinctness (Richards 2001). However, it is possible for the object to remain in a postverbal position without affecting the scopal judgments as long as the moved subject is marked with IK2 and the rest of the sentence is left stressless.

138

Elena Titov

It is not surprising that Russian contrastive foci can move. In a wide range of languages, contrastive elements undergo A’-movement. What is surprising is that the position into which contrastive foci reconstruct should be as low as it seems to be. That is, in an all-focus sentence, a subject outscopes an object in Russian because its position c-commands (and precedes) the object position.21 However, when subjects are fronted as contrastive foci, they scope under the object, suggesting that the fronting operation is launched from a position below the object. This observation falls into place if the rule in (10) is taken into consideration: a backgrounded object in Russian is expected to A-scramble across a focused subject, explaining the scopal properties of fronted foci. In other words, an A-scrambled structure, as in (7b), where a backgrounded object is interpreted as more prominent than a focused subject in line with (10), serves as input to A’movement of the focused subject when the latter is contrastive (see (19b)). The above analysis extends to object-across-object A-scrambling found in ditransitive constructions. These also exhibit surface scope in the absence of contrast. That is, an object quantifier that constitutes new information focus scopes under the other object quantifier that belongs to the background, in accordance with (10): (20)

a. Ivan podaril dvum studentam KAŽDUJU Ivan gave two students.DAT every ‘Ivan gave two students every postcard.’

OTKRYTKU postcard.ACC >; *>

b. Ivan podaril dve otkrytki KAŽDOMU STUDENTU Ivan gave two postcards.ACC every student.DAT ‘Ivan gave two postcards to every student.’ >; *>

Whenever the objects are A’-fronted as contrastive foci, the pattern of surface scope once again breaks down. That is, an object with the interpretation of contrastive focus is consistently interpreted below an object that belongs to a background, suggesting that the launching site for movement of the former is the position where NIF surfaces:

21 Due to space limitations, I am ignoring all-focus constructions where the relative prominence of arguments is established on the basis of interpretations not discussed in the present chapter and where A-scrambling is licensed by these interpretations.

Chapter 4 Encoding focus and contrast in Russian (21)

a. KAŽDUJU OTKRYTKU1 every postcard.ACC

ja I

xoþu, want

þtoby that

139

Ivan Ivan

podaril dvum studentam t1, (a ne každuju knigu) gave two students.DAT (and not every book) ‘I want Ivan to give two students every postcard (not every book).’ >; *> b. KAŽDOMU STUDENTU1 every student.DAT

ja I

xoþu want

þtoby that

Ivan Ivan

podaril dve otkrytki t1 (a ne každomu docentu) gave two postcards.ACC (and not every lecturer) ‘I want Ivan to give two postcards to every student (not every lecturer).’ >; *>

There is an ongoing debate concerning the unmarked order of Russian objects, but – regardless of the position taken – the examples in (21a) and (21b) cannot both involve reconstruction to an unmarked thematic position. Therefore, one of the examples in (21) must involve reconstruction of a focused object below an A-scrambled backgrounded object in accordance with (10). The above data were presented in terms of the features [focus] and [contrast]. In fact, this is more than a presentational convenience: the Russian data support a decompositional view of contrastive foci. In order to explain why contrastive foci and new information foci share an underlying position, we must assume that they share some attribute. In order to explain why only contrastive foci move, we must assume that they have an additional property. This is exactly what the features [focus] and [contrast] are meant to express. The observation that Russian foci can surface in a variety of syntactic positions depending on the presence/absence of a contrastive reading but are consistently interpreted below background in accordance with (10) undermines the idea that a syntactic constituent obtains the interpretation of focus via movement to a designated focus projection. If the cartographic approach discussed in the introduction to this chapter is assumed, where non-contrastive foci move to SpecFocusP, contrastive categories move to the higher SpecKontrastP and the remnant backgrounded IP to SpecTopP that follows the latter but precedes the former, it is unclear how the abovediscussed scope readings obtain. Crucially, neither surface scope nor reconstruction to thematic positions would give the required effect. The former would result in the widest scope for contrastive focus, whereas the latter would be reliant on the thematic interpretations and therefore the

140

Elena Titov

grammatical functions of the moved arguments rather than their information-structural interpretations. Moreover, as will be argued in the next section, considerations such as binding can override structural encoding of interpretative prominence resulting in focus being realised in a position other than the default clausefinal position. If correct, this further undermines the idea of a designated syntactic position for focus.

4.

Binding

Although any structure in Russian aims at being faithful to (2), this cannot always be achieved. Even unmarked structures often fail to be generated despite adhering to (2). To be precise, whenever there is an independent requirement for a prominent argument to occur in the scope of a nonprominent argument, this overrides any type of information structural encoding including focus-background encoding: (22)

[Kogo ljubjat ego roditeli?]CONTEXT Who is loved by his parents? a.

IVANA1 ljubyat ego1 Ivan.ACC love his ‘Ivan is loved by his parents.’

b. * Ego1 his

roditeli parents

ljubyat love

roditeli parents

OVS

IVANA1 Ivan.ACC

SVO

In (22a), the pronoun embedded in the discourse-anaphoric subject is bound by the object.22 In (22b), however, the pronoun is not c-commanded by the object, rendering binding impossible. Moreover, the pronoun cannot refer to ‘Ivan.ACC’ through coreference either in (22b). First, the structure in (22b) does not obey the General pattern of anaphoric dependence (Williams 1997) given in (23) below, as the pronoun neither follows

22 That (22a) involves binding is supported by the observation that a sloppy reading is available (and in fact favored) in a structure involving VP-ellipsis, as in (i) below: (i) IVANA1 ljubyat ego1 roditeli i BORISA tože Ivan.ACC love his parents and Boris.ACC too ‘Ivan is loved by his parents and Boris is too (=Boris is loved by his own parents not Ivan’s).’

Chapter 4 Encoding focus and contrast in Russian

141

its antecedent nor is it in a subordinate clause. That is, the structure in (22b) corresponds to the ill-formed structure in (23b). Second, the object in (22) is discourse-new and it is the constituent that forms the focus of the sentence and bears the main sentence stress. It can therefore not be interpreted as referring back to a discourse-antecedent that also serves as an antecedent for the embedded pronoun. (23)

General pattern of anaphoric dependence (Williams 1997 : 26) a. [ … pro … ]subord [ … antec … ]subord b. * [ … pro … ]matrix [ … antec … ]matrix c. [ … antec … ]matrix [ … pro … ]subord d. [ … antec … ]subord [ … pro. … ]matrix

Interestingly, the sentence in (22b) is ill formed even when the object is destressed, as in (24a). (24)

a. * Ego1 his

roditeli parents

LJUBJAT love

Ivana1 Ivan.ACC

SVO

b.

Ivana1 Ivan.ACC

LJUBJAT love

EGO1 his

RODITELI parents

O[VS]FOC

c.

Ivana1 Ivan.ACC

ljubjat love

EGO1 his

RODITELI parents

OV[S]FOC

The English variant of (24a) is grammatical presumably because the destressed object can be interpreted as discourse-anaphoric simply in virtue of not forming a part of the constituent carrying the main sentence stress. It is therefore plausible that there is a discourse-antecedent for the object in the English variant that also serves as an antecedent for the pronoun (see also Williams 1997 for the same conclusion). In Russian, a backgrounded object can occur in constructions with different information structure but none of them have the word order in (24a). That is, the object can either occur in a sentence where it is the only backgrounded argument, as in (24b) and (24c), or, the focus can be expressed on the verb, with both arguments belonging to the background. Let’s consider both options. In the former case, the backgrounded object has to A-scramble across the focused subject in accordance with (10), as in (24b) and (24c). The SVO word order, as in (24a), can therefore not capture this information structure.

142

Elena Titov

In the case with a focused verb, both arguments belong to the background but for the object to follow the subject containing the pronoun, the sentence has to occur in a context that already contains an ungrammatical sentence: (25)

* [Ego1 his

roditeli parents

* Net, no

ego1 his

nenavidjat hate

roditeli parents

Ivana1]CONTEXT Ivan.ACC

LJUBJAT love

Ivana1 Ivan.ACC

SVO

This is because Russian requires a parallelism between the word order within the background of a sentence and the word order in the immediately preceding context. Changing the argument order within the background results in a degraded acceptability: (26)

[Ivana1 Ivan.ACC a. ??/* Net, no b.

nenavidjat hate ego1 his

ego1 roditeli]CONTEXT his parents roditeli parents

LJUBJAT love

Net, Ivana1 LJUBJAT ego1 no Ivan.ACC love his ‘Ivan is loved by his parents.’

Ivana1 Ivan.ACC

SVO

roditeli parents

OVS

The condition that disallows any inconsistency between the argument order within a background and the argument order in the immediately preceding context is captured in (27). (27)

Argument Order Consistency Condition The order of arguments within a background has to reflect the order of arguments in the immediately preceding context

The examples in (25) and (26) contain a contrastive focus. As can be seen from (28), the condition in (27) also rules out inconsistent structures with NIF on the verb. (28)

[Kak otnosjatsja k Ivanu ego roditeli?]CONTEXT How is Ivan treated by his parents? a. ??/* Ego1 his b.

roditeli parents

Ivana1 LJUBJAT Ivan.ACC love

Ivana1 ego1 roditeli Ivan.ACC his parents ‘Ivan is loved by his parents.’

LJUBJAT love

SVO OVS

Chapter 4 Encoding focus and contrast in Russian

143

As can be seen from the above examples, all the constructions where a pronoun embedded in a subject precedes a discourse-anaphoric object that is coreferenced with the pronoun are ruled out in Russian for independent reasons. Since neither coreference nor binding are possible for (22b), the unmarked structure is replaced by the scrambled one in (22a) whenever the pronoun and the object denote the same individual. The grammatical sentence in (22a), however, does not reflect the relative prominence of arguments, as the focused object NP linearly precedes the backgrounded subject NP. Note that the rule in (10) demands a specific interpretation for an A-scrambled structure only when no syntactic restrictions are placed on the generation of an unmarked structure with the same truthconditional interpretation. However, we have just established that it is impossible to have an unmarked structure where the pronoun embedded in a subject denotes the same individual/entity as the object in Russian. Consequently, generation of a scrambled structure for this truthconditional interpretation is the only option, as no simpler structure with this interpretation can be created. The A-scrambled structure in (22a) must therefore be licensed not by (10) but by other considerations. By hypothesis, the requirement for the subject to occur in the scope of the focused object in order to be bound by it (or to linearly follow it in order to be linked to it via coreference) licenses the scrambled structure in (22), albeit at the cost of not reflecting the relevant prominence of arguments. Importantly, focused objects that surface in an A-scrambled position are interpreted in this position as well. That is, an A-scrambled construction licensed by binding considerations also exhibits surface scope: (29)

[Kogo posetili dva ego rodstvennika]CONTEXT Who was visited by two of his relatives? KAŽDOGO STUDENTA1, pocetili dva ego1 every student.ACC visited two his ‘Every student was visited by two of his relatives’.

(30)

rodstvennika relatives >; *>

[Kogo posetil každyj ego rodstvennik]CONTEXT Who was visited by every relative of his? DVUX STUDENTOV1, pocetil každyj ix1 rodstvennik two students.ACC visited two their relatives ‘Two students were visited by every relative of theirs.’ >;*>

144

Elena Titov

In (29), the existential quantifier fails to take scope over the universal presumably because the former embeds a pronoun, which forces a distributive reading. In (30), on the other hand, the distributive interpretation is unavailable and the universal must take scope under the indefinite. As discussed in the previous section, A’-moved contrastive foci are as a rule interpreted in clause-final position, which is the position where new information focus typically surfaces. However, armed with the findings of this section we can now predict that moved contrastive foci involved in binding would not reconstruct to clause-final position, that is, below the constituent embedding the bound pronoun. This prediction is indeed borne out: (31)

[KAŽDOGO STUDENTA]1, ja every student.ACC I

xoþu want

þtoby t1 that

pocetili visited

dva ego1 rodstvennika, (a ne každogo docenta) two his relatives (and not every lecturer) ‘Every student I want to be visited by two of his relatives (not every lecturer)’. >; *>

In (31), the A’-moved quantifier has to reconstruct above the constituent embedding the pronoun in order to bind it. Reconstruction to a clausefinal position inevitably results in a WCO violation. WCO effects can also be observed in split-scrambled constructions, as in (32a). Split scrambling involves A’-movement of a constituent with the interpretation of contrastive focus out of a larger syntactic constituent leaving a remnant that is construed as belonging to a background.23 The presence of a remnant in a clause-final position forces the constituent conveying the interpretation of contrastive focus to reconstruct below the embedded pronoun in (32a), resulting in ungrammaticality of the sentence under coreferential reading. The sentence in (32b) demonstrates that the ungrammaticality of (32a) is indeed due to a WCO violation. The sentence in (32c), on the other hand, can be analysed as an A-scrambled structure licensed by binding considerations, where the entire focused object NP krasivuju devoþku ‘beautiful girl’ A-scrambles in order to bind the pronoun embedded in the subject NP.

23 Split scrambling can also target contrastive topics, in which case the remnant has the interpretation of new information focus.

Chapter 4 Encoding focus and contrast in Russian

145

Interestingly, the A-scrambled structure in (32c) can serve as an underlying structure for split scrambling (see (32d)). In fact, the sentence in (32d) is the only structure that allows both split scrambling and coreference, supporting the hypothesis that A-scrambling can be licensed by binding considerations, albeit at a cost of disobeying (2).24 (32)

a. * DEVOýKU1 girl.ACC

eë1 her

mama mum

ljubit loves

[krasivuju t1] beautiful

(a ne babušku) (and not grandma) b.

DEVOýKU1 Ivan ljubit [krasivuju t1] (a ne babušku) girl.ACC Ivan loves beautiful (and not grandma) ‘Ivan loves a beautiful girl (and not a beautiful grandma).’

c.

[Krasivuju DEVOýKU]1 ljubit eë1 beautiful girl.ACC loves her ‘A beautiful girl is loved by her mum.’

mama mum

d.

DEVOýKU1 girl.ACC

mama mum

krasivuju t1 beautiful

ljubit loves

eë1 her

OVS

(a ne babušku) (and not grandma) ‘A beautiful girl is loved by her mum (and not a beautiful grandma).’

The observation that focused binders do not obey (2) extends to ditransitive constructions involving object-across-object A-scrambling. That is, objects that are involved in binding surface in an A-scrambled position even when they are new information foci. Moreover, they are also interpreted in this position: (33)

[ýto Anna otoslala dvum ego avtoram?]CONTEXT What did Anna send to two of its authors? Anna otoslala KAŽDYJ ROMAN1 dvum ego1 avtoram Anna sent every novel.ACC two its authors.DAT ‘Anna sent every novel to two of its author.’ >; *>

24 I am not using long-distance movement in (26d) because split scrambling becomes less felicitous when the distance between the landing site and the trace is not minimal. That is, even within one clause, split scrambling can become impossible when this clause contains phonologically complex constituents.

146 (34)

Elena Titov [Komu Anna otoslala dva ego romana?]CONTEXT Who did Anna send two of his novels? Anna otoslala KAŽDOMU AVTORU1 dva Anna sent every author.DAT two ‘Anna sent every author two of his novels.’

ego1 romana his novels.ACC >; *>

When binding objects are fronted as contrastive foci, they are interpreted in a scrambled position: (35)

KAŽDYJ every

ROMAN1, novel.ACC

ja I

xoþu, want

þtoby that

Anna otoslala t1 dvum ego1 avtoram (a ne každuju stat’ju) Anna sent two its authors.DAT (and not every article) ‘I want Anna to send every novel to its author (not every article).’ >; *> (36)

KAŽDOMU AVTORU1, every author.DAT

ja I

xoþu, want

þtoby that

Anna otoslala t1 dva ego romana (a ne každomu redaktoru) Anna sent two his novels (and not every editor) ‘I want Anna to send every author two of his novels (not every editor).’ >; *>

The findings of this section support the hypothesis that Russian A-scrambling can be licensed by a variety of considerations, with the requirement of one argument to occur in the scope of another overriding clause-final focus encoding. That is, although focus in Russian is typically encoded in clause-final position, it can in fact be realised in a non-clause-final position, as long as binding considerations demand it. Such an outcome strongly suggests that an analysis that sees Russian constituents as moving to a specific syntactic position to obtain the interpretation of focus cannot be sustained. Instead, Russian is better analysed as aiming at matching representations in information structure with syntactic representations, whenever stronger constraints such as binding do not prevent this correspondence. We have seen that the mapping rule in (10) successfully captures the observation that in an A-scrambled construction a focus typically follows a background and that an A’-fronted contrastive focus is interpreted below a background in the default case. However, as contrastive foci tend to surface in non-final A’-scrambled positions in Russian, the grammar of

Chapter 4 Encoding focus and contrast in Russian

147

the language must include an additional mapping rule that associates syntactic structures with a moved focus with a particular representation in information structure that makes a reference to contrast. Moreover, given the optional status of A’-scrambling, the aforementioned mapping rule must also be analyzed as applying optionally. The next section introduces the relevant mapping rule and proposes an analysis that captures the optional status of its application.

5.

Optionality of A’-scrambling

The argument put forward in the present chapter is that the syntax of languages like Russian is sensitive to information structural notions like background, focus and contrast. Thus, A-scrambled orders are licensed by the interpretative need for discourse-prominent background to linearly precede non-discourse-prominent focus. Whenever a focus is enriched to yield contrastive interpretation, it is allowed to undergo A’-movement but is interpreted in the launching site dictated by (10). On the other hand, A’-scrambling in Russian, although much favored, is optional. That is, it is possible to interpret a clause-final focus in (37), below, as contrastive as long as IK2 is placed on it.25 (37)

[Ivan kupil Volkswagen?]CONTEXT ‘Did Ivan buy a Volkswagen?’ a. Net, Ivan kupil [TOYOTU]CF (a ne Volkswagen) no Ivan bought Toyota.ACC (and not Volkswagen) ‘No, Ivan bought a Toyota, (not a Volkswagen).’ b. Net, Volkswagen kupil [BORIS]CF no Volkswagen.ACC bought Boris ‘No, Boris bought a Volkswagen, (not Ivan).’

(a ne Ivan) (and not Ivan)

To account for the optionality of A’-scrambling, I would like to partly adopt the theory introduced by Bobaljik and Wurmbrand (2008) and Wurmbrand (2008) that they refer to as the ¾ signature. The core idea behind this theory is that there exist ‘soft’ (violable) constraints (economy conditions) that value a particular type of correspondence between LF

25 Sentences like in (37) are judged by native speakers as less felicitous than

the A’-scrambled variants (Brun 2001).

148

Elena Titov

and PF representations. For instance, UG includes an economy condition that favors isomorphism between LF (scope) and PF (linear order) representations, which is dubbed by the authors as ScoT (for Scope Transparency). These constraints are uni-directional: LF (broadly construed) is calculated first, and determines PF (surface word order). The interaction of these constraints yields a ‘signature effect’, which Bobaljik and Wurmbrand call the ¾ signature. That is, taking one LF property and one PF property, what we frequently find is that three of the four logical combinations are grammatical, which results in the occurrence of optionality. To remain faithful to the present analysis, instead of adopting ScoT, I would like to argue that the faithfulness constraint operative in Russian is the one that maps representations in information structure onto syntactic structures. The isomorphism condition between syntax and IS has already been touched upon when A-scrambled structures regulated by the rule in (10) were discussed. In this case, the isomorphism between syntax and IS can be said to be fully satisfied, with no optionality present. That is, an unmarked structure that fails to satisfy (2) is consistently replaced by a marked structure that obeys (10) and (2).26,27 Assuming that movement is a syntactic process, an A’-scrambled structure can also be analyzed as matching a certain IS representation. This can be once again stated in the form of a mapping rule (see Neeleman et al 2009, Neeleman and van de Koot 2009 and the introduction to this volume for an analysis of A’-scrambling as marking contrastive scope): (38)

[Contrast]-marking mapping rule Interpret an A’-fronted focus as structurally marking its contrastive scope via A’-movement

The rule in (38) captures the observation that any contrastively focused element in Russian aims at leaving clause final position, with A’-scrambling licensed by the need to structurally distinguish CF from NIF. The optionality of such displacement, on the other hand, results from the fact 26 We have seen that this observation holds unless an unmarked structure for a given truth-conditional interpretation fails to be generated altogether. 27 Although the scope of the present chapter prevents me from discussing the parametric variation with respect to the presence or absence of optionality of A-scrambling in greater detail, the current analysis predicts that there must be languages that have an option of producing unmarked structures that do not match IS along with marked structures that do.

Chapter 4 Encoding focus and contrast in Russian

149

that movement is a costly operation. Following Chomsky 1995, I am assuming that a structure involving movement is more costly than the one that does not, as the former contains an additional movement operation. The costly nature of movement can be expressed in a form of a syntactic constraint that disallows any type of syntactic displacement, e.g. *MOVE. On the other hand, movement is permitted when it provides a better reflection of some aspect of interpretation than the sentence would without movement. That is, a more costly structure receives the interpretative license given in (38), as it serves at discriminating CF from NIF. Note, that (38) says nothing about the interpretation of focus in clausefinal position. Consequently, such focus should admit both contrastive and non-contrastive readings. In other words, a clause-final focus is interpretatively ambiguous. Although other tools such as prosody can be used to disambiguate the interpretation of a clause-final focus, they are not always available. Thus, in the written language, the interpretation of focus as regards the presence or absence of a contrastive reading can only be disambiguated by means of structure and to a certain extent context. There is therefore a tendency to interpret a clause final focus as non-contrastive unless there is an adequate overt indication that such focus is enriched to yield a contrastive interpretation. An A’-fronted focus, on the other hand, can only be interpreted as contrastive. By hypothesis, this is because A’-scrambling marks the contrastive scope of the moved focus. If so, it is expected that a NIF can never undergo A’-scrambling: a non-contrastive focus cannot mark a contrastive scope. Hence, movement provides a better reflection of the information structure of the sentence distinguishing contrastive focus from non-contrastive new information focus by placing the former in a position where the non-contrastive reading is impossible, but the trade-off is a costly structure. Under Bobaljik and Wurmbrand’s approach, such a trade-off generally results in the appearance of optionality. Conversely, if the sentences in (37) contained new information foci, as in (39) and (40), there would be no trade, so movement would be unmotivated, and hence disallowed:28 (39)

[ýto Ivan kupil?]CONTEXT ‘What did Ivan buy?’

28 The replies in (39b) and (40b) are possible only under the emphatic interpretation, which is analyzed here as contrastive.

150

Elena Titov a.

Ivan kupil Ivan bought ‘ Ivan bought a Toyota.’

b. # Ivan Ivan (40)

[TOYOTU]NIF1 Toyota.ACC

[TOYOTU]NIF Toyota.ACC kupil t1 bought

[Kto kupil Volkswagen?]CONTEXT ‘Who bought a Volkswagen?’ a.

Volkswagen kupil Volkswagen.ACC bought ‘No, Boris bought a Volkswagen.’

b. # Volkswagen Volkswagen.ACC

[BORIS]NIF Boris

[BORIS]NIF1 Boris

kupil t1 bought

Table (1), below, illustrates the Russian focus paradigm. The information structural constraint here distinguishes contrastive focus from non-contrastive focus via A’-movement. Syntax produces structures that either reflect this interpretative difference or not and therefore either satisfy (38) or not. On the other hand, *MOVE forces syntax to produce simple structures. Following Bobaljik and Wurmbrand (2008), I assume that A’scrambling is “free” (not feature-driven or required for convergence), but costly (*MOVE).

# T: 1

IS

Syntax

(38)

*Move

(a)

9

[CP … NIF]

[CP … NIF]

9

9

(b)

* (A’-scrambling)

[CP … NIF]

[CP NIF1… t1]

*

*

(c)

9 (A’-scrambling)

[CP CF1… t1]

[CP CF1… t1]

9

*

(d)

9

[CP CF1… t1]

[CP … CF]

*

9

The ¾ paradigm demonstrated in Table (1) results from three out of four combinations satisfying at least one of the two constraints. Thus, a structure with a clause final NIF in (a) satisfies both constraints. That is, (38) is vacuously satisfied, as it has nothing to say about the interpretation of focus that has not undergone A’-movement; and *Move is satisfied because the structure has no additional A’-movement operation. A struc-

Chapter 4 Encoding focus and contrast in Russian

151

ture with a moved NIF in (b), conversely, violates both constraints, as it does not only contain movement, it contains movement that does not obey (38), as a moved non-contrastive focus cannot mark a contrastive scope. A structure with a moved contrastive focus in (c) violates *Move but satisfies (38), whereas a structure with a clause final CF satisfies *Move but at the cost of not structurally reflecting the information structural representation of the sentence. The data in (41) illustrates the emerging ¾ paradigm: (41)

a.

Ivan kupil Ivan bought ‘ Ivan bought a Toyota.’

b. # Ivan Ivan

[TOYOTU]NIF1 Toyota.ACC

[TOYOTU]NIF Toyota.ACC kupil t1 bought

c.

Ivan [TOYOTU]CF1 kupil t1 (a ne Volkswagen) Ivan Toyota.ACC bought (and not Volkswagen) ‘ Ivan bought a Toyota, (not a Volkswagen).’

d.

Ivan kupil [TOYOTU]CF (a ne Volkswagen) Ivan bought Toyota.ACC (and not Volkswagen) ‘No, Ivan bought a Toyota, (not a Volkswagen).’

Therefore, the ¾ signature can successfully account for the fact that contrastive focus can optionally undergo movement in Russian, in line with (38), whereas new information focus always remains in clause final position, in accordance with (2) and (10): (42)

a. b. c. d.

Clause final non-contrastive elements *A’-moved non-contrastive elements A’-moved contrastive elements Clause final contrastive elements

The rule in (38) states that a focused category is disambiguated as contrastive via A’-movement. However, it says nothing about the position targeted by movement. In Russian, contrastive focus tends to move to a position immediately before an inflected verb. However, it is also possible for it to move further to the left within the clause, as well as undergo longdistance movement. Although, it can be postulated that a displaced contrastive constituent can in principle adjoin to any maximal projection (e.g. VP or CP), it is still unclear what licenses a longer chain, considering that movement to a VP-adjoined position already serves the purpose of disambiguating this constituent as contrastive. Ideally, structures with a longer

152

Elena Titov

A’-chain should be interpretatively distinct from structures with a shorter A’-chain. Otherwise, one is expected to block the other by economy. Admittedly, it is not easy to find any interpretative difference between a structure with a longer A’-chain and the one with a shorter chain, as all A’-scrambled foci are interpreted in clause-final position in Russian.29 Moreover, any A’-scrambled sentence is truth-conditionally identical to its unmarked variant. On the other hand, if there were structures that exhibited interpretative differences depending on the length of the A’-chain, such structures would justify the existence of chains of different lengths. After all, if different-sized chains are needed to encode different interpretations in such structures, grammar must be able to produce structures with chains of a different size. At the same time, for the structures with different-sized A’chains to show any interpretative differences, the scope of A’-scrambling must be overtly marked. Since A’-scrambled contrastive foci reconstruct in Russian, the scope of movement of such foci can only be marked by surface scope markers, such as, for instance, focus sensitive operators. It is debatable whether focus sensitive operators can form a constituent with the focus in Russian but there is an obvious requirement for them to c-command a focus and to be maximally close to it. Arguably, this requirement predicts which phrase focus sensitive operators adjoin to (Büring and Hartmann 2001). Following Büring and Hartmann (2001), I will assume that a focus sensitive operator can adjoin to a maximal projection immediately dominating a focused constituent. Moreover, the adjunction operation can target a maximal projection above an A’-scrambled focus. Consequently, an A’-fronted contrastive focus is expected to reconstruct for scope to clause-final position but a focus sensitive operator that modifies it will overtly mark the scope of movement. If the data involving focus sensitive operators demonstrate that a shorter A’-chain is incompatible with a reading that a longer chain is compatible with and vice versa, this interpretative difference can be used to explain why there is no blocking of one structure by the other; that is, both structures are allowed to exist as they are not consistently linked to identical interpretation. The data in (43) and (44) show that this is indeed the case. Here I am using contexts – borrowed from Wagner (2008a) – that require a particular scope relation between two such operators. The first of these is given in (43). 29 The concept of DoC-marking, discussed in the introduction to the book, may shed light on the interpretive effect of the length of the A’-chain formed by focus movement (see also Neeleman and van de Koot 2008 for a related discussion).

Chapter 4 Encoding focus and contrast in Russian (43)

153

Context I: even > only: Except for Boris, the kids in this summer camp have no respect for animals and the potential dangers, which makes them take too many risks, including with poisonous snakes. a.

Daže even

samaja jadovitaja zmeja most poisonous snake

b.

Daže samaja jadovitaja zmeja tol’ko [BORISA]1 pugaet t1 even most poisonous snake only Boris.ACC scares

c. # Tol’ko [BORISA]1 only Boris.ACC

pugaet scares

tol’ko only

Borisa Boris.ACC

daže samaja jadovitaja zmeja pugaet t1 even most poisonous snake scares

d. # Tol’ko Boris daže [SAMOJ JADOVITOJ ZMEI]1 boitsja only Boris even most poisonous snake.GEN fears e.

Daže [SAMOJ JADOVITOJ ZMEI]1 tol’ko Boris boitsja t1 even most poisonous snake.GEN only Boris fears

As can be seen from (43a), in a structure without movement, focus sensitive operators exhibit surface scope. The sentence in (43b) demonstrates that A’-scrambling to an intermediate position is compatible with the context, whereas a structure with a longer chain, as in (43c) is not. Unsurprisingly, when the order of operators is reversed, as in (43d) and (43e), it is the sentence with a longer chain, as in (43e), that is compatible with the context but not the one with a shorter chain, as in (43d). Identical observations can be made with respect to a context that forces a reversed scope relation between the operators, as in (44). (44)

Context II: only > even: The kids in the summer camp are afraid of snakes to some degree, but it depends on how dangerous they are. Everyone is afraid of rattlesnakes, since they are really poisonous, but almost everyone is ok with some less poisonous snake. a.

Tol’ko Boris boitsja daže samoj bezopasnoj zmei only Boris fears even most harmless snake.GEN

b.

Tol’ko Boris daže [SAMOJ BEZOPASNOJ ZMEI]1 boitsja t1 only Boris even most harmless snake.GEN fears

c. # Daže [SAMOJ BEZOPASNOJ ZMEI]1 tol’ko Boris boitsja t1 even most harmless snake.GEN only Boris fears d. # Daže samaja bezopasnaja zmeja tol’ko [BORISA]1 pugaet t1 even most harmless snake only Boris.ACC scares e.

Tol’ko [BORISA]1 daže samaja bezopasnaja zmeja pugaet t1 only Boris.ACC even most harmless snake scares

154

Elena Titov

The above data demonstrate that structures with different lengths of A’chains can be linked to different interpretations, some of which are incompatible with a particular reading and therefore with a particular context. The existence of such data is sufficient to justify the existence of longer chains. The optional status of the movement of contrastive focus undermines the idea that such fronting operation targets a KONTRAST-position in the left periphery and that it is triggered by a syntactic K-feature (Molnár 2001). If focus had to move to the specifier of the KONTRAST-projection in order to check the K-feature, it would be impossible for a focused constituent to be associated with contrastive interpretation in the absence of movement to this position, as in (41d), where the focus follows the background. Moreover, the fact that movement of contrastive focus can target a variety of positions cannot be easily accounted for under the cartographic approach. Even if a ‘low left periphery’ above the vP (Belletti 2001) is postulated to account for the difference between (45b) and (45c), the sentence in (45a) will still present a problem. (45)

a. Ivan dal Ivan gave

[KNIGU]CF1 Borisu t1 book.ACC Boris.DAT

b. Ivan [KNIGU]CF1 Ivan book.ACC c. [KNIGU]CF1 book.ACC

6.

(a ne gazetu) (and not newspaper)

dal Borisu t1 gave Boris.DAT

(a ne gazetu) (and not newspaper)

Ivan dal Borisu t1 Ivan gave Boris.DAT

(a ne gazetu) (and not newspaper)

Conclusion

The above analysis strongly supports the hypothesis that contrastive focus in Russian is interpretatively characterized by the features [focus] and [contrast], with the former licensing its underlying position below a background, in line with (10), and the latter licensing A’-fronting, in accordance with (38). A theory that analyzes clause-final focus encoding in Russian as a result of an interpretive effect at the interface between syntax and information structure successfully captures the observation that this encoding can be overridden by stronger considerations that are also capable of licensing an A-scrambled structure. Conversely, a theory relying on a designated syn-

Chapter 4 Encoding focus and contrast in Russian

155

tactic position for focus would require further stipulations to explain the fact that focus in Russian can be encoded in a variety of positions. A’-scrambling is considered by the present analysis as taking place in order to structurally encode a contrastive reading. Plausibly, the position to which a CF moves is not fixed precisely because this type of displacement is not aimed at targeting a specific syntactic position in order to obtain a particular interpretation linked to this position but takes place solely to disambiguate the interpretation of a focused category as contrastive via A’-movement. The optionality of A’-scrambling is accounted for by making a reference to a competition of constraints (economy conditions) at the interface between information structure and syntax that value a particular type of correspondence between representations within these two modules. Thus, a constraint that forces simple structures competes with the requirement for syntactic structures to reflect a particular IS interpretation, with an outcome that a costly structure only survives this competition if it serves to maximally achieve an interpretation that a simple structure cannot disambiguate. The existence of longer A’-chains is justified using data involving focus sensitive operators, which overtly mark contrastive scope of a moved focus.

Chapter 5 Against FP Analyses of Clefts* Matthew Reeve 1.

Introduction

Two questions which figure highly in current debates about the relation between syntax and information structure are: (i) how many different types of focus need to be recognised, and (ii) how are these different types of focus encoded syntactically (if at all)? A common position that is taken on the first question is that we must at least distinguish the type of focus that merely encodes new information (É. Kiss’s information focus) from the type of focus that picks out a member of a set established by the discourse, and asserts that this member, in contrast to at least one other member of this set, makes the sentence true (É. Kiss’s identificational focus). The answer to the second question is often taken to be related to the first, in that only identificational focus is represented syntactically, and furthermore that this representation involves a focus-related functional head (let us call it F) somewhere in the clausal spine. This distinction between syntactically represented identificational focus and syntactically inert information focus is then taken to be responsible for the series of properties that distinguishes the two types of focus. The aim of this paper is to argue that too much has been made to hinge on this distinction between syntactically represented and syntactically inert focus, and that the evidence for positing a universal functional head encoding identificational focus is weak. The empirical focus of the paper is on so-called cleft constructions in English in Russian, as contrasted with sentences in these languages in which a focused constituent occurs in a displaced position or in situ. It is sometimes argued that both clefts and focus-movement instantiate movement of a focused constituent to SpecFP, * Thanks to Klaus Abels, David Adger, Caroline Heycock, Ad Neeleman, Elena Titov, Hans van de Koot and Reiko Vermeulen, as well as audiences at the Cleft Workshop 2008 at ZAS, Berlin, the LAGB Annual Meeting 2008 at the University of Essex and the Potsdam/UCL Workshop on Information Structure 2010 at UCL, for comments, encouragement and other forms of input in relation to this paper. This work was funded by an Arts and Humanities Research Council Doctoral Studentship.

158

Matthew Reeve

and that the feature structure of F is responsible for the different semantic/pragmatic effects to which these constructions give rise. I will argue, however, that clefts must be structurally distinguished from focus-fronting in both English and Russian, and that there is no reason to think that the particular semantic/pragmatic properties associated with clefts, vis-à-vis focus-fronting, result from anything other than the interplay between the equative semantics of specificational copular sentences and the definiteness of the subject pronoun. In other words, the evidence that a functional head F has any syntactic or semantic role to play in these languages is weaker than is often claimed. In section 2 I briefly outline the FP analysis of English clefts provided in É. Kiss (1998), noting some initial problems with it. I then consider how the FP analysis might be applied to Russian clefts, once again pointing out potential difficulties. Then, in section 3 I outline an alternative ‘copular’ analysis of both English and Russian clefts which makes no use of a focusrelated functional head. I claim that, despite the syntactic differences between the two constructions, both involve the same syntax-semantics mapping, and that the semantics of the constructions is equative. Section 4 presents a series of novel empirical arguments against the FP analysis and in favour of the copular analysis: the non-expletivehood of the subject pronoun, the patterning of the presuppositions of clefts, the properties of Russian ‘wide-focus clefts’, the requirement for ‘adjacency’ between the pronoun and the focus in Russian narrow focus clefts, the possibility of focus-movement in clefts, the distinction between optional and obligatory contrastive focus, and the fact that English clefts seem to involve an extraposition relation between the cleft clause and clefted XP. Section 5 is the conclusion. 2.

Focus head analyses of clefts

The constructions to be discussed in this paper are illustrated in (1-2):1 (1)

a. It was JOHN that Mary hit. b. It was ON THE PLANE that Mary hit John. c. It was BLACK AND BLUE that John’s eyes were.

(2)

Maria vypila. a. Èto VODKU this vodka-ACC Maria-NOM drank ‘It was vodka that Maria drank.’

1 Throughout, I use small capitals to indicate the intended focus of the sentence.

Chapter 5 Against FP Analyses of Clefts

159

b. Èto V PARKE Maria vypila vodku. this in park-PREP Maria-NOM drank vodka-ACC ‘It was in the park that Maria drank vodka.’ c. Èto P’JANOJ Maria sidela v parke. this drunk-INSTR Maria-NOM sat in park-PREP ‘It was drunk that Maria sat in the park.’

The examples in (1) illustrate the English cleft construction, which appears to be a copular sentence with it as subject, a postcopular constituent which is typically focused (the clefted XP), and what looks like a relative clause (the cleft clause) representing the ‘background’ of the focus.2 In (2), we find examples of a Russian construction which is also sometimes called a ‘cleft’. This differs syntactically in various ways from the English cleft: in particular, there is no evidence of copular structure or a relative clause. Nevertheless, the Russian cleft is apparently introduced by a pronoun (èto, which independently functions as a neuter singular proximal demonstrative), and the information structure is roughly the same as in English – a focused XP followed by an ‘open clause’ interpreted as the background of the focus. In both languages, the clefted XP can (at least) be a DP, a PP or an AP.3 In her paper on identificational and information focus, É. Kiss (1998) argues that the position occupied by the clefted XP in English is the structural and semantic equivalent of the preverbal focus position in Hungarian, as illustrated in (3a), argued by É. Kiss to have the structure in (3b). Here, the focused DP Marinak surfaces in the specifier of a focus head F, to which the finite verb raises: (3)

a. Tegnap este MARINAK mutattam be Pétert. last night Mari.DAT introduced.1SG PERF Péter.ACC ‘It was to Mary that I introduced Peter last night.’ b.

[TopP tegnap este [FP MARINAK [F mutattami] [VP ti be Pétert]]]

There are several advantages of adopting this structure for the Hungarian preverbal focus position. First, the assumption of V-to-F movement ensures the strict adjacency found between the focus and the verb (an adjacency found in many other languages), and the fact that particles

2 I follow the characterisations of focus and background given in Chapter 1. 3 See Emonds (1976), Delahunty (1981) and Heggie (1988) for different claims about the possible categories of the clefted XP.

160

Matthew Reeve

which normally occur immediately before the verb occur immediately after it when there is a preverbal focus. It also accounts for the fact that identificational focus appears to take scope relative to quantifiers in the sentence. É. Kiss then proposes that the cleft construction is the English equivalent of the Hungarian preverbal focus construction – that is, the cleft is the structural means of encoding exhaustive identificational focus in English. She accordingly proposes a syntactic structure for English clefts as in (4), a biclausal structure in which the FP hosting the clefted XP is selected for by the matrix I, which takes an expletive it as subject, and the cleft clause is a CP selected for by F. The copula is an expletive which ‘lexicalises’ the F head, as required by FP theory of Brody (1990). The copula then raises obligatorily to matrix I, resulting in the structure in (4):4 (4)

[IP It [I [F was]i I] [FP JOHN ti [CP that Mary hit]]].

This parallel analysis of Hungarian focus and English clefts is primarily designed to capture the interpretative parallels between them: the existential and uniqueness presuppositions to which both give rise (and the related ban on certain types of quantifier phrase appearing in focus position), and the fact that identificational focus enters into scope relations with quantifiers in the same clause. Because the analysis involves A’movement of the focused phrase into SpecFP, it also accounts for the wellknown ‘connectivity’ effects that clefts display – that is, the clefted XP appears to reconstruct into the trace position in the cleft clause for properties such as binding and scope. Furthermore, É. Kiss entertains the additional possibility of base-generating the focus in SpecFP and linking it to a relative pronoun; this would also accounts for certain ‘anti-connectivity’ effects (see Pinkham and Hankamer 1975, Reeve 2010).

4 Hedberg (2000 : 913) points out a potential problem with the idea that be lexicalises F and then raises to I. In clefts with modals, such as (i), the copular verb would still have to raise to some functional head position above F to locate it in its correct linear position: (i) It could be Clinton who wins. However, in contrast to the raising of the finite copula, which is independently motivated, there is no obvious reason why the copula in (i) should raise. The present analysis avoids this problem, since it treats be as a main verb which only raises if finite.

Chapter 5 Against FP Analyses of Clefts

161

Under the FP analysis, the structure of clefts is divorced from the structure of specificational sentences (in the sense of Higgins 1973), of which clefts are sometimes considered a subclass (e.g., Akmajian 1970, Percus 1997). Compare the clefts in (5) with the specificational equivalents in (6): (5)

a. It was JOHN that Mary hit. b. It was ON THE PLANE that Mary hit John. c. It was BLACK AND BLUE that John’s eyes were.

(6)

a. The one that Mary hit was JOHN. b. The place where Mary hit John was ON THE PLANE. c. What John’s eyes were was BLACK AND BLUE.

The specificational sentences in (6a-c) are semantically equivalent to the clefts in (5a-c) respectively; both constructions specify the postcopular focused phrase as the correct value for a variable (represented by a gap or relative pronoun) in an open proposition. In addition, they share the same ‘presuppositions’ – thus, for example, both (5a) and (6a) presuppose that Mary hit someone and that John was the only person hit by her, in contrast to the non-copular equivalent Mary hit John. Yet there are obvious problems with analysing (6a-c) in terms of FP, not least the fact that the focused phrase would have to be extracted out of a complex DP to reach its surface position. It is not clear how the base-generation alternative would work either, as the relative pronoun in the (surface) subject in (6ac) is already related to a ‘head noun’. Although this is not a decisive argument against analysing clefts using FP, it does at least suggest that an alternative should be sought which provides for a unified analysis of clefts and specificational sentences. One alternative approach of this type involves treating it and the cleft clause as a discontinuous definite description. According to this analysis, in (5a-c) it and that Mary hit at some point form a definite description equivalent to the one that Mary hit, thus capturing the semantic parallels between (5a-c) and (6a-c), and the mechanism of relating this definite description to the focus will then be the same in both clefts and specificational sentences (e.g., Wirth 1978, Percus 1997).5 5 For example, specificational sentences could involve ‘inverse predication’ – that is, the surface subject is a predicate which takes the postcopular XP as its subject (e.g., Adger and Ramchand 2003, Den Dikken 2006) – or it could involve equation – that is, the surface subject and postcopular XP are asserted to have the same reference (e.g., Heycock and Kroch 1999). In this paper I take the position that specificational sentences involve equation, a position I defend in more detail in Reeve (2010).

162

Matthew Reeve

Russian clefts provide an interesting contrast to English clefts, as they apparently do not submit to a unified analysis with specificational sentences. Rather, they seem to have more in common with focus-fronting sentences. As King (1993) shows, they do not allow a matrix copula, and hence a copular structure seems out of the question. In addition, relative pronouns do not appear in Russian clefts, which suggests that the pronoun and the cleft clause cannot be treated as a discontinuous definite description. Russian clefts, then, might seem to be a more promising candidate than English clefts for an FP analysis. King proposes a structure of this general type, illustrated in (7). Here, the focus-related functional head F hosts the pronoun-like element èto in its specifier. F takes as its complement a phrase (YP) headed by an empty functional head, and the focused phrase overtly moves to SpecYP: (7)

[FP èto F [YP this

VODKUi

vodka-ACC

[Maria Maria-NOM

vypila ti]]]. drank

The clear problems with this structure from the point of view of É. Kiss’s theory are that the specifier of F is not occupied by the focus, and there is no material to lexicalise F. The first of these can easily be remedied by putting F in place of Y and assigning èto to some higher position: (8)

[ZP èto [FP this

VODKUi F vodka-ACC

[Maria Maria-NOM

vypila ti]]]. drank

However, this (straw man) analysis would still suffer from the problem of what lexicalises F. Although it is not inconceivable that a functional head could be empty – and there are of course numerous such cases in the literature – one of the strengths of Brody’s and É. Kiss’s analyses of Hungarian and English is that the requirement for overt lexicalisation of F accounts for particle-movement in Hungarian and the position of the copula in English. One might instead propose that èto itself is the lexicaliser of F, and is thus a head, but this means we are once again faced with the problem that (7) posed. Alternatively, we could imagine that èto starts in F in (8) and undergoes head-movement to Z, giving the correct surface ordering. This would most closely parallel É. Kiss’s analysis of English clefts, in which the copula moves from F to I. However, while head-movement in English clefts would be independently motivated, since the copula generally moves to I in English, the putative head-movement in (8) would not, since there is no evidence for a parallel movement elsewhere in the language.

Chapter 5 Against FP Analyses of Clefts

163

To summarise, then, despite the attractive features of analysing English and Russian clefts in terms of a Focus Phrase, the FP analysis faces a number of problems. Furthermore, while it seems difficult to account for the properties of Russian clefts using the specificational approach, this approach is well-suited to the properties of English clefts. In this paper, however, I want to argue in favour of a partially unified specificational analysis of English and Russian clefts, one which accounts for their interpretative and configurational similarities while capturing their undoubted syntactic differences.

3.

Towards a unified specificational analysis of clefts

What is interesting about English and Russian clefts is that, despite the differences between them in more fine-grained syntactic properties, they can both be analysed in terms of the following gross constituent structure, where (i) D is a definite singular pronoun, E is a focus and Jis its background, (ii) D c-commands J and (iii) Jm-commands D: (9)

[XP D [YP EJ@]

In Reeve (2010) I argued that the schema in (9) is realised as in (10a,b) for English and Russian respectively: (10)

a. [IP it [VP [VP VCOP focus] background]] b. [IP1 èto [IP2 focus [IP2 background]]]

That is, English clefts, as in (10a), are truly biclausal structures in which the copular verb takes the focus as its complement, and the cleft clause (which is a true restrictive relative clause) is adjoined to its VP projection; on the other hand, Russian clefts are monoclausal in the sense that the two IPs belong to the same extended verbal projection (i.e., there is no copular verb ‘belonging’ to IP1), but biclausal in having two subjects, èto and the subject of IP2. In addition, the structure in (10b) allows us to preserve the insight that the material following èto in clefts behaves just like a standard finite clause with focus-movement to clause-initial position (assuming this movement adjoins the focus to IP, as in Sekerina 1997, Stepanov 1998). Two questions then arise: how does the ‘background’ link up with the pronoun to form a definite description if it is not base-generated as a constituent with it, and how is the overall specificational meaning derived? On the first question, there are essentially two approaches that have been

164

Matthew Reeve

proposed to derive the semantic identity of clefts and specificational sentences in English. The most popular of these, originally due to Akmajian (1970), involves an underlying D+NP+CP constituent whose CP then undergoes extraposition to clause-final position (see also Wirth 1978, Percus 1997). The second approach takes the cleft clause to be base-generated in its surface position and linked up interpretatively with the cleft pronoun, either in a parallel manner to right-dislocation (Gundel 1977) or making use of principles which have been proposed in the literature to link base-generated extraposed relatives with their ‘antecedents’ (Hedberg 1990). The approach I propose in Reeve (2010, 2011) is similar to this last approach in that the cleft clause never forms a constituent with the cleft pronoun. I show, however, that using a mechanism such as that proposed by Hedberg leads to incorrect empirical predictions with respect to the surface position of the cleft clause. Instead I propose that the principle linking the cleft clause interpretatively with the cleft pronoun is parasitic on the fact that a similar structural relation exists between the determiner and DP-internal relatives. Analysing the cleft pronoun as an NP-less D, I propose that the rule in (11) allows the cleft pronoun to saturate the thetarole borne by the cleft clause (via Higginbotham’s (1985) ‘theta-binding’):6 (11)

Thematic licensing condition For theta-binding to take place, the constituent bearing the ș-role to be bound must be c-commanded by the theta-binding determiner and must m-command that determiner.

Thus, what licenses the restrictive relation between the cleft clause and pronoun is not the specific mode of combination of the individual items, but the fact that a particular locality requirement holds between the two, and that this is parallel to the locality requirement between a determiner and a relative clause modifying (the NP of) that DP. Consider the following structures for clefts:7

6 I assume, following May (1985), Chomsky (1986) and many others, that the segment-category distinction is relevant for c/m-command. Thus, in (12a), VP does not dominate CP, since only one of its segments dominates CP. This means that the closest maximal projection dominating CP is IP. 7 In Reeve (2010) I argue that the cleft clause in (12a) starts out as an adjunct to the clefted XP and undergoes extraposition to adjoin to VP. This extraposition is obligatory, since otherwise the cleft clause would not be local enough to the pronoun for theta-binding to take place. Although it is difficult to demonstrate

Chapter 5 Against FP Analyses of Clefts (12)

a.

b.

IP DP it

I’ I+Vi

ti

IP1 DP èto

VP VP

165

I’ I

IP2 DPi vodku

CP

DP that Mary saw John

IP2

DP Boris

I’ I

VP V vypil

Despite their different structures, in both (12a) and (12b) the cleft clause is syntactically local to the initial pronoun, in that (i) the pronoun c-commands the cleft clause (CP in (12a), lower IP2 in (12b)), and (ii) the cleft clause m-commands the pronoun. Now, a parallel can be drawn with DPs modified by a restrictive relative, both syntactically and semantically. Assuming the structure in (13) for such DPs, the determiner c-commands the relative clause, and the relative clause m-commands the determiner:

the obligatoriness of cleft clause extraposition in English, it is obvious in Dutch and German, which have similar cleft constructions to English: (i) a.

b.

Jutta sagt, dass es DIESER WAGEN war, Jutta says that it this car was den sie kaufen wollte. which she to.buy wanted ‘Jutta says that it was this car that she wanted to buy.’ *Jutta Jutta den which

sagt, says sie she

dass that kaufen to.buy

es it

DIESER

this wollte, wanted

WAGEN, car war. was

(Smits 1989)

The crucial thing is that relative clause extraposition is normally optional in these languages, so there seems to be no independent reason why cleft clause extraposition should be obligatory.

ti

166

Matthew Reeve

(13)

DP D the

NP NP dog

CP that Mary saw

Semantically speaking, the relative clause is a predicate whose predicatehood is ‘satisfied’ by the determiner – that is, the determiner operates on the intersection between the set of dogs and the set of things that Mary saw, returning an individual. Assuming a semantic parallel between clefts and specificational sentences, the situation is exactly the same in clefts. In (12a), it can be viewed as a definite determiner operating on the set of things that Mary hit and returning an individual (i.e., the one that Mary hit). The claim, then, is that the syntax-semantics interface in clefts is parallel to that in DPs such as (13). The second question relates to the semantics of copular sentences more generally. I follow Heycock and Kroch (1999) in taking specificational sentences (including clefts) to be a type of equative. This distinguishes them from predicational copular sentences, which simply involve a predication relation between the two XPs. Thus, in a sentence like (14a), in which the second DP cannot be interpreted as a predicate applied to the first DP (since the second DP is a name, and names cannot be predicates), the semantics is roughly as in (14b), where the definite article translates as the Russellian iota-operator. On the other hand, in the predicational copular sentence in (14c), there is no equation – the property represented by the second DP is simply predicated of the first DP, as in (14d): (14)

a. b. c. d.

The man that I like is John. Țx.like(x)(i) = j Mary is a teacher. teacher(m)

In specificational sentences, then, equative semantics must be encoded somewhere in the copular clause. I propose that this role is fulfilled by a functional head Eq in the ‘clausal spine’, which (like É. Kiss’s F) must be lexicalised by phonologically overt material, either in the head position or in its specifier. For English, this lexicalisation is carried out by the copular verb (an underlying V), which moves to Eq overtly. The structure for a SCS is English can thus be updated as in (15a), and the structure for a cleft as in (15b):

Chapter 5 Against FP Analyses of Clefts (15)

a.

b.

TP

DPi T’ the one that T+Eq+V EqP Mary hit ti

TP DPi it

T’

T+Eq+V

EqP ti

Eq’ tEq+V

Eq’ tEq+V

VP tV

167

DP John

VP VP

tV

CP

DP that Mary saw John

The Eq analysis of specificational sentences has the advantage of capturing the asymmetry between specificational and predicational sentences with respect to the obligatoriness of be in ‘small clause’ complements (e.g., Moro 1997, Den Dikken 2006): (16)

a. I consider the murderer *(to be) John. b. I consider John (to be) the murderer.

In the predicational (16b), the copula in the complement clause is optional. This is generally thought to be due to the semantic inertness of the copular verb. On the other hand, in (16a) the copula is obligatory. Under the present analysis, this is because the embedded clause is forced to be equative (since it cannot be predicational), and hence Eq must appear in the clause. Because Eq must be ‘supported’ by overt material, the copula must be present in order that it can fulfil this function. In contrast to English clefts, Russian clefts do not feature a copular verb. Given that Russian copulas are null in the present tense, however, this is not in itself a bar to a copular analysis of clefts. However, King (1993) notes that an overt copula is never possible, even when non-present forms would be possible in the equivalent English sentence. She takes this (and other facts) to be fatal to a copular analysis. Note, however, that if we take equative semantics to be divorced from the copular verb, this may not pose such a problem. Recall that Eq must be ‘lexicalised’ by overt material. Given the structure proposed for Russian clefts, the lexicaliser cannot be the verb, which in Russian does not raise to I (see, e.g., Slioussar 2007). My claim is therefore that the pronoun èto lexicalises Eq (which

168

Matthew Reeve

corresponds to IP1 in (12b)) by appearing in its specifier. Thus, under this analysis, not only do we predict that a copular verb is never possible, but we allow for an equative analysis of clefts. The structures I propose for specificational sentences and clefts in Russian are given in (17a,b) respectively:8 (17)

TP

a.

DPi xrabryj soldat T

T’ EqP DP èto

Eq’ Eq

I

VP V

b.

DP Ivan

TP pro

T’ T

EqP DP èto

I

Eq’ Eq

TP DPi vodku

TP DP Boris

T’ T

VP V vypil

ti

8 The structure for the cleft in (17b) includes pro in the higher SpecTP. This is because èto cannot move to this position, being the overt licenser for Eq. Although Russian is only a partial pro-drop language, it does arguably allow

Chapter 5 Against FP Analyses of Clefts

169

In addition to allowing for an equative interpretation for Russian clefts, this analysis has the advantage of capturing the parallel between clefts and specificational sentences. As in English, predicational copular sentences in Russian make use of the copular verb, the difference being that it is normally null in the present tense – compare (18a,b). In present tense specificational/equative sentences, on the other hand, èto must intervene between the two DPs where there is no copular verb, as shown in (18c): (18)

a. Ivan – idiot. Ivan idiot ‘Ivan is an idiot.’ b. Ivan *(byl) idiotom. Ivan was idiot-INSTR ‘Ivan was an idiot.’ c. Zevs *(èto) Jupiter. Zeus this Jupiter ‘Zeus is Jupiter.’ d. Zevs *(byl) Jupiter. Zeus was Jupiter ‘Zeus was Jupiter.’

These data receive a parallel explanation to the data in (35).9 Predicational sentences do not make use of Eq, so the copular verb is not required. On the other hand, specificational/equative sentences must contain Eq, and this must be overtly lexicalised. Since there is no present tense copula available, èto is inserted in the specifier of Eq.10 expletive pro, assuming that Russian T has the EPP property (see, e.g., Slioussar 2007). I argue in Reeve (2010) that èto is exceptional among DPs in Russian in not needing Case. It also does not trigger subject-verb agreement. Thus, there is no requirement for a featural relationship between èto and the higher T. This means that the only requirement that needs to be satisfied on the higher T is the EPP, and hence that expletive pro can occupy this SpecTP. 9 Alternative analyses can be found in Geist and Błaszczak (2000) and Geist (2007). 10 It is worth noting that many of the world’s languages other than English and Russian differentiate specificational/equative and predicational copular sentences in this way; that is, specificational/equative sentences contain some overt element which is absent or optional in predicational sentences (e.g., Hebrew – Rapoport 1987; various Atlantic creoles – Holm et al. 1999). Nevertheless, in many of these cases there is good reason to think that the overt

170

Matthew Reeve

There is one final aspect of Eq that must be made clear. The structures proposed for specificational sentences in English and Russian do not, as it stands, allow for a straightforwardly compositional interpretation of Eq. If it is assumed that Eq must take two arguments and equate their denotations, then the problem is that the second argument (i.e., the focus, in most cases) would not be the complement or specifier of Eq. The proposal I put forward in Reeve (2010) is that Eq is in fact a focus-sensitive particle along the lines of only or even – that is, it ‘associates’ with the focus of its clause, taking it as an argument. The remaining argument is provided by the XP in the specifier of Eq. (See Reeve 2010 for argumentation in favour of this.) I have now outlined the structures and interpretation I am assuming for clefts in English and Russian. For more detailed justification, see Reeve (2010, 2011). What is crucial for the purposes of this paper is that there is no appeal to a focus-related functional head to encode the properties of clefts. In the following sections, I will provide a number of empirical arguments which favour the ‘copular’ analysis of clefts over the FP analysis.

4.

Empirical arguments against FP approaches

4.1.

The pronoun has semantic content

Under FP analyses of clefts, cleft pronouns are necessarily expletives – that is, devoid of semantic content, and inserted merely to satisfy a structural requirement. This leads to various problems which are solved if clefts are treated as copular sentences with a semantically contentful subject.

element in question is not a verb (e.g., the ‘pronominal copula’ of Hebrew, and of course èto in Russian). Even in English, as Heycock and Kroch (1999 : 381) point out, equative semantics seems to be independent of the copular verb, since we have ‘equative small clauses’ such as (i) and verbs other than be cooccurring with equative semantics, as in (ii). (Of course, (i) raises the question of what would lexicalise Eq in this case, for which I currently have no answer.) (i) But if what you say is true, that would make the real murderer John! (ii) The real problem remains what to do next. The point is that equative semantics is not tied to the copular verb itself, and that there is a generalisation (not without exceptions) that specificational/ equative clauses often require some overt (possibly non-verbal) element where predicational clauses do not.

Chapter 5 Against FP Analyses of Clefts

171

For example, cleft pronouns pattern with semantically contentful DPs with respect to obligatoriness under S-V inversion in Germanic, the possibility of appearing in Italian Aux-to-Comp constructions, and blocking experiencer raising in French. Clefts in these languages all share features of the English construction, the main one of which is that the cleft clause is formally identical to a restrictive relative clause. It therefore seems reasonable to treat these as variants of the same construction. In the verb-second Germanic languages, a phrasal constituent other than the subject in matrix clause-initial position forces ‘inversion’ of subject and verb: (19)

a. Gestern hat Johann yesterday has Johann ‘Yesterday, John saw Mary.’

Maria Maria

gesehen. seen

b. *Gestern yesterday

Maria Maria

gesehen. seen

Johann Johann

hat has

Interestingly, in some of these languages the behaviour of expletive pronouns diverges from the behaviour of arguments in that expletives are optional or even disallowed as inverted subjects. For example, ‘extraposition’ pronouns in German, while obligatory in SV orders, are optional in VS orders, as shown in (20). This is also true of Icelandic, which in addition disallows ‘weather’ pronouns in VS orders while requiring them in SV orders, as shown in (21) (all examples from Vikner 1995): (20)

(21)

a. Es ist it is

gut, good

b. Natürlich of.course

ist is

a. Það er it is

gott good

b. Að sjálfsögðu of.course c. Það it

dass that

du you

(es) gut, dass it good that að that er is

þú you

(það) it

rigndi. rained

d. Í gær yesterday

gekommen come

rigndi (*það). rained it

ert are

du you

bist. are

gekommen bist. come are

kominn. come

gott að þú ert good that you are

kominn. come

172

Matthew Reeve

On the other hand, cleft pronouns in these languages are always obligatory in both SV and VS orders ((22a-c) from Vikner 1995; (22d) from P. Svenonius, p.c.): (22)

a. Es war DIESER WAGEN, den sie kaufen it was this car which she to.buy ‘It was THIS CAR that she wanted to buy.’

wollte. wanted

b. Gestern war *(es) DIESER WAGEN, den sie kaufen wollte. yesterday was it this car which she to.buy wanted c. Það var JÓN sem ég hitti it was Jón that I saw ‘It was JÓN that I saw in town.’ d. Í gær var *(það) yesterday was it

í in

JÓN sem ég Jón that I

bænum. the.town hitti saw

í in

bænum. the.town

The patterning of cleft pronouns with arguments is clearly unexpected under an FP analysis of clefts, under which cleft pronouns are necessarily expletives. Another case in which clefts appear to pattern with arguments rather than expletives involves the ‘Aux-to-Comp’ construction in Italian, illustrated in (23) (Rizzi 1982): (23)

Ritengo [aver Ria risolto molti believe.1.SG to.have Ria solved many ‘I believe Ria to have solved many problems.’

problemi]. problems

Here, the normal SAux order of Italian clauses is reversed in the embedded clause, a state of affairs which is usually analysed as involving movement of the Aux to C. Italian differs from the Germanic languages in that expletives are always represented by pro. Rizzi (1986) notes that while Aux-to-Comp clauses tolerate true expletive and ‘extraposition’ pro, as in (24a-b), they disallow referential pro, as shown in (24c). Once again, clefts pattern with the referential case – Aux-to-Comp clauses featuring cleft pro are degraded in acceptability, as shown in (24d): (24)

a. Ritengo essere pro probabile che believe.1.SG to.be pro likely that ‘I believe it to be likely that S.’ b. Ritengo essere pro troppo believe.1.SG to.be pro too ‘I believe it to be too late for S.’

tardi late

S. S per for

S. S.

Chapter 5 Against FP Analyses of Clefts c. *Ritengo essere pro believe.1.SG to.be pro ‘I believe him to be nice.’

173

simpatico. nice

d. ??Ritengo essere pro GIANNI che ha rotto il vaso. believe.1.SG to.be pro Gianni that has broken the vase ‘I believe it to be GIANNI that has broken the vase.’

The final argument in this section concerns ‘experiencer blocking’ in French. In this language, a referential DP may not undergo raising across an experiencer argument of the matrix predicate, as in (25b). On the other hand, raising of an expletive across an experiencer is permitted, as in (25d) (Boškoviü 2002 : 196–7). As expected, cleft pronouns pattern with referential DPs in being blocked from raising across experiencers, as in (25f): (25)

a. Deux soldats semblent manquer à la caserne. two soldiers seem to.miss at the barracks ‘Two soldiers seem to be missing at the barracks.’ b. *Deux soldats semblent au general t manquer à la caserne. two soldiers seem to.the general to.miss at the barracks ‘Two soldiers seem to the general to be missing at the barracks.’ c. Il semble y avoir deux soldats manquants à la caserne. it seems there to.have two soldiers missing at the barracks ‘There seem to be two soldiers missing at the barracks.’ d. Il semble au général y avoir deux soldats manquants it seems to.the general there to.have two soldiers missing à la caserne. at the barracks ‘There seem to the general to be two soldiers missing at the barracks.’ e. Ce semble être JEAN que Marie a vu. it seems to.be Jean that Marie has seen ‘It seems to be JEAN that Marie saw.’ f.

*Ce semble au general être JEAN que Marie a vu. it seems to.the general to.be Jean that Marie has seen ‘It seems to the general to be JEAN that Marie saw.’

The evidence surveyed in this section suggests that cleft pronouns have semantic content, since they pattern with referential/argument DPs rather than with expletives. This clearly favours a copular analysis of clefts over an FP analysis.

174

Matthew Reeve

4.2.

Factors behind exhaustivity and existential presuppositions

A second problem with treating cleft pronouns as expletives is that the choice of subject pronoun determines the presuppositional properties of the construction as a whole. It is well-known that clefts in various languages give rise to so-called ‘presuppositions’ of two types. Thus, (26a) ‘presupposes’ that there exists someone who John hit, and that the only person that John hit was Mary. The same is true of the specificational equivalent in (26b). On the other hand, these presuppositions are commonly felt to be absent from the non-copular equivalent in (26c): (26)

a. It was MARY that John hit. b. The one that John hit was MARY. c. John hit MARY.

The first type of presupposition is often referred to as an ‘existential presupposition’ and the second as a ‘uniqueness presupposition’ or ‘exhaustivity’. The question of what is responsible for the presuppositional difference between (26a-b) and (26c) has been much discussed in the literature (e.g., Halvorsen 1978, Atlas and Levinson 1981, Delahunty 1981, Horn 1981, Percus 1997, É. Kiss 1998, Rooth 1999). For those who take the cleft pronoun to be semantically contentful (such as Percus 1997), they are straightforwardly reduced to the presuppositions of definite articles. For those who take the cleft pronoun to be an expletive, accounting for the difference is less straightforward. Earlier analyses such as Halvorsen (1978) and Delahunty (1981) claimed that they were implicatures not derivable from the Logical Form of clefts. For É. Kiss (1998), they were due to features present on the focus head, as we will see below. A comparison of clefts with different types of surface subjects strongly suggests that the ‘definiteness’ approach is the correct one. There is a subtype of cleft in English, little discussed in the literature on clefts (though see Davidse 2000), in which there rather than it appears as the subject: (27)

a. There’s JOHN who’s causing us trouble. b. There’s only HUMPTY DUMPTY that’s that shape.

It seems intuitively clear that there-clefts lack the existential presupposition and exhaustivity found in their it-cleft equivalents (the exhaustivity in (27b) being entirely due to the presence of only).11 That exhaustivity is 11 It is quite difficult to test whether the existential presupposition is absent from there-clefts, besides using intuitions. The classic test for this presupposition is

Chapter 5 Against FP Analyses of Clefts

175

absent can be tested for by adding ‘additive’ focus particles such as also and even to the matrix clause, which are incompatible semantically with exhaustivity. É. Kiss (1998) notes that it-clefts become ungrammatical if such particles are added, as in (28a). On the other hand, adding such particles to there-clefts does not give rise to ungrammaticality, as shown in (28b): (28)

a. It’s MARY who’s causing us trouble, and it’s ??also/*even JOHN that’s causing us trouble. b. Well, THERE’S MARY who’s causing us trouble, and there’s also/even JOHN who’s causing us trouble.

These facts can in principle be accommodated by a copular analysis of clefts. The presuppositional properties of it-clefts can be accounted for under such analyses by treating it as semantically equivalent to a definite determiner. Since definite determiners themselves give rise to exhaustivity, the exhaustivity of clefts is built into the pronoun already. Furthermore, if the semantics of the copula involves equation, then the cleft says that the focus of the cleft is identical in reference to whatever is denoted by the subject. It is thus presupposed that nothing distinct from the focus can be identical to the denotation of the subject. In there-clefts, on the other hand, a copular analysis would predict the absence of exhaustivity. If this there is the same there that appears in existential sentences (as it seems reasonable to suppose), then it must either have no semantic content at all (i.e., be a true expletive), or be some sort of ‘existence’ predicate (e.g., McNally 1992). Either way, there would not encode exhaustivity in the same way as it. Compare this treatment with what would be required under FP analyses of clefts. For example, under É. Kiss’s approach, the presuppositional properties of the

to focus negative quantifiers. If the result seems contradictory, this indicates that there is a presupposition. This distinguishes clefts from their non-cleft equivalents and from focus-fronting sentences: (i) John hit NO ONE. (ii) NO ONE, John hit. (iii) ?*It was NO ONE that John hit. However, the fact that the there-cleft in (iv) is acceptable with no one in focus does not tell us anything about the presuppositions of the there-cleft, since this could also be an existential sentence: (iv) There was NO ONE that John hit.

176

Matthew Reeve

focus are determined by feature values of F. These could be determined by the head selecting F(P). Thus, if F is unselected or selected by C, as in focus-fronting clauses, F could be stipulated to bear the feature values [+ contrastive, – exhaustive]: (29)

a. [FP JOHN F[+con,-exh] [CP that Mary hit]] b. I said [CP [C that] [FP JOHN F[+con,-exh] [IP Mary hit]]]

On the other hand, if F is selected for by I, as in clefts, then F could be said to the feature values [+ contrastive, + exhaustive]: (30)

[IP it [I was] [FP JOHN F[+con,+exh] [CP that Mary hit]]]

This would incorrectly predict that whatever appears in the matrix SpecIP should not make a difference to the presuppositions of the cleft, as F would be selected by I regardless of what occupies SpecIP. In order to capture the facts correctly, the analysis would have to be modified so that the feature values of F would be determined not by what head selects for F, but by what occupies SpecIP – clearly not a very appealing analysis.

4.3.

Narrow vs. wide focus clefts

Another problem for the FP account is posed by the fact that clefts do not always express narrow focus. First, Prince (1978) noted a type of cleft which she called ‘informative-presuppositional’, in which the whole of the material following the copula appears to be in focus. In English, this possibility appears to be restricted to the case where the clefted XP is a VPadjunct of some type, as in (31a): (31)

It was just about 50 years ago that Henry Ford gave us the weekend. (Prince 1978 : 898)

In some languages, however, wide focus is possible even if the clefted XP is an argument. This is an option in both French clefts, which like English clefts have a relative clause-like syntax, and Russian clefts, which do not:12

12 Although I cannot present a full analysis of French wide-focus clefts here, it seems reasonable to treat them as true copular sentences. Unlike English, French also has cases in which a relative clause modifies a DP, but the resulting

Chapter 5 Against FP Analyses of Clefts (32)

Q: Tu sembles inquiète. Qu’ est-ce qui se you seem worried what is-it which self ‘You seem worried. What happened?’

177

passe? happens

A: C’ est le petit qui est tombé dans l’ escalier. it is the small who is fallen in the stairs ‘The young one fell down the stairs.’ (Clech-Darbon et al. 1999 : 84) (33)

(There was a knock on the dinner table.) Èto Nikanor Ivanoviþ uronil ložku na kleënku. this Nikanor Ivanovich dropped spoon on oilcloth ‘Nikanor Ivanovich had dropped the spoon on the oilcloth.’ (Junghanns 1997 : 176)

As regards the French case, note that under the É. Kiss account, there must still be movement of the argument (le petit in (32)) or the associated relative pronoun to some left-peripheral position. The question then arises of why just the DP subject moves if the entire IP is in focus.13 In Russian, the wide-focus case arises when there is no movement to the left periphery. In (33), the material following èto follows the SVO pattern, which is generally thought to be the unmarked word order in Russian (e.g., Slioussar 2007, contra King 1993). Thus, (33) would be able to have IP focus whether or not èto was present. Under the FP account, presumably, the whole IP would have to move to the left-peripheral position in such cases. As is well-known, however, IPs are generally immobile (e.g., Den Dikken 1995), and so the possibility of IP-movement in these cases is entirely unexpected. Furthermore, there is strong evidence that the IP does not in constituent can occur as the complement of a perception verb (so-called ‘pseudorelatives’; see Radford 1975): (i) J’ ai vu Marie qui courait à I have seen Marie who ran at ‘I saw Marie running at full speed.’

toute all

vitesse. speed

Here, what is seen is not the referent of the DP itself, but rather the event in which the DP participates. Similarly, in the wide-focus cleft case in (31), we can give the sentence the same structure as a standard cleft, in which the DP is the complement of the copula, but the DP+relative denotes an event rather than an individual – making the cleft a natural answer to the question ‘What happened?’. 13 There are cases documented in the literature of languages in which movement of part of a phrase can signal focus of the entire phrase, but they remain exceptional. One example is Kîîtharaka (Abels and Muriungi 2008).

178

Matthew Reeve

fact move in wide-focus ‘clefts’ such as (33). Focus-movement in Russian involves an obligatory contrastive interpretation of the focus – it cannot be used to express mere new information. Thus, (34b) is not a possible response to the wh-question in (34a). The same is true of narrow focus clefts, as shown in (34c) – one reason to think they involve obligatory A’movement of the focused XP: (34)

a. ýto Maria kupila? what Maria bought ‘What did Maria buy?’ b. #ŠLJAPU ona kupila. hat she bought ‘A hat, she bought.’ c. #Èto ŠLJAPU ona kupila. this hat she bought ‘It was a hat that she bought.’

In a context such as (35a), however, both focus-fronting and narrow focus clefts are acceptable: (35)

a. Maria kupila ŠLJAPU. Maria bought hat ‘Maria bought a hat.’ b. Net, KURTKU ona kupila. no coat she bought ‘No, a coat, she bought.’ c. Net, èto KURTKU ona kupila. no this coat she bought ‘No, it was a coat that she bought.’

‘Wide focus clefts’ such as (33) differ from narrow focus clefts and focusfronting in that they can be used to answer a wh-question – that is, the wide focus need not be contrastive: (36)

a. ýto sluþilos’? what happened ‘What happened?’ b. Èto Nikanor Ivanoviþ uronil ložku na kleënku. this Nikanor Ivanovich dropped spoon on oilcloth ‘Nikanor Ivanovich dropped the spoon on the oilcloth.’

Chapter 5 Against FP Analyses of Clefts

179

How could an FP analysis deal with such cases? It would be odd to claim that the IP moves to the focus position, both for ‘IP immobility’ reasons and because of the lack of obligatory contrastivity. It would have to be stated that the focus position induces obligatory contrastivity unless the constituent occupying the position is IP – clearly an undesirable stipulation. On the other hand, if the IP did not move to the focus position in (36b), it is not clear why the focus must move in narrow focus cases. Again, an IP-specific stipulation would have to be introduced. Under the present copular analysis, wide-focus clefts in Russian can be treated as the equation of a question with its answer, as has been proposed by several authors for English pseudoclefts (e.g., Ross 1972, Den Dikken et al. 2000, Schlenker 2003, Romero 2005), and also by Geist and Błaszczak (2000) and Markman (2008) for Russian clefts in general. That is, in (36b) the pronoun èto is anaphoric to the question What happened?, and the IP following èto represents its answer. Given that, under the present proposal, èto is the specifier of the equative head Eq, and IP is its complement, the denotations of the two can be straightforwardly equated. Crucially, there is no need for movement of the IP in order to achieve this result. Of course, this raises the question of why narrow foci do need to move in Russian clefts; this question will be addressed in the next subsection.

4.4.

Adjacency

All else being equal, if focus constructions in a particular language involve a focus-related functional category, then we might expect focus-related movement to behave in a parallel fashion in both clefts and non-clefts. Under É. Kiss’s approach, this cannot be the case, since clefted XPs precede the complementiser that while focus-moved XPs must follow it: (37)

a. b. c. d.

It was Mary that John hit. *It was that Mary hit John. (with narrow focus on Mary) You said that Mary John hit. *You said Mary that John hit.

A copular analysis of clefts, which does not rely on focus-movement in (all) clefts, at least has the advantage that we expect the clefted XP to precede the complementiser, just as the head noun of a relative clause must precede the relative complementiser. It turns out that languages such as Russian provide an even more serious problem for FP analyses. I

180

Matthew Reeve

showed in Reeve (2010) that clefted XPs in Russian involve the same type of movement as focus-moved XPs (adjoined to SpecIP; a similar analysis of focus-movement in clefts is given in Junghanns 1997), regardless of the category of the clefted XP. However, the two diverge in that clefted XPs must normally precede their ‘background’ – that is, they must generally appear adjacent to èto in a narrow focus context: (38)

A:

Maria uvidela Borisa. Maria saw Boris ‘Maria saw Boris.’

a. B:

Net, èto Ivana Maria no this Ivan Maria ‘No, it was Ivan that Maria saw.’

uvidela. saw

b. B:

?#Net, no

uvidela. saw

èto this

Maria Maria

Ivana Ivan

On the other hand, non-cleft focus-movement is much freer: in fact, the standard position for a contrastively focused XP is the pre-VP position (which can be analysed as adjoined to SpecVP), though pre-IP is also possible (e.g., Krylova and Khvronina 1984, King 1993): (39)

A:

Maria uvidela Maria saw ‘Maria saw Boris.’

Borisa. Boris

a. B:

Net, Maria Ivana uvidela (a ne no Maria Ivan saw and not ‘No, Maria saw Ivan (not Boris).’

Borisa). Boris

b. B:

Net, Ivana Maria uvidela (a ne no Ivan Maria saw and not ‘No, Maria saw Ivan (not Boris).’

Borisa). Boris

Any FP analysis of these facts would have to (i) posit at least two FPs, one pre-IP and one pre-VP, and (ii) state that when èto is present, only the preIP position can be used. Although (i) is feasible, (ii) would be an undesirable stipulation unless independently motivated. How does the present analysis capture these facts? First, I assume, following Sekerina (1997) and Stepanov (1998), that focus-movement is left-adjunction to a verbal maximal projection. Thus, all else being equal, focus-movement is free in principle to adjoin to VP or IP. The difference between clefts and non-clefts resides in the fact that in clefts, the pro-

Chapter 5 Against FP Analyses of Clefts

181

noun èto needs to be interpreted as a discontinuous definite description with the backgrounded material. Recall that under the present analysis, this is achieved via the thematic licensing principle in (11), repeated as (40): (40)

Thematic licensing condition For theta-binding to take place, the constituent bearing the ș-role to be bound must be c-commanded by the theta-binding determiner and must m-command that determiner.

In a structure in which the clefted XP moves to adjoin to IP, this condition can be satisfied – the lower segment of IP m-commands èto and is c-commanded by it, as shown in (41): TP

(41) pro

T’ T

EqP DP èto

Eq’ Eq

TP DPi vodku

TP DP Boris

T’ T

VP V vypil

ti

The condition clearly would not be satisfied, however, if vodku in (41) adjoined to VP instead. In fact, the background would not even form a constituent. In the case of non-cleft focus movement, the thematic condition does not apply, and so focus-movement is free to adjoin to either IP or VP. The present analysis thus avoids the disadvantages of the FP analysis mentioned above while capturing the common structural properties of English and Russian clefts – namely, that they involve theta-binding of a ‘background’ by a pronoun.

182

Matthew Reeve

4.5.

Focus movement and contrastivity

Since FP approaches posit focus-movement of the clefted XP, they predict that further focus-movement of clefted XPs should always be impossible.14 This is because of the commonly-noted fact that A’-movements of a particular type do not iterate (e.g., Abels 2008); thus, focus-movement of an XP should not be able to followed by focus-movement of the same XP (or part of it). In fact, this is an incorrect prediction: clefted XPs can often be focus-moved to the left periphery of the matrix clause. The availability of such movement is partly dependent on the category of the clefted XP: DPs and PPs can move relatively easily, while other XPs are harder to move: (42)

a. b. c. d.

JOHN it was that Mary saw. ?*GREEN it was that her eyes were. ?IN LONDON it was that I saw a rat. ??TO JOHN it was that I gave the vodka.

As expected if wh-movement and focus-movement are distinct types of A’-movement (as Abels argues), wh-movement of a clefted XP is available in all the above cases: (43)

a. b. c. d.

Who was it that Mary saw? What colour was it that her eyes were? In which city was it that you saw a rat? To whom was it that you gave the vodka?

In Reeve (2010), I argued that the contrasts in (43) were due to the availability of a derivation of DP-clefts and certain PP-clefts in which the clefted XP is base-generated in its surface position (a ‘dual derivation’ analysis that has its roots in Pinkham and Hankamer 1975). This means that the DP/PP in postcopular position has not already undergone movement licensed by a focus interpretation, and so is free to undergo focusmovement to the matrix left periphery. On the other hand, clefted APs, for example, must raise from the relevant position inside the cleft clause to the postcopular position, a movement which is licensed by focus, and thus cannot undergo further focus-movement. In Russian clefts, on the other hand, I proposed a single-derivation analysis – the clefted XP always moves regardless of its category. As

14 I am grateful to Ad Neeleman (p.c.) for suggesting this argument.

Chapter 5 Against FP Analyses of Clefts

183

expected, then, focus-movement of the clefted XP over èto is impossible even in DP-clefts, while wh-movement of the DP is possible: (44)

a. *VODKUi èto ti Boris vodka-ACC this Boris.NOM drank ‘VODKA it was that Boris drank.’ b. Ètoi èto ti Boris what.ACC this Boris.NOM ‘What was it that Boris drank?’

vypil

ti.

vypil ti? drank

As noted above, the FP account, in contrast to the equative account, predicts that further focus-movement should never be possible. This is because, regardless of whether the clefted XP is base-generated in SpecFP or moves there, it is already in a focus-licensing position, and so should not be able to move to a second focus-licensing position; that is, we would always end up with the following structure: (45)

*[FP XPi [F´ F[CP C [IP … I [FP ti [F´ F [CP … (ti) …]]]]]]]

A related fact which is problematic for the FP account is that it is not true that all clefted XPs in English are only optionally contrastive. Recall that É. Kiss’s analysis assigns the features [+contrastive, +exhaustive] to clefted XPs and would have to assign [+contrastive, +exhaustive] to focus-fronted XPs. This predicts, then, that anything moved to SpecFP of clefts need not be interpreted contrastively. As noted by authors including Heggie (1988), however, AP-clefts must be interpreted contrastively. If, as is standard, we take wh-questions to exemplify a context in which the focus in the reply is non-contrastive, then (46a) shows that DP-clefts can be used non-contrastively, as captured by É. Kiss’s proposal. The ‘dialogue’ in (46b), on the other hand, shows that they can also be used contrastively. (46)

a. A: B: b. A: B:

Who did Mary hit? It was JOHN that Mary hit. I think that Mary hit BILL. No, it was JOHN that Mary hit.

Now consider a parallel case with AP-clefts provided by Heggie (1988): (47)

a. A: B: b. A: B:

What colour are her eyes? #It’s GREEN that her eyes are. Her eyes are green. No, it’s BLUE that her eyes are, not GREEN.

184

Matthew Reeve

In contrast to DP-clefts, AP-clefts are infelicitous in the wh-question (i.e., non-contrastive) context, and require explicit contrast, as in (47b), to be felicitous. How could this be accounted for under the FP approach? It would surely be unnatural to propose that F has a different feature setting for [contrastive] if an AP moves to it than if a DP moves to it, and it is certainly not an option if the feature settings depend on selection by a higher head, as FP would presumably be selected for by I in both cases. Under the copular approach, as noted above, this contrast can be reduced to the availability of a non-movement derivation for DP-clefts as opposed to AP-clefts, which must involve movement. That is, obligatory contrastivity is a result of A’-movement, a phenomenon which has often been pointed out in other languages (e.g., Molnár 2006, Neeleman and van de Koot 2008). Recall that the main point of É. Kiss’s paper is that there is a distinction between syntactically represented identificational focus (represented by F) and syntactically inert information focus. However, given that the ‘cleft position’ can express both information focus and (contrastive) identificational focus in her terms, this casts doubt on the conceptual reasoning behind the FP analysis of clefts, since SpecFP would not be a position uniformly representing identificational focus.

4.6.

Clefts and extraposition

There is a further set of facts which suggest that the FP approach is on the wrong track. These provide evidence that the relation between the clefted XP and the cleft clause is not one of a focused XP sitting in the left periphery of a clause, as under the FP account, but a relation between a restrictive relative clause (the cleft clause) and its ‘host’ DP (the clefted XP). The most obvious indication of this is the optional presence of a relative operator in the cleft clause. However, as many authors have questioned the idea that cleft clauses are identical to restrictive relatives (e.g., Delahunty 1981 : 130, Rochemont 1986), it seems that this is not necessarily the most convincing evidence either way. On the other hand, there is evidence that the cleft clause obeys locality restrictions on extraposition that would not be expected if it were merely a CP selected for by a higher clausal head. Consider first the following data from Culicover and Rochemont (1990) (see also Taraldsen 1981) involving relative clause extraposition (note that in all the examples the relative clause is to be construed as modifying a man):

Chapter 5 Against FP Analyses of Clefts (48)

a. A man was believed t to have arrived that I had insulted. b. ?A man was believed [t to have arrived] by everybody that I had insulted. c. *A man was believed [t to have arrived that I had insulted] by everybody.

(49)

a. ??It seemed to heri [that a man had arrived that Maryi knew from school]. b. A man seemed to heri [t to have arrived] that Maryi knew from school.

185

(48a) contains a biclausal structure with a matrix subject which originates in the embedded clause. Furthermore, this subject is linked interpretatively to an extraposed restrictive relative clause. Notice that it is not clear from (48a) whether the relative surfaces in the matrix clause or the embedded clause. If, however, material clearly belonging to the matrix clause follows the relative, as in (48c), the sentence becomes strongly degraded, whereas if this material precedes the relative, as in (48b), the sentence is relatively acceptable. What this suggests is that the relative clause must be in the matrix clause if its ‘host’ DP is. A similar point can be made regarding binding. In (49a), her and Mary cannot normally be coreferential, which can be attributed to a Condition C violation. In other words, the relative clause in (49a) is inside the embedded clause, and is hence (almost-)c-commanded by her. On the other hand, if the embedded subject (with which the relative containing Mary is interpretatively linked) is raised to matrix subject position, as in (49b), then coreference becomes possible. This can be explained by saying that the relative appears in the matrix clause when its ‘host’ DP does, thus escaping c-command by her. The conclusion Culicover and Rochemont draw from this is that there is a surface (S-Structure, in their terms) requirement for an extraposed relative clause to belong to the same clause as the DP with which it is construed. Now consider clefts. Interestingly, the tests above indicate that the cleft clause obeys the surface locality requirement with respect to the clefted XP (and not to it, as theories such as those of Hedberg (1990) and Percus (1997) would predict). The relevant data are given in (50–52): (50)

a. ??It was believed [t to be JOHN] by everybody that Mary saw. b. It was believed [t to be JOHN that Mary saw] by everybody.

(51)

a. ?*It seemed to heri [that it was JOHN that Maryi saw]. b. ?*It seemed to heri [t to be JOHN that Maryi saw].

186 (52)

Matthew Reeve a. Who did it seem to heri [t to be t] that Maryi saw? b. *Shei wondered [who it seemed [t to be t] that Maryi saw].

In (50a), when matrix material (a passive by-phrase) precedes the cleft clause, forcing it to appear in the matrix clause, the result is degraded. By contrast, when the matrix material follows the cleft clause, as in (50b), the result is acceptable. If the cleft clause must bear the same locality relation to John in (50) as the relative clause in (48) must to a man, then this contrast can be accounted for. In (51), raising of cleft it in (51b) does not ameliorate the Condition C violation, which is expected given that John appears in the same position in both (51a,b). (52) illustrates the fact that wh-movement of the clefted XP can ameliorate Condition C violations, provided the target of wh-movement is not c-commanded by the relevant pronoun. The contrast between (52a,b) is accounted for as follows. In (52a) the clefted XP is wh-moved into the matrix SpecCP, forcing the cleft clause to appear in the matrix clause too. This means that the pronoun no longer c-commands into the cleft clause, preventing a Condition C violation. On the other hand, wh-movement in (52b) proceeds only as far as the embedded SpecCP, forcing the cleft clause to remain inside the embedded clause, where it is still c-commanded by the matrix pronoun. The present analysis can account for these facts, because it claims that the cleft clause originates as a modifier of the clefted XP in a purely syntactic sense, though it comes to modify it semantically via extraposition. Thus the locality conditions on extraposition are calculated with respect to the clefted XP. However, the extraposition relationship between the cleft clause and clefted XP is completely unexpected under an FP analysis, according to which the cleft clause is a CP selected for by matrix I. If this were the case, we would not expect wh-movement of the clefted XP to force the CP to extrapose.

5.

Conclusion

In this paper, I have argued against the postulation of a focus-related functional head from the perspective of cleft constructions in English and Russian. The arguments in this paper, which are primarily empirical, also contribute towards the general conceptual argument against focus heads, which finds expression in some of the other contributions to this volume. Clefts and cleft-like constructions have been made to bear quite a lot of the empirical burden for focus-related functional categories. If it can be shown that in many of these cases, a copular approach (or still another

Chapter 5 Against FP Analyses of Clefts

187

approach) is in fact superior on empirical grounds, this weakens considerably the motivation for postulating F in this first place: its only purpose will be to deal with the perceived theoretical pitfalls of not encoding focus (and other discourse-related notions) in the syntactic derivation. This is not to say that the job of undermining FP approaches has been done – there are still some interesting cases (such as the focus particles in the Gbe languages; see, e.g., Aboh 2004) for which a copular analysis is highly implausible, and a focus head approach seems unavoidable. Yet as long as we take seriously the idea that postulating functional categories as part of Universal Grammar must have some empirical basis, the stark difference between focus-marking (not often expressed syntactically, if this paper and others in this volume are correct) and, for example, wh-marking (seemingly universal) should at least motivate a reconsideration of the FP approach.

Chapter 6 Focus movement can be destressing, but it need not be* Kriszta SzendrĘi This paper has two parts. First, I investigate the possibility of focus movement inside the noun phrase, reanalysing some of the data I discussed in SzendrĘi (2010), reaching partially different conclusions. I follow Neeleman et al (2009) in treating focus movement as an instance of movement marking the domain of contrast. I propose that adjective reordering in English should be analysed in these terms. Some of the evidence for this analysis comes from the possibility of reconstruction in such cases. In contrast, the Greek polydefinite construction, which is often argued to involve DP-internal focus movement, does not mark the domain of contrast and does not involve movement. Rather, it seems to involve givenness marking (Lekakou and SzendrĘi 2012). This raises the theoretical question whether contrastive focus marking and givenness marking are two sides of the same coin, as suggested by Schwarzschild (1999) and more recently by Wagner (2006, 2012). The second part of the paper is devoted to this issue. I present some new theoretical and empirical arguments against this unified position, supporting earlier proposals along these lines by Neeleman and Reinhart (1998), Neeleman and SzendrĘi (2004), Féry and Samek-Lodovici (2006), Krifka (2006b) and Reinhart (2006).

* This paper would have never taken shape without the perfect combination of gentle pressure and seemingly inexhaustible patience of the editors, Ad Neeleman and Reiko Vermeulen. I am very grateful to them. I also thank them for helpful suggestions and thorough discussions of the material. For suggestions, data, comments and discussions I am also indebted to Ingo Feldhausen, Berit Gehrke, Hans van de Koot, Timothy Leffel, Marika Lekakou and audiences of the Prosody colloquium at the University of Frankfurt (1 February 2012) and the graduate course on information structure at Potsdam University (20 February-2 March 2012). This paper includes material published in SzendrĘi (2010). Lingua’s permission to reproduce this material is gratefully acknowedged here.

190

Kriszta SzendrĘi

1.

Introduction

SzendrĘi (2010) reviewed a series of phenomena where it has been previously proposed that topic and focus movement takes place DP-internally, targeting designated left-peripheral high functional heads in the DP. I argued there that the DP, being argumental and not propositional, is illsuited for topic/comment or focus/background partitioning (see also Sánchez 2010 : 130–131 for the same claim not only for regular DP arguments but also for nominalised clauses lacking independent tense in Quechua). For this reason, no such partitioning can take place DP-internally. I looked at two sets of data in more detail: adjective reordering and polydefinites in Greek.1 For the former, I claimed, the reordering serves scope considerations; for the latter, I suggested that givenness rather than focussing was the correct trigger. In the light of recent proposals about focus movement in Neeleman et al (2009) and Wagner (2005, 2012), I would like to re-examine these data. I will conclude that focus movement marking its sister as its domain of contrast is possible inside the noun phrase. Arguably, this takes place in the case of noncanonical adjective placement. In the absence of movement, the domain of contrast is not explicitly marked in the syntax. This is the situation, I will argue, in the Greek polydefinite construction, whose function is to deaccent its nominal head and thus mark it as anaphorically given. Next, I will consider Wagner’s (2007, 2012) proposal that focussing a constituent is the same thing as marking its sister as given and vice versa. I will argue that this ‘see-saw’ theory of focusing and givenness, although very attractive, cannot be the whole story: it must be supplemented by an independent stress strengthening (i.e. focussing) mechanism. The arguments presented fall in line with much previous work by Neeleman and SzendrĘi 2004, Féry and Samek-Lodovici (2006), Krifka (2006) and Reinhart (2006). Many languages show discourse-related word order variation. It is generally accepted that examples like (1) involve a topic/ comment structure, while (2) involves a focus/ background structure. (1)

This tie, FRED bought.

(Cormack and Smith 2000 : 390)

(2)

NOTHING I ate for breakfast.

(Cormack and Smith 2000 : 397)

1 At the end of SzendrĘi (2010), I also reviewed claims about alleged topic movement inside the Hungarian DP involving dative possessors. I hope to have shown that there is nothing topical about them. I will not repeat the arguments here as this paper concentrates on focus movement.

Chapter 6 Focus movement can be destressing, but it need not be

191

Reinhart (1981), following Strawson (1964), proposed that when an utterance is assessed in context, this process involves checking ‘predication’, where one expression in the sentence is taken as the argument and the rest as the predicate. The argument is the topic of the utterance in the given context; the predicate is the comment. Syntactic considerations may constrain what the topic (i.e. the argument of the predication) may be. For instance, in passives, the topic must be the subject and in clitic-left dislocation, the topic is always the dislocated element. In utterances like (1), the topic is distinguished by its position, the rest of the utterance constitutes the comment. Focus is the part of an utterance that provides an answer to a corresponding (implicit) wh-question. This can be implemented with structured meanings (Jacobs 1983, Von Stechow 1990, Krifka 2006b) as well as alternative semantics (Rooth 1992). Either way, what is important is that the background associated with the focus determines the set corresponding to the implicit wh-question. In other words, the background is an open proposition, which can be matched to a set of alternatives. In utterances like (2), the focus is syntactically displaced, and the background is the open proposition constituted by the rest of the utterance. Neeleman et al (2009) propose an account based on the idea that A’movement of a contrastive topic or focus determines its scope, much like A’-movement of other quantificational elements. The movement marks what material is included in the scope of this operator, which they call the domain of contrast (DoC). The proposal is formalized as below. (3)

(4)

DoC Marking The sister of a moved contrastive focal (or topical) constituent, XP, is interpreted as the domain of contrast for XP. (adapted from Neeleman et al 2009 ex. 10) N1 [M #] XP[contrast] N2 [M ] (Neeleman et al 2009 ex.11)

If a contrastive topic or focus remains in situ, the domain of contrast is not marked. In interpreting the sentence, the hearer must therefore construe an appropriate domain of contrast based on contextual clues. This can be the sister of the contrastive category, but it can also contain more material and as a result be discontinuous. In other words, the DoC for XP in (5) can be its sister, as in (5b) or YP minus XP itself, as in (5a). A’-movement of

192

Kriszta SzendrĘi

the contrastive category has a disambiguating effect, as it creates a match between the syntactic representation and the information structure associated with the sentence, as shown in (5b). (5)

a.

YP

b. XP[contrast]

(DoC)

XPDoC

XP[contrast] (DoC)

tXP

To illustrate, in (6), the contrastive focus XP FLOWERS has been fronted, marking its sister John has given tXP to Mary as the domain of contrast. The specific semantic representation Neeleman et al propose for (6) is given in (7), with the DoC underlined. In plain terms, in (6), flowers are contrasted with other entities that John could have given to Mary. (6)

FLOWERS John has given to Mary.

(7)

a. < Ox[John has given x to Mary], flowers, {flowers, chocolate,…}> b. y [y  {flowers, chocolate, …} & ™[John has given y to Mary]].

Neeleman et al also discuss cases where the focal element does not move to the left-periphery, but to a position below the subject. This is not allowed in English, but Dutch, for instance, allows focus movement to target the so-called middle field of the sentence. The semantics they propose for such cases is given in (8). Here, the DoC is not an open proposition. When this is the case, they argue, existential closure applies to the DoC (underlined). This gives rise to a meaning where flowers are contrasted with other entities that someone could have given to Mary. (8)

a. < Ox[John has given x to Mary], flowers, {flowers, chocolate,…}> b. y [y  {flowers, chocolate, …} & z ™[z has given y to Mary]].

This solution follows Schwarzschild (1999)’s proposal. Wagner (2012) argues for a somewhat different solution involving universal closure. The technical details of the solution need not concern us here. What matters is that the DoC may be semantically something other than an open proposition, and by closure, an open proposition can be obtained. This paves the way for a treatment of DoC-marking DP-internally. In particular, what

Chapter 6 Focus movement can be destressing, but it need not be

193

Neeleman et al argue is that the material that is not part of the domain of contrast is simply filled in from the context. So, in (8), the giving event will be understood to involve John as its subject if John is the most easily accessible individual in the previous context. Similarly, if domain of context marking takes place inside the DP argument, the predicate and the remaining arguments will be represented by existentially bound variables and specified further by the preceding context. Wagner (2007, 2012) looked at cases involving noncanonical stress placement inside the DP, such as (9) and (10). He argued that destressing the noun convertible is only possible if it is contextually given and its sister constituent, the adjectival modifier, can be felicitously contrasted with the modifier of the antecedent. So, (9b) is inappropriate because blue does not contrast with high-end, while (10b) is well-formed because cheap does. (9)

Sally’s uncle, who is incredibly rich and produces high-end convertibles, came to her wedding. I wonder what he brought as a present. a. Guess what: He brought a blue CONVERTIBLE. b. ?# Guess what: He brought a BLUE convertible. (Wagner 2012 : 13 ex 26)

(10)

Sally’s uncle, who is incredibly rich and produces high-end convertibles, came to her wedding. I wonder what he brought as a present. a. ?# Guess what: He brought a cheap CONVERTIBLE. b. Guess what: He brought a CHEAP convertible. (Wagner 2012 : 13-4 ex 27)

Based on examples like these, he argued for what I will call the ‘see-saw’ theory of focus and givenness. He follows Williams (1997) in assuming that accent placement can be represented by assigning Strong or Weak labels to binary branching sisters. Marked accent placement is the result of a Strong-Weak swap on a pair of sisters. In particular, Wagner claims, assigning Weak to the noun convertible, by virtue of it being given in the preceding discourse, automatically means assigning Strong to its sister node, the adjectival modifier. As a direct consequence of this, the adjective will be interpreted contrastively (invoking a set of alternatives). Thus, it is only if the arising contrast is meaningful (i.e. cheap vs. highend but not blue vs. high-end) that destressing on the noun can take place. I will come back to a more thorough discussion of this proposal in section 3. Recall that Neeleman et al (2009) argue that the sister of a moved focal element is marked as its domain of contrast. Thus, focus movement is

194

Kriszta SzendrĘi

scope-driven movement. Wagner (2007, 2012) argued that givenness marking and focus marking are two sides of the same coin. Although he does not discuss cases of movement, a straight forward extension of his proposal would mean that focus movement involves assigning a Strong label to the moved constituent, and by the see-saw, assigning Weak to its sister node, as given schematically in (11). (11)

Extension of Wagner (2012) to focus movement: [XPStrong [Weak … tXP … ]]

In principle, it is possible that both Neeleman et al (2009) and Wagner (2007, 2012) are correct and the moved constituent marks its sister as the domain of contrast and given at the same time. Indeed, this will hold in many cases. Contrastive focus is often used in discourse situations where the domain of contrast is given. Take for instance an instance of correction, as in (12). (12)

A: She took the S7 to Wannsee. B: No. The S1 she took.

However, this does not mean that this always holds, or that it must hold. In what follows I will look at a series of data where the two theories potentially diverge.

2.

Adjective reordering inside DP: A-bar movement marking the domain of contrast

2.1

Adjective reordering marks the domain of contrast

The evidence presented in favour of a DP-internal topic or focus position in the literature involves several different lines of argumentation. I will start by looking at what I believe is the most robust set of data: adjective reordering associated with contrastive focus. In the next section, I will turn to the Greek polydefinite construction, which has been argued to involve DP-internal focus movement in many proposals (see e.g. Ntelitheos 2004, Kariaeva 2004, Ioannidou and Den Dikken 2009). Truswell (2005) is concerned with certain discourse-related word order variations inside the DP. (See also Laenzlinger (2000, 2005)). He admits that a syntactic parallel between the clausal left-periphery and the DP is

Chapter 6 Focus movement can be destressing, but it need not be

195

less than straight forward, but nevertheless claims that data like (12) give evidence for focus movement inside the DP.2 (13)

My friends all drive big cars, but only I drive a BLACK big car. (Truswell 2005)

Here the normal ordering between a subsective adjective like big and an intersective adjective like black is reversed and the intersective adjective bears heavy, contrastive stress and pitch accent. Truswell notes that such reordering is only possible if the intersective adjective is contrasted. He thus concludes that the adjective is fronted inside the DP in order to be marked for contrastive focus. But note that reordering is not necessary for a contrastive focus reading. In (14) the adjectives follow the normal order and the contrasting intersective adjective is only marked prosodically. So, focus movement inside the DP, if this is what this example contains, is optional. (14)

My friends all drive big cars, but only I drive a big BLACK car.

To see what the real import of the noncanonical adjective ordering is, consider an utterance like (15), which is a simplified version of (13). By Neeleman et al’s proposal, the domain of contrast is the NP big car, which is the sister of the accented adjective BLACK. So, (15) is appropriate in the discourse context 1, given in (16a) (also illustrated by Scenario 1). In this context there is a given set of big cars (known to the speaker and the hearer) one of which is black. But the same utterance is not felicitous in discourse context 2 (illustrated in Scenario 2), where there is a set of black cars one of which is big alongside other sets of cars of different colours where one of each set is big. (15)

I drive a BLACK big car.

(16)

a. Context 1:

Scenario 1:

Context 1: 9

Context 2: 

In this car park you can see my friends’ cars and my car. There is a bunch of big cars here. They are of many different colours. BLACK RED BLUE WHITE BLUE YELLOW

2 Note that Truswell (2009) concludes that adjective reordering is often scopally motivated, rather than information structurally. Similar data is discussed below (cf. the collective reading of (22)).

196

Kriszta SzendrĘi b. Context 2:

Scenario 2:

In this car park you can see my friends’ cars and my car. There are cars of many different colours. Some of them are small, but there are big cars of every colour. black BLACK black black red red red RED red red …

So, the import of noncanonical adjective placement seems to be marking the modified noun, big car in our example, as the domain of contrast. The adjective BLACK is contrastively focussed. But it is not focusing the adjective that distinguishes this order from the canonical order, but marking the modified noun big car as the domain of contrast. As expected, the canonical order, given in (17), is compatible with either scenarios in (16). This is because this utterance contains canonical order and stress shift to the adjective BLACK. The latter operation has the effect of focusing the adjective and also, by the see-saw, marking its sister, the noun car, as given. This is compatible with Context 2. In addition, (17) is also appropriate in Context 1, where the domain of contrast is determined by the modified NP big car. But since no focus movement took place, the domain of contrast need not directly correspond to a syntactic constituent. It can be discontinuous. Hence the appropriateness of (17) in Context 1 follows. (17)

I drive a big BLACK car.

Context 1: 9

Context 2: 9

The following example shows that contrast on the moved adjective is necessary. If the adjective black is not contrastive, it cannot appear in a noncanonical position.3 (18)

*All my friends drive a black car and I drive a black BIG car.

To sum up, the domain of contrast for the noncanonically placed focused adjective is its sister. What licences the noncanonical placement is that this allows for marking the modified noun phrase big car to be the domain of contrast.

3 This needs to be qualified in the light of scope data discussed below. There we will see that the adjective can also appear in a noncanonical position if that expresses a special scope reading.

Chapter 6 Focus movement can be destressing, but it need not be

2.2

197

Adjective reordering involves A-bar movement

Next we should establish whether noncanonical adjective placement involves movement or just a noncanonical base-generated order. To see this, we have to briefly examine how focus movement interacts with scope in the clausal domain. As Neeleman & van de Koot (2007) demonstrate, languages like Dutch allow two different types of scrambling. So-called A-scrambling, interacts with A-binding and secondary predication, does not give rise to weak crossover effects and never reconstructs for scope. In contrast, A-bar scrambling, does not affect binding or secondary predication, gives rise to weak crossover effects and obligatorily reconstructs for scope. The two also have distinct discourse properties: A-scrambling affects discourse anaphoric elements; elements that undergo A-bar scrambling are either (contrastive) topics or foci. These opposing sets of properties led Neeleman & van de Koot (2007) to offer different syntactic analyses for the two different types of scrambling. They claim that A-scrambling is the result of different base-generated orders (see also Ruys 2001), while A-bar scrambling is essentially topic or focus movement, triggered by the aim to create a syntactically continuous comment or background. Since here we are interested in the different scopal properties of the two constructions, I give the relevant data in (19) and (20). (See Neeleman & van de Koot (2007) and references there for the rest of the data.) As (19) shows, an indefinite noun phrase that has been A-scrambled across an adverb cannot reconstruct for scope under the adverb. (19)

a. dat ik waarschijnlijk iemand uit New York zal uitnodigen that I probably someone from New York will invite (i) ?someone > probably; (ii) probably > someone ‘that I will probably invite someone from New York’ b. dat ik iemand uit New York waarschijnlijk zal uitnodigen that I someone from New York probably will invite (i) someone > probably; (ii) *probably > someone ‘that I will probably invite someone from New York’

In contrast, as (20) shows, A-bar scrambling allows for reconstruction. Here the reading where ‘most’ outscopes ‘at least one’ is available.4 4 In fact, Neeleman & van de Koot (2007) present arguments that in sentences in which focus movement spans a clause boundary, reconstruction is obligatory but this need not concern us here.

198 (20)

Kriszta SzendrĘi dat [ tenminste ÉÉN artikel over syntaxis] de meeste studenten tDP wel that at-least one article about syntax the most students indeed gelezen zullen hebben read will have (i) at least one > most; (ii) most > at least one ‘that most students will at least have read one article about syntax’

The syntactic analysis proposed by Neeleman & van de Koot (2007) captures the scope reconstruction data in a straightforward manner: A-scrambling is analysed as a different base-generated order, so it is not surprising that it does not allow reconstruction; A-bar scrambling, which involves A-bar movement, unsurprisingly, allows reconstruction for scope. Now the question is whether the adjective reordering data patterns with A-scrambling or with A-bar scrambling with respect to scope reconstruction (the other tests Neeleman & van de Koot (2007) enumerate are not replicable in the DP domain). Going back now to our constructed scenarios in (16), the inappropriateness of the example with noncanonical adjective order (15) in the Context 2, where cars of no particular size (or colour) were mentioned shows that the adjective does not reconstruct for scope. As argued above, lack of reconstruction is the property associated with A-scrambling and in turn with different base-generated orders. At first blush, this suggests that (15) (and also (13)) does not involve DPinternal adjective movement. Rather, such utterances have an atypical base-generated adjective sequence. In the noncanonical order, the order of the adjectives reflects their scopes. Unfortunately, however, this reasoning is faulty. The point is that in my proposal, which follows Neeleman et al’s (2009) proposal, the function of the noncanonical adjective placement is to allow for the domain of contrast to be marked syntactically as the modified NP big car. So, why should the adjective scopally reconstruct to a smaller domain of contrast (i.e. car)? If movement is to mark DoC explicitily, it is not going to reconstruct for DoC even if noncanonical adjective placement is indeed the result of A-bar movement. In order to test for reconstruction we need to look at examples that mark scope in different ways. One such example is (21). As Scott (2002 : 113) notes, the utterance may be uttered in a context where it has been established that Carol has twelve children, six of them horrible and another six, nice. He takes this to be indicative of focus movement inside the DP. In our terms, the noncanonical adjective placement would be marking the numeral-noun complex six children as the domain of contrast.

Chapter 6 Focus movement can be destressing, but it need not be (21)

199

Carol’s horrible six children made life miserable for her second husband. (Andrews 1983 : 697)

But this utterance is also perfectly felicitous in a context where Carol has six children altogether, all of them horrible. The noncanonical adjective placement is then licensed by the quantificational reason that ‘it is the group of six children rather than their cardinality that is horrible’ (Andrews 1983 : 697, cited in Scott 2002 : 113). Note that under this collective reading the adjective horrible need not even be stressed: (22)

Carol has six children. a. As a group, her horrible six children make life miserable for her second husband. b. #Her horrible six children each make life miserable for her second husband.

It is now possible to construct an example where two kinds of quantificational considerations can be combined: the domain of contrast and the collective reading. In (23), the context indicates that the domain of contrast for the adjective horrible is the NP six children. As (23a) and (23b) show the domain of contrast can be explicitly marked in syntax by the noncanonical placement of the adjective HORRIBLE. The relevant example is (23b), as it shows, that this time, the distributive reading is available, indicating that the adjective can reconstruct for scope under the numeral. (23)

Context: Carol has 12 children, 6 horrid, 6 kind. (DoC: six children) a. As a group, her HORRIBLE six children make life miserable for her second husband. collective reading: 9 b. Her HORRIBLE six children each make life miserable for her second husband. distribuitive reading: 9

Note that it is not entirely obvious that (23) is completely parallel to an example with adjective reordering as (15). It is possible that when an adjective appears in front of a numeral it moves from a lower position, while it is simply base-generated in a higher position when it scopes over another adjective. So, we need to consider such cases separately. Arguably, such a case can be constructed involving adjectives as in (24). (24a) is a pragmatically unnatural order, as its reading suggests that the chicken in question was first frozen and then sliced. In contrast (24b) sim-

200

Kriszta SzendrĘi

ply means sliced chicken with a temperature below zero. (24c) is ok with a contrastive reading, as indicated. Nevertheless, it need not mean the pragmatically odd reading that would require a chainsaw. Rather, it seems possible to contrast (24c) under the natural reading also examplified in (24b). (24)

a. #some sliced frozen chicken

reading: sliced > frozen

b. some frozen sliced chicken

reading: frozen > sliced

c. some SLICED frozen chicken (not, say, some MINCED frozen chicken) reading: sliced > frozen: 9 frozen > sliced: 9

A further case has been cited by Scott (2002 : 113). He noted that examples like (25a) and (25b) have distinct scopal orders. (25a) involves noncanonical adjective ordering. To the extent that this adjective placement not only marks the domain of contrast, but also determines the scope of alleged, there seems to be no reconstruction, as (24a) does not seem allow a reading where the person was not simply an alleged baron but an alleged English baron. But judgments here are very subtle. (25)

a. an ENGLISH alleged baron b. an alleged ENGLISH baron

A more transparent case was suggested to me by Ad Neeleman and Reiko Vermeulen (p.c.). In the context provided, the utterance in (26) is wellformed. Given this context, it must refer to British passports in the sense of documents that can prove British citizenship. This is because the context actually specifies that the documents in question were made in the DDR, so they could be referred to as German fake British passports. So, the adjective British can occupy a noncanonical position, marking the NP fake passports as its domain of contrast, while it reconstructs for scope under the adjective fake.5 5 In fact, one native speaker that I have consulted marginally accepted BRITISH German fake passports in the above context. But in a revised context where fake passports of various nationalities made in different countries of the former Eastern Block, i.e. Czech fake British passports, Czech fake French passports, … East German fake British passports, East German fake French passports, …, are available, she preferred the order GERMAN BRITISH fake passports to refer to fake British passports made in the DDR. This is in accordance with the proposal. In this revised context country of origin is no longer

Chapter 6 Focus movement can be destressing, but it need not be (26)

201

Context: We’re in a museum looking at a series of fake passports produced by the East German secret service in the 1970s. There are fake passports from many countries: fake French passports, fake Dutch passports, fake British passports, and so on. You turn to guide and say: I have a question about the BRITISH fake passports

To conclude this section, my claim is that adjective reordering is triggered by scope requirements: the need to mark the post adjectival constituent as the domain of contrast for the adjective. The operation, in line with other A-bar movement operations (see Bobaljik and Wurmbrand 2010 for a similar claim), allows reconstruction. Although it is possible to account for this by assuming a DP-internal designated functional FocusP inside DP, this would not adequately express the fact that the position of the fronted adjective does not seem to be predetermined by a specific focus position in the cartographic sense, i.e. in the sense that the position would always be preceded and followed by the same functional heads. Rather, as the interpretative effect of the reordering is not focus on the fronted adjective, but rather scope, marking the domain of contrast, the target position is variable depending on the size of the domain of contrast: it can involve other adjectives or the numeral.

3.

The Greek polydefinite construction: givenness marking

3.1

No A-bar movement, no domain of contrast marking

Having shown that adjective reordering in English involves DP-internal A-bar movement marking the domain of contrast for the contrastively focused adjective, I will now proceed to analyse the so-called polydefinite construction in Greek. I will argue against analyses that treat this as an instance of DP-internal focus movement (e.g. Ntelitheos 2004). Instead, I will argue that the function of the construction is givenness marking. In particular, I will propose that the syntax (and semantics), proposed by Lekakou and Szendroi (2012), coupled with Wagner’s (2007, 2012) account of givenness, captures the data. Thus, under the present proposal, the construction does not involve A-bar movement and consequently no domain of contrast marking in the sense of Neeleman et al (2009). part of the domain of contrast, but rather a point of contrast itself, as there are passports from many different countries in this context. So, placement of BRITISH above German is not justified.

202

Kriszta SzendrĘi

3.2

The syntax of the Greek polydefinite construction

The Greek polydefinite is a combination of at least one adjective and a noun where each features its own determiner, as in (27). (27)

a. to megalo to spiti the big the house b. to spiti to megalo the house the big ‘the big house’

Polydefinites co-exist in the language with monadics like (28), i.e. modification structures where only one determiner is present—although polydefinites have special syntactic, semantic and pragmatic properties (see Kolliakou (2004) and Campos and Stavrou (2004)). (28)

a. to megalo spiti the big house b. *to spiti megalo the house big ‘the big house’

Ntelitheos (2004) argued that the polydefinite construction involves DPinternal focus movement (see also Ioannidou and Den Dikken (2009) for a similar claim). Following Kolliakou (2004), I will try to show that a better interpretation of the data is that it relies on givenness, not focus. Focus only arises by the prosodic see-saw of Wagner (2012). Following Lekakou and SzendrĘi (2012), I will present a syntactic analysis of the construction where the different orders are due to base-generation, rather than movement. But let us review Ntelitheos’ proposal first. Ntelitheos (2004) proposes an analysis that treats discontinuous DPs, NP-ellipsis and polydefinites in a parallel structure. In Greek, the fronted part of a discontinous NP is focused: (29)

to kokkino idha to forema. the red saw-1S the dress ‘It is the red dress that I saw.’

In addition, the second part can easily undergo NP-ellipsis, even if the first part remains in situ.

Chapter 6 Focus movement can be destressing, but it need not be (30)

203

to KOKKINO idha. the red saw-1S ‘It is the red one that I saw.’

Moreover, the first part of the DP need not move, giving rise to what is called the polydefinite construction (Kolliakou 2004): (31)

idha to kokkino to forema. saw-1S the red the dress ‘It is the red dress that I saw.’

Ntelitheos (2004 : 10) proposes that all three rely on a common structure, which involves focus movement inside the DP, with the fronted part moving to a DP-internal FocusP, and the elided part moving into a DP-internal TopicP. This is the analysis of the polydefinite in (31). In the NP-ellipsis case in (30), the TopicP inside the DP undergoes deletion. In the discontinuous DP case in (29) the DP-internal FocusP undergoes further movement to the clausal FocusP. Although potentially far-reaching, this unified treatment of the data is undermined by the following problem. There are languages that allow discontinuous NP-topicalisation and NP-ellipsis, but where DP-internal focus fronting is not possible. This is unexpected in a theory where both these constructions rely on the availability of DP-internal focus fronting. Take Hungarian. In (32a) we see a case of discontinuous DP-topicalisation6; (32b) illustrates NP-ellipsis. (32c) and (32d) illustrate that focus movement inside the DP is impossible, with (32c) involving movement of the N over the A, and (32d) involving movement of the A over a possessor. (The brackets around the accusative markers indicate that the problem with these examples is not due to the presence of double accusative marking.) (32)

a. Bicikliket, a nagyokat vettem. bikes-acc the big-pl-acc bought-I ‘Bikes, I bought the big ones.’

6 In (32a) the adjectival part of the discontinous DP is in the focus position, while the nominal part is in a contrastive topic position. The parallel with the Greek data would be neater if the nominal part was in situ, but Hungarian does not allow B-accented contrastive topics to remain in situ and it seems that the nominal part of discontinuous DPs must be contrastive topics in this language.

204

Kriszta SzendrĘi b. A nagyokat vettem. the big-pl-acc bought-I ‘I bought the big ones.’ c. *[DP A biciklik(et) nagyok(at) tN] vettem meg. the bikes-acc big-pl-acc bought-I prt ‘I bought the big bikes.’ d. *[DP A nagy(okat) Péternek tA bicikliei(t)] vettem meg. the big-pl-acc Peter-dat bikes-poss3sg-pl-acc bought-I prt ‘I bought Peter’s big bikes.’

It is of course possible to analyse NP-ellipsis and discontinuous NP fronting in ways that do not rely on the availability of DP-internal focus fronting. But the Hungarian data suggests that the merits of Ntelitheos’ proposal must be evaluated only with respect to the polydefinite construction. So, the analysis boils down to the question whether or not this construction involves DP-internal focus fronting. If this turned out to be the correct analysis of the data, that would constitute an argument in favour of the cartographic approach. In contrast, if it turned out that polydefinites can be reduced to NP-ellipsis, which as the Hungarian data shows is needed independently, that would make DP-internal focus movement superfluous. In what follows, I will demonstrate that this line of thinking is feasible. Lekakou and SzendrĘi (2012) treat polydefinites as a case of close apposition, as in (33) from Greek and (34) from English: (33)

a. o aetos to puli the eagle the bird b. to puli o aetos the bird the eagle ‘the eagle the bird (not the symbol)’

(34)

a. Burns the b. the poet

poet Burns

Since in close apposition both nominal parts contribute to the determination of reference, Lekakou and SzendrĘi (2012) suggest that both DPs involved in the construction are referential DPs. This is taken to mean that both DPs have an R(eferential) role in the sense of Williams (1981, 1989), Higginbotham (1985), Zwarts (1992), Baker (2003). In Williams’ system, which is adopted by Lekakou and SzendrĘi, when a nominal occupies an argument position, its R-role is bound by a thematic role of the selecting

Chapter 6 Focus movement can be destressing, but it need not be

205

predicate, whereas when the nominal occurs as a predicate, it assigns the R-role to its subject. Lekakou and SzendrĘi (2012) propose that in close apposition an operation takes place which identifies the R-roles of two DPs. This operation, which can be thought of as complex argument formation, is schematically illustrated in (35): DP1,2 [R1=R2]

(35)

DP1[R1] DP2[R2]

Applied to close appositives, theta-identification amounts to identification of two R-roles. This creates a syntactically symmetric structure:7 DP1,2 [R1=R2]

(36)

DP1[R1] D o

DP2[R2] D

NP

aetos to

puli

NP

That close appositives involve a symmetrical syntactic structure is strongly suggested by agreement facts. An adjective in predicative position can agree in gender with either DP (provided of course that it can sensibly apply to either DP), as shown, in (37). (37)

a. o aetos to puli ine megaloprepos/ megaloprepo. the.m eagle(M) the.n bird(N) is majestic.M/ majestic.N b. to puli o aetos ine megaloprepos/ megaloprepo the.n bird(N) the.m eagle(M) is majestic.M/ majestic.N ‘The eagle the bird is majestic.’

As far as polydefinites are concerned, Lekakou and SzendrĘi’s (2012) proposal is that polydefinites are an instance of close apposition. They are 7 R-role identification solves a potential theta-theoretic problem that arises by assuming the availability of multiple R-roles within a CA: the occurrence of two (potentially) argumental DPs in the presence of a single theta-role assigner should violate the Theta Criterion. The actual mechanism is more complex than suggested in the text here. The reader is referred to Lekakou and SzendrĘi (2012) for details.

206

Kriszta SzendrĘi

only special in that they involve NP-ellipsis in one of their DP-subparts (cf. Panagiotidis (2005)): (38)

a. [DP [DP

to the

b. [DP [DP

to the ‘the big house’

spiti] [DP house

to the

megalo ‡]] big

megalo ‡ big

[DP

to the

spiti]] house

The symmetric structure proposed for polydefinites/close appositives is perfectly consistent with their ordering freedom: Since the proposed structure is multiheaded, i.e. the two DPs are sisters, they can appear in either order. This explains the word order freedom exhibited by the construction. The presence of the multiple determiners is also accounted for, as in this analysis each heads its own DP projection. See Lekakou and SzendrĘi (2012) for more details. Although there is some disagreement over whether the noun may bear stress at least in some orders (see Kariaeva 2004) most authors agree that the noun is usually destressed. The adjectives, in contrast, bear stress: (39)

a. [ [Weak to the

spiti] house

[Strong to megalo ‡]] the big

b. [ [Strong to megalo ‡ [Weak to spiti]] the big the house ‘the big house’

3.3

The discourse role of the Greek polydefinite construction: givenness marking on the noun

Kolliakou (2004), who studied the Greek polydefinite construction in detail, claims that the pragmatic difference between the monadic construction and its polydefinite counterpart is not focusing the adjective, but deaccenting the noun. If this turned out to be correct, that would undermine a DP-internal focus movement analysis in a fundamental way. She proposed the data in (40) as characteristic of the discourse context in which polydefinites may occur. As (40d) shows, the polydefinite is licensed if the noun pena ‘pen’ is accessibly given information. Deaccenting the noun in the monadic construction, as in (40d’), is also possible in this context.

Chapter 6 Focus movement can be destressing, but it need not be (40)

a. Zoe:

207

Ti pires tu Yanni gia ta christugena ? ‘What did you get Yiannis for Christmas? ’

b. Daphne: (Tu pira) tin asimenia PENA. (I got him) the silver pen. b’. Daphne: #(Tu pira) tin ASIMENIA pena. #(I got him) the silver pen. b’’. Daphne: #(Tu pira) tin pena tin asimenia. I bought the pen the silver #(I got him) the silver the pen c. Zoe:

Ti pires tis Marias? ‘What did you get Maria? ’

d. Daphne: (Tis pira) tin pena ti CHRISI. (I bought her) the golden the pen ‘I got her the golden pen.’ d’. Daphne: (Tis pira) ti CHRISI pena. (I got her) the golden pen. d’’ Daphne: #(Tis pira) ti chrisi PENA. #‘(I got her) the golden pen.

Note, however, that the above data can be interpreted in another way. One could argue that what licenses stress shift in (40d’) and the polydefinite in (40d) is not the givenness of the noun, but rather the contrast on the adjective chrisi ‘golden’ with the previously mentioned adjective asimenia ‘silver’. In this case, it would be focus, rather than givenness that licences the marked constructions. In Wagner’s theory, the two analyses are actually indistinguishable: deaccenting the noun is only possible if the adjective can bear focus. Lekakou and SzendrĘi (2012) argued that the adjectival part of a polydefinite is in fact a full DP involving NP ellipsis. It is well-known that there is a disanaphora requirement on the remnant of ellipsis. So, there are two possibilities: the polydefinite is licensed by focus on the elliptical DP; or the pragmatic function of the construction is destressing the DP containing the overt noun. In this latter case, the focal stress on the adjective is the direct consequence of Wagner’s see-saw: the sister of a destressed constituent bears stress, and thus, focus. Being an elliptical DP, it will always be focusable. The following set of data attempts to show that even though the adjectival part of a polydefinite is focal, the polydefinite is not licensed unless the nominal part is given.

208

Kriszta SzendrĘi

The relevant set of data can be constructed using unexpected contrastive stress in contexts where there is no corresponding given constituent, such as Rooth’s (1992) example in (41). (41)

An AMERICAN farmer was talking to a CANADIAN farmer. (Rooth 1992)

Féry and Samek-Lodovici (2006) argue that even if givenness can account for the deaccenting of the noun and the stressing of the adjective in the DP a Canadian farmer, it is impossible that stress on the adjective in the first DP, an American farmer, can be justified by the givenness of the noun. Rather, it must be the contrastive focus on the adjective that makes stress shift necessary. Although, Wagner (2012) argues against this conclusion and claims that examples like (41) involve cataphora. In other words, the first instance of the noun farmer is destressed in anticipation of the identical second occurrence of the same noun later. But the Greek data presented below argues against this conclusion. In (42), which is based on Rooth’s (1992) example, the contrast between the adjectives does not license the polydefinite. The fact that such cases can only be expressed by stress shift within the nominal in a monadic construction and not by the polydefinite, supports the idea that the polydefinite is licensed by givenness of the nominal part, rather than by focussing the adjectival part. (42)

Anigo tin tileorasi ke ti vlepo? switch.on the television and what see.1sg? Ton AMERIKANO (*ton) proedro na sinomili me ton IRANO the american the president subj talk.with with the iranian (*/??ton) proedro. the president ‘I switch on the telly and what do I see? The AMERICAN president is talking to the IRANIAN president.’

A similar datum can be constructed based on Krifka’s (2006b) example. In (43), the noun aftokinito ‘car’ is not given, and consequently the polydefinite is not allowed. The contrast between the adjectives (kokkino ‘red’ vs. ble ‘blue’) is not in itself enough to license the construction. (43)

O Janis ithele ena metaforiko meso ja tis diakopes tu. the janis wanted a means-of-transport for the holidays his. Pije sto garage tu patera tu. went to.the father’s garage

Chapter 6 Focus movement can be destressing, but it need not be

209

Epidi to KOKINO (*to) aftokinito ihe idi pulithi pire as the red the car was already taken took to BLE ((to) aftokinito). the blue the car ‘Janis wanted a vehicle for his holidays. He went to his father’s garage. As the RED car was already taken, he took the BLUE car.’

We can conclude that although there is a clear effect of focus on the adjectival part of the polydefinite, this is due to the fact that the construction involves NP-ellipsis, rather than DP-internal focus fronting. In carefully constructed examples where the nominal part is not given even though the adjectival part is contrastively focused, the polydefinite cannot be used. So, as Kolliakou (2004) argued, the pragmatic import of the polydefinite seems to be the deaccentuation and therefore givenness marking of the nominal part, rather than the focussing of the adjectival part. A further argument in favour of a givenness approach as opposed to an approach based on focusing I would like to mention that even though data about different orders and especially about the relative pragmatic import of the different orders is notoriously subtle it seems to be the case that most authors agree that the noun is destressed in most orders (although see Kariaeva 2004). This can be nicely accommodated in the syntactic system proposed by Lekakou and SzendrĘi (2012) as they argue that the subparts of the polydefinite are appositive DPs. Based on a symmetric structure, it is understandable that the Strong and Weak labels can be assigned in any order, given that Weak-Strong ordering often reflects head-complement status (see Wagner 2007, 2012).

3.4.

No domain of contrast marking in the polydefinite

The proposal here is that Neeleman et al’s (2009) A-bar movement marking the domain of contrast can also apply in the DP domain. However, contrary to many analyses (e.g. Ntelitheos 2004), I do not think the Greek polydefinite construction involves DP-internal focus movement. Rather, its function is to mark its nominal component as given. If this is the case, then we do not expect that the construction marks the domain of contrast of the focal adjective. This seems to be on the right track. Kyriakaki (2010 : 5–6) argues that the adjectival part bears focus in the DADN order but that this can be new information focus and need not be contrastive.

210

Kriszta SzendrĘi

(44)

Xriazome [TO

KOKKINO to

Need-1SG the-NEU red-NEU ‘I need the RED dress.’

forema]. (Kyriakaki 2010 : 6 ex 13) the-NEU dress-NEU

In contrast, the a contrastive focus reading is the only one available if the polydefinite undergoes split with the adjectival DA-part fronted to the clausal focus position:8 (45)

TO KOKKINO xriazome (?to) forema, oxi the-NEU red-NEU Need-1SG the-NEU dress-NEU not to ble the-NEU blue ‘It’s the red one I need from the dresses, not the blue one.’ (Marika Lekakou p.c.)

This is in line with the analysis proposed here: focus movement is associated with domain of contrast marking (Neeleman et al 2009). Without the clausal split, the polydefinite simply marks the noun as given, and by the see-saw the adjectival part as focused. But contrast is not necessarily available and the domain of contrast is not explicitly marked syntactically unless the polydefinite undergoes DP-split with the adjectival DA-part fronted to the clausal focus position. The case of the Greek polydefinite construction highlights the need to distinguish form and function. Because of the way the prosodic operation of Strong-Weak assignment operates on sisters (i.e. the see-saw) the adjectival part appears to be stressed and focused, and the nominal part destressed and thus marked as given. But it is not necessarily the case that the function of the construction is to mark the nominal part as given and to mark the adjectival part as focused. These may diverge: as I have shown by the examples in (42)-(43) above, the function of the Greek polydefinite is to mark the nominal part as given. It is simply a by-product of the syntax of the construction and the see-saw nature of stress assignment rule that the adjective is focused. In fact, focussing and givenness marking need to be distinguished more generally. This is the subject of the next section. 8 The question mark in (45) indicates the judgment of Marika Lekakou, who finds data where both the split and the remnant DP have a definite article marginal in general. To the extent that the example is allowed, it has the contrastive reading indicated in the text.

Chapter 6 Focus movement can be destressing, but it need not be

4.

In defense of two systems: focussing does not equal givenness marking

4.1

No Occam’s razor when you need two systems

211

Wagner (2007, 2012) following Schwarzschild (1998) developed a theory of marked contrastive focus as givenness marking of the sister of the focal constituent. His theory is based on the idea that the see-saw-like property of stress assignment suggests that destressing and stress strengthening is one and the same operation, and consequently focussing a constituent is the same thing as destressing its sister, and vice versa. We saw above that this is indeed true in many cases. His theory involves no duplication (an improvement on Schwarzschild’s original proposal): there is only one way to ensure that a constituent is focused, i.e. by applying destressing to its sister. In contrast, in Neeleman et al’s (2009) system marked contrastive focus does not necessarily equal givenness marking of the sister of the focal constituent. Rather, following Neeleman and Reinhart’s (1998) original proposal, they take destressing and stress strengthening to be separate prosodic operations. I will now argue that this duplication is empirically necessary. This has the consequence that no Occam’s razor-type arguments apply (see also Neeleman and Reinhart 1998, Neeleman and SzendrĘi 2004, Féry and Samek-Lodovici 2006, Krifka 2006b and Reinhart 2006). No matter how much simpler and therefore a priori more appealing a system without duplications is, if one can show that both operations are empirically necessary, then arguments of elegance and simplicity do not apply. I concede that Wagner (2012) was right about destressing triggering a see-saw. However, we still need stress strengthening as a separate operation. In contrast to destressing, stress strengthening does not trigger destressing on its sister by the see-saw.

4.2

Marked focus can be destressing … but it need not be

Wagner’s (2012) proposal to equate focussing and givenness marking rests on the see-saw nature of stress assignment: if a Strong label is changed to Weak in order to account for its anaphoric nature, the Weak label on its sister must automatically be changed to Strong. And conversely, a Weak label can be changed to Strong in order to account for its contrastively focused nature, but then its sister must be marked Weak, and therefore

212

Kriszta SzendrĘi

given. I would like to raise the question whether these two cases are in fact truly parallel, and equally motivated. Let us take the case of destressing first. Consider the state of affairs schematically illustrated in (51a). Assume that the Node 1 undergoes anaphoric destressing because the constituent corresponding to that node is given. This node is then assigned a Weak label. This is illustrated in the second tree diagram in (51b). As (51b) shows, this leads to an impossible state of affairs. Two Weak nodes cannot be dominated by a Strong node. In theory, two possible escape routes are open. First, as shown in the third diagram in (51c), one may change the Strong label on Node 3 to Weak as well. But this would mean that the constituent corresponding to this node, Node 3, is anaphorically given, which may or may not be the case, depending on the status of the material under Node 2. But even if material under Node 2 was in fact given, then Node 3 should have been the one that was marked as given instead of Node 1 in the first place. So, this route is in fact only a theoretical possibility. The only way out of this, then is to apply the see-saw to Nodes 1 and 2, and assign a Strong label to Node 2. So, Wagner’s see-saw is motivated in case anaphoric destressing targets a particular Strong-labelled node. (46)

a.

b.

… 4W

3S 2W

c.

*… 4W

1S

… 4W

3S 2W

2W

1W d.

3W 1W

… 4W

3W 2S

1W

The situation does not seem to me to be quite the same with stress strengthening. Assigning a Strong label to an otherwise Weak node would create a Strong-Strong sister pair, but that is not necessarily a prosodic violation. Why should two accents be disallowed? Even if the configuration of the tree is such that the two accents end up on adjacent words, in violation of the Obligatory Contour Principle, this can be remedied on the surface, by inserting a short pause, or by downstepping one of the accents. The upshot is that unlike destressing, which seems to have automatic consequences for its sister, stress strengthening is not necessarily understood

Chapter 6 Focus movement can be destressing, but it need not be

213

on a node in relation to its sister. It can be, and then the see-saw can take care of that. But it need not be.

4.3

When it IS: givenness marking triggers the see-saw

Consonant to Wagner’s (2012) proposal, anaphoric destressing triggers the see-saw. In fact, it is potentially possible to see this example as a case of applying contrastive focus to the verb. Crucially, the domain of contrast, which is the open proposition {My neighbour did x to a desk} is not marked explicitly in the syntax. So, Neeleman et al’s (2009) proposal does not apply. Instead, Wagner’s see-saw applies, ensuring that a desk is destressed and thus marked as given, while the verb is contrasted. (47)

A: Has your neighbour bought a desk already? (Neeleman and Reinhart 1998) a. B: #My neighbour is building a DESK. b. B: My neighbour is BUILDING a desk.

In fact Wagner (2012) is correct in arguing that destressing must be allowed to apply to constituents larger than DP.9 In the following example, arguably the whole extended VP is building a desk is anaphorically given and therefore destressed. By the see-saw, this has the effect of contrasting the subject, MY NEIGHBOUR. (48)

4.4

A: Last week, there was a lot of noise because Bill was building a desk. What’s the noise today? a. # B: [F My neighbour is building a DESK]. b. B: [F MY NEIGHBOUR is building a desk]. (Wagner 2012: ex 96)

When it is NOT

4.4.1 Adjective reordering as givenness marking? Recall from Section 1 that Wagner (2007, 2012), following Schwarzschild (1998), argued that marking a noun given by destressing automatically makes its sister focused. So, previous mention of a red convertible allows

9 The see-saw, at least in English, can also apply to domains smaller than the DP: see Wagner’s examples included here as (9)–(10) above.

214

Kriszta SzendrĘi

subsequent occurrence of the NP a BLUE convertible with noncanonical stress placement. Wagner’s point was that convertible is given in the discourse, so can be marked as Weak, so long as its sister, BLUE, can be understood as felicitously contrasting with the previous modifier (in this case red). So, the adjective BLUE bears Strong by what I called the seesaw rule of prosodic stress assignment. In this section I would like to show that his theory cannot explain the adjective reordering data discussed above. Recall also our example from Section 2, repeated here for convenience: (15)

I drive a BLACK big car.

(16)

a. Context 1:

Scenario 1: b. Context 2:

Scenario 2:

Context 1: 9

Context 2: 

In this car park you can see my friends’ cars and my car. There is a bunch of big cars here. There are many different colours. BLACK RED BLUE WHITE BLUE YELLOW In this car park you can see my friends’ cars and my car. There are cars of many different colours. Some of them are small, but there are big cars of every colour. black BLACK black black red red red RED red red

…

Under Wagner’s (2012) proposal: (15) is predicted to be appropriate if the NP big car was given in the previous discourse and the adjective black contrasts felicitously, i.e. if big cars of various colours are available in the context. This is in fact true in both contexts. The difference between the contexts lies in whether BLACK applies to a set of cars or a set of big cars; in the domain of contrast. So, Wagner’s theory fails to capture the fact that (15) is only felicitous in Context 1, but not in Context 2. It is too permissive. So, it would have to be supplemented by something along the lines of the present proposal, namely, with the idea that the domain of contrast is not determined by the prosodic see-saw, but by A-bar movement.

4.4.2 Postfocal but not given In other cases, Wagner’s theory is too restrictive. His see-saw operation makes the sister of every Weak-marked constituent Strong and conversely, the sister of any constituent receiving a marked Strong label, will automatically receive Weak, by the see-saw. This predicts that the sister of a

Chapter 6 Focus movement can be destressing, but it need not be

215

moved and focused constituent, which bears Strong by assumption, will be marked as Weak, and will therefore be interpreted as anaphorically given. So, he predicts that the material on the right of the moved adjective should be destressed. But, as the following examples (from Ad Neeleman and Reiko Vermeulen p. c.) illustrate, this is not necessarily the case. (46) allows adjective fronting inside the DP even though the noun BUS is not anaphorically destressed. (47) shows that the movement, as in (15), is optional. (46) I’d really like a big car. But a RED big BUS would be fine too. (47) I’d really like a big car. But a big RED BUS would be fine too.

But, Neeleman et al’s (2009) proposal gives the right prediction. In (46), the domain of contrast for RED is the set of big buses. This is the scope of the moved element. In terms of accenting, the DP bears two accents, arguably both contrastive. So, Wagner’s (2012) predictions do not hold. This time, it undergenerates. There are, in fact, many cases where the postfocal domain is not given or fully destressed. Take for instance Neeleman and Reinhart (1998)’s example in (48), which involves focusing on the noun milieu-fanaat ‘environmental fanatic’ in the scope of zelfs ‘even’ even though the NP een auto ‘a car’ is not destressed. To obtain marked stress on the former, while retaining the original accent on the latter, an independent stress strengthening operation is necessary. The see-saw rule would not work. (48)

Zelfs die milieu-fanaat heeft nu een auto gekocht even that environment-fanatic has now bought a car (Neeleman and Reinhart 1998, ex. 61)

There are also such cases in English. A few random examples from the British National Corpus (my caps) illustrate the same point. In (49) two is stressed and focused, while eyes need not be given or destressed. Further, the contrast on above and below licences focal stress while water surface need not be given or anaphorically destressed. In (50) stress and focus is required on nomadic and tribal but the discussion could be about pigments in which case rugs need not be previously mentioned or anaphorically destressed. (49)

It only has TWO eyes but they are divided for viewing ABOVE and BELOW the water surface.

216 (50)

Kriszta SzendrĘi Strong or dark shades are normally only found in NOMADIC and TRIBAL rugs.

So, cases involving postfocal but not given material seem to be systematically available and cannot be argued to be sporadic.

4.4.3 Icelandic Wagner’s unified proposal predicts that if a language or construction lacks anaphoric destressing it should disallow stress strengthening too.10 This is because these are two sides of one and the same prosodic operation, a Strong-Weak swap. Italian is a case in point: generally, it allows neither anaphoric destressing nor stress strengthening. But as Dehé (2009 : 20–21 ex. 18) reports, Icelandic has obligatory stress strengthening but only optional anaphoric destressing. To illustrate, she elicited utterances with various focal patterns determined by a preceding question. She found that the focal element always bore a pitch accent; so Icelandic seems to be similar to English when it comes to focussing. (52)

a. (Q: Hverjum gaf Stéfan appelsínuna? ‘To whom did Stéfan give the orange?’) A: Stéfan gaf [Elínu]Foc appelsínuna. Stéfan gave Elín orange.DEF ‘Stéfan gave the orange to Elín.’ b. (Q: Hverjum sendi Björg bæklinginn? ‘To whom did Björg send the booklet?’) A: Björg sendi [Elínu]Foc bæklinginn. Björg sent Elín booklet.DEF ‘Björg sent the booklet to Elín.’ c. (Q: Hverjum gaf Hildur eplið? ‘To whom did Hildur give the apple?’) A: Hildur gave [Ástu]Foc eplið. Hildur gave Ásta apple.DEF ‘Hildur gave the apple to Ásta.’

10 In fact, Wagner also predicts that there should not be a language (or a construction) where destressing is possible but stress strengthening is not. As far as I can tell, this seems correct: I know of no language that has anaphoric destressing but no stress shift for focus.

Chapter 6 Focus movement can be destressing, but it need not be

217

Pitch (Hz)

But as she reports, although all speakers accented the focal indirect object in all cases, ‘of the five speakers, three speakers failed to deaccent the direct object (DO) appelsinuna in (52a), two speakers failed to deaccent the DO baeklinginn in (52b), and two speakers failed to deaccent the DO eplið in (52c). In this data set, all post-focus pitch accents were of the same type as the focus-marking pitch accent and were downstepped with respect to the focus-marking pitch accent.’ (Dehé 2010 : 20–21). An instance of an utterance involving the material in (52c) with no deaccenting is given in Figure 1. (Dehé refers to Nolan & Jónsdóttir (2001) for corroborating the findings about the lack of deaccenting.)

300 250 200 150 100 Hil

dur

gaf

0

H*L Ás Time(s)

tu

!H* L e

L% plið 1.2761

Figure 1: Accenting of contextually given information: given information eplið (see (52c)) fails to be deaccented by a (female) speaker. (from Dehé 2010 : 21)

So, Icelandic seems to go agianst the purported generalisation of Wagner (2012) that languages that apply marked prosodic operations for focussing also apply deaccenting. In Icelandic, stress shift for focussing seems to be obligatory, while deaccenting given material seems to be optional.

4.4.4. Language acquisition Arguably, a similar pattern emerges in language acquisition. It has been found in different experiments that typically developing children always stress focal constituents, but sometimes fail to destress given ones (see e.g. Baltaxe 1984, Chen 2010). Children with certain pathologies such as autism seem to show the effect in a more marked way (Baltaxe 1984). This is in line with the idea that destressing and stress strengthening are two

218

Kriszta SzendrĘi

different operations. Children seem to use stress strengthening correctly earlier than they reliably use anaphoric destressing.11 I would like to conclude then, that in the light of this evidence it is perhaps not so problematic to have in our grammar both a stress strengthening and a destressing operation. It would not be a straight forward duplication, but rather, two operations with overlapping and sometimes not identical effects.

4.5

Superman rules

Wagner (2012 : 137–145) discusses extensively why Reinhartian approaches to stress shift and deaccenting are wrong. In particular, he discusses and re-analyses in his local alternatives approach so-called ‘Superman’ sentences put forward by Neeleman and SzendrĘi (2004) in defense of stress strengthening. I would like to reproduce his arguments against Neeleman and SzendrĘi’s (2004) position and counter them. I will try to show that contrary to his claim, Wagner’s local alternatives approach cannot handle Superman sentences, so stress strengthening is still necessary to account for them. As I already discussed in section 4.3 Wagner (2012) argues that the contrast in (53) vs. (54) suggests that (53) involves anaphoric destressing, based on the presence of the VP was building a desk in the preceding context. This is because contrastive stress on the subject DP MY NEIGHBOUR is only felicitous if the VP is building a desk is explicitly and accessibly given in the previous context. (53)

A: Last week, there was a lot of noise because Bill was building a desk. What’s the noise today? a. # B: [F My neighbour is building a DESK]. b. B: [F MY NEIGHBOUR is building a desk]. (Wagner 2012: ex 96)

(54)

A: Last week, Bill was visiting. What’s going on this week? B: # [F MY NEIGHBOUR is building a desk]. (Wagner 2012: ex 99)

11 It is of course a possibility that there is an independent reason why children (especially autistic children) would have a problem applying destressing even though the same grammatical operation derives focussing and deaccenting. For instance, it could be that the pragmatic import of deaccenting, i.e. anaphoric givenness, is something they apply reliable only at an older age. I will have to leave this possibility open here.

Chapter 6 Focus movement can be destressing, but it need not be

219

Indeed, this contrast convincingly shows that anaphoric destressing must be allowed to apply to constituents larger than lexical items. Wagner (2012) offers a reanalysis of Neeleman and SzendrĘi’s (2004) Superman sentences in terms of anaphoric destressing. This is necessary in his approach, which does not make a difference between stress strengthening and destressing. Let us retell the original argument here. Neeleman and SzendrĘi (2004) presented an utterance with complex, nested focus. Arguably, the context provided ensures that the utterance in (55), with prosodic prominence on the direct object Superman, involves contrastive focus on the direct object, contrasting it with ‘decent books’, the VP reading Superman to some kid, contrasting it with ‘doing his homework’. Furthermore, the whole utterance answers the context question ‘What happened?’, so it also involves focus on the whole sentence. (55)

Suppose father comes home from work and finds mother in obvious distress. Then the following discourse may take place: Father: What happened? Mother: You know how I think our children should read decent books. Well, when I came home, rather than doing his homework, [IP Johnny was [VP reading [DP SUPERMAN] to some kid]].

Wagner (2012 : 142) claims that in his local alternatives approach he ‘could analyze this example as having the constituent ‘read x to some kid’ being marked as given relative to the ‘Superman.’’ As he acknowledges in a footnote, this would also require some nontrivial syntactic movements to take place as Superman is not actually the sister of ‘read x to some kid’. But this is a technical problem. More importantly, the constituent ‘read x to some kid’ is not given in the context. Neeleman and SzendrĘi (2004) designed the context with the specific aim in mind that it would not be. So, Wagner has to assume that it is accommodatable. He says that this is plausible based on (56), where, he claims ‘x made the best seller list’ is also accommodated. (56)

Context borrows the negative view of Superman comics from the original example: Father: What happened? Mother: I thought good fiction is valued in the country. Well, I just read that A SUPERMAN comic made the best seller list. (Wagner 2012: ex 100)

220

Kriszta SzendrĘi

But what makes ‘x made the best seller list’ accessible when the context mentions that good fiction is valued in the country? Or what makes ‘read x to some kid’ accessible in a context where Johnny was expected to do his homework? Importantly, Wagner has to answer these questions with a proposal about accommodation that is not too powerful: he has to make sure that whatever forces are at work in (55) and (56), they are not applicable to (54). In other words, he has to make sure that ‘is building a desk’ does not become accessible in the context given in (54) mentioning Bill’s previous visit. Without such a theory of accommodation, the contrast between (55) and (56) on the one hand and (54) on the other is not actually explained, merely noted. In my view, we may explain this contrast in a theory with two operations: (54) is inappropriate because anaphoric destressing cannot apply unless the destressed constituent is given, as Wagner himself proposed. (56) is indeed parallel to (55). It involves stress strengthening and complex (nested) focus; it is a ‘Superman’ sentence. Thus, no accommodation needs to be assumed. In particular, in (56) the subject a Superman comic contrasts with good fiction and the entire sentence A Superman comic made the best seller list contrasts with the previous utterance Good fiction is valued in this country. The question that we now face is why an example like (54) cannot involve stress strengthening and complex focus. In other words, how do we turn (54) into a Superman sentence? To see this let us first dissect the original Superman sentence in (55). As I have already claimed, it is a crucial part of the example that the indirect object, following the direct object, is not given. It is also important, however, that the indirect object is not itself contrasted. If it was, there would be two accents: one on the direct object and one on the indirect object. Witness (57), from Neeleman and SzendrĘi (2004), where the context invited contrast on both the direct object and the indirect object, in addition to the contrast on the VP (and the additional focus on the IP). (57)

You know how I want our children to read decent books and how I think it is important that Johnny plays with kids his own age. Well, when I came home, rather than doing his homework, [IP Johnny was [VP reading [DPSUPERMAN] to [DP some SIXTEEN-YEAR OLD]]]. (Neeleman and SzendrĘi 2004 : 153, ex 12)

This is why an indefinite like to some kid is ideal as an indirect object in a Superman sentence. Thus, in order to turn (54) into a Superman sentence, we need to make sure that there is a contrast on the subject DP my neigh-

Chapter 6 Focus movement can be destressing, but it need not be

221

bour, and on the entire sentence, but, crucially, not on the VP following the subject. The VP also cannot be given. Wagner’s example (56) achieves precisely that. In (56), the subject a Superman comic contrasts with good fiction and the entire sentence A Superman comic made the best seller list contrasts with the utterance Good fiction is valued in this country. At the same time, the two VPs, made the best seller list and is valued in this country do not contrast. This is why a unique accent on the subject is so easily available. In order to turn (54) into a Superman sentence, we need to do the same thing. However, this means that we would have to find a context where my neighbour is contrasted with some other DP, the utterance My neighbour is building a desk is contrasted with some previous utterance, while the VP is building a desk is neither contrasted nor given. The examples I created, in (58), involves some changes compared to (54) because I found it hard to include a VP is building a desk without it being either given or contrastive. In (58) there is a contrast on the subject my neighbour with other people who are not so close to the speaker. There is also a contrast between My neighbour is buying a car and the context sentence ‘I thought that people in this ecovillage were environmentally conscious.’ The latter contrast expresses the speaker’s dismay at his situation. This example is not as neat as Wagner’s (56) in that here the VP ‘is buying a car’ could be contrasted with ‘were environmentally conscious’ but apparently it is not necessary to do that, so main stress on the subject is permissible. Note that there is a secondary stress on the object ‘a car’, but that is simply due to the fact that it is a nonspecific indefinite. The focus “projects” from the accent on the subject. (58)

A: I thought people in this ecovillage were environmentally conscious. But things have been getting worse and worse … Now [F MY NEIGHBOUR is buying a car]!

Another example showing the same phenomenon is in (59). Here the expectation is built up in the context that the place is not as socialist as it used to be. Again, a contrast is built up between my neighbour and other people living further away. What upsets the speaker is that his neighbour is becoming individualistic. Here, perhaps it is less obvious how the VP ‘is building a fence’ would directly contrast with any of the predicates mentioned in the context. Again, the focus responsible for contrasting the whole proposition “projects” from the main prominence on the subject. The secondary stress on the object is just to mark the DP ‘a car’ as nongiven.

222 (59)

5.

Kriszta SzendrĘi A: People in this kibbutz used to live in a real socialist community: everything we owned would be shared by everyone. Over the last couple of years, some more materially conscious people moved in. I didn’t like that, but I also didn’t care that much. But now [F MY NEXT DOOR NEIGHBOUR is building a fence]!

Focussing and givenness: a complete typology at the clause-level and inside the DP

Let us take stock. I reviewed both some syntactic and some prosodic options avilable in the grammar to ensure that a particular element is focused or marked as given. In this paper, I concentrated on how these operations apply inside the DP. In particular, I showed that adjective reordering involves A-bar movement marking the domain of contrast in the sense of Neeleman et al (2009). In contrast, the Greek polydefinite instantiates anaphoric destressing. It does not involve A-bar movement, and consequently does not serve to mark the domain of contrast. Destressing triggers Wagner’s (2012) see-saw and may apply at various structural levels. A separate prosodic operation, stress strengthening is necessary to account for instances of multiple focus and postfocal nongiven material. I showed that this operation is also necessary to account for Neeleman and SzendrĘi’s (2004) ‘Superman’ sentences and for those Icelandic speakers that allow focus strengthening in the absence of anaphoric deaccenting. This operation does not obligatorily trigger the see-saw effect. Stress strengthening may also apply in the nominal domain, as illustrated above by sentences such as the English example I’d really like a big car. But a RED big BUS would be fine too. Here, two accents occur inside the subject DP: presumably, the accent on BUS would be assigned by the nuclear stress rule and the stress on RED by stress strengthening. Having seen that givenness can be marked by deaccenting, we may ask whether there is a syntactic option available in language whose function is to mark anaphoric givenness. In the present proposal, which follows Neeleman et al (2009) in assuming that A-bar movement associated with with focusing serves to mark the domain of contrast, we do not expect that givenness marking would be associated with A-bar movement. If givenness marking does not have the same quantificational, scope-taking properties as A-bar movement associated with focussing does, then no movement marking the domain of contrast should be associated with it. At same time, arguably, A-scrambling in languages like Dutch and German seems to mark givenness. As Neeleman and Reinhart (1998) and Neele-

Chapter 6 Focus movement can be destressing, but it need not be

223

man and Van de Koot (2007) and many others have argued, the import of A-scrambling is that it allows a DP to be merged in a position where it would not receive main stress, and consequently focal interpretation, by the nuclear stress rule. So, apart from anaphoric destressing, a further way to mark a DP as given is by A-scrambling.12 Given the logic of Wagner’s proposal, in contrast with prosodic destressing, A-scrambling is not expected to have an automatic stress strengthening effect. Rather, main stress assigned by the nuclear stress rule will simply fall on another constituent. No marked prosodic operation (i.e. Weak-Strong swap) takes place, so there should be no requirement that the constituent that bears main stress be interpreted contrastively with respect to a previous mention. This seems to be true. (60) involves Dutch data from Neeleman and Reinhart (1998: 343, ex 79, 81) involving object scrambling across an adverb. Here, the nuclear stress rule assigns main stress to the verb, but as the focal possibilities indicate, this need not be contrastive focus on the verb itself. (60)

a. dat Jan het boek gisteren gelézen that John the book yesterday read Scrambled structure

heeft has

b. Focus set: {IP, VP, V} c. Object: Destressed

Similarly, in German, a direct object may scramble across a dative object as in (62), with main stress now falling on the indirect object. Scrambling destresses the direct object (cf. (61) with (62)), but it does not automatically focus the indirect object. Rather, the focus can project from the neutral stress position to the VP or the whole IP. (German data from B. Gehrke p. c.) (61)

a. Kai sagte, dass Verena ihrer Mutter die BROMBEERE Kai said that Verena her mother.DAT the blackberry. ACC geben würde give would ‘Kai said that Verena would give the blackberries to her mother.’ b. Focus set: {IP, VP, DPacc}

12 Another candidate for syntactic givenness marking is left- and especially right-dislocation in Romance languages (SzendrĘi (2002 : 297–300), DelaisRoussarie et al. (2004 : 528), Astruc (2005), Feldhausen 2010: ch.5). I leave this issue open here.

224 (62)

Kriszta SzendrĘi a. Kai sagte, dass Verena die Brombeere ihrer Kai said that Verena the blackberry.ACC her MUTTER geben würde mother.DAT give would ‘Kai said that Verena would give the blackberries to her mother.’ b. Focus set: {IP, VP, DPdat} c. Direct Object: Destressed

Interestingly, as (63) shows, if the scrambled direct object is a nonspecific indefinite, and thus the trigger for scrambling may not be its need to be destressed, the indirect object must bear narrow focus. This would be a case of focus-induced scrambling in the sense that the function of the scrambling is not destressing the scrambled object, but focusing the constituent that scrambling crosses.13 (63)

a. dass Verena eine that Verena a

Brombeere ihrer MUTTER gegeben hat blackberry.acc her mother.dat given has

b. Focus set: {DPdat}

Although they did not discuss such cases, Neeleman and Reinhart’s (1998) economy principles could perhaps be evoked to account for the obligatory narrow reading: since scrambling does not serve the function of deaccenting the direct object it must have some other interpretative consequence, such as narrow focus on the indirect object. It is not clear how the contrast between (62) and (63) could be explained in a theory, such as Wagner’s, which does not distinguish focussing from givenness marking. The final issue to be considered then is whether there are also cases of A-scrambling DP-internally. In other words, whether noncanonical orders in the DP are sometimes licensed by the need to mark a subpart of the DP as given. Although I cannot offer a full account of these cases, I would like to suggest that an example by (Simik and Wierzba 2012) instantiates this option in Czech. In (64), the noun can optionally undergo DP-internal reordering to the left in order to avoid stress.

13 Note that (63) is of course also grammatical with contrastive focus on the indefinite accusative object, but in that case it would constitute a case of A-bar scrambling and thus focussing, and marking of the domain of contrast for the accusative object.

Chapter 6 Focus movement can be destressing, but it need not be (64)

What kind of fish do you like? a. Jím pĜedevším [DP MOěSKÉ eat:1sg mainly sea b. Jím eat:1sg

pĜedevším [DP mainly

ryby fish

225

ryby] fish

MOěSKÉ] sea

To conclude, I argued that we need to distinguish focus marking from givenness marking, challenging Wagner’s (2012) unified proposal. Thus we arrive at a four-way typology: the syntactic ones are A-bar movement for focussing and A-scrambling for givenness marking and the prosodic ones are stress strengthening and destressing. It seems that all of these options can be instantiated at the clausal level as well as inside the noun phrase.

Chapter 7 Types of Focus and their Interaction with Negation* Ad Neeleman and Reiko Vermeulen 1.

Introduction

We proposed in Chapter 1 that contrast contributes an additional, negative component of meaning based on the focus semantics. This approach is very much in line with the standard analyses of focus-sensitive particles such as only, also and even. This chapter deals with data involving negation which seem to pose a challenge to this proposal. In particular, negative sentences with some types of focus-sensitive particles yield a meaning that is unexpected and seems to falsify the general approach that the component of meaning introduced by the focus-sensitive particle is computed based on the focus semantics of the sentence. We argue that the data do not in fact require adjustment of the standard semantic theory of focussensitive particles, but can be explained if elements contributing an additional, negative component of meaning are treated as positive polarity items. As a result, the semantics of contrast, as well as the idea that contrastive focus should be decomposed into [contrast] and [focus], proposed in Chapter 1, can remain intact. We can illustrate the issue with examples containing negation and DPs prefixed by only and even. When these appear in the c-command domain of negation, it appears that their scope is different, in that only cannot take scope over negation, while even must, as indicated in (1) and (2). (1)

John didn’t invite only PIA. Pia is such that John didn’t invite her

NOT: Only

* We have benefitted from presentations of parts of this chapter at the University of Manchester (LAGB 2011), the University of Leiden (Leiden Syntax Lab, 2011) and the University of Frankfurt (Semantics Colloquium, 2012). Special thanks are due to Rick Nouwen for detailed comments on an early draft. We would further like to thank Daniel Büring, Will Harwoord, Larry Horn, Marika Lekakou, Rachel Nye, Matthew Reeve, Rob Truswell and Ede Zimmermann for comments, questions, judgments and other help.

228 (2)

Ad Neeleman and Reiko Vermeulen John didn’t invite even PIA. | Even Pia is such that John didn’t invite her

It is easy to underestimate this problem. After all, the contrast between (1) and (2) could be explained by assuming that even-DPs undergo quantifier raising over negation, but only-DPs do not. Quantifier raising exists independently, and it is known from the literature that different quantifiers have distinct scope-taking properties. There is no reason why this should not be true of only and even. However, an analysis along this line is unlikely to be on the right track, given that the same pattern is found when negation combines directly with only or even, as in the examples in (3) and (4). As in our earlier examples, only cannot scope over negation, while even must do so. In the case at hand, quantifier raising is unhelpful, because DPs in general are islands for quantifier raising, and subjects are islands for any kind or movement. (3)

Not only JOHN invited Pia. John is such that he didn’t invite Pia

NOT: Only

(4)

Not even JOHN invited Pia. | Even John is such that he didn’t invite Pia

As we proceed, we will encounter a further problem with reducing the asymmetry between only and even to an asymmetry in quantifier raising. As it turns out, only-DPs and even-DPs can undergo long distance quatifier raising in contexts of the type discussed by Taglicht (1984 : 150), but neither can raise across negation. This fact might help us understand why the scope of only is confined to below negation, but it makes the contrasting behaviour of even quite mysterious. The above problems relate to the assumption that even-DPs undergo quantifier raising. There is also a problem with the assumption that onlyDPs remain in situ: the interpretation of (1) is not what one would expect if only is simply in the scope of negation. That is, (1) implies that John did in fact invite Pia, but this should not be the case if the sentence is a simple negation of its positive counterpart (John invited only Pia). The effect is sometimes described in terms of presupposition projection. On some accounts of only (e.g., Horn 1969), John invited only Pia carries the presupposition that John invited Pia, and presuppositions survive negation. However, this cannot work, because John did not invite only Pia is not equivalent to It is not the case that John invited only Pia. The former implies that John invited Pia (and also someone other than Pia), but the latter does not. The following context makes the contrast clear:

Chapter 7 Types of Focus and their Interaction with Negation (5)

229

A: Everyone invited many people to Bill’s birthday party, but John invited only Pia. B: That’s not true. It’s not the case that John invited only PIA, because he actually didn’t invite anyone / because he actually didn’t invite Pia – he invited Sue and Frida. B’: That’s not true. #John didn’t invite only PIA, because he actually didn’t invite anyone / because he actually didn’t invite Pia – he invited Sue and Frida.

The fact that a presupposition survives negation of the form it is not the case that... can be seen from examples such as It is not the case that John regretted Bill came, which still carries the presupposition that Bill came. If the interpretation of (1) is simply due to presupposition projection, one would not expect the contrast in (5).1 So, the difference in the interpretation of even and only with respect to negation remains unexplained. In the following section, we will reformulate the above puzzle in terms of the general idea that the semantics of sentences with focus-sensitive particles such as even and only consist of two components. Our proposal is that there is variation in the component onto which negation is mapped.

1 One qualification is in order. It is possible to deny presuppositions. In general, this is helped by stressing negation, as illustrated below: (i) a. b.

Mary DIDN’T discover that Bill was happy, because he wasn’t happy in the first place. Mary DIDN’T stop smoking, since she never smoked in the first place.

The same effect can be observed with sentences containing only. In the following example, the presupposition of the first conjunct (i.e., that John invited Pia) is denied by the second conjunct. (ii) John DIDN’T invited only PIA, because he didn’t invite anyone in the first place. This reading is not generally available with stress on Pia, indicating that this is a genuine case of denial of a presupposition. This conclusion is further supported by the fact that (ii) is compatible with the insertion of in the first place. As illustrated in (i), this adjunct appears with denial of presuppositions, but not with denial of assertions: (iii) a. b.

A: B: A: B:

John read this book. #No, he didn’t read it in the first place. John invited only PIA. #No, he also invited Mary in the first place.

230

Ad Neeleman and Reiko Vermeulen

This variation captures the data discussed in this section, as well as a range of further observations.

2.

Reformulation of the problem

In the previous section, we have described the empirical problem in terms of scope and quantifier raising. However, there is another way of approaching the data, which we will outline in this section. Traditionally, the semantics of sentences with a focus-sensitive particle is described as having two components of meaning. This line of analysis goes back at least to Horn 1969 and is still widely assumed (see Beaver & Clark 2008 and references therein). The example in (6), for instance, states that John invited Pia and also that John did not invite any relevant alternative to Pia. The first component is often called the prejacent: it is the meaning of the sentence without the focus-sensitive particle only (see (6a)). We propose that this component consists of both the ordinary value and the focus value of the sentence (in the sense of Rooth 1985, 1992). For this reason, we will call this component the ‘focus component’. The second component is the interpretation that only contributes. It is given in (6b) and provides further information about the members of the set of alternatives. We will call this component the ‘alternatives component’. We now consider even. The sentence in (7) has the same focus component as (6), but the content of the alternatives component is different. It does not contain a negative operator, but introduces a scale.2 (The examples in (6) and (7) are adapted from Krifka (1999). Here and below small caps are used to indicate focus.) (6)

John invited only PIA. a. [John invited PIA] b. ™FcFcz Pia, [John invited Fc]

2 We restrict our discussion here and below to focus-sensitive particles attached to DPs. We are aware that focus-sensitive particles may also attach to predicates, which may be preferred under certain circumstances, and that their interpretational effects in that case may be different. For example, I invited even Pia implies that the speaker invited someone else in addition to Pia, but I even invited Pia need not. The one exception we make is for also, which does not attach to a DP, so that we can only consider cases where it is attached to a predicate.

Chapter 7 Types of Focus and their Interaction with Negation (7)

231

John invited even PIA. a. [John invited PIA] b. FcFczPia, [[[John invited Pia] AgrOP). However, this is not sufficient to explain the second property, that is, the fact that the order of constituents interpreted as topic and focus is free when they remain in situ, but rigid as soon as one of them moves. As we have shown, a topic can move out of a constituent containing an in situ focus, but a focus cannot move out of a constituent containing an in situ topic. The problem this raises for a ‘relativized cartography’ approach is that moved topics and moved foci would occupy positions in the topic-focus hierarchy, while topics and foci that remain in situ would occupy positions in the argumental hierarchy. Hence, ordering restrictions must be formulated that involve positions in more than one hierarchy, showing that a tier-based account is insufficient. This is the core of the argument. However, we can be more precise. Neeleman and Van de Koot 2008 demonstrate that the following assumptions must be made about the grammar of Dutch if moved topics and foci are to be licensed in the specifier of TopP and FocP, respectively: iii. iv. v.

The position of TopP and FocP is free (at least in Dutch). (This captures the observed variation in landing sites.) Projection of either [contrast] or [topic] and [focus] is optional (at least in Dutch). (This allows in-situ topics and foci.) Heads containing [contrast] mark their complements as the domain of contrast. (This, in conjunction with restrictions on information structure, captures the observed ordering restrictions.)

Each of these assumptions sacrifices an assumption central to the cartographic framework. Assumptions (iii) and (iv) imply that the idea that there is a fixed clausal skeleton must be abandoned. Assumption (v) implies that the idea that movement is triggered solely by properties of the specifier/attracting head must be abandoned. These sacrifices seem considerable. But the Dutch data are even more damaging. The assumptions in (iii) and (iv) work if joint projection of [contrast] and [topic]/[focus] as features of the same head is allowed. However, within the cartographic framework, the independent syntactic effects of [topic], [focus] and [contrast] require an account in terms of three separate functional projections: TopP, FocP and ContrastP. This is because cartography strives for a oneto-one relation between syntactic position and interpretive effect, and the data motivate three separate interpretively relevant features. Cross-linguistic variation would be captured by variation in the extent to which

Chapter 8 Concluding Remarks

273

these projections trigger displacement. In Japanese, topics appear in the specifier of TopP. In Russian, foci appear in the specifier of FocP, while contrastive foci move on to the specifier of ContrastP. Finally, in Dutch, contrastive topics and foci move to the specifier of ContrastP. The addition of ContrastP to the topic-focus hierarchy requires an additional ordering statement. On the assumption that TopP dominates FocP, there are three logical possibilities: (7)

a. ContrastP > TopP > FocP b. TopP > ContrastP > FocP c. TopP > FocP > ContrastP

The Russian data suggest that ContrastP dominates FocP, because contrastive foci in this language move from the position in which foci in general are licensed. This rules out (7c). We are thus left with (7a) and (7b), but it can be demonstrated that neither of these orders is compatible with the Dutch data. If ContrastP dominates TopP, the landing site for contrastive topics and foci is identical and it is therefore impossible to capture the ordering restrictions associated with topic and focus movement. In particular, this arrangement implies that moved foci always cross the position in which topics are licensed. As a result, it becomes hard, if not impossible, to explain why – at an observational level – topics may move across foci but foci may not move across topics. This rules out (7a). If ContrastP occupies a position between TopP and FocusP, one might attempt to capture the Dutch data by saying that contrastive topics move on to the specifier of TopP, while contrastive foci surface in ContrastP. We should then require TopP to trigger movement, which in turn would lead to expect displacement of aboutness topics, contrary to fact. This rules out (7b) (and also (7a)). These considerations suggest that, although [contrast] has syntactic effects that can be distinguished from those of [topic] and [focus], a cartographic decomposition into three separate functional projections is not possible. But since the logic of cartography is such that it requires featural decomposition, the data discussed in this book argue against the cartographic view.

References Abels, Klaus 2008 Towards a restrictive theory of (remnant) movement. Linguistic Variation Yearbook 7 : 53–120. Abels, Klaus 2012 The Italian left periphery: A view from locality. Linguistic Inquiry 43. Abels, Klaus and Peter Muriungi 2008 The focus marker in Kîîtharaka: Syntax and semantics. Lingua 118 : 687–731. Aboh, Enoch Oladé 2004 The Morphosyntax of Complement-Head Sequences: Clause Structure and Word Order Patterns in Kwa. Oxford/New York: Oxford University Press. Abrusán, Márta and Kriszta SzendrĘi 2011 Experimenting with the King of France: Topics and verifiability and definite descriptions. Ms., Somerville College, University of Oxford/ Lichtenberg Kolleg and University College London. Ackema, Peter and Ad Neeleman 1998 Optimal questions. Natural Language and Linguistic Theory 16 : 443–490. Adamec, Pržemysl 1966 Porjadok slov v sovremennom russkom jazyke [Word order in contemporary Russian]. (Rozpravy ýeskoslovenske Akademie VƟd, ěada spoleþenskych vČd 76 [Debates of the Czech Academy of Sciences, Social Sciences Series 76].) Prague: Academia. Adger, David and Gillian Ramchand 2003 Predication and equation. Linguistic Inquiry 34 : 325–359. Akmajian, Adrian 1970 On deriving cleft sentences from pseudo-cleft sentences. Linguistic Inquiry 1 : 149–168. Andrews, Avery 1983 A note on the constituent structure of modifiers. Linguistic Inquiry 14 : 695–697. Aoyagi, Hiroshi and Sachiko Kato 2009 On information packaging of topicalized and scrambled sentences in Japanese. Current Issues in Unity and Diversity of Languages: Collection of Papers Selected from the International Congress of Linguists 18, 1–19. Seoul: The Linguistic Society of Korea. Ariel, Mira 1987 Referring and accessibility. Journal of Linguistics 24 : 65–87. Ariel, Mira 1990 Accessing Noun-Phrase Antecedents. London: Routledge.

276

References

Arregui-Urbina, Karlos 2002 Focus on Basque Movements. Ph.D. dissertation, Massachusetts Institute of Technology. Astruc, Lluïsa 2005 The Intonation of Extra-Sentential Elements in Catalan and English. Ph.D. dissertation, University of Cambridge. Atlas, Jay and Stephen Levinson 1981 It-clefts, informativeness, and logical form. In: Peter Cole (ed.), Radical pragmatics, 1–62. New York: Academic Press. Bailyn, John F. 2004 Generalized inversion. Natural Language and Linguistic Theory 22 : 1–49. Baker, Mark 2003 Lexical Categories. Cambridge: Cambridge University Press. Baltaxe, Christiane 1984 Use of contrastive stress is normal, aphasic, and autistic children. Journal of Speech and Hearing Research 27 : 97–105. Barker, Chris 2001 Integrity: A syntactic constraint on quantificational scoping. In: Karine Megerdoomian and Leora Anne Bar-El (eds.), Proceedings of the 20th West Coast Conference on Formal Linguistics, 101–114. Somerville, MA: Cascadilla Press. Barwise, Jon and Robin Cooper 1981 Generalized quantifiers and natural language. Linguistics and Philosophy 4 : 159–219. Beaver, David I. and Brady Zack Clark 2008 Sense and Sensitivity: How Focus Determines Meaning. Oxford: Wiley-Blackwell. Beaver, David I., Brady Zack Clark, Edward Stanton Flemming, T. Florian Jaeger and Maria Wolters 2007 When semantics meets phonetics: Acoustical studies of second occurrence focus. Language 83 : 245–276. Beck, Sigrid 1996 Quantified structures as barriers for LF movement. Natural Language Semantics 4 : 1–56. Beck, Sigrid 2006 Intervention effects follow from focus interpretation. Natural Language Semantics 14 : 1–56. Belletti, Adriana 2001 Inversion as focalization. In: Aafke Hulk and Jean-Yves Pollock (eds.), Subject Inversion in Romance and the Theory of UG, 60–90. Oxford: Oxford University Press. Benincà, Paola and Cecilia Poletto 2004 Topic, focus and V2: Defining the CP sublayers. In: Luigi Rizzi (ed.),

References

277

The Structure of CP and IP: The Cartography of Syntactic Structures, 52–75. Oxford: Oxford University Press. Bianchi, Valentina and Mara Frascarelli 2010 Is topic a root phenomenon? Iberia 2 : 43–88. Bobaljik, Jonathan 1999 Adverbs: The hierarchy paradox. Glot International 4.9/10. Bobaljik, Jonathan and Susi Wurmbrand 2008 Word order and scope: Transparent interfaces and the 3/4 signature. Ms., University of Connecticut, Storrs. To appear in Linguistic Inquiry. Bolinger, Dwight L. 1961 Contrastive accent and contrastive stress. Language 37 : 83–96. Boškoviü, Željko 2002 A-movement and the EPP. Syntax 5 : 167–218. Brody, Michael 1990 Some remarks on the focus field in Hungarian. UCL Working Papers in Linguistics 2 : 201–225. Brody, Michael 1995 Focus and checking theory. In I. Kenesei (ed.), Approaches to Hungarian, vol. V, 29–44. Szeged: JATE. Bruening, Benjamin 2001 QR obeys superiority: Frozen scope and ACD. Linguistic Inquiry 32 : 233–273. Brun, Dina 2001 Information structure and the status of NP in Russian. Theoretical Linguistics 27 : 109–135. Brunetti, Lisa 2009 On links and tails in Italian. Lingua 119 : 756–781. Bryzgunova, E.A. 1971 O smyslorazlichitel’nyh vozmozhnostyah russkoj intonatsii [On the interpretative effects of Russian intonation]. Voprosy jazykoznanija [Issues in Linguistics] 4 : 42–51. Bryzgunova, E.A. 1981 Zvuk i intonatsija russkoj rechi [Sound and Intonation of Russian Speech]. Moscow: Russkij Yasyk. Büring, Daniel 1997 The Meaning of Topic and Focus: The 59th Street Bridge Accent. London: Routledge. Büring, Daniel 2003 On D-trees, beans, and B-accents. Linguistics and Philosophy 26 : 511–545. Büring, Daniel and Katharina Hartmann 2001 The syntax and semantics of focus-sensitive particles in German. Natural Language and Linguistic Theory 19 : 229–281.

278

References

Campos, Hector and Melita Stavrou 2004 Polydefinites in Greek and Aromanian. In: Olga Tomic (ed.), Balkan syntax and semantics, 137–173. Amsterdam: John Benjamins. Chafe, Wallace L. 1976 Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In: Charles N. Li (ed.), Subject and Topic, 25–56. New York: Academic Press. Charlow, Simon 2009 Can DP be a scope island? Proceedings of the European Summer School for Logic, Language and Information 2009, 148–159. Chen, Aoju 2010 Is there really an asymmetry in the acquisition of the focus-toaccentuation mapping? Lingua 120: 1926–1939. Cho, Jai-Hyoung 1994 Scrambling in Korean: Crossover, reconstruction and binding theory. Ph.D. dissertation, University of Connecticut. Choe, Hyon Sook 1995 Focus and topic movement in Korean and licensing. In: Katalin É. Kiss (ed.), Discourse Configurational Languages, 269–334. Oxford: Oxford University Press. Choi, Hye-Won 1997 Topic and focus in Korean: The information partition by Phrase structure and morphology. In: Ho-min Sohn and John Haig (eds.), Japanese/Korean Linguistics 6, 545–560. Stanford: Center for the Study of Language and Information. Choi, Hye-Won 1999 Optimizing Structure in Context: Scrambling and Information Structure. Stanford: Center for the Study of Language and Information. Chomsky, Noam 1976 Conditions on rules of grammar. Linguistic Analysis 2 : 303–351. Chomsky, Noam 1986 Barriers. Cambridge, MA: MIT Press. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: MIT Press. Cinque, Guglielmo 1999 Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: OUP. Clech-Darbon, Anne, Georges Rebuschi and Annie Rialland 1999 Are there cleft sentences in French? In: Georges Rebuschi and Laurice Tuller (eds.), The Grammar of Focus, 83–118. Amsterdam: John Benjamins. Constant, Noah 2006 English rise-fall-rise: A study in the semantics and pragmatics of intonation. M.A. thesis, University of California, Santa Cruz.

References

279

Cormack, Annabel and Neil Smith 2000 Fronting: The Syntax and Pragmatics of ‘Focus’ and ‘Topic’. In: Corinne Iten and Ad Neelemann (eds), UCL Working Papers in Linguistics 12, 387–417. London: UCL. Culicover, Peter, W. 1991 Topicalization, inversion, and focus in English. In: Denis Delfitto, Martin Everaert, Arnold Evers and Frits Stuurman (eds.), Going Romance and Beyond: Fifth Symposium on Comparative Grammar (OTS Working Papers 91-002), 1–43. Utrecht: University of Utrecht. Culicover, Peter W. and Michael S. Rochemont 1990 Extraposition and the Complement Principle. Linguistic Inquiry 21 : 23–47. Daneš, František 1974 Functional sentence perspective and the organization of the text. In: František Daneš (ed.), Papers on Functional Sentence Perspective, 106–128. Prague: Academia. Davidse, Kristin 2000 A constructional approach to clefts. Linguistics 38 : 1101–1131. Dehé, Nicole 2002 Particle verbs in English: Syntax, Information Structure, and Intonation. Amsterdam: John Benjamins. Dehé, Nicole 2009 An international grammar for Icelandic. Nordic Journal of Linguistics 31(1) : 5–34. Delahunty, Gerald 1981 Topics in the Syntax and Semantics of English Cleft Sentences. Ph.D. dissertation, University of California, Irvine. Delais-Roussarie, Elisabeth, Jenny Doetjes and Petra Sleeman 2004 Dislocations in French. In: Corblin Francis and Henriëtte de Swart (eds.), Handbook of French semantics, 501–528. Stanford: CSLI. den Dikken, Marcel 1995 Particles: On the Syntax of Verb-Particle, Triadic, and Causative Constructions. New York: Oxford University Press. den Dikken, Marcel 2006 Relators and Linkers: The Syntax of Predication, Predicate Inversion, and Copulas. Cambridge, MA: MIT Press. den Dikken, Marcel, André Meinunger, and Chris Wilder 2000 Pseudoclefts and ellipsis. Studia Linguistica 54 : 49–89. Dretske, Fred I. 1972 Contrastive statements. The Philosophical Review 81 : 411–437. Ebert, Cornelia and Stefan Hinterwimmer 2010 The Interpretation of Topical Indefinites as Direct and Indirect Aboutness Topics. In: Malte Zimmermann and Caroline Féry (eds.), Information Structure: Theoretical, Typological and Experimental Perspectives, 89–114. Oxford: Oxford University Press.

280

References

É Kiss, Katalin 1998 Identificational focus versus information focus. Language 74 : 245– 273. Emonds, Joseph 1976 A Transformational Approach to English Syntax. New York: Academic Press. Emonds, Joseph 2004 Unspecified categories as the key to root constructions. In: David Adger, Cécile De Cat, and Georges Tsoulas (eds.), Peripheries: Syntactic Edges and their Effects, 75–120. Dordrecht: Kluwer. Endo, Yoshio 2007 Locality and Information Structure: A Cartographic Approach to Japanese. Amsterdam: John Benjamins. Endriss, Cornelia 2009 Quantificational Topics: A Scopal Treatment of Exceptional Wide Scope Phenomena. (Studies in Linguistics and Philosophy.) Dordrecht: Springer. Ernst, Thomas 2001 The Syntax of Adjuncts. Cambridge: CUP. Erteschik-Shir, Nomi 1997 The Dynamics of Focus Structure. Cambridge: Cambridge University Press. Feldhausen, Ingo 2010 Sentential Form and Prosodic Structure of Catalan [Linguistik Aktuell/Linguistics Today 168]. Amsterdam: John Benjamins. Féry, Caroline 2007 The prosodic basis of topicalization. In: Kerstin Schwabe and Susanne Winkler (eds.), On Information Structure, Meaning and Form, 69–86. Amsterdam/Philadelphia: John Benjamins. Féry, Caroline and Vieri Samek-Lodovici 2006 Focus projection and prosodic prominence in nested foci. Language 82 : 131–50. Fiengo, Robert and William McClure 2002 On how to use -wa. Journal of East Asian Linguistics 11 : 5–41. Firbas, J. 1964a From comparative word-order studies. Brno Studies in English 4 : 111–128. Firbas, J. 1964b On defining the theme in functional sentence analysis. Travaux linguistiques de Prague 1 : 267–280. Fox, Danny 1999 Reconstruction, binding theory, and the interpretation of chains. Linguistic Inquiry 30 : 157–196. Fox, Danny 2000 Economy and Semantic Interpretation. Cambridge, MA: MIT Press.

References

281

Frascarelli, Maria and Roland Hinterhölzl 2007 Types of topics in German and Italian. In: Kerstin Schwabe and Susanne Winkler (eds.), On Information Structure, Meaning and Form, 87–116. Amsterdam/Philadelphia: John Benjamins. Frey, Werner 2001 About the whereabouts of indefinites. Theoretical Linguistics 27 : 137–161. Frey, Werner 2004 A medial topic position for German. Linguistische Berichte 198 : 153–190. Fukui, Naoki 1995 The principles-and-parameters approach: A comparative syntax of English and Japanese. In: Masayoshi Shibatani and Theodora Bynon (eds.), Approaches to Language Typology, 327–372. Oxford: Oxford University Press. Geist, Ljudmila 2007 Predication and equation in copular sentences: Russian vs. English. In: Ileana Comorovski and Klaus von Heusinger (eds.), Existence: Semantics and syntax, 79–105. Berlin: Springer. Geist, Ljudmila and Joanna Błaszczak 2000 Kopulasätze mit dem pronominalen Elementen to/èto im Polnischen und Russischen. Zentrum für Allgemeine Sprachwissenschaft Papers in Linguistics 16, 115–139. Geurts, Bart and Rob van der Sandt 2004 Interpreting focus. Theoretical Linguistics 30 : 1–44. Gill, Kook-Hee and George Tsoulas 2004 Peripheral effects without peripheral syntax: The left periphery in Korean. In: David Adger, Cécile De Cat, and George Tsoulas (eds.), Peripheries: Syntactic Edges and their Effects, 121–141. Kluwer: Dordrecht. Giusti, Giuliana 2006. Parallels in clausal and nominal periphery. In: Mara Frascarelli (ed.),Phases of Interpretation. Berlin: Mouton de Gruyter. Givón, Talmy 1976 Topic, pronoun and grammatical agreement. In: Charles Li (ed.), Subject and Topic, 149–188. New York: Academic Press. Givón, Talmy 1983 Introduction. In: Talmy Givón (ed.), Topic Continuity in Discourse, 5–41. Amsterdam/Philadelphia: John Benjamins. Gruber, Jeffrey S. 1965 Studies in lexical Relations. Ph.D. dissertation, Massachusetts Institute of Technology. Gundel, Jeanette 1974 The role of Topic and Comment in Linguistic Theory. Ph.D. disserta-

282

References

tion, University of Texas at Austin. Distributed by Indiana University Linguistics Club, Bloomington, IN, 1977. Gundel, Jeanette 1977 Where do cleft sentences come from? Language 53 : 543–559. Haegeman, Liliane 2002 Anchoring to speaker, adverbial clauses and the structure of CP. In: Simon Mauck and Jenny Mittelstaedt (eds.), Georgetown University Working Papers in Linguistics 2 : 117–180. Haider, Hubert 2000 Adverb placement – Convergence of structure and licensing. Theoretical Linguistics 26 : 95–134. Haider, Hubert and Inger Rosengren 1998 Scrambling. Sprache und Pragmatik 49 : 1–104. Hajiþova, Eva 1974 Meaning, presupposition and allegation. Philologica Pragencia17 : 18–25. Hajiþová, Eva, Barbara. H. Partee, and Petr Sgall 1998 Topic-Focus Articulation, Tripartite Structures, and Semantic Content. Dordrecht: Kluwer. Halliday, M.A. K. 1967 Notes on transitivity and theme in English II. Journal of Linguistics 3 : 199–244. Halvorsen, Per-Kristian 1978 The syntax and semantics of cleft sentences. Ph.D. dissertation, University of Texas at Austin. Hamblin, Charles 1973 Questions in Montague English. Foundations of Language 10 : 41–53. Han, Chung-hye 1998 Asymmetry in the interpretation of -(n)un in Korean. In: Noriko Akatsuka, Hajime Hoji, Shoichi Iwasaki, Sung-Ock Sohn, and Susan Strauss (eds.), Japanese/Korean Linguistics 7, 1–15. Stanford: Center for the Study of Language and Information. Hara, Yurie 2006 Grammar of knowledge representation: Japanese discourse items at interfaces. Ph.D. dissertation, University of Delaware. Hara, Yurie 2007 Dake-wa: Exhaustifying assertions. In: Takashi Washio, Ken Satoh, Hideaki Takeda, and Akihiro Inokuchi (eds.), New Frontiers in Artificial Intelligence: Proceedings of the 20th Annual Conferences of the Japanese Society for Artificial Iintelligence Workshops 2006, 219– 231. Berlin/Heidelberg: Springer. Hara, Yurie and Robert van Rooij 2007 Contrastive topics revisited: A simpler set of topic-alternatives. Presentation at the North East Linguistics Society 38, University of Ottawa, 26–28 October.

References

283

Hartmann, Katharina and Malte Zimmermann. 2006 In place – out of place?: Focus strategies in Hausa. In: Kerstin Schwabe and Susanne Winkler (eds.), On Information Structure, Meaning and Form, 365–403. Amsterdam/Philadelphia: John Benjamins. Hedberg, Nancy 1990 The Discourse Function of Cleft Sentences in English. Ph.D. dissertation, University of Minnesota. Hedberg, Nancy 2000 On the referential status of clefts. Language 76 : 891–920. Heggie, Lorie 1988 The syntax of copular structures. Ph.D. dissertation, University of Southern California. Heim, Irene and Angelika Kratzer 1998 Semantics in Generative Grammar. Oxford: Blackwell. Herburger, Elena 2000 What Counts: Focus and Quantification. Cambridge, MA: MIT Press. Hetland, Jorunn 2007 The Korean particle nun, the English fall-rise accent and thetic/categorical judgements. In: Kerstin Schwabe and Susanne Winkler (eds.), On Information Structure, Meaning and Form, 117–127. Amsterdam/Philadelphia: John Benjamins. Heycock, Caroline and Anthony Kroch 1999 Pseudocleft connectedness: Implications for the LF interface. Linguistic Inquiry 30 : 365–397. Heycock, Caroline 2006 Embedded root phenomena. In: Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell Companion to Syntax, Volume 2, 174–209. Oxford: Blackwell. Heycock, Caroline 2007 Japanese -wa, -ga, and information structure. In: Mamoru Saito and Shigeru Miyagawa (eds.), Oxford Handbook of Japanese Linguistics, 54–83. Oxford: Oxford University Press. Higginbotham, James 1985 On semantics. Linguistic Inquiry 16 : 547–593. Higgins, Roger 1973 The pseudo-cleft construction in English. Ph.D. dissertation, Massachusetts Institute of Technology. Hinds, John, Senko Maynard, and Shoichi Iwasaki (eds.) 1987 Perspectives on Topicalization: the Case of Japanese wa. Amsterdam: John Benjamins. Hoji, Hajime 1985 Logical form constraints and configurational structures in Japanese. Ph.D. disseration, University of Washington.

284

References

Hole, Daniel 2011 A new intervention effect with ‘only’ – additional evidence for a distributed syntax-and-semantics of scalar ‘only’. Presentation at the Generative Linguistics in the Old World 34 workshop on intervention effects from a semantic perspective, University of Vienna, 27 April. Holm, John et al. 1999 Copula patterns in Atlantic and non-Atlantic creoles. In: Suzanne Romaine (ed.), Creole Genesis, Attitudes and Discourse: Studies Celebrating Charlene J. Sato, 97–119. Amsterdam: John Benjamins. Horn, Laurence 1969 A presuppositional analysis of only and even. Chicago Linguistics Society 5 : 97–108. Horn, Laurence 1972 On the semantic properties of logical operators in English, Ph.D. dissertation, University of California, Los Angeles. Distributed by Indiana University Linguistics Club, Bloomington, IN, 1976. Horn, Laurence 1981 Exhaustiveness and the semantics of clefts. In: Victoria A. Burke and James Pustejovsky (eds.), Proceedings of the North East Linguistic Society 11, 125–142. Amherst, MA: University of Massachusetts, Graduate Linguistics Student Association. Horn, Laurence 1989 A Natural History of Negation. Chicago: University of Chicago Press. Horn, Laurence 2002 Assertoric inertia and NPI licensing. Chicago Linguistics Society 38: The Panels: 55–82. Horvath, Julia 2000 Interfaces vs. computational system in the syntax of focus. In Martin Everaert and Eric Reuland (eds.), Interface strategies, 183–206. Amsterdam: Royal Netherlands Academy of Arts and Sciences. Horvath, Julia 2010 “Discourse features”, syntactic displacement and the status of contrast. Lingua 120 : 1346–69. Huang, Cheng-Teh James 1982 Logical relations in Chinese and the theory of grammar. Ph.D. dissertation, Massachusetts Institute of Technology. Hwang Jackson, Kyuseek, Amy Schafer and William O’Grady 2010 Contrastive focus facilitates scrambling in Korean sentence processing. In: Shoishi Iwasaki, Hajime Hoji, Patricia M. Clancy, and Sung-Ock Sohn (eds.), Japanese/Korean Linguistics 17, 167–181. Center for the Study of Language and Information: Stanford.

References

285

Ioannidou, Alexandra and Marcel Den Dikken 2009 P-drop, D-drop and D-spread. MIT Working Papers in Linguistics 57: 393–408. Ionin, Tania 2001 Scope in Russian: Quantifier Movement and Discourse Function. Ms., Massachusetts Institute of Technology. Isaþenko, Alexander V. 1976a Frazovoe udarenie i porjadok slov [Phrasal stress and word order]. In: Alexander V. Isaþenko, Opera Selecta, 373–382. Munich: Fink. Isaþenko, Alexander V. 1976b O grammaticheskom porjadke slov [On syntactic word order]. In: Alexander V. Isaþenko, Opera Selecta, 365–372. Munich: Fink. Ishizuka, Tomoko To appear Alienable-Inalienable Asymmetry in Japanese and Korean Possession. UPenn Working Papers in Linguistics 16. Jackendoff, Ray 1972 Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jackendoff, Ray 1996 The Architecture of the Language Faculty. Cambridge, MA: MIT Press. Jacobs, Joachim 1983 Fokus und Skalen. Zur Syntax und Semantik der Gradpartikel im Deutschen. Tübingen: Niemeyer Jacobs, Joachim 1984 Funktionale Satzperspektive und Illokutionssemantik. Linguistische Berichte, 91 : 25–58. Jacobs, Joachim 1991 Negation. In: Armin von Stechow and Dieter Wunderlich (eds.), Semantics: An International Handbook of Contemporary Research, 560–596. Berlin: Walter de Gruyter. Jacobs, Joachim 1997 Bemerkungen zur I-Topikalisierung. Linguistische Berichte 168 : 91–133. Jun, Sun-Ah 1993 The phonetics and phonology of Korean prosody. Ph.D. dissertation, The Ohio State University. Junghanns, Uwe 1997 On the so-called èto-cleft construction. In: Martina Lindseth and Steven Franks (eds.), Proceedings of the Sixth Annual Workshop on Formal Approaches to Slavic Linguistics, 166–190. Stanford: Center for the Study of Language and Information. Kariaeva, Natalia 2004 Determiner spreading in Modern Greek: split-DP hypothesis. Ms. Rutgers University.

286

References

Kartunnen, Lauri and Stanley Peters 1979 Conventional implicature. In: Choon-Kyo Oh and David A. Dinneen (eds.), Syntax and Semantics 11: Presupposition, 1–56. New York: Academic Press. Katz, Jonah and Lisa Selkirk 2011 Contrastive focus vs. discourse-new: Evidence from phonetic prominence in English. Language 87 : 771–816. Kim, Jieun 2009 A unified theory of contrastive topic and RFR. In: Proceedings of the 2nd International Conference on East Asian Linguistics. King, Tracy Holloway 1993 Configuring topic and focus in Russian. Ph.D. dissertation, Stanford University. King, Tracy Holloway 1995 Configuring Topic and Focus in Russian. Stanford: Center for the Study of Language and Information. Kishimoto, Hideki 2009 Topic prominency in Japanese. The Linguistic Review 26 : 465–513. Kiss, Tibor 2001 Configurational and relational scope determination in German. In: W. Detmar Meurers and Tibor Kiss (eds.), Constraint-Based Approaches to Germanic Syntax, 141–175. Stanford: Center for the Study of Language and Information. Kitagawa, Yoshihisa 1990 Anti-Scrambling. Ms. University of Rochester. Koch, Karsten 2008 Focus projection in NËe?kepmxcin (Thompson River Salish). In: Charles B. Chang and Hannah J. Haynie (eds.), Proceedings of the 26th West Coast Conference on Formal Linguistics, 348–356. Somerville, MA: Cascadilla Proceedings Project. Kolliakou, Dimitra 2004 Monadic definites and polydefinites: their form, meanning and use. Journal of Linguistics 40 : 263–333. König, Ekkehard 1991 The Meaning of Focus Particles: A Comparative Perspective. London: Routledge. Koster, Jan 1991 Variable binding without c-command. In: Language and Cognition 1, 187–193. Groningen: University of Groningen. Kratzer, Angelika 2004 Interpreting focus: Presupposed or expressive meanings? A comment on Geurts and van der Sandt. Theoretical Linguistics 30 : 123– 136.

References

287

Krifka, Manfred 1993 Focus and presupposition in dynamic interpretation. Journal of Semantics 10 : 269–300. Krifka, Manfred 1998 Scope Inversion under the Rise-Fall Contour in German. Linguistic Inquiry 29 : 75–112. Krifka, Manfred 1999 Additive particles under stress. Proceedings of Semantics and Linguistic Theory 8, 111–128. Cornell: CLC Publications. Krifka, Manfred 2001 For a structured meaning account of questions and answers (revised version). In: Caroline Féry and Wolfgang Sternefeld (eds.), Audiatur Vox Sapientia. A Festschrift for Arnim von Stechow. (Studia Grammatica 52.), 287–319. Berlin: Akademie Verlag. Krifka, Manfred 2006 Association with focus phrases. In: Valerie Molnar and Susanne Winkler (eds.), The Architecture of Focus, 105–136. Berlin: Mouton de Gruyter. Krifka, Manfred 2006b Can focus accenting be eliminated in favor of deaccenting given constituents? In: Proceedings of the 9th Synposium on Logic and Language (Besenyötelek, Hungary, August 24–26, 2006), 107–119. Krifka, Manfred 2008 Basic notions of information structure. Acta Linguistica Hungarica 55 : 243–276. Krylova, Olga and S.A. Khavronina 1984 Porjadok slov v russkom jazyke [Word Order in the Russian Language]. Moscow: Russkij Yazyk. Krylova, Olga and S.A. Khavronina 1988 Word Order in Russian. Moscow: Russkij Yazyk. Kyriakaki, Maria 2010 What Greek DET’s do: The restrictive DP. In Melinda Heijl (ed.), Proceedings of the 2010 annual conference of the Canadian Linguistic Association. Kuno, Susumu 1973 The Structure of the Japanese Language. Cambridge, MA: MIT Press. Kuroda, S.-Y. 1965 Generative Grammatical Studies in the Japanese Language. Ph.D. dissertation, Massachusetts Institute of Technology. [reproduced in 1979 by Garland, New York.] Kuroda, S.-Y. 1992 Judgment Forms and Sentence Forms. In: S.-Y. Kuroda, Japanese Syntax and Semantics: Collected Papers, 13–77. Dordrecht: Kluwer.

288

References

Kuroda, S.-Y. 2005 Focusing on the matter of topic: A study of wa and ga in Japanese. Journal of East Asian Linguistics 14 : 1–58. Ladusaw, William 1979 Polarity sensitivity as inherent scope relations. Ph.D. dissertation, University of Texas, Austin. Laenzlinger, Christopher 2000 French adjective ordering: perspectives on DP-internal movement types. Generative grammar in Geneva 1 : 55–104. Laenzlinger, Christopher 2005 Some notes on DP-internal movement. Generative Grammar in Geneva 4 : 227–260. Lakoff, George 1972 On Generative Semantics. In: Danny D. Steinberg, and Leon A. Jakobovits (eds.), Semantics, 232–296. Cambridge: Cambridge University Press. Lambrecht, Knud 1994 Information Structure and Sentence Form. Cambridge: Cambridge University Press. Larson, Richard K 1985 Quantifying into NP. Ms., Massachusetts Institute of Technology. Lasersohn, Peter 1993 Existence presuppositions and background knowledge. Journal of Semantics 10 : 113–122. Lechner, Winifried 2009 Evidence for survive from covert movement. In: Mike Putnam (ed.), The Survive Principle in a Crash-Proof Syntax, 231–254. Amsterdam: John Benjamins. Lee, Chungmin 1999 Contrastive topic: A locus of the interface evidence from Korean and English. In: Ken Turner (ed.), The Semantics/Pragmatics Interface from Different Points of View, 317–342. Oxford: Elsevier. Lee, Chungmin 2003 Contrastive topic and/or contrastive focus. In: William McClure (ed.), Japanese/Korean Linguistics 12. Stanford: Center for the Study of Language and Information. Lee, Chungmin 2006 Contrastive (Predicate) Topic, intonation, and scalar meanings. In: Chungmin Lee, Matt Gordon and Daniel Büring (eds.), Topic and Focus: Cross-linguistic Perspectives on Meaning and Intonation, 151–175. Dordrecht: Springer. Lee, Mina 2006 Overt focus movement and minimal information marking. Ph.D. dissertation, University of Southern California.

References

289

Lee, Youngjoo 2005 Exhaustivity as agreement: the case of Korean man ‘only’. Natural Language Semantics 13 : 169–200. Lekakou, Marika and Kriszta SzendrĘi 2012 Polydefinites in Greek: Ellipsis, close apposition and expletive determiners. Journal of Linguistics 48 : 107–149. Mahajan, Anoop Kumar 1990 The A/A-bar distinction and movement theory. Ph.D. dissertation, Massachusetts Institute of Technology. Maki, Hideki, Lizanne Kaiser, and Masao Ochi 1999 Embedded topicalization in English and Japanese. Lingua 109 : 1–14. Markman, Vita G. 2008 Pronominal copula constructions are what? Reduced specificational pseudo-clefts! In: Charles B. Chang and Hannah J. Haynie (eds.), Proceedings of the 26th West Coast Conference on Formal Linguistics, 366–374. Somerville, MA: Cascadilla Proceedings Project. Mathesius, Vilém 1964 On some problems of the systematic analysis of grammar. In: Josef Vachek (ed.), A Prague School Reader in Linguistics, 306–319. Bloomington, IN: Indiana University Press. May, Robert 1977 The grammar of quantification. Ph.D. dissertation, Massachusetts Institute of Technology. May, Robert 1985 Logical Form. Cambridge, MA: MIT Press. McCawley, James D. 1970 Where do noun phrases come from? In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 166–183. Waltham, MA: Ginn and Co. McCoy, Svetlana 2003 Connecting information structure and discourse structure through “Kontrast”: The case of colloquial Russian particles -to, že, and ved’. Journal of Logic, Language and Information 12 : 319–335. McGinnis, Martha 2008 A new look at Japanese and Korean scrambling. Toronto Working Papers in Linguistics 28 : 239–258. McNally, Louise 1998 Existential sentences without existential quantification. Linguistics and Philosophy 21 : 353–392. McNally, Louise 1998 On recent formal analyses of topic. In: Jonathan Ginzburg, Zurab Khasidashvili, Carl Vogel, Jean-Jacques Lévy, and Enric Vallduví (eds.), The Tbilisi Symposium on Language, Logic, and Computa-

290

References

tion: Selected Papers, 147–160. Stanford: Center for the Study of Language and Information Publications. Meinunger, André 1998 A monoclausal structure for (pseudo-)cleft sentences. In: Pius N Tamanji and Kiyomi Kusumoto (eds.), Proceedings of the North East Linguistic Society 28, 283–298. Amherst, MA: University of Massachusetts, Graduate Linguistics Student Association. Miyagawa, Shigeru 1997 Against optional scrambling. Linguistic Inquiry 281 : 1–25. Miyagawa, Shigeru 2006 On the ‘undoing’ property of scrambling: A response to Boškoviü. Linguistic Inquiry 37 : 607–624. Molnár, Valéria 2001 Contrast from a contrastive perspective. In: Ivana Kruijff-Korbayova and Mark Steedman (eds.), European Summer School in Logic, Language and Information 2001 Workshop on Information Structure, Discourse Structure and Discourse Semantics, 99–114. Molnár, Valéria 2002 Contrast – from a contrastive perspective. In: H. Hallelgard, Stig Johansson, Bergljot Behrens, and Cathrine Fabricius-Hansen (eds.), Information Structure in a Cross-Linguistic Perspective, 147–161. Amsterdam/New York: Rodopi. Molnár, Valéria 2006 On different kinds of contrast. In: Valéria Molnár and Susanne Winkler (eds.), The Architecture of Focus, 197–233. Berlin: Mouton de Gruyter. Montague, Richard 1974 The proper treatment of quantification in ordinary English. In: Richard H. Thomason (ed.), Formal Philosophy, 247–270. New Haven, C.T.: Yale University Press. Moro, Andrea 1997 The Raising of Predicates. Cambridge: Cambridge University Press. Munakata, Takashi 2006 Japanese topic-constructions in the minimalist view of the syntax semantics interface. In: Cedric Boeckx (ed.), Minimalist Essays, 115–159. Amsterdam: John Benjamins. Nagahara, Hiroyuki 1994 Phonological phrasing in Japanese. Ph.D. dissertation, University of California, Los Angeles. Nakanishi, Kimiko 2001 Prosody and information structure in Japanese: A case study of topic marker wa. In: Noriko Akatsuka, Susan Strauss, and Bernard Comrie (eds.), Japanese/Korean Linguistics 10, 434–447. Stanford: Center for the Study of Language and Information.

References

291

Nakanishi, Kimiko 2007 Prosody and scope interpretations of the topic marker wa in Japanese. In: Chungmin Lee, Matt Gordon, and Daniel Büring (eds.), Topic and Focus: Cross-linguistic Perspectives on Meaning and Intonation, 177–193. Dordrecht: Springer. Neeleman, Ad 1994 Complex predicates. Ph.D. dissertation, University of Utrecht. Neeleman, Ad and Tanya Reinhart 1998 Scrambling and the PF interface. In: Miriam Butt and Wilhelm Gueder (eds.), The Projection of Arguments, 309–353. Stanford: Center for the Study of Language and Information. Neeleman, Ad and Kriszta SzendrĘi 2004 Superman Sentences. Linguistic Inquiry 35 : 149–159. Neeleman, Ad and Elena Titov 2009 Focus, contrast, and stress in Russian. Linguistic Inquiry 40 : 514– 524. Neeleman, Ad, Elena Titov, Hans van de Koot, and Reiko Vermeulen 2009 A syntactic typology of topic, focus and contrast. In: Jereon Van Craenenbroeck (ed.), Alternatives to Cartography, 15–51. Berlin: Mouton de Gruyter. Neeleman, Ad and Hans van de Koot 2007 Theta theory. Ms. UCL. Neeleman, Ad and Hans van de Koot 2008 Dutch scrambling and the nature of discourse templates. The Journal of Comparative Germanic Linguistics 11 : 137–189. Neeleman, Ad and Hans van de Koot 2009 Scope Inversion. Ms., University College London. Neeleman, Ad and Hans van de Koot 2011 Information-structural restrictions on A’-scrambling. The Linguistic Review 27 : 365–385. Neeleman, Ad and Kriszta Szendröi 2004 Superman sentences. Linguistic Inquiry 35(1) : 149–159. Nilsen, Øystein 2003 Eliminating Positions: Syntax and Semantics of Sentential Modification. Ph.D. dissertation, Utrecht University. Nolan, Francis and Hildur Jónsdóttir 2001 Accentuation patterns in Icelandic. In: Wim A. van Dommelen and Thorstein Fretheim (eds.), Nordic Prosody: 8th Conference, Trondheim 2000, 187–198. Frankfurt, Berlin & New York: Peter Lang. Ntelitheos, Dimitrios 2004 Syntax of elliptical and discontinuous nominals. MA thesis, UCLA. O’Connor, J. D. and G. F. Arnold 1973 Intonation of Colloquial English, 2nd edition. London: Longmans.

292

References

Oh, Chisung 2007 Topic and focus constructions in spoken Korean. Ph.D. dissertation, University of Texas at Austin. Oshima, David Y. 2008 Morphological vs. phonological contrastive topic marking. Chicago Linguistics Society 41 : 371–384. Panagiotidis, Phoevos 2005 Determiner spreading as DP-predication. Ms. Cyprus College. Palmer, Frank Robert 1997 Negation and modality in the Germanic languages. In: Toril Swan and Olaf Jansen Westvik (eds.), Modality in Germanic Languages: Historical and Comparative Perspectives, 133–149. Berlin: Mouton de Gruyter. Percus, Orin 1997 Prying open the cleft. In: Kiyomi Kusumoto (ed.), Proceedings of the North East Linguistic Society 27, 337–351. Amherst, MA: University of Massachusetts, Graduate Linguistics Student Association. Pereltsvaig, Asya 2004 Topic and focus as linear notions: Evidence from Italian and Russian. Lingua 114 : 325–344. Perlmutter, David 1972 Evidence for shadow pronouns in French. In: Paul M. Perantueau, Judith N. Levi and Gloria C. Phares (eds.), The Chicago Which Hunt, 73–105. Chicago: Chicago Linguistic Society. Pierrehumbert, Janet 1980 The phonology and phonetic of English intonation. Ph.D. dissertation, Massachusetts Institute of Technology. Pinkham, Jessie and Jorge Hankamer 1975 Deep and shallow clefts. Chicago Linguistics Society 11 : 429–450. Polinsky, Maria and Eric Potsdam 2001 Long-distance agreement and topic in Tsez. Natural Language and Linguistic Theory 19 : 583–646. Portner, Paul 2007 Instructions for Interpretation as Separate Performatives. In: Kerstin Schwabe and Susanne Winkler (eds.), On Information Structure, Meaning and Form, 407–426. Amsterdam/Philadelphia: John Benjamins. Portner, Paul and Katsuhiko Yabushita 1998 The semantics and pragmatics of topic phrases. Linguistics and Philosophy 21 : 117–157. Portner, Paul and Katsuhiko Yabushita 2001 Specific indefinites and the information structure theory of topics. Journal of Semantics 18 : 271–297. Prince, Ellen F. 1978 A comparison of wh-clefts and it-clefts in discourse. Language 54 : 883–906.

References

293

Prince, Ellen F. 1981 Topicalization, Focus-Movement, and Yiddish-Movement: A Pragmatic Differentiation. Proceedings of the Seventh Annual Meeting of the Berkeley Linguistics Society, 249–264. Radford, Andrew 1975 Pseudo-relatives and the unity of subject-raising. Archivum Linguisticum 6 : 32–64. Rapoport, Tova 1987 Copular, nominal and small clauses: a study of Israeli Hebrew. Ph.D. dissertation, Massachusetts Institute of Technology. Reeve, Matthew 2010 Clefts. Ph.D. dissertation, University College London. Reeve, Matthew 2011 The syntactic structure of English clefts. Lingua 121 : 142–171. Reicher, Maria 2010 Nonexistent objects. Edward N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Fall 2010 Edition). http://plato.stanford.edu/ archives/fall2010/entries/nonexistent-objects/ Reinhart, Tanya 1981 Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27 : 53–94. Reinhart, Tanya 1983 Anaphora and Semantic Interpretation. London: Croom Helm. Reinhart, Tanya 1993 Elliptic conjunctions – non-quantificational LF. In: Asa Kasher (ed.), The Chomskyan Turn, 360–384. Cambridge, MA: Blackwell. Reinhart, Tanya 1995 Interface Strategies. Ms., Utrecht University. Reinhart, Tanya 1998 Wh-in-Situ in the framework of the Minimalist Program. Natural Language Semantics 6 : 29–56. Reinhart, Tanya 2006 Interface Strategies. Cambridge, MA: MIT Press. Repp, Sophie 2009 Defining ‘contrast’ as an information-structural notion in grammar. Lingua 120 : 1333–1345. Richards, Norvin 2001 A distinctness condition on linearization. Ms., Massachusetts Institute of Technology. Richards, Norvin 2008 Can A-scrambling reorder DPs? Ms., Massachusetts Institute of Technology. Rizzi, Luigi 1982 Issues in Italian syntax. Dordrecht: Foris.

294

References

Rizzi, Luigi 1986 Rizzi, Luigi 1990 Rizzi, Luigi 1997 Rizzi, Luigi 2004

Null objects in Italian and the theory of pro. Linguistic Inquiry 17 : 501–557. Relativized Minimality. Cambridge, MA: MIT Press. The fine structure of the left periphery. In: Liliane Haegeman (ed.), Elements of Grammar. Dordrecht: Kluwer.

Locality and left periphery. In: Adriana Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures, Volume 3, 223– 251. Oxford: Oxford University Press. Roberts, Craige 1996 Information structure in discourse: towards an integrated formal theory of pragmatics. In: Jae-Hak Yoon and Andreas Kathol (eds.), OSU Working Papers in Linguistics 49: Papers in Semantics, 91–136. Rochemont, Michael S. 1986 Focus in Generative Grammar. Amsterdam: John Benjamins. Rochemont Michael S. and Peter W. Culicover 1990 English Focus Constructions and the Theory of Grammar. Cambridge: Cambridge University Press. Romero, Maribel 2005 Concealed questions and specificational subjects. Linguistics and Philosophy 28 : 687–737. Rooth, Mats 1985 Association with focus. Ph.D. dissertation, University of Massachusetts, Amherst. Rooth, Mats 1992 A theory of focus interpretation. Natural Language Semantics 1 : 75–116. Rooth, Mats 1999 Association with focus or association with presupposition? In: Peter Bosch and Rob van der Sandt (eds.), Focus: Linguistic, Cognitive, and Computational Perspectives, 232–244. Cambridge: Cambridge University Press. Ross, John R. 1972 Act. In: Donald Davidson and Gilbert Harman (eds.), Semantics of Natural Language, 70–126. Dordrecht: Reidel. Rullmann, Hotze 1988 Referential dependency. M.A. thesis, University of Groningen. Ruys, Eddy 2001 Dutch Scrambling and the Strong-Weak Distinction. Journal of Comparative Germanic Linguistics 4 : 39–67. Sæbø, Kjell Johan 1997 Topic, Focus and Quantifier Raising. In: Paul Dekker, Martin Stokhof, and Yde Venema (eds.), Proceedings of the 11th Amsterdam

References

295

Colloquium, 67–72. Amsterdam: Institute for Logic, Language and Computation. Saito, Mamoru 1985 Some asymmetries in Japanese and their theoretical implications. Ph.D. dissertation, Massachusetts Institute of Technology. Saito, Mamoru 1989 Scrambling as semantically vacuous A’-movement. In: Mark Baltin and Anthony Kroch (eds.), Alternative Conceptions of Phrase Structure, 182–200. Chicago: University of Chicago Press. Saito, Mamoru 1992 Long distance scrambling in Japanese. Journal of East Asian Linguistics 1 : 69–118. Saito, Mamoru 2003 A derivational approach to the interpretation of scrambling chains. Lingua 113 : 481–518. Sakai, Hiromu 1994 Complex NP constraint and case conversion in Japanese. In: Masuru Nakamura (ed.), Current Topics in English and Japanese: Proceedings of the Second English/Japanese Comparative Syntax Workshop, 179–203. Tokyo: Hituzi Syoboo. Samek-Lodovici, Vieri 2005 Prosody-syntax interaction in the expression of focus. Natural Language and Linguistic Theory 23 : 687–755. Samek-Lodovici, Vieri 2009 Topic, focus, and background in Italian Clauses. In: Andreas Dufter and Daniel Jacob (eds.), Focus and Background in Romance Languages. (Studies in Language Companion Series 112.), 333–358. Amsterdam: John Benjamins. Sánchez, Liliane 2010 The Morphology and Syntax of Focus and Topic: Minimalist Inquiries in the Quechua periphery. Amsterdam: John Benjamins. Sauerland, Uli 2000 Syntactic economy and quantifier raising. Ms., University of Tübingen. Sauerland, Uli 2005 DP is not a scope island. Linguistic Inquiry 36 : 303–314. Schlenker, Philippe 2003 Clausal equations: a note on the connectivity problem. Natural Language and Linguistic Theory 21 : 157–214. Schwarz, Bernhard and Rajesh Bhatt 2006 Light negation and polarity. In: Raffaella Zanuttini, Héctor Campos, Elena Herburger, Paul H. Portner (eds.), Cross-linguistic Research in Syntax and Semantics: Negation, Tense and Clausal Architecture, 175–198. Washington, D C: Georgetown University Press.

296

References

Schwarzchild, Roger 1999 Givenness, avoidF and other constraints on the placement of accent. Natural Language Semantics 7 : 141–177. Scott, Gary-John 2002 Stacked Adjectival Modification and the Structure of Nominal Phrases. In: Guglielmo Cinque (ed.), Functional Structure in DP and IP: The Cartography of Syntactic Structures, Volume 1, 91–120. Oxford/New York: Oxford University Press. Sekerina, Irina 1997 The syntax and processing of split scrambling constructions in Russian. Ph.D. dissertation, The City University of New York. Selkirk, Elisabeth 1984 Phonology and Syntax: The Relation between Sound and Structure. Cambridge, MA: MIT Press. Selkirk, Elisabeth 1995 Sentence prosody: intonation, stress, and phrasing. In: John A. Goldsmith (ed.), The Handbook of Phonological Theory, 550–569. Cambridge, MA/Oxford, U. K.: Blackwell. Selkirk, Elisabeth 2008 Contrastive focus, givenness and the unmarked status of “discourse new”. Acta Linguistica Hungarica 55 : 331–346. Sgall, Petr 1972 Topic, Focus and the Ordering of Elements of Semantic Representations. Philologica Pragensia 15 : 1–14. Shibatani, Masayoshi 1990 The Languages of Japan. Cambridge: Cambridge University Press. Shimojo, Mitsuaki and Hye-Won Choi 2001 On asymmetry in topic marking: The case of Japanese wa and Korean nun. Chicago Linguistics Society 36 : 454–467. Šimík, Radek and Marta Wierzba 2012 Some notes on Czech scrambling. IS-Syntax seminar, Linguistics Department, Potsdam University, 18 January 2012. Slioussar, Natalia 2007 Grammar and information structure: a study with reference to Russian. Ph.D. dissertation, Universiteit Utrecht and St Petersburg State University. Smits, Rik 1989 Eurogrammar: the relative and cleft constructions of the Germanic and Romance languages. Dordrecht: Foris. Starke, Michal 2001 Move dissolves into merge: A theory of locality. Ph.D. dissertation, Université de Genève. Steedman, Mark 2000 Information structure and the syntax-phonology interface. Linguistic Inquiry 31 : 649–689.

References

297

Stepanov, Arthur 1998 On wh-fronting in Russian. In: Pius N. Tamanji and Kiyomi Kusumoto (eds.), Proceedings of the North East Linguistic Society 28, 453–467. Amherst, MA: University of Massachusetts, Graduate Linguistics Student Association. Strawson, Peter F. 1964 Identifying reference and truth-values. Theoria 30 : 96–118. Svenonius, Peter 1996 The optionality of particle shift. Working Papers in Scandinavian Syntax 57 : 47–75. Szabolcsi, Anna 1981 The semantics of topic–focus articulation. In Jeroen Groenendijk, Theo Janssen, and Martin Stokhof (eds.), Formal Methods in the Study of Language, 513–41. Amsterdam: Mathematical Centre. SzendrĘi, Kriszta 2002 Stress-focus correspondence in Italian. In: Beyssade, C., Reineke Bok-Bennema, Frank Drijkoningen, Paola Monachesi (eds.), Romance languages and linguistic theory 2000. Selected papers from ‘Going Romance’ 2000, Utrecht, 30 November–2 December, 287– 303. Amsterdam: John Benjamins. SzendrĘi, Kriszta 2003 A stress-based approach to the syntax of Hungarian focus. Linguistic Review 20 : 37–78. Tada, Hiroaki 1993 A/A-bar partition in derivation. Ph.D. dissertation, Massachusetts Institute of Technology. Taglicht, Josef 1984 Message and Emphasis: On Focus and Scope in English. London: Longman. Takahashi, Shoichi 2003 More than two quantifiers. In: Makoto Kadowaki and Shigeto Kawahara (ed.), Proceedings of the North East Linguistic Society 33, 405–424. Amherst, MA: University of Massachusetts, Graduate Linguistics Student Association. Taraldsen, Knut Tarald 1981 The theoretical interpretation of a class of marked extractions. In: Adriana Belletti, Luciana Brandi and Luigi Rizzi (eds.), Theory of Markedness in Generative Grammar, 475–616. Pisa: Scuole Normale Superiore. Tateishi, Koichi 1994 The Syntax of Subjects. Stanford: Center for the Study of Language and Information. Titov, Elena 2007 Pragmatic effects on Russian word (dis)order: Evidence from syn-

298

References

tactic and semantic reconstruction. M.A. thesis, University College London. Tomioka, Satoshi 2007 Pragmatics of LF intervention effects: Japanese and Korean whinterrogatives. Journal of Pragmatics 39 : 1570–1590. Tomioka, Satoshi 2010 Contrastive topics operate on Speech Acts. In: Malte Zimmermann and Caroline Féry (eds.), Information Structure: Theoretical, Typological, and Experimental Perspectives, 115–138. Oxford: Oxford University Press. Torregrossa, Jacopo 2012 Encoding Topic, Focus and Contrast: Informational Notions at the Interfaces. PhD dissertation, Verona. Truckenbrodt, Hubert 1999 On the relation between syntactic phrases and phonological phrases. Linguistic Inquiry 30 : 219–55. Truswell, Robert 2005 Non-restrictive Adjective Interpretation and Associacion with Focus. In: Richard Ashdowne and Thomas Finbow (eds.), Oxford Working Papers in Linguistics, Phonetics and Philology, 9 : 133–154. Truswell, Robert 2009 Attibutive Adjectives and Nominal Templates. Linguistic Inquiry 40(3) : 525–533. Tsoulas, George 1999 On long-distance scrambling in Korean. In: Harvard Studies in Korean Linguistics 8 : 217–302. Ueyama, Ayumi 1998 Two types of dependency. Ph.D. dissertation, University of Southern California. Vallduví, Enric and Elisabet Engdahl 1996 The linguistic realization of information packaging. Linguistics 34 : 459–519. Vallduví, Enric and Maria Vilkuna 1998 On rheme and kontrast. On rheme and kontrast. In: Peter W. Culicover and Louise McNally (eds.), Syntax and Semantics 29: The Limits of Syntax, 79–108. San Diego: Academic Press. Vallduví, Enric 1992 The Informational Component. New York: Garland. Vanden Wyngaerd, Guido 1989 Object shift as an A-movement rule. MIT Working Papers in Linguistics 11 : 256–271. Van Hoof, Hanneke 2003 The rise in the rise-fall contour: Does it evoke a contrastive topic or a contrastive focus? Linguistics 41 : 515–563. Van Riemsdijk, Henk 1978 A Case Study in Syntactic Markedness. Lisse: Peter de Ridder.

References

299

Vermeulen, Reiko 2009 Topic, contrast and contrastive topics in Japanese. Reiko Vermeulen and Ryosuke Shibagaki (eds.), Proceedings to the Workshop on Formal Altaic Linguistics 5, 361–372. Vermeulen, Reiko 2011 Non-topical wa-phrases in Japanese. In: Rafaella Folli and Christiane Ulbrich (eds.), Interfaces in Linguistics: New Research Perspectives, 135–148. Oxford: Oxford University Press. Vermeulen, Reiko To appear ‘Topic’ and ‘contrast’: A comparative approach to Japanese and Korean. In: Bjarke Frellesvig and Peter Sells (eds.), Japanese/Korean Linguistics 20. Stanford: Center for the Study of Language and Information. Vermeulen, Reiko To appear b On the position of topics in Japanese. The Linguistic Review 30(1). Vikner, Sten 1995 Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press. von Fintel, Kai 2004 Would you believe it? The King of France is back! (Presuppositions and truth-value intuitions.) In: Anne Bezuidenhout and Marga Reimer (eds.), Descriptions and Beyond: an Interdisciplinary Collection of Essays on Definite and Indefinite Descriptions and other Related Phenomena, 315–341. Oxford: Oxford University Press. Wagner, Michael 2005 Prosody and Recursion. Ph.D. dissertation, Massachusetts Institute of Technology. Wagner, Michael 2007 Givenness and locality. In: Jonathan Howell and Masayuki Gibson (eds.), Proceedings of Semantics and Linguistic Theory 16, 295–312. Ithaca: CLC Publications. Wagner, M. 2008a. A compositional analysis of contrastive topics. In: Muhammad Abdurrahman, Anisa Schardl, and Martin Walkow (eds.), Proceedings of the North East Linguistic Society 38, 415–428. Amherst, MA: University of Massachusetts, Graduate Linguistics Student Association. Wagner, Michael 2008b Contrastive topics decomposed. Ms., Cornell University. Wagner, Michael 2012 Focus and Givenness: A Unified Approach. In: Ivona Kuþerová and Ad Neeleman: Information Structure. Contrasts and Positions. Cambridge University Press

300

References

Wallenberg, Joel C. 2009 Antisymmetry and the conservation of c-command. Ph.D. dissertation, University of Pennsylvania. Wallenberg, Joel C. To appear Antisymmetry and Heavy NP Shift across Germanic. In: Theresa Biberauer and George Walkden (eds.), Peripheries Past and Present: The Interaction of Information Structure, Prosody and Syntax. Oxford: Oxford University Press. Watanabe, Akira 1992 Subjacency and S-Structure movement of wh-in-situ. Journal of East Asian Linguistics 1 : 255–291. Williams, Edwin 1981 Argument structure and morphology. The Linguistic Review 1 : 81–114. Williams, Edwin 1989 The anaphoric nature of theta-roles. Linguistic Inquiry 20 : 425–256. Williams, Edwin 1994 Thematic Structure in Syntax. Cambridge, MA: MIT Press. Williams, Edwin 1997 Blocking and anaphora. Linguistic Inquiry 28 : 577–628. Williams, Edwin 2003 Representation Theory. Cambridge, MA: MIT Press. Williams, Edwin 2006 Circumstantial evidence for dative shift. In: Hans Broekhuis, Norbert Corver, Riny Huybregts, Ursula Kleinhenz, and Jan Koster (eds.), Organizing Grammar: Linguistic Studies in Honor of Henk van Riemsdijk, 661–668. Berlin: Mouton de Gruyter. Wirth, Jessica 1978 The derivation of cleft sentences in English. Glossa 12 : 58–81. Wurmbrand, Susi 2008 Word order and scope in German. Ms., University of Connecticut, Storrs. Yokoyama, Olga T. 1986 Discourse and Word Order. Amsterdam: John Benjamins. Zimmermann, Malte 2008 Contrastive Focus and Emphasis. Acta Linguistica Hungarica 55 : 347–360. Zubizarreta, Maria Luisa 1998 Prosody, Focus and Word Order. Cambridge,Mass.:MIT Press. Zwart, Jan-Wouter C 1993 Dutch syntax: A minimalist approach. Ph.D. dissertation, University of Groningen. Zwarts, Joost 1993 X-bar Syntax, S-bar Semantics. Ph.D. dissertation, University of Utrecht.

Index ±VLJQDWXUH± $DFFHQW± $EHOV.IQ $ERK( $EUXViQ0 DFFRPPRGDWLRQ $FNHPD3IQ $GDPHF3± DGMHFWLYHUH RUGHULQJ $GJHU'IQ DGMDFHQF\ DGYHUELDOV $NPDMLDQ$ DOWHUQDWLYHV± DOWHUQDWLYHVFRPSRQHQW DVDVVHUWHGRUSUHVXSSRVHG± µORFDO¶DOWHUQDWLYHV± SUDJPDWLFVHWRI± DOWHUQDWLYHVHPDQWLFV DQDSKRUDUHVROXWLRQDVDWHVWIRUWRSLFKRRG ± $QGUHZV$ $R\DJL+ DSSRVLWLRQFORVHDSSRVLWLRQ ± DUJXPHQWSURPLQHQFH $ULHO0 $UQROG*) $UUHJXL8UELQD. DVIRUDVDWHVWIRUWRSLFKRRG  DVVHUWLRQ±± DVVHUWLRQRSHUDWRUIQ $VWUXF/IQ $WODV- %DFFHQW±IQIQ EDFNJURXQGIQ± ± %DLO\Q- %DNHU0 %DOWD[H& %DUNHU&

%DUZLVH- EDVHJHQHUDWLRQ %HDYHU'±IQ %HFN6IQ %HOOHWWL$ %HQLQFD3 %KDWW5 %LDQFKL9IQ ELQGLQJ± DQDSKRULF± %áDV]F]DN-IQ %REDOMLN-±± %ROH:HVW&KDGLF  %ROLQJHU' %RãFRYLü= %URG\0 %UXHQLQJ%IQIQ %UXQ' %UXQHWWL/ %U\]JXQRYD($ %ULQJ'IQ  &DPSRV+ FDUWRJUDSK\± &DWDODQ &KDIH:/ &KDUORZ6 &KHQ$ &KR-+IQ &KRH+6IQ  &KRL+:  &KRPVN\1IQ &LQTXH*± &ODUN%±IQ FODXVHLQLWLDOSRVLWLRQ±± FOHIWVHQWHQFHVII FORVHGVHW FROOHFWLYHUHDGLQJ FRPPHQW± FRPSOH[DUJXPHQWIRUPDWLRQ

302

Index

FRQGLWLRQ&± &RQVWDQW1 FRQWH[W FRQVWUXDEOH FRQWUDVW±± DFFRPPRGDWLRQ± DQGVFRSH± FDQFHOODELOLW\ FRQWUDVWLYHIRFXV±± ±± ± FRQWUDVWLYHWRSLF± ±IQ GRPDLQRIFRQWUDVWRUVFRSHRI FRQWUDVW± VHHDOVRVFRSH FRSXOD± SUHGLFDWLRQDOFRSXODUVHQWHQFH ± VSHFL¿FDWLRQDOFRSXODUVHQWHQFH ± &RRSHU5 &RUEHWW*IQ FRUHIHUHQFH± FRUUHFWLRQFRQWH[W DVDWHVWIRUFRQWUDVWLYHIRFXV  DVDWHVWIRUFRQWUDVWLYHWRSLF &66FRQGLWLRQRQVFRSHVKLIW  DQGLQGH[SHUFRODWLRQ± DQGLQWHUDFWLRQEHWZHHQVFRSH VKLIWDQG$¶VFUDPEOLQJ DQGUHVWULFWLRQVRQTXDQWL¿HU VFRSH± DQGWRSLFIRFXVLQWHUDFWLRQ± &XOLFRYHU3± &]HFK 'DQHã) GHDFFHQWXDWLRQ GH¿QLWHGHVFULSWLRQ 'HKp1± 'HODKXQW\*IQ 'HODLV5RXVVDULH(IQ 'HQ'LNNHQ0IQ 

GHVWUHVVLQJ± DFTXLVLWLRQRI ,FHODQGLF GLVFRQWLQXRXV13WRSLFDOL]DWLRQ GLVFRXUVHDQDSKRULFLW\VHHROGLQIRUPDWLRQ GLVMXQFWLYHTXHVWLRQFRQWH[W GRPDLQRIFRQWUDVWVHHXQGHUFRQWUDVW 'UHWVNH), GXSOLFDWLRQ 'XWFK±IIIQ ±± (EHUW& e.LVV.± ± (PRQGV-IQ (QGR