The Grammar of Focus (Linguistik Aktuell Linguistics Today)

THE GRAMMAR OF FOCUS LINGUISTIK AKTUELL This series provides a platform for studies in the syntax, semantics, and pra...

Author: Georges Rebuschi | Laurice Tuller

34 downloads 1204 Views 2MB Size Report

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

DOWNLOAD PDF

THE GRAMMAR OF FOCUS

LINGUISTIK AKTUELL This series provides a platform for studies in the syntax, semantics, and pragmatics of the Germanic languages and their historical developments. The focus of the series is represented by its German title Linguistik Aktuell (Linguistics Today). Texts in the series are in English. Series Editor Werner Abraham Germanistisch Instituut Rijksuniversiteit Groningen Oude Kijk in ’t Jatstraat 26 9712 EK Groningen The Netherlands E-mail: [email protected] Advisory Editorial Board Guglielmo Cinque (University of Venice) Günther Grewendorf (University of Stuttgart) Liliane Haegeman (University of Geneva) Hubert Haider (University of Salzburg) Christer Platzack (University of Lund) Ian Roberts (University of Stuttgart) Ken Safir (Rutgers University, New Brunswick NJ) Höskuldur Thráinsson (University of Iceland, Reykjavik) Lisa deMena Travis (McGill University) Sten Vikner (University of Stuttgart) C. Jan-Wouter Zwart (University of Groningen)

Volume 24 Georges Rebuschi and Laurice Tuller (eds) The Grammar of Focus

THE GRAMMAR OF FOCUS Edited by

GEORGES REBUSCHI University of Paris III - Sorbonne nouvelle

LAURICE TULLER University of Tours

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM / PHILADELPHIA

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.

Library of Congress Cataloging-in-Publication Data The Grammar of Focus / edited by Georges Rebuschi, Laurice Tuller. p. cm. -- (Linguistik aktuell / Linguistics today, ISSN 0166-0829; v. 24) Includes bibliographical references and index. 1. Rebuschi, Georges. II. Tuller, Laurice. III. Series: Linguistik aktuell ; Bd. 24. P299.F63673 1999 415--dc21 isbn 90 272 2745 4 (eur) / 1 55619 908 2 (us) (Hb; alk. paper)

99-22380 CIP

© 1999 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O.Box 75577 · 1070 AN Amsterdam · The Netherlands John Benjamins North America · P.O.Box 27519 · Philadelphia PA 19118-0519 · USA

Table of Contents

The Grammar of Focus: An Introduction Georges Rebuschi and Laurice Tuller Aspects of the Syntax of Focus in Portuguese Manuela Ambar Bound Focus or How can Association with Focus be Achieved without Going Semantically Astray? Josef Bayer Are There Cleft Sentences in French? Anne Clech-Darbon, Georges Rebuschi & Annie Rialland

1 23

55 83

Focus Structure and Scope Nomi Erteschik-Shir

119

The Interaction between Focus and Tone in Bantu Larry Hyman

151

The Syntax of the P-Focus Position in Turkish Sarah D. Kennelly

179

Word Order and Focus Positions in Universal Grammar Ayesha Kidwai

213

Focus in Wolof: A Study of What Morphology May Do to Syntax Alain Kihm

245

Focus in Somali Jacqueline Lecarme

275

Focus in Basque Jon Ortiz de Urbina

311

vi

TABLE OF CONTENTS

Focus and Arabic Clefts Jamal Ouhalla

335

Index

361

List of Contributors

365

The Grammar of Focus An Introduction* Georges Rebuschi University of Paris III

Laurice Tuller University of Tours

“There is no reason to suppose that a satisfactory characterization of focus and presupposition can be given in purely grammatical terms, but there is little doubt that grammatical structure plays a part in specifying them.” Chomsky (1970: footnote 22)

1.

Preliminary Remarks

The grammar of focus has been studied in generative grammar from its inception. It has been the subject of intense, detailed cross-linguistic investigation for over 20 years, particularly within the Principles and Parameters framework, resulting in a large body of empirical and theoretical contributions which cover a wide variety of languages and types of focus. It is appropriate at this point, therefore, to take stock. Appraisal at this particular point is all the more legitimate because it comes at a time of general evaluation of the results of the profound activity that has characterized the Principles and Parameters framework. This general assessment has produced a radical new direction within that framework. How does this bear on our understanding of the grammar of focus? How does the grammar of focus help us determine the validity of this new vision of syntactic theory? What, old or new, focus problems remain to be solved? The core innovation of the reformulation of syntactic theory proposed by the Minimalist Program is that, aside from economy principles, syntactic structures are entirely determined by constraints on the two interface levels PF and LF. The

2

GEORGES REBUSCHI AND LAURICE TULLER

number of levels of syntactic representation is reduced to this minimal number because this is what is imposed by the interaction between the language faculty and other cognitive systems: the ‘articulatory-perceptual’ system on the one hand and the ‘conceptual-intentional’ system on the other. The constraints themselves are minimal in that they are limited in scope to those imposed by the requirements of these performance systems. Since the form of linguistic representations at the interfaces is relevant only to the performance systems (the cognitive systems which touch the language faculty), the minimal assumption is that the output of the language faculty contains what is relevant to these systems, and nothing else. All syntactic computation is reduced on this basis to checking of features so that the only elements remaining in the interface representations are those that are interpretable to the relevant cognitive system. All features then are either plus or minus interpretable to the performance systems. So, for example, categorical features are presumably used by the ‘conceptual-intentional’ system and thus are LF [+interpretable], whereas Case features are not and thus are LF [−interpretable]. Phonological features, on the other hand, are clearly PF [+interpretable], etc. One of the consequences of this (ambitious) research program is therefore that the pertinence to the performance systems of various features that have been argued to be part of the language faculty must be carefully (re)evaluated. The feature [+Focus] is one of these. Indeed, the grammatical phenomena related to focus interpretation constitute a particularly fertile domain of investigation. Focus has a discourse function, but is unquestionably a product of grammar. Its grammatical manifestation may be phonological, morphological, or syntactic, and there is evidence in many languages that it is quantificational. It would thus seem to be related to various “components” of grammar. The question of where the processes leading to its interpretation take place is thus central. Focus, from this perspective, provides a test case for the proper characterization of the interface levels, whether or not the specific tenets of Minimalism are accepted. An appraisal of the grammar of focus is the purpose of this collection of papers, which grew out of the 1996 Paris workshop of the same name. This introductory chapter aims to provide an outline for this assessment. We start with an overview of the evolution of the study of focus in generative grammar. The reader will observe that we do not attempt to impose a definition of focus here. Indeed, it is our feeling that the syntactic feature focus may be interpreted differently in different languages, and, furthermore that its status is largely dependent on the particular state of syntactic theory within which it is embedded.


3

Our goal here is to illustrate some of this theory dependent variation by placing the study of focus in its historical context within generative grammar. This overview is followed by a recapitulation of the principal issues associated with focus that have been raised over the years, how they are addressed by the contributors to this volume, and what we believe these contributions have helped to forward. The filling in of this sketch is what the remaining chapters of this book are all about.

2.

Overview of the Evolution of the Study of Focus in the Generative Tradition1

2.1 If we momentarily accept to define ‘focus’ as ‘prosodic salience at the sentential level’, we can safely say that, although the word itself was not used, the very notion of focus was already on the agenda in LSLT (Chomsky [1955] 1975: 445–448), where examples like (1) were discussed: (1)

a. He ⁄ come b. *He did/does 

As Chomsky notes (on p. 446), do “must occur with heavy stress”, a property it does not usually have in other positions (i.e. in interrogative or negative sentences), and which it does not share with other auxiliaries, which do not  carry such heavy stress — at least if  , whose status is declared “unclear” (ibid., footnote 11) is left aside; hence the introduction of a morpheme “A [for] accentuation” among the primes, which, if selected, will trigger Do- just as not will, for instance. (In Chomsky 1957, §7.1, A, despite the wary note cited above, is redefined as “a morpheme A of contrastive stress”.) Since the main concern was form, and given the state of the art in linguistic semantics at that time, it was only too natural not to ask what possible  A(c) had — obviously, it did not denote anything. However, it is now possible to ask rather, “What is the contribution of A(c) to the overall meaning of the sentence?” — and to answer, “It underscores whatever is in its scope, thereby excluding alternatives” (this is what will later on be known as the E L effect). If a lexical item is in its scope, then other lexical items will be excluded, but if the item in its scope is the (future) functional head I0, then it is the positive character of the assertion that is underlined, hereby excluding doubt or negation (see Selkirk 1995 for a fairly similar characterization). Hence A(c)

4


following a modal auxiliary will be ambiguous, since either the lexical content of that aux., or the position it happens to occupy, can constitute its scope, as shown by (2c), which can be a response to either (a) or (b): (2)

a. b. c.

Peter can do it. Peter won’t do it. (Yes,) he  (do it)

In other words, it is not because A(c) is automatically present alongside do in positive assertions, and only optionally present with other inflected auxiliaries, that it does not have the same semantic import in both cases — on the contrary, this approach gives independent support to the reality of the morpheme A(c), which might otherwise look like an ad hoc contrivance.2 2.2 Fifteen years later, the question of the – articulation (henceforth PFA) of English sentences, a more general phenomenon than emphasis on the assertion itself, was brought up when the controversy over the relevance of surface structure for semantic interpretation was at its height. At this time (prior therefore to the postulation of LF and PF), it was the surface structure of a sentence that contained its “intonation center” (what do in (1a) above undoubtedly is), and the question was whether its PFA was determined by the latter, or whether it was determined at a deeper level of representation.3 Using examples like (3), Chomsky (1970) showed that it was simpler to derive its PFA from the surface structure, for two distinct reasons. (3)

He was [awarned to [b look out for [g an ex-convict with a [d red [e ]]]]]

For one thing, the independently established bracketing of (3) allows one to compute the focus of the sentence, whose intonation center is borne by the last lexical item, as corresponding to just  constituent (identical to or) containing that lexical item — what will come to be known as the   problem.4 It directly follows that “the focus is [one of] the phrase[s] containing the intonation center, and the presupposition is determined by replacement of the focus by a variable” (C [1970] 1972: 91). Any deep-structure based account would, needless to say, require  distinct partitionings in the case of sentence (3). Second, Chomsky asks what sort of “deep structure” could formally incorporate the PFA, and suggests (4) (his (74)), where F (focus) and P (presup-


5

position) “are arbitrary structures and S′ functions as the initial symbol of the categorial component of the base”: (4)

S → S′ F P

A “filtering rule” would be then necessary, that would make sure that “the focus and presupposition, as determined from surface structure, are identical with F and P”. For instance, if red is the intonation center, and the constituent [a red shirt] appears under F of (4), the filter will rule the sentence out. But, of course, such a device would be totally redundant, and thus unacceptable.5 The mechanism described and criticized here crucially relies on an implicit restriction barring the possibility for just any PS rule to contain an optional abstract morpheme  analogous to the A(c) morpheme. Apart from the ensuing risk of overgeneration (which could be handled by performance considerations), a natural objection to such a device could simply consist in noting that there is no other formative that can appear absolutely anywhere. The distinction between focus and presupposition can also be seen from a slightly different perspective, with the focused constituent interpreted as “the predicate of [a] dominant sentence” as in the paraphrase (5b) of (5a) (adapted from Chomsky’s (42) and (44)): (5)

a. b.

John writes poetry in his . The place where John writes poetry is in his study

However, (5b) cannot be a  source for (5a), even if its focus projects no higher than the adverbial phrase,6 since there is no independent evidence that (5a) consists of more than  clause ([1970] 1972: 91); the idea that a focused item can thus be regarded as the predicate of an identificational sentence will however be exploited from an  point of view later on.7 2.3 A slightly different (and technically somewhat more sophisticated) approach is developed by Jackendoff (1972, Chapter 6).8 He starts from the following principle (his 6.58): (6)

If a phrase P is chosen as the focus of a sentence S, the highest stress in S will be on the syllable of P that is assigned highest stress by the regular stress rules.

This explains his recourse to “one artificial construct […]: a syntactic marker F which can be associated with any node in the surface structure” (p. 240). Here

6


lies the source of the (post-surface) “focus structure” which will be utilized by various scholars (Selkirk 1986; Ertechik-Shir, this vol.), and which serves as input both to the phonology and to the semantic representation on the sentence. On the phonological side of the description, the Emphatic Stress Rule (7) (his 6.67) assigns   to the most salient syllable of the phrase PF of (6): (7)

V → [emph stress] / [X [1stress] Y]F

and the output will “not weaken on successive cycles, as do other stresses” (241–2). On the semantic side, the derivation proceeds as follows (p. 245). First, [PresuppS(x)] is formed (as in C 1970) by replacing the Focus (the surface material dominated by F) by a variable in the sentence. Next a  , defined as “the set of values which, when substituted for x in [PresuppS(x)] yield a true proposition”: [lx PresuppS(x)] is built, which must have the (pragmatically) determined property of being for instance “under discussion”. Finally, the “assertion of a declarative sentence claims that the focus is a member of the presuppositional set”: (8)

Focus ∈ lx PresuppS(x)

Belonging to a set and being identical to the referent of a definite expression are clearly two different things: looking at (5a) again, according to Chomsky’s paraphrase, there’s only one place where John writes poetry, namely, his study, whereas according to Jackendoff’s theory, the study will be one among several such places; the opposition between Rooth’s and Krifka’s semantics of focus may well have its roots here (see Pulman 1997 for a recent review of semantic approaches to focus.) 2.4 In ‘Conditions on Rules of Grammar” (C 1976), a “rule ” is introduced which “informally” yields the “partially developed logical forms” (10a,b) when applied to (9a,b) (Chomsky’s (115–118)): (9)

a. b.

Bill likes J B likes John

(10)

a. b.

the x such that Bill likes x — is John the x such that x likes John — is Bill

The semantic intuition of the 1970 paper (see (5) above) is kept intact. More interestingly, the same abstract structure is used, in association with the principle (11) (Chomsky’s (105)):


(11)

7

A variable cannot be the antecedent of a pronoun to its left

to rule out a coreferential reading of he and John in (12)   NP [John]    : (12)

The woman he loved betrayed John

Why this should be is clear; if the rule applies, the ensuing logical form will be (13), where (11) will prevent any rule from replacing he by x: (13)

the x such that the woman he loved betrayed x — is John

On the contrary, if it is the verb that bears the focal accent, the logical form of (13) will be quite different: crucially, it will contain no variable corresponding to the object NP, so that coreference will be permitted. 2.5 Guéron’s (1980) article is the first widely acknowledged generative study that stressed the difference between contrastive and non-contrastive focus in (14a,b), (her (70)): (14)

a. b.

Georges loves M (contrastive or noncontrastive) G loves Martha (contrastive only)

Guéron also distinguishes between logical and intonational foci; the logical focus is the last argument in the c-command domain of the verb (or else the VP), so that when the two foci do not coincide, the contrast exemplified in (14b) automatically follows.9 Consider now the following sentences (after her (115a)): (15)

a. b.

His wife mistreats J His wife  John

In (a), John (like Martha in (14a)), can be a non-contrastive focus. To account for the impossibility of coindexing his and John here, Guéron’s hypothesis is that this effect is not due to sentence grammar,10 but arises from a clash in discourse semantics: the semantic function of unmarked focus is to introduce a new entity in the universe of discourse; but if John is such a new entity, its coindexation with his will result in a contradiction, since his can only represent old information (on the other hand, if the verb is the intonational focus, then John is old information, whence its ability to function as the possessive’s antecedent).11

8


Guéron does not address the question of WCO effects as induced in the  interpretation of (15a) in her paper. Two analyses seem possible; on the one hand, being contrastive, the NP John could be considered as being “recycled” as new information,12 even if it had been pronounced before, in which case the argument that precedes would directly carry over to this case; on the other hand, we could also suppose that Chomsky’s 1976 rule of Focus (now restricted to contrastive foci) would take care of that reading, with the LF adjunction of John to S mimicking May’s QR (see below). 2.6 Very little heed is paid to focus in Chomsky (1981): the “rule of focus” still belongs, along with “the rule of quantifier movement” and “the LF rule of whmovement” to the list of abstract, post-s–s, instances of Move-a. However, its specific output is slightly different, as shown by the association of the s–s (15a) and the LF (16b) (196: (34v,vi); see also p. 238: (20)–(22)):13 (16)

a. b.

his mother loves J (“J with focal stress”) for x = John, his mother loves x

In fact, (16b) can be regarded as too far away from s–s to be the real LF, since the syntactic object “trace” is replaced by the occurrence of a variable x; following Koopman & Sportiche’s (1982) influential paper (in particular, p. 155 (39)) we could replace (16b) by (17), which is at least a necessary intermediary step towards the former anyhow: (17)

Ji [[his mother] [loves ti]]

Given K&S’s own definition of ‘variables’, ti will be interpreted as one, thereby disallowing coindexation between his and the raised focused phrase, either as a consequence of their Bijection Principle or as a consequence of the Leftness Principle (11) above. The crucial issue now becomes whether (16b) is the correct gloss for (17). If, following Guéron’s tack, we do not treat non-contrastive focus by adjunction at LF, that may well not be the case: in order to avoid reading (17) as a mere case or ordinary l-abstraction over the subject, it is now possible to interpret the raised NP as carrying some quantification along, as in: (18)

for x = John    x = John [his mother loves x]


9

Radically distinguishing between contrastive and non-contrastive focus and allowing only the former to be treated by Raising at LF has at least one merit: the abstract movement thus postulated has a visible counterpart in some natural languages, among which Hungarian. Thus, Horvath (1981/86), Kiss (1981), and Szabolcsi (1981) agree on at least one empirical fact, namely that, in this language, an item left-adjacent to the finite verb and prosodically salient with respect to it must be interpreted as contrastive focus (i.e. exhibiting   properties), and that any element that is prosodically salient must immediately precede the (correlatively destressed) finite verb and be interpreted as a contrastive focus. Kiss and Horvath also agree that this left adjacent position is (for Kiss) or may be (for Horvath) filled by a movement transformation, and that wh-phrases have priority over others to fill it.14 However, the identification of this position was (and still is) a much debated issue. 2.7 The Government and Binding research program was in full swing in the mid-eighties, with 1986 a landmark. As far as focus is concerned, we must at least mention the publication (i) of Horvath’s revised version of her 1981 dissertation, in which the term “focus” is now highlighted, (ii) of Abraham and de Meij’s collection, where the “middle field” of both German and Hungarian is investigated in great detail, (iii) of Rochemont’s book, an attempt at giving a semantically unified, but syntactically working, definition of focus as new information, and (iv) of Selkirk’s study of intonation. The latter takes up Jackendoff’s hypothesis of an information or focus structure distinct from s–s and LF, strongly argues in favor of a radical distinction between grammatically determined stress and the assignment of a   to focused items, and proposes a concomitant a “Phrasal Focus Rule” which allows the percolation of the focus property of a focused head to either its phrasal projection or to an internal  of that head; Selkirk also proposes a principle establishing a clear divide between linguistic and metalinguistic use of pitch prominence when the pertinent unit is the syllable rather than the word (a problem noted, but not solved, in C 1955 — recall Section 2.1): “Perhaps the generalization is that pitch accents can be assigned to anything of level word or below, but that a pitch-accent-bearing element is only interpreted along the lines of a normal focused constituent when it has an identifiable separate meaning. When the pitch-accent-bearing element cannot be interpreted in this way, the presence of pitch accent is interpreted instead in metalinguistic term.” (p. 271)

10


1986 is also the year when “two books by  hit the newsstand”:15 Barriers (C 1986a) and Knowledge of Language (C 1986b). An intriguing fact must be noted: the Rule of Focus has disappeared. In itself, this does not mean anything, since Barriers hardly mentions LF at all (except to borrow Huang’s hypothesis that post s–s movement is not constrained by Subjacency). But LF is quite present in C (1986b), in particular in the sketchy typology (pp. 75–76) which opposes “English type languages”, in whose grammars wh-movement takes place before s–s, and “Chinese-Japanese type languages”, in which the same movement takes place after s–s. Since Chomsky cites Kiss’s work in Barriers (footnote 5),16 it is to be wondered why “Hungarian-type languages” are not introduced, which would differ from the other two types by having Focus movement applying before s–s rather than in between s–s and LF. It is quite possible that Chomsky was already doubting whether (visible or abstract) syntax was the proper locus for dealing with focus-related questions (see 2.8). In spite of this, Barriers proved very important in the analysis of focus, especially insofar as languages with a visible focal position were concerned. For example, Ortiz de Urbina (1986) fully exploited the generalization of X-bar theory to S and S′, now IP and CP, and dealt with focalization in Basque in terms of the revised V2 framework, with the focused phrase raising to Spec,CP (see also his paper in this volume). In 1989, Marácz defended a dissertation on Hungarian syntax in which he took the same stance (for instance explicitly characterizing “long wh-movement as an instance of long Focus-movement”). In both works, island effects, subjacency and the ECP are central concerns — just as in Rochemont & Culicover (1990), whose major goal is “to eliminate stylistic rules altogether” from English syntax, and in particular in the treatment of Directional/Locative inversions and Presentational There Insertion. Work on focus during this period of the Principles and Parameters framework, as was typical of that time, witnessed an unprecedented explosion in terms of cross-language empirical coverage. Important descriptive work entered the generative mainstream via the elaboration of the grammar of focus. Horvath’s analysis of previous work by Watters (1979) on the Bantu language Aghem is but one example. One result emerging from these studies (see Kiss 1995 for extensive references) is that the  of focus has a life of its own. This contributed to the motivation for the splitting of the functional category C0 (following that of I0), and the idea that the feature [Focus] has a (specific, or shared with other elements expressing “point of view”) syntactic projection (see work by Uriagereka, Brody, Choe, Laka, Tsimpli, among others). This hypothesis


11

is still very much alive (see recent publications such as Rizzi 1997, and Szabolcsi 1997). 2.8 Moving now directly into the most recent developments of generative syntax, at the beginning of Chapter 4, Chomsky (1995: 220) writes: Notice that I am sweeping under the rug questions of considerable significance, notably, questions about what in the earlier Extended Standard Theory (EST) framework were called “surface effects” on interpretation. These are manifold, involving -  - , figureground properties, effects of adjacency and linearity, and many others. Prima facie, they seem to involve some         , postmorphology but prephonetic, accessed at the interface along with PF (phonetic Form)  LF (Logical Form). If that turns out to be correct, then the abstraction I am now pursuing may require some qualification. [emphasis ours, G.R. & L.T.]

At least two questions must be asked: (a) Why did Chomsky decide to banish rheme and focus (probably now understood as  focus, if ‘rheme’ is to have any specific content) from the realm of syntactic computation/derivation? (b) Is that ban necessary within the Minimalist Program? As we saw in Section 2.1, when he was so to speak inventing generative grammar, Chomsky had doubts concerning the exact status of the notions of emphasis and contrast (whether expressed by pitch or stress) — and he had very good reasons for that. Firstly, there was no principle like Selkirk’s, quoted above, that drew the line between metalinguistic and emotive pitch marking on the one hand, and quasi-quantificational focusing on the other hand; secondly, even if he had had such a principle, we must remember that, at that time, one of Chomsky’s primary aims was to show that if (syntactic)  did influence meaning, there was very little evidence that meaning, a fuzzy notion if any, affected form. A third possible reason is due to the fact that sentences with contrastive focus are typically uttered in polemical contexts. To understand why this is important, let us turn to ‘On wh-Movement’ (Chomsky 1977: 81). There, Chomsky discusses rules like the one that is concerned with the interpretation of resumptive pronouns, which violate about every known syntactic constraint: “CNPC, the whisland constraint, and subjacency”. Hence the comment: “So interpreted, the rules in question fall completely outside the framework I have so far discussed and are not subject to any of the conditions cited, as seems to be the case.” Interestingly, he immediately adds:

12

GEORGES REBUSCHI AND LAURICE TULLER The same is true of rules that are not rules of sentence grammar at all, e.g. VP-deletion, which, as observed by Sag and Hankamer […], can apply across speakers and, correspondingly, is not subject to principles of sentence grammar, cf. [(21)]: [(21)]

a. b.

Speaker 1: John didn’t hit a home run Speaker 2: I know a woman who did John didn’t hit a home run, but I know [a woman who did – ]

Thus, in spite of the fact that VP deletion can take place with the antecedent under the same root node, as shown in (21b), it does not belong to sentence grammar; and, on this subject, Chomsky will not change his mind, as witnessed by the last section of Chomsky & Lasnik (1993; in C 1995: 125–126). Now note that the words theme and presupposition describe information shared by the speaker and the hearer, whereas rheme and focus refer to information communicated by the speaker to the hearer; from the  viewpoint, then, presupposition is old information, and focus, new information. It is therefore rational to define the opposition between presupposition (or theme) and focus (or rheme) uniquely in terms of what the hearer, rather than the speaker, knows. Moreover, as shown by (21a,b), ellipsis can be interpreted as deletion of whatever the hearer (as well as the speaker) already knows. In other words, whatever good reasons there are to decide that “VP-deletion […] is not subject to principles of sentence grammar” automatically carry over to the couple presupposition-rheme, and even more perspicuously to the couple presupposition–(contrastive) focus. Is the argument compelling? Probably not, since negative sentences and, above all, questions, are typically phenomena that imply both a speaker and a hearer — but they’re not rejected outside of sentence grammar, nor outside of the computation that leads to LF, as opposed to PF. If this reasoning is correct, then even though the suggestion is certainly worth pursuing (see Kidwai’s paper in this volume), at least two other options are available: (i) assume that there may be more than two interfaces, and build a specific “focus structure” interface (see 2.3 above, Vallduví 1992 and Erteschik-Shir’s contribution); (ii) postulate the existence of an abstract morpheme (or feature) F, taken to be both PF- and LF-interpretable. The latter approach has been chosen by many of the contributors to this volume. As we shall see in Section 2, the implementation of this option is not obvious, and is susceptible to several interpretations, among which, for instance, the localization of [F] in a specific functional head (mentioned above), or its optional (and possibly iterated) selection in a Numeration before Merge applies — note that the bottom-to-top building of syntactic complexity allowed by the relinquish-


13

ing of PS rules, the Projection Principle, and d-structure more generally, provides a more simple means of introducing F (almost) anywhere in a derivation than was possible in the Standard Theory or its immediate successors (see 2.2).

3.

Focus Types and Issues

As the preceding discussion of the evolution of the study of focus in generative grammar has highlighted, there are a series of recurrent issues surrounding the grammar of focus. These issues are taken up in the remaining chapters of this book, where various grammatical means of marking focus (as well as grammaticalization of focus marking) are analyzed in a wide variety of languages. As usual, language diversity increases chances of arriving at the level of abstraction required to characterize properly the grammar of the domain under study. The languages studied here represent six major phyla and various families within these: Afroasiatic (Standard Arabic, Moroccan Arabic, Bade, Tangale, Somali), Altaic (Turkish), Dravidian (Malayalam), Indo-European (English, German, French, Portuguese, Hindi-Urdu), Niger-Congo (lu-Haya, lu-Ganda, Wolof), and Basque (isolated). Phonological markers of focus and their effects are discussed by Hyman in his analysis of various tonal phenomena found in the Bantu languages lu-Haya and lu-Ganda and by Erteschik-Shir in her study of scope construal in English sentences involving focal stress; the intonation of French clefts is argued by Clech, Rebuschi and Rialland to offer clues for the proper syntactic and semantic analysis of focus clefting. Lexical focus markers such as German and English adverbial focus particles, the copula in languages such as Portuguese, English and Wolof, and term and predicate focus markers in Somali are discussed in the chapters by Bayer, Ambar, Kihm, and Lecarme, respectively. Various syntactic markers of focus are attested in the languages studied here, often in combination with phonological or morphological marking of focus: Cleft (and pseudo-cleft) constructions are examined in Arabic, English, French, Wolof, in the chapters by Ouhalla, Clech et al., and Kihm; a fixed position for focused constituents in the left periphery of the sentence, as in Arabic (Ouhalla), or in a position in/toward the right periphery, as in so-called free inversion, in Portuguese, or heavy NP-shift or stylistic inversion, in French; a fixed position immediately to the  of V, as in Basque (Ortiz de Urbina), Turkish (Kennelly),

14


Hindi and Malayam (Kidwai), or to the  of V, as in lu-Ganda and lu-Haya (Hyman) or the Chadic languages Bade and Tangale (Kidwai). The surface typological diversity illustrated in the focus constructions of these languages is associated with a number of inter-related issues concerning the proper characterization of focus. The global issue is clearly that of the interaction/mapping between information structure and grammatical focus marking, and between grammatical focus marking and semantic interpretation. This central problem of the articulation between pragmatic function, grammatical marking, and semantic interpretation entails several specific questions given the diversity just mentioned. What is the proper relation between focus and topic? Of the difference between presentational and contrastive focus? What is the connection between focus, Case-marking and the semantic/referential properties of a DP? How is structural focus related to configurationality? What is the nature of the link between focus and wh-interrogatives — why do both seem to display V2 effects and/or utilization of the left or right periphery? How should we characterize the proximity-to-V property found in focus constructions in many languages? (Is it Case adjacency? A V2 effect?) What is the status of [+focus]? (Is it a PF/LF interpretable feature? Does it have a specific syntactic projection or is it tied to another functional head?) The studies assembled in this volume shed important light on these issues. Various answers to these individual questions, and, more interestingly, analyses which offer answers to several of them at the same time can be found in these contributions. Hyman examines the relationship between the tonal effects characteristic of focus constructions and the pragmatic function of focus in Bantu, arguing that this relation is never a direct one. In other words, focus is so thoroughly entrenched in these grammatical systems that the association between a particular grammatical means of marking focus and the pragmatic function of focus is not (any longer) biunique. Lecarme also discusses a language — Somali — in which grammatical focus marking has become disassociated from pragmatic focus. Lecarme argues that in Somali the IP domain contains only pronominal arguments. Overt NPs may be either adjoined to the sentence as topics (coindexed with pronominal arguments) or appear in a focus position adjacent to C (where morphological focus markers are argued to be merged). The focus system in this language in this view is a Case-related device whose purpose is to make configurational, Case positions available for certain types of arguments (nonspecifics, wh-phrases and non-referring nominals). Kennelly likewise argues that the interpretive similarity between nonspecifics, wh-phrases and focused arguments


15

in Turkish follows from their position immediately to the left of V. The unambiguous wide scope construal of these elements is analyzed as their being rightadjoined to VP, while the verb is in I0. This adjacency to V follows from the view that focused elements are derived predicates. Erteschik-Shir also bases her study on scope relations. In her model of informational structure, truth values are assigned to sentences on the basis of fstructures (structural descriptions in which Topic and Focus constituents are identified) in such a way that topic quantifiers always take wide scope, and, more generally, scope relations are transparent. She argues that f-structure is a component of grammar and that therefore it interacts significantly with phonology, syntax, semantics, and pragmatics. The close links between information structure, phonology and syntax are accounted for by Kidwai by postulating a distinct level of interpretation, but which is located at the edge of the PFcomponent — so-called Domain D(iscourse). Focus is argued thus to be a PF [+ interpretable] feature, as it can be licensed at any of the levels internal to PF: PF movement (scrambling, of which focus movement is argued to be a case, on the basis of data from Hindi-Urdu, Malayalam, Bade and Tangale), morphology (cliticization, as in Hindi-Urdu) and phonology (prosody). This analysis is extended to languages which are not taken to be focus-position languages, in the spirit of Zubizarreta (1998), with the result that residual V2 effects in English receive an account parallel to that given to proximity-to-V focus position in other languages. The V2/proximity-to-V parallel is also examined closely by Ortiz de Urbina in a study which attempts to account for the distributional similarities between wh-words and foci in Basque, a particularly challenging problem given that Basque is right-headed and focus and other operators are  peripheral. Standard Arabic is another language in which foci and wh-words are left peripheral. Foci may also be in situ, the difference being the latter are presentational focus, while the former are contrastive focus. Ouhalla suggests that the distinction between the two stems from the latter being associated with the feature [+f] under the functional head F (taken to be a position for all categories expressing information regarding the propositional content of the sentence, as in Culicover 1991), whereas presentational focus is not. Only contrastive focus phrases therefore must move to Spec,F (either in the syntax or at LF) for interpretation. Ambar likewise shows that contrastive focus in Portuguese involves raising to the Spec of a functional projection, whereas presentational focus involves checking by the verb of constituents that remain in situ. In untangling a very complex array of data, she argues that focus structures must be

16


distinguished from evaluative structures (which are taken to have, like focus, an associated feature and functional projection in the left periphery). Her analysis is developed to extend to other types of focus structures (BE-focus, as well as (pseudo-)clefts). Ouhalla also goes on to analyze in detail cleft and pseudo-cleft constructions in Standard and Moroccan Arabic (as well as English) arguing that the same interpretive mechanism — existential closure over a choice function — can account for the interpretation of focus-preposing and in situ focus as well as (pseudo-)clefts. Clech-Darbon, Rebuschi and Rialland, on the other hand, argue that the semantic interpretation of clefts in French can be read directly off the S-structure. The post-focal constituent is argued to be adjoined to an ordinary copular or identificational IP, an analysis which permits a maximally simple account of both the interpretive and the intonational properties of clefts. The result is that there is no separate cleft construction at all, but merely the amalgamation of independently occurring types of identificational sentences and relative clauses. Comparing their analysis with that of Ouhalla for clefts in Arabic and English and that of Kihm for clefts in Wolof, these authors suggest that clefts may not be a unitary syntactic or semantic phenomena, even though they share a semantic core. The particularity of focus in Wolof, argues Kihm, lies in the presence of a predicational copula which is a multicategorial word: what is a single word in the surface string represents an entire VP — the verb, its external arguments and its complements — so that non-pronominal arguments are topics (cf. Lecarme’s analysis of Somali). The presence of this morpheme in focus constructions suggests that these are in fact clefts. An analysis of the extraposed CP is developed in terms of copying so that what distinguishes pseudo-clefts and clefts is which copy is pronounced. Bayer also makes crucial use of the minimalist interpretation of trace theory as involving copying of moved elements in his analysis of focusing particles such as only and even in English and German (though he, also crucially, rejects movement as attraction of features only). A unified syntactic and interpretive analysis of focus particles is developed which is based on the necessity of having a configuration of operator-variable binding, by movement or by reconstruction. We believe that the studies presented here are representative of the “state of the art” with respect to the elaboration of the grammar of focus in generative grammar. These studies attempt to provide unified analyses of focus in which the various focus issues reviewed here are correlated to a significant degree, though this correlation is accomplished in (sometimes radically) different ways. There


17

thus appears to be consensus on what the focus issues are, which surely amounts to progress in our understanding of this grammatical phenomenon. The exploration of the various avenues open to the construction of a global analysis of these issues, we think, will bring us closer to a grammar of focus that will play an important role in our comprehension of the interaction between the language faculty and adjacent cognitive systems, a long-standing goal of the generative enterprise itself, independently of the particular research programs that have so far instantiated it.

Notes * Thanks to the participants in the International Workshop on the Grammar of Focus which took place in Paris in February 1996 for stimulating discussion of the ideas and issues reviewed here. Particular thanks to Jacqueline Guéron for comments on an early draft. Remaining errors of interpretation and analysis remain ours solely. 1. For reasons of space, we limit ourselves principally to works investigation focus in English and only two other languages: Hungarian and Basque; this choice is justified by the fact that they have a designated focus position, that this position was acknowledged by traditional grammar long before the emergence of generative grammar and that generative research on this and related phenomena began very early (Kiefer 1967; de Rijk 1969). See Kiss (1995: §4) for a history of generative treatment of languages having a designated structural position for focus — so-called “focus-prominent” languages — as part of an essay on the broader idea that there is a widespread language type (termed “discourseconfigurational” languages) in which sentence organization is a result of discourse-semantic, rather than Theta-role or Case, considerations. 2. Furthermore, in languages where V0 and I0 are always amalgamated, heavy stress on the finite verb will be ambiguous. In Hungarian, for instance, (i.a) will, according to the context, translate either as (b) or (c), thereby confirming that the  element is at work in both English counterparts (after Kálmán 1985): (i)

a. b. c.

Péter  Marit Peter  love Mary (contrary to what you said/seem to assume) Peter  Mary (he doesn’t  her)

3. It is worth noting that, according to Kiefer (1970), a paper by Lu was published as early as 1965, in which (i) was given (ii) as its underlying structure: (i)

J bought a book

(ii)

[[John past buy a book]  [ past buy a book]]

4. The focus does not project if another item bears the intonational center, e.g. red in (3). 5. De Rijk (1969) and Donzeaud (1972) are early attempts to apply this interpretive approach to Basque.

18


6. If focus projection does take place, (5b) will no longer correctly gloss (5a); see footnote 10 for more on this. 7. See 2.4 and 2.6 below, and Clech-Darbon et al. (this vol.) for more details on the history of the treatment of cleft sentences. 8. Interestingly, the same chapter contains a discussion of ellipsis: Jackendoff is probably the first generative linguist to have noticed that focusing and ellipsis are very close to being the two faces of the same phenomenon, a fact whose significance will appear in Section 2.8. Recall also that this 6th chapter contains the first discussion of the phenomenon of (operator) association with focus. 9. In such cases, Guéron’s system also permits another interpretation, dubbed “backgrounding of the logical focus”, as in her example (74): (i)

It was a beautiful day. The  was shining. The  were singing…

10. Although she does not say it explicitly, if John is non-contrastive, the focus property can percolate to the VP node (recall (3)). Then, either no movement takes place, and the coindexation cannot be ruled out, or the  VP undergoes the Rule of Focus, yielding (i) in the case of (15), — and (ii) in the case of (12): (i)

(for P=) [mistreat John]i [his wife (₎ ti]

(ii)

(for P=) [betray John]i [the woman he loved (₎ ti]

In neither case can the trace of the VP and the pronoun ever be coindexed; it follows that (11) will never even  all cases of WCO effects. Something like Guéron’s analysis is therefore  to supplement the formal analysis of C (1976) or any modified version thereof. 11. See Horvath & Rochemont (1986), Rochemont (1986) and Williams (1997) for a refinement of those notions. 12. According to Rochemont (1986: 44), Schmerling (1976: 77) was the first scholar to point out that “the focused/nonfocused distinction cuts across the factive/nonfactive one”, so that, in certain contexts, it is quite natural to focalize a constituent in the scope of a factive verb like realize. This remarks naturally carries over to cases where “old information” is focused, hence “recycled”, as in Rooth’s well-known example (i): (i)

‘Does Ede want tea or coffee?’ ‘Ede wants .’

13. In C (1976), representations like (16b) were used as LFs for - sentences, as below (op. cit.: 193 (70–71)): (i)

John seems [t to be a nice fellow]

(ii)

For x = John, x seems [x to be a nice fellow]

For all practical purposes, (ii) then simply glosses a GQ approach to the semantics of NPs: (iii) [lx[x seems [x to be a nice fellow]]](John)


19

14. De Rijk (1979) had reached the same conclusions concerning Basque. Note however that, as indicated by Abraham (1997), focusing need not always be associated with V2 in languages that have visible wh-movement, since both of the following sentences are grammatical in German: (i)

a.

b.

Wir haben gestern   gesehen we have yesterday the girl seen ‘We sax (lit. have seen) THE GIRL yesterday’   haben wir gestern gesehen

Moreover, Abraham also indicates that a prosodically focused XP may be in initial position when contrast is implied, cf. (ii), but an NP representing the answer to a wh-question may not be initial, cf. (iii): (ii)

Dein BUCH mubt du Paul geben (nicht deine TASCHE) Your BOOK must you Paul give (not your BAG)

(iii)

a. b.

Was hast du GELESEN? ‘What have you read?’ #Dein BUCH habe ich gelesen

See Kennelly’s contribution (this volume) for similar data in Turkish. 15. After Guéron (1980: 664). 16. The paper, cited as Kiss (1985), was published two years later (Kiss 1987).

References Abraham, W. 1997. “The base structure of the German clause under discourse functional weight: contentful functional categories vs. derivative functiona l categories”. In W. Abraham & E. van Geldersen (eds.), German: Syntact ic Problems — Problematic Syntax. Tübingen: Niemeyer, 11–42. Abraham, W., & S. de Mey (eds.). 1986. Topic, Focus, and Configurationality. Amsterdam: John Benjamins. Chomsky, N. [1955] 1975. The Logical Structure of Linguistic Theory. New York: Plenum. Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton. Chomsky, N. 1970. “Deep Structure, Surface Structure and Semantic Interpretation”. in R. Jakobson & S. Kawamoto (eds.), Studies in General and Oriental Linguistics […], Tokyo: TEC, 52–91. Also in N. Chomsky, 1972, Studies on Semantics in Generative Grammar, The Hague: Mouton, 62–119. Chomsky, N. 1976. “Conditions on Rules of Grammar”. Linguistic Analysis 2(4).303–351. Also in N. Chomsky, 1977, Essays on Form and Interpretation, New York: North-Holland, 163–210.

20


Chomsky, N. 1977. “On wh-Movement”. In P. Culicover, T. Wasow & A. Akmajian (eds.), Formal Syntax. New York: Academic Press, 72–132. Chomsky, N. 1980. Rules and Representations. Oxford: Blackwell. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986a. Barriers. Cambridge (Mass.): MIT Press. Chomsky, N. 1986b, Knowledge of Language. New York: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, N., & H. Lasnik. 1993. “Principles and Parameters Theory”. In J. Jacobs et al. (eds.), Syntax: Ein internationales Handbuch zeitgenossischer Forschung. Berlin: de Gruyter, 506–569. Culicover, P. 1991. “Focus and Grammar”. In P. Ackema and M. Schoorlemmer (eds.), Proceedings of the Workshop on the Semantic and Syntactic Analysis of Focus. Utrecht: OTS, 1–18. Donzeaud, F. 1972. “The Expression of Focus in Basque”. Anuario del Seminario de Filología Vasca ‘Julio de Urquijo’ 6.35–45. Guéron, J. 1980. “On the Syntax and Semantics of PP Extraposition”. Linguisti c Inquiry 11(4).637–678. Horvath, J. 1981. Aspects of Hungarian Syntax and the Theory of Grammar. Doctoral dissertation, UCLA. Horvath, J. 1986. Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Horvath, J. & M. Rochemont. 1986. “Pronouns in Discourse and Sentence Grammar”. Linguistic Inquiry 17(4).760–765. Jackendoff, R. 1972. Semantic Interpretation in Generative Grammar. Cambridge, Mass.: MIT Press. Kálmán, L. 1985. “Word Order in Non-Neutral Sentences”. In I. Kenesei (ed.), Approaches to Hungarian, 1: Data and Descriptions (Szeged, JATE), 25–37. Kiefer, F. 1967. On Emphasis and Word Order in Hungarian. Bloomington: Indiana University Publications (Uralic and Altaic series, vol. 76). Kiefer, F. 1970. “On the Problem of Word Order”. In M. Bierwisch & K. E. Heidolph (eds.), Progress in Linguistics. The Hague: Mouton, 127–142. Kiss, K.É. 1981. “Structural Relations in Hungarian, a ‘Free’ Word Order Language”. Linguistic Inquiry 12(2).185–213. Kiss, K.É. 1987. “Is the VP Universal?”. in I. Kenesei, (ed.), Approaches to Hungarian, 2: Theories and Analyses (Szeged: JATE), 13–85. Kiss, K.É. 1995. “Introduction”. In K.É. Kiss (ed.), 3–27. Kiss, K.É. (ed.) 1995. Discourse Configurational Languages. Oxford: OUP.


21

Lasnik, H. and M. Saito. 1992. Move-a:: Conditions on its Application and Output. Cambridge, Mass.: MIT Press. Lu, J.H.T. 1965. Contrastive Stress and Emphatic Stress. The Ohio State University, Project Report, #10. Marácz, L. 1989. Asymmetries in Hungarian. Doctoral dissertation, University of Groningen. [Published as Supplement to ASJU #22, San Sebastian, 1994]. Ortiz de Urbina, J. 1986. Parameters in the Grammar of Basque. Doctoral dissertation, University of Illinois at Urbana-Champaign. [Revised version: Dordrecht: Foris, 1989.] Pulman, S.G. 1997. “Higher Order Unification and the Interpretation of Focus” Linguistics and Philosophy 20(1).73–115. de Rijk, R.P.G. 1969. “Is Basque an SOV Language?” Fontes Linguae Vasconum 1.319–351. de Rijk, R.P.G. 1978. “Topic Fronting, Focus Positioning, and the Nature of the Verb Phrase in Basque”. In F. Jansen (ed.), Studies in Fronting. Lisse (Leiden): Peter de Ridder Press, 81–112. Rizzi, L. 1997. “The Fine Structure of the Left Periphery”. In L. Haegeman (ed.), Elements of Grammar: Handbook of Generative Syntax. Dordrecht: Kluwer, 281–337. Rochemont, M. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins. Rochemont, M. & Culicover, Peter W. 1990. English Focus Constructions and the Theory of Grammar. Cambridge: CUP. Rooth, M. 1996. “Focus”. In S. Lappin (ed.), The Handbook of Contemporary Semantic Theory. Oxford: Blackwell, 271–297. Schmerling, S. 1976. Aspects of English Sentence Stress. Austin: University of Texas Press. Selkirk, E. 1986. Phonology and Syntax. Cambridge, Mass.: MIT Press. Selkirk, E. 1995. “Sentence Prosody: Intonation, Stress, and Phrasing”. In J. Goldsmith (ed.), The Handbook of Phonological Theory. Oxford: Blackwell, 550–569. Szabolcsi, A. 1981. “The Semantics of Topic-Focus Articulation”. In J. Groenendijk et al. (eds.), Formal Methods in the Study of Language 2, Amsterdam: Matematisch Centrum, 503–540. Szabolcsi, A. 1997. “Strategies for Scope Taking”. In A. Szabolcsi (ed.), Ways of Scope Taking. Dordrecht: Kluwer, 109–154. Vallduví, E. 1992. The Informational Component. New York: Garland.

22


Watters, J.R. 1979. “Focus in Aghem”. In L.M. Hyman (ed.) Aghem Grammatical Structure. USC: Southern California Occasional Papers in Linguistics 7. Williams, E. 1997. “Blocking and Anaphora”. Linguistics Inquiry 28(4).577–628. Zubizarreta, M.-L. 1998. Prosody, Focus and Word Order. Cambridge, Mass.: MIT Press.

Aspects of the Syntax of Focus in Portuguese* Manuela Ambar Universidade de Lisboa

Abstract Taking as a starting point the description of Portuguese data, this paper aims at a unifying treatment of focus. We will first discuss some of the analyses that have tried to encode the notions of topic and focus in the grammar in a purely formal way, still maintaining the distinction made on the basis of the dichotomic notions of “old” and “new” information. We will then present the distribution of focus constructions in Portuguese in order to establish the real distinction between so-called contrastive focus, and presentational focus. Finally we will see what light Portuguese data shed on the discussion brought to linguistic research on this phenomenon by two opposite views on the treatment of focus: the one which defends that different manifestations of focus constructions in different languages are the effect of movement operations to a focus projection where focus is licensed (cf. Brody 1990, among others) and the one which claims that there is not any such movement and rather that a focused constituent has to stay in situ (Zubizarreta 1993, among others) in the more embedded position to which the focus stress is assigned (cf. Cinque 1980). We will try to show that evidence drawn from Portugese focus structures, seems to support the first claim.

1.

Introduction

This paper is a reduced version of a larger work on focus in Portuguese. Limitations of space preclude the presentation of all the empirical and theoretical motivation underlying the analysis we will present, namely, motivation found in (pseudo-) clefts and in ergative structures.

24

MANUELA AMBAR

As will become clearer in the text, the crucial object of our research is word order, which highlights the large syntactic component involved in Focus. Given the extension and complexity of this phenomenon, it is impossible to present a systematic comparative approach here. However other languages, namely Italian, Hungarian, English and French were taken into account, as we developed our system.1 As we proceeded with the description, the establishment of data, requiring more and more refinements and care, became a complex task — in fact, focus involves pragmatics, semantics, prosody and syntax, this making the necessary abstractions difficult to attain. The system proposed is intended to explain the following facts: contrastive focus (restrictive and non-restrictive) presentational focus, focus involved in ergative structures, expletive BE structures, cleft and pseudo-cleft structures and tense restrictions on the latter constructions.

2.

Topic and Focus — different views

The grammatical notions of “Topic” and “Focus” have been distinguished on the basis of the dichotomic discourse notions of “old” and “new” information respectively. Several analyses tried to encode these notions in the grammar in a formal way. Guéron (1980) proposes the distinction between “predication” and “presentation” structures, as represented in (1a) vs. (1b) respectively: (1)

a. b.

(S (NP) (VP)) (S VERBj (S (NP) (…tj))) (Guéron 1980: 651)

(1b) is the LF configuration for English constructions introducing a new referent for the subject in the Universe of Discourse, whereas in (1a) the subject is already known and there is a predication on it. (1) represents two main issues in the treatment of Focus: (i) the opposition between topic and focus in terms of old vs. new information, respectively in (1a) and (1b), and (ii) the relation the focused phrase establishes with the verb in these constructions. The first issue was developed in different works by different authors using different terminology for the dichotomy;2 the second point, i.e. the relation the verb establishes with the focused constituent, has also received different interpretations depending on the theories constructed and on the languages

ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE

25

studied. For Horvath (1986) the verb assigns the feature F to the focused constituent, but for Tuller (1992) it is Infl that plays the role. Ambar (1988) assumes that in order to check the focus feature borne by the focus constituent the verb canonically governs the focused element at S-structure (overtly) or at LF (covertly). In Ortiz de Urbina (1989) focus is licensed through a Spec-Head agreement relation in CP; for Culicover and Rochemont (1990) the verb governs, but does not q-mark the focused element; for Brody (1990), Uriagereka (1992), Rouveret (1992, 1996) and Rizzi (1995) there is a functional projection for focus, where focus-licensing takes place. This projection is SP for Laka (1990) and OP (Operator Phrase) for Raposo (1994). For Zubizarreta (1993). however, such a projection does not exist, the focused element rather stays in situ, occupying the rightmost embedded position, to which the stress accent is assigned (cf. Cinque 1993) and has to be c-commanded by the Aspect node at LF. It is not possible to mention here in detail all the proposals that have been made for the treatment of focus constructions. Throughout our description of data, we will crucially make use of the diagnostics for focus hinted at by Cullicover and Rochemont (1990: 19): “In a well formed simple question-answer sequence, all and only the information provided in the response that is not contained in the question is focused.” This diagnostics will guide the establishment of data and, once systematically applied, will lead us to conclude that some of the so-called ‘focus constructions’ have been misleadingly interpreted.

3.

Focus in Portuguese — the data

Taking the way information is introduced in the Universe of Discourse as the leading criterion for the classification, two major groups of Focus constructions exist in Portuguese as in other languages: (i) contrastive focus — the one where new information is viewed in contrast with other specific old or new information, and (ii) presentational focus — the one where the focused constituent simply introduces new information without contrasting it with any other type of information, either old or new. If, however, we take structural criteria to characterize focus, a finer classification emerges according to the formal properties of each construction: (i) those in which just a marked focal stress is assigned to the focused element and no visible movement shows up — both the verb and the focused element

26

MANUELA AMBAR

appear in their canonical order: (2)

A MARIA beijou o Pedro MARY kissed Peter

(3)

A Maria beijou O PEDRO

(4)

A Maria BEIJOU o Pedro

The interpretation is the one of contrastive focus associated with an exclusive/ restrictive value; the focused element is contrastive in the sense that the new information introduced in the Universe of Discourse presupposes an opposition with other possible old information, against which the new information is placed3 and it is restrictive/exclusive because the new information is exclusive or restricted to the entity or entities the focused element denotes. I will not be concerned with this type of focus here.4 (ii) those in which, besides a particular stress assigned to the focus constituent, focus-licensing is crucially syntactic, i.e. movement is visible. Two situations can then occur: A. Contrastive focus — restrictive vs. non-restrictive the focused element — either the subject or the complement — has to precede the verb in the visible syntax. Focus is then contrastive, but, depending on the type of relation it establishes with the verb, either the exclusive/restrictive or the non-exclusive/non-restrictive interpretations (a concept to be made clearer below) are available; B. Presentational focus — restrictive those in which the verb has to precede the focused element — VS order is then exhibited in a SVO language, whenever the focused element is the subject. The focus interpretation is presentational and in presentational focus only the exclusive/restrictive value is available; no particular stress is assigned to the focused element. Let us observe the data concerning these two cases. Consider first a simple declarative sentence: (5)

A Maria beijou o Pedro Mary kissed Peter

With unmarked stress,5 the whole sentence can receive focus; its underlying


27

question would be What happened? But it can also be an answer to questions such as: What has Mary done? and Who has Mary kissed? — the focused constituents are then the VP and the Object respectively. With this canonical order and this unmarked stress, sentence (5) does not allow either presentational or contrastive focus readings for the subject alone. Focus on the subject triggers movement. One of our goals is to understand why this is so. Let us examine the relevant properties of presentational and contrastive focus, tested in question-answer pairs. Consider the following paradigm: (6)

Quem comeu a tarte? Who ate the pie?

(7)

a. b. c. d. e.

Comeu a Joana. Ate the Joana A tarte comeu a Joana *A Joana comeu *A Joana comeu a tarte *Comeu a Joana a tarte6

Assuming Portuguese is a SVO language,7 and comparing sentence (5) with the   paradigm in (7), the following observations follow: (i) presentational focused objects apparently appear in their base position — the VO order is respected; inversely, presentational focus on the subject entails a change in the canonical order — instead of the SV order the VS one is required, i.e. the verb has to precede the focused subject; (ii) contrary to what happens with focus, a constituent that represents old information — roughly a topic — does not appear, in the unmarked case8, in post-verbal position (cf. 7d–e). (7c), with a special intonation, is an adequate  answer, but not a  one, to (6), as (8) exemplifies: (8)

A Joana … comeu (about the others I don’t know) Joana … ate

In (8) -⁄- 9 focus on a Joana is at stake. Concerning interpretation, (7a) differs from (8) in that focus is ⁄ in the former and -⁄- in the latter. An element has an exclusive/restrictive focus reading if the correlation between members of a set of entities is not explicitly established, i.e. if the speaker  the new information to the entity denoted by the focused element. In (7a), for instance, he does not think of other entities that may have

28

MANUELA AMBAR

eaten the pie and he does not consider any shared information with the hearer about them; in other words, the focused element is not placed against other entities that constitute its complement in the set of entities to which it belongs. Inversely, an element has a -⁄- focus reading if the correlation between members of the set is established, i.e. among the entities of the set, the speaker chooses one, which he identifies as the focused element, against other entities that belong to the same set. In (8) the speaker knows, or, more precisely, he pretends to show he knows, that other entities may have eaten the pie and to some extent he considers that there exists some knowledge shared by him and by the hearer about that: he takes one of those entities and identifies it with the new information — the focused element can be viewed both as a topic and as a focus and the speaker only cares about it, not about the others. It is why a -⁄-  is a  (non saturated) answer to the addressed question — it does not saturate the reference to the entities that range in the set which the focused element belongs to; in (7a), or in restrictive/exclusive readings in general, this reference is saturated, i.e. the speaker assumes the information as , without considering other entities. Observe that, contrary to presentational focus, contrastive focus is incompatible with inversion: (9)

*Comeu a Joana

in (9) the reading described for (8) is lost. On the other hand, the constituent that refers to old information cannot appear in the answer without a marked pause, as it can in (7b) above: (10)

a. *A tarte a Joana comeu b. A tarte // a Joana comeu The pie (concerning the pie) Joana ate

The same observations hold for presentational focus vs. contrastive focus on object constituents: (11)

Que comeu a Maria? What did Mary eat?

(12)

a.

Comeu a tarte Ate the pie b. A Maria comeu a tarte c. *A tarte comeu


29

d. *A tarte comeu a Maria e. *Comeu a Maria a tarte Again the focused element (now the object) has to be preceded by the verb in presentational focus and, similarly to paradigm (7), an element referring to old information (presumably a topic, as a Maria in (12d-e)) cannot follow it. We conclude that presentational focus behaves uniformly with respect to word order: the subject/object asymmetry that characterizes SVO languages of the Portuguese type disappears, instead we get everywhere a VX order, where X stands for the presentational focused element and V for the verb. Inversely, contrastive focus with a non-exclusive/non-restrictive reading requires the focused element to precede the verb:10 (13)

a.

A tarte… a Maria comeu … (outras coisas não sei…) The pie… Mary ate … (what more she ate, I don’t know…) b. *A Maria comeu a tarte Mary ate the pie (*What else she ate, I don’t know…)

Once again, inversion is excluded from this context — with the reading in question, (14) is ruled out: (14) *A tarte comeu a Maria A unified analysis of focus should be able to explain why presentational focus is in complementary distribution with contrastive focus of the type described and at the same time should make the correct predictions on other focus contexts, namely focus in the context of three place predicates, ergative structures, (pseudo)cleft and BE-Focus structures. In the following sections we present the crucial assumptions of our proposal and describe its predictions concerning some of these facts.

4.

The proposal — FocusP vs. TopicFocusP

The analyses that have considered the existence of a focus projection have claimed, according to minimalist requirements, that the focused constituent goes to [Spec,FP] to check its focus feature.11 For Portuguese, Rouveret (1992, 1996), following Brody (1990), proposes that this language projects FocusP and that preposed focused constituents have to be in [Spec,FP] at the Spell-out. The author takes sentences like those in (15) to illustrate his proposal:

30

MANUELA AMBAR

(15)

a. b.

ISTO fazem os reis quando viajam This is what kings do when they travel O José disse que ISTO fazem os reis quando viajam Jose said that this is what kings do when they travel

According to Rouveret, the focused element is ISTO in both sentences and it is in Spec,FP at the Spell-Out. The author also observes that “in Focus constructions, subject-verb inversion systematically occurs and, when a clitic is present, we always get the clitic-verb order”, as in (16) below: (16)

a. b.

ISSO lhe disse eu This is what I told him ATÉ À MARIA o apresentaram eles They presented him even to Maria

Also Uriagereka (1992) proposed that in Gallician focused elements occupy Spec,FP in sentences like (17): (17)

Moitas cousas lhe eu dixera

Martins (1995), following Uriagereka (1992), assumes for Portuguese that quantified focused elements, as in (17), go to Spec of FP, an instance of Laka’s (1990) SP. However, if we take the dichotomy old/new information in a question/ answer pair as a diagnostics to define topic and focus respectively, then in (15) the element that introduces new information is the subject — reis (kings) — rather than the object ISTO ‘this’: neither   — which, as we saw above, entails VX order — nor   — in which context inversion is not possible — are available for ISTO in this environment. As for (16), things are more complex, since in those cases it is not clear that the subject in post-verbal position bears a presentational focus interpretation either, in spite of the fact that it is still receiving a focal-type accent. We will see below that the sentences in (16) are  . The data described so far have crucially showed that a focused phrase can appear in two positions in sentence structure: (i) preceding the verb (contrastive focus) and (ii) following it (presentational focus). Assuming, as we are doing, that an analysis based on a licensing focus projection is more adequate for a  treatment of focus — see, for instance, (pseudo-)cleft constructions, where clearly there is movement, then the question turns out to be how to make the existence of those two positions compatible with their licensing in only one functional projection.


31

For concreteness, suppose we assume the following representation, where F means Functional: (18)

[CP [ [F [F′ [IP …]]]]]

Abstracting now from the presence of CP, how could we derive the VX order of presentational structures, while still being able to derive the XV order of contrastive focus? Consider first presentational focus, exemplified in (7a–b) and in (15): in the former what is presupposed is a tarte ‘the pie’, a topic-like element functioning as the subject of the focused element o João (John). Note that (7a) — which exemplifies the so-called ‘free inversion’ phenomenon12 — is a grammatical sentence if and only if it is interpreted as a presentational focus structure in the context of a question/answer pair; it is then natural to assume that the structural representation of this sentence will somehow include the element which refers to old information, its presence being crucial for interpretability and grammaticality. In (15) what is presupposed is Isto, which is then a topic-like element, not a focus one. The question now is how to determine which positions these elements occupy in representation (18). One possibility is to assume that the constituent referring to old information is in Spec of a Topic Phrase, above FP, that the verb moves to the head of FP and that the focused element stays inside IP13, the complement of the focus head — in this case, Spec,FP would be empty; another hypothesis is to assume that the constituent that refers to old information goes to Spec,FP or is represented there by an anaphoric operator referring to the discourse or to an explicit topic element, all the rest being equal. The idea of a null operator relaying the topic element in structures like (7a) above is not new. Ambar (1988) proposed the following representation for these structures: (19)

[CP OPi [C′ comeuv [IP o João tv ti]]]

In that analysis, it was claimed that: (i) a focus feature was assigned to the focus element — o João; (ii) the verb had to raise to COMP position in order to canonically govern the focus feature, for identification reasons; (iii) the operator in Spec,CP, restating information given in the question — a tarte —, would bind its variable inside IP; it could be null as in (7a) or phonetically realized as in (7b). Both the correct interpretation of the sentence and the fact that an Rexpression — functioning as a topic-like element — cannot appear in post-verbal position were predicted (cf. 7d–e).14

32

MANUELA AMBAR

Let us explore the hypothesis outlined in representation (19). A first observation concerns exactly the relation between topic and focus. In (19) CP is a projection where a - element is the subject of a predicate that introduces the focused element as its complement, i.e. it is a projection where the relation between topic and focus is met; this relation is necessary for focus licensing. Let us rather call this projection a TF P and hypothesize that focus obligatorily involves a - element in order to be licensed15. If we look at the data described above, we see that focus is always introduced in a given relation with old information — represented either by a - element that functions as the subject of the focused constituent, as in  focus (cf. 7a–b) or by a -- element that, in a sense, functions as its own subject (its own topic) in spite of being able to introduce new information, a focus function, as in (8). This topic-like element is not to be confused with true topics: the former has the function of introducing focus. Consequently TopicFocusP, the projection where focus is licensed, is distinct from TopicP, the projection for topic licensing. (20) below would then substitute for (19): (20)

[TopicP [Topic′ [TopicFocusP Opi [TopicFocus′ comeuv [IP o João tv ti ]]]]]

Further evidence for this proposal can be drawn from the distribution of clitics. In structures like the ones in (7a–b, 15) a clitic connected to the topic-like element does not occur. Considering Cinque’s 1990 proposal for CLLD constructions in Romance vs. topicalization structures of the English type, we could attribute the absence of the clitic to the presence of the operator in the local domain of its variable in these constructions. But then, since Portuguese is a Romance language, the parameter opposing English to Romance cannot reside in the availability of such an operator in English vs. its unavailability in Romance as defended by Cinque (1990) and Rizzi (1995). Notice that Portuguese and Italian behave alike concerning structures like the one in (7a) — the so-called free inversion structure (cf. Rizzi 1982), but that they differ with respect to other related structures, this meaning that the typology cannot be stated in terms of Romance vs. non-Romance. Another structure where the presence of the clitic is dispensed with both in Italian and in Portuguese is the one of contrastive focus, where a clitic does not appear related to the contrastive focused element as observed in Rizzi (1995). However, there are structures where the presence of the clitic connected to a topic element is obligatory in Italian, and in Portuguese for some speakers sentences are improved if the clitic is present16:


(21)

33

O teu vestido, a Joana vestiu-o Your dress, Joana put it on

With these contrasts in mind, which show that the - element — a tarte and isto respectively in (7a-b) and (15) — differs from the   in (21) — o teu vestido, given the absence of the clitic in the former vs. its presence in the latter, let us explore the hypothesis outlined in representation (20). Contrary to what has been assumed in the literature, TopicFocusP, rather than FocusP, would then be the projection where focus is uniformly licensed in Portuguese.17 Let us observe how this hypothesis accommodates the facts described. Assume that TopicFocusP, a mixed projection, has topic and focus features. Suppose further that the verb is endowed with the capacity of licensing focus — a capacity inherited from its Event component.18 The idea that Event is involved in focus licensing seems a natural one, since Focus is nothing else but the identification of an event. This is crucial for our analysis of (pseudo-)cleft constructions. If there is focus on a given phrase, TopicFocusP is projected and its features have to be checked. I will designate the features of the - element occupying the Spec of this projection as [topicf] features and the features of  , i.e. topics that occupy the Spec,Topic Phrase, as [topic] features, in order to distinguish them. Similarly, I will call the features of a - element that moves to TopicFocusP [focust] in order to distinguish it from a   (the one which will stay inside IP, instead of moving to that projection — presentational focus). I will claim that  focused constituents, which I have been assuming have both topic and focus properties, have two features: a [focust] and a [topicf] feature. Only [topicf] and [focust] elements can go to Spec,TopicFocusP. All the other focused elements will be complements of the TopicFocus head; since they have [focus] features instead of [focust] features, they cannot move to TopicFocusP; they will then stay inside IP, the complement of the TopicFocus head. This type of focus is  focus. As will become clearer in what follows, in order to have presentational focus on a constituent it does not suffice for this element to be inside IP; three requirements have to be met: (i) TopicFocusP has to be projected (ii) its focus feature has to be checked by the verb19 and (iii) the projection has to contain phonetically realized material20 — either one or both of its positions have to be lexically filled. Now we understand representation (20): Spec,TopicFocusP is filled by a [topicf] element expressing old information from the discourse. This element will check the topic feature of the head, but not its focus feature. In order for the

34

MANUELA AMBAR

focus feature to be checked, the verb has to move to the head of TopicFocusP.21 Observe that in (20) IP becomes the location for presentational focus and only contains the presentational focused element, which occupies the rightmost embedded position, satisfying Cinque’s (1993) algorithm for focal stress assignment. Against minimalist assumptions, I am considering that in these structures movement is triggered by the interpretative topic and focus features, which will remain visible at LF and will determine the interpretation not only of the category bearing them but also of their complements (see Rizzi (1995) for a similar assumption). Note that the operator in (20) can be overt, providing the adequate derivation for (7b) and that the same analysis holds for presentational focused objects exemplified in (12) whose representation is given in (22) below: (22)

a.

[TopicP[Topic″ [TopicFocusP OPi / A Mariai [TopicFocus′ comeuv [IP ti tv a tarte ]]]]]

Take now contrastive focus structures as in (8), where A Joana bears contrastive focus associated to a non-restrictive/non-exclusive interpretation. I have suggested above that these elements are TopicFocus elements; they have then [topicf] and [focust] features; therefore they move to Spec,TopicFocusP, where they check both the topic and the focus feature of the head. Once the focus feature is checked by the contrastive focus constituent in Spec,TopicFocusP, it is not necessary for the verb to move to TopicFocus0; consequently it cannot move. Sentences (8) and (13a), where the contrastive focused phrases are respectively the subject and the object, receive representations (23a) and (23b): (23)

a. b.

[TopicP OPk/a tartek [Topic′[TopicFocusP A Joanai [TopicFocus′ [IP ti comeuv [ek]]]]]] [TopicP [Topic′[TopicFocusP A tartei [TopicFocus′ [IP a Joana comeuv [tk]]]]]]

Now we understand: (i) why inversion is incompatible with the contrastive focus interpretation: the verb cannot move to the head of TopicFocusP, because the focus feature was checked by the contrastive focused element in Spec, TopicFocusP; (ii) why the presence of a topic element requires a pause in these structures but not in presentational focus (cf. (7b) vs. (10)): in the former the topic element is outside TopicFocusP but in the latter it is inside it; (iii) why a clitic appears in these structures (cf.(21)), in contrast with presentational focus: its antecedent is outside TopicFocusP.


35

Note that the gap in object position in (23a) cannot be bound by an anaphoric operator in Spec,TopicFocusP, since this position is filled by the contrastive TopicFocus subject. Arguably then this gap does not qualify as a variable, because the operator restating information from the question is located outside its local domain — it is not in Spec,TopicFocusP, but in Spec of TopicP. I will therefore assume that this gap has to be a resumptive pronoun, which can be null in Portuguese, but not in Italian. I will take this contrast between Italian and Portuguese as an instance of a more general phenomenon: the null object parameter which opposes languages of the Portuguese type to languages of the Spanish or Italian type. Whichever the analysis of the null object parameter turns out to be, for our purposes here, I will simply assume that in (23) the object gap has a different status from the one in (20): here we have a variable bound in its local domain by the empty vs. lexical operator in Spec,TopicFocusP; in (23) this empty category is a pro-like element, which can be bound by an antecedent outside its local domain; the difference between Italian and Portuguese seems then to lie in the fact that Italian does not allow a long distance identification of pro, whereas Portuguese does, this correlating to the null object parameter22. Recall now that sentences represented in (20) receive a restrictive/exclusive reading whereas the ones in (23) have a non-restrictive/non-exclusive interpretation.23 How can we derive this contrast from our proposal? I would like to suggest that the restrictive/exclusive interpretation is available whenever the focused element is under the scope of the verb, more precisely under the scope of its Event component, and that, inversely, non-restrictive/non-exclusive readings obtain just in case the focused element is not under its scope. Assuming that scope can be defined in terms of spec-head agreement and c-command, the facts described above receive a simple explanation and provide evidence for our analysis: in (20) and (22) the verb has raised to the head of TopicFocusP, the result being that the focused element is c-commanded by the verb, but in (23) the non-restrictive interpretation will be the only possible: since the verb does not raise, the focused element does not end up either in a spec-head or in a c-command relation with the verb. This hypothesis successfully carries over to other contexts, namely to (pseudo-)cleft constructions.24

36 5.

MANUELA AMBAR

Other Predictions

5.1 Three-place predicates Observe how the predictions of our system are borne out in structures of threeplace predicates, where the focused element is the  . Given a question like (24), different combinations are possible: (25a) through (25e) are their representations. As expected, (25f) is excluded: (24)

A quem ofereceu o Pedro as flores?

(25)

a. b. c. d. e. f.

[TopicP Opj/i [Topic′ [TopicFocusP OPi/j [TopicFocusP ofereceuv [IP ei tv ej à Joana]]]]] [TopicP o Pedroi [Topic′ [TopicFocusP as floresj [TopicFocusP ofereceuv [IP proi tv tj à Joana]]]]] [TopicP as floresj [Topic′ [TopicFocusP o Pedroi [TopicFocus′ ofereceuv [IP ti tv proj à Joana]]]]] [TopicP o Pedroi [Topic′ ofereceuv [TopicFocusP as floresj [TopicFocusP t′v [IP proi tv tj à Joana]]]]] [TopicP as floresj [Topic′ ofereceuv [TopicFocusP o Pedroi [TopicFocusP t′v [IP ti tv à Joana]]]]] *[TopicP Opj [Topic′ ofereceuv [TopicFocusP o Pedroi [TopicFocusP t′v [IP ti tv as floresj à Joana]]]]]

We could add more elements (e.g. adverbs) to sentences in (25); they still would be well-formed. Then, we had to assume that other TopicPs were projected. A first basic intuition underlying those representations is that there can be several topic projections, but only one structural focus position, in the spirit of Rizzi (1997). However, although this is so, more than one constituent can be focused  sentence. Suppose, instead of question (24), we had (26), then the natural answer would be the one in (27): (26)

Quem ofereceu flores a quem?

(27)

[TopicP [Topic′ [TopicFocusP Opj / floresj [TopicFocusP ofereceuv [IP o Pedro tv tj à Joana ]]]]]

Again, as expected, the two focused constituents — the subject and the indirect


37

object — would occupy their respective positions inside IP. Another empirical effect of (25) is that, depending on the pragmatic choice by the speaker in topicalizing one given constituent but not the other, as a true topic in a TopicP, different word orders obtain. From these different orderings, it results in turn that some constituents, but not others, e.g. those that occupy the Spec,TopicFocus position, but not those that are in TopicP, are in stricter relation with the focus phrase, functioning as their subjects. The null non-quantificational anaphoric operators in (25a) are the empty counterpart of the topic-like element in (25b) — o Pedro and as flores —, and will bind their variables inside IP as the lexical elements do.25 Note that the pro element in subject position within IP cannot be an overt resumptive pronoun. This is expected. Since it occupies a position inside IP, if it was phonetically realized it would qualify as focused element, in conflict with its topic status. Inversely, the object pro can be bound by an overt clitic on the verb: this is allowed, because the  does not remain inside IP. Since Portuguese does not have subject clitics, the same strategy is not available for the subject. The ungrammaticality of (25f) also follows: the structure is ruled out because a constituent that is a topic-like element is in a position where only presentational focus constituents are licensed. I leave the reader with the task of considering other focused constituents and of evaluating the predictions of our hypotheses on them26. 5.2 BE-Focus structures The analysis presented so far has accounted for structures where focus is licensed through movement — either movement of a [+topicf, +focust] element to Spec, TopicFocusP ( ) or verb movement from IP to the head of TopicFocus ( ). Interestingly, then, in some of these structures, the verb ser ‘to be’ — and no other verb — appears in pre-focus position, emphasizing the ⁄  interpretation (the only possible). These structures have in common with the so-called (pseudo-)cleft constructions the fact that in both the presence of BE emphasizes the exclusive/restrictive reading of the focused constituent, but they differ in several respects, namely, contrary to (pseudo-)clefts, in these structures — henceforth BE-Focus structures, BE is not associated with the complementizer que ‘that’. Let us examine these constructions. Take first  focus structures. BE is excluded from this context, from all positions, as exemplified below:

38

MANUELA AMBAR

(28)

a.

A tarte … a Maria comeu (o que mais comeu não sei…) The pie … Mary ate (what else she ate, I don’t know…) b. *Foi a tarte … a Maria comeu. c. *A tarte foi … a Maria comeu

But  structures allow the presence of BE in a pre-focus position if the focused element is the object (cf.(29a)) but not if it is the subject (cf.(29b)) unless a clitic is present (cf.(29c)): (29)

a. A Maria comeu foi a tarte b. *A tarte comeu foi a Maria c. A tarte, comeu-a foi a Maria the pie ate it was Mary

Why should the insertion of BE be possible in structures like (29a,c) but not in (29b)? Suppose we assume that in these structures BE is a kind of expletive that can appear in a chain headed by a lexical verb. Then, a first requirement for it to occur is the presence of an available position; a second one is that it has to be in a given relation with the main verb — it has to be coindexed with and preceded by it, very plausibly through c-command or Spec-head agreement. Before BE insertion, contrastive and presentational focus on the object receive representations in (30) and (31) respectively: (30)

[TopicP [Topic″ [TopicFocusP a tartei [TopicFocus′ [IP a Maria comeu tj]]]]]

(31)

[TopicP [Topic″ [TopicFocusP a Mariai [TopicFocus′ comeuv [IP ti tv a tartei ]]]]]

BE cannot appear in (30) — the only position available for BE would be the head of TopicFocusP, but this position is not connected to the verb that has stayed in situ and therefore would not c-command BE. In (31), however, the lexical verb has raised, consequently it c-commands its trace inside IP. One possibility is to assume that this trace is lexicalized as a resumptive ‘proverb’. However, we would prefer BE to not occupy a position inside IP, since this is the location for presentational focus elements and foi ‘was’ is not a focused constituent, but rather an element that introduces focus. Before having a decision on this, let us observe the subject/object asymmetry exemplified in the minimal pair (29a)–(29b). Compare (31) above with (32) below, the representation of (29b): (32)

[TopicP [Topic″ [TopicFocusP a tartei [TopicFocus′ comeuv [IP a Maria tv ti ]]]]]


39

In (32) there is no available position for BE that would allow BE to precede the focused element a Maria.27. We derive the ungrammaticality of (29b). But things are still more interesting when we compare (29b) with (29c). Recall that the presence of the clitic in these structures is a diagnostics for deciding which position the left dislocated element occupies. According to our analysis,28 two hypotheses are available for deriving (29c): VP movement 29to Spec,TopicFocusP in one step, followed by movement of the object to TopicP, as in (33) below, or movement of the object followed by movement of the verb in different steps as in (34): (33)

[TopicP a tartei [Topic″ [TopicFocusP [ [comeuv -a [proi]]k [TopicFocus′ foiv [IP a Maria tk ]]]]]

(34)

[TopicP a tartei [Topic″ comeuv -ai[TopicFocusP [TopicFocus′ foiv [IP a Maria tv proi ]]]]]

In both structures the object is a true topic element and ends up in TopicP being then reduplicated by the resumptive clitic. Whichever of these analyses proves to be the best in terms of economy30, from both it seems possible to derive (29c) — the activation of TopicP provides an available position for the copula. In (33) TopicFocus0 is empty; therefore BE can be inserted and it will be coindexed with the verb in Spec,TopicFocusP, through Spec-Head agreement. As for (34), we have to assume that further verb movement to TopicP is necessary; the expletive will then be the lexicalization of the trace in the head of TopicFocusP, which will be c-commanded by and coindexed with the verb. Once movement of the object is visible and objects in TopicP are reduplicated by clitics, these structures provide empirical evidence for movement of the topic-like element to TopicP, in BE-focus structures where the focused element is the object:31 (35)

[TopicP a Mariai [Topic″ comeuv [TopicFocusP [TopicFocus′ foi v [IP proi tv a tartei ]]]]]

(35) differs from (12a–b): whereas in the latter the topic-like element can be null, in (35) the acceptability is improved if that element is phonetically realized, this confirming its topic nature and its function of identifying pro in subject position. If foi is a resumptive form coindexed with the lexical verb, we predict that their tenses will match — a prediction empirically borne out. We also successfully account for the intriguing distribution of BE in these structures; these facts provide then further empirical evidence for our unified treatment of focus.

40

MANUELA AMBAR

We have suggested that to focalize a constituent means to identify an Event and that in the grammar, this is accomplished by a given relation the verb, more precisely the Tense carried by the verb, establishes with the focused element. Suppose we formulate this intuition in the following terms: (36)

   (FTI): At the latest at LF, a full exclusive/restrictive focused element has to be under the scope of a verb, more precisely under the scope of its Event component, where scope is defined either in terms of c-command or of Spec-Head agreement and Event is carried by Tense on the verb.32

The only verbal form that can reduplicate the lexical verb is BE, since its intrinsic lexical properties do not block the transmission of the lexical properties of the main verb, necessary for focus. In other words, the only verb that can accomplish this task must be a   verb, which has nothing but identificational properties, provided by the Tense marks it bears, which are needed for focus identification, as (36) requires. Notice finally that, as desired, in all these structures the focus interpretation is the restrictive one. We have already suggested that this reading obtains whenever the focused element is under the scope of the verb. In all these sentences, the focused element is under the scope of the copula which has the same properties the main verb has, transmitted through matching (coindexation) of their Tense marks. The analysis of BE-Focus structures presented in this section, namely the content of (36), receives further motivation when applied to the treatment of the so-called (pseudo-)cleft constructions and of focus in ergative structures. Unfortunately, severe restrictions on the extention of this paper preclude the presentation of these analyses here, which, nevertheless strongly support the proposal outlined so far. A detailed treatment of different intriguing, but interesting, aspects of those constructions, namely Tense restrictions on (pseudo-) clefts, can be found in the extended version of the present article (cf. Ambar 1997). As already announced, the next and last section deals with some of those structures that have been considered in the literature as focus structures and that I will suggest correspond to another type of phenomenon, whose licensing is accomplished in another Functional Projection located above FocusP, but below CP. I will call that projection Evaluative Phrase.


6.

41

TopicFocusP vs. EvaluativeP

Recall what we observed about (15)–(16). Observe (37): (37)

Isto fazem os reis This do the kings

What is focused is the subject reis; this sentence is an adequate answer to question Quem faz isto? ‘Who does this?’, where the wh-constituent is the subject, but not to Que fazem os reis? ‘What do the kings?’, where the whconstituent is the object. Notice further that isto ‘this’ in (37) cannot receive contrastive focus interpretation, which, as we have seen, is incompatible with inversion. (37) is then a presentational focus structure in which: (i) isto is a topic-like element in Spec,TopicFocusP; (ii) os reis is a presentational focused element in Spec,IP and (iii) the verb has raised to the head of TopicFocusP. If a clitic is present, enclisis is obligatory: only (38a) is an adequate answer to Quem lhe disse isto? ‘Who them said this?’: (38)

a.

Isto disseram-lhe os reis This said them the kings b. *Isto lhe disseram os reis

In order for isto in pre-verbal position to receive focus, it would have to receive non-exclusive/non-restrictive contrastive focus, but then inversion would be excluded and still enclisis would be required: (39)

a.

Isto … os reis disseram-lhe This … the kings have said to them (what more they may have said, I don’t know…) b. *Isto … disseram lhe os reis c. *Isto … os reis lhe disseram

We conclude that, contrary to what has been assumed in the literature, Focus structures — both presentational and contrastive — trigger . A second observation concerns sentences like (38b) which, in other contexts, and with a different interpretation, are well-formed: (40)

a. b.

Isto lhe fazem os reis! This him do the kings Isso lhe disse eu! That him said I

42

MANUELA AMBAR

c. d.

Muitos livros lhe ofereceu o Pedro! Many books him offered Peter Até ao Pedro o apresentaram eles! Even to Peter him introduced they

The paradigm above has some similarity with the one in (41) below: in both cases, an exclamative value is present, expressing an evaluation by the speaker on a given state of affairs: (41)

a. b.

Belo trabalho me fizeste tu! Nice work to me did you Linda casa lhe comprou o pai! Beautiful house him bought the father

Here the subject in post-verbal position does not bear the presentational focus interpretation as it does in (38a). This difference can clearly be seen in the following minimal pair: (42)

a. b.

Isso This Isso This

disse-lhe said him lhe disse him said

eu I eu! I

Only (42a) is an adequate answer to question Q lhe disse isso? ‘Who said him that?’ — where eu ‘I’ introduces a new referent in the Universe of Discourse; in (42b) if something  is introduced is the relation of predication established between object-subject-predicate, which expresses an appreciation by the speaker. However, also in (42b) a focal-type stress is assigned to the subject, just as in presentational focus structures. (42) illustrates then two different constructions: (42a) is a   structure; (42b) an  construction. Limitations of space preclude the detailed treatment of these constructions, in this work. I will consequently just point out some of their crucial properties. Note that elements that enter evaluative structures like the ones in (41) and (42b) are submitted to lexical restrictions, whereas focus structures are not: (43a) being evaluative needs an  element in the dislocated phrase, but (43b) being a   does not need it: (43)

a. *A casa lhe comprou o pai! A casa comprou-lhe o pai


43

Observe now some evidence drawn from quantification. Barwise and Cooper (1981) consider many (= muitos) as a non-logic determiner, as opposed to every (= todos) — a logic quantifier, insofar as its definition cannot be stated independently of the evaluation models. Assuming this and that the structures we are calling  have to be distinguished from focus constructions, the following contrasts follow: (44)

Muitos livros lhe ofereci eu! Many books him offered I

(45)

a. *Todos os livros lhe ofereci eu! All the books him offered I b. *Ambos lhe apresentei eu! Both to him introduced I

Structures in (44)–(45) are  constructions. This is why evaluative elements, like adjectives, are required. The contrast between (44) and (45) finds an explanation: muitos ‘many’, but not todos ‘every’ or ambos ‘both’, qualifies as evaluative and can occur in constructions of this type.33 The question now turns out to be why sentences like (45) cannot be syntactically derived, whereas (44) can. Suppose we assume that there exists a projection in the syntactic representation of the sentence where - elements are licensed (checked) siting above IP but below CP, as in (46): (46)

[CP [C′ [EvaluativeP [Evaluative′ [TopicFocusP [TopicFocus′ [IP ]]]]]]]

Assume that the features of E(valuative)P have to be checked against evaluative features of lexical items. If EP is projected, then the appropriate lexical constituent — e.g. constituents headed by muitos, which have the relevant evaluative feature, but not by todos, which does not have such a feature — has to raise to it for checking reasons. Take for instance (44) above, where movement of muitos livros is visible (given its object status); its representation would be as in (47): (47)

[CP [C′ [EvaluativeP Muitos livrosi [Evaluative′ [TopicFocusP ti [TopicFocus′ lhe ofereciv [IP eu tv ti ]]]]]]]

Evidence drawn from the distribution of adverbs and subjects is crucial for establishing all the properties of this projection.34 Note that the subject in post verbal position is the rightmost embedded position (cf. Cinque 1993) to which a focal-type stress is assigned — this explains why these constituents have been confused with focused elements; it is

44

MANUELA AMBAR

also why they become emphatic elements. Therefore, I will adopt Rouveret’s (1996) proposal concerning the distribution of enclisis vs. proclisis:35 “At [LF] QPs can and must trigger agreement with a head hosting an operator feature. Whatever the identity of this head is, it can be assumed that, like other operator features, its feature can only be satisfied by a verbal unit, hence by a clitic-verb combination, not by a verb-clitic combination.” (Rouveret 1996: 29) Focus constructions have recourse to null or overt non-quantificational anaphoric operators, which do not have to be related to a verbal unit, and therefore enclisis is expected to occur. As for evaluative phrases, I will assume that they behave as QPs and, consequently, have to be related to a verbal unit, and that therefore enclisis is excluded.

7.

Conclusion

The main goal of the analysis presented here was to give Focus a unified treatment. As we went through the analysis of data, the large array of apparently diverse facts started becoming coherent and principled motivated. We were then led far beyond our initial research purposes. The submission of the description to the systematic use of the same diagnostics for clearly defining focus has permitted us not only to introduce different relevant concepts — e.g. non-exclusive/non-restrictive contrastive focus, opposed to the exclusive/restrictive one, but also enabled us to uncover new data and to clarify some misleading interpretations of some so-called focus constructions, namely, to precise the distinction between focus structures and evaluative structures. We have named the projection where evaluative features are checked Evaluative Phrase. The consideration of such a projection in the left periphery of the sentence will lead, as we expect, to a finer definition of what CP and TopicFocusP (or FocusP in other languages) are. Concerning focus structures, we have reached the following conclusions. In Portuguese, Focus is uniformly checked in a projection we called TopicFocusP; true Topics are checked in TopicP. TopicFocusP has both topic and focus features; focused constituents can be either subjects or complements of that projection. If a constituent is marked [+topicf +focust] (the relevant features concerning TopicFocusP) it will raise to Spec,TopicFocusP to check both features, with the consequence that inversion does not show up in this context —


45

the case of  . If a constituent is marked [topicf] it will be able to check just the topic feature of the head; then the focus feature has to be checked by the Verb (the only available head with that capacity); once the focus feature of the head is checked by the verb, all the elements inside IP (the complement of TopicFocus°) will qualify as focused constituents, with the empirical effect that subject-verb inversion will show up whenever the focused element is the subject — this is  . Two focus interpretations are available: (i) exclusive/restrictive and (i) nonexclusive/non-restrictive. The first obtains whenever the focused element is under the scope of the verb, scope being defined in terms of spec-head and of ccommand relations; the second one whenever none of those relations is satisfied. Presentational focus will always be exclusive/restrictive, because the focused element is always c-commanded by the verb. Contrastive focus can be either exclusive or non-exclusive. BE-Focus structures and the so-called (pseudo-)cleft constructions are focus structures. BE will never carry event specifications of its own, but it inherits them from the lexical verb. In the first type of structures, BE will occur as a resumptive in a chain headed by the lexical verb; in the second one, it is the carrier of Tense which introduces the focus structure. Condition (36) (  ) has stated that restrictive focused constituents have to end up under the scope of the verb (specifically, under the event specifications of the verb) at the latest, at LF. We hope that our proposal will help to identify the parameters of variation of Focus constructions across languages. When discussing, for example, the differences in the distribution of clitics between Italian and Portuguese in the socalled CLLD structures, we observed that, our analysis states that what Italian seems to lack is a long distance identification of pro. Further research on the distribution of adverbs, testing word order possibilities, will certainly shed light on the adequacy of the analysis presented for focus structures, on the scope of a projection like Evaluative Phrase and, consequently, on its contribution for a more precise definition of other related projections, namely CP.

Notes * I thank the audience of the International Workshop on Focus, which took place in Paris in February 1996 for important comments on an earlier version of this paper. For relevant support

46

MANUELA AMBAR and discussion I am very grateful to Jacqueline Guéron, Hans Obenauer, Iris Pereira, João Peres, Jean-Yves Pollock, Alain Rouveret, Laurie Tuller, and Rita Veloso (last but not least — without her systematic help and insight, this work would not have been ready in time). This work was funded by the Fundaçad para a Ciência e Tecnologia (Project PCSH/C/LIN/936/95).

1. When this work was almost concluded, I presented in Budapest the paper “Focus and Movement in Portuguese vs. Hungarian”, departing from work by Kiss (1996). Limitations of space have precluded the inclusion of that study in this work (to appear in the Proceedings of the 20th Anniversary of Portuguese Teaching in the Universities of Hungary, Budapest). 2. For instance, topic/comment (Hockett 1958); thema/rhema (Halliday 1967); categorical/thetic judgments (Kuroda 1972, 1992); declarative/existential (Babby 1980); declarative/presentational (Suñer 1982) etc.. 3. This is why there is an implicit negation of what was previously said. 4. Since no movement is visible, I call this type of focus  , even though I think that very plausibly movement also applies in these structures, but covertly (cf. Ambar 1988). 5. For the distinction between  and  , see Guéron (1980) and Cinque (1993). 6. Sentences like (7e), where the topic element appears at the right of the verb have been considered in the literature as well-formed sentences (cf. Zubizarreta 1993, for Spanish). In Portuguese they are clearly excluded. 7. For a discussion on word order in Portuguese, see Ambar (1988), where different argumentation is presented departing from works by Greenberg (1963) and Ross (1970). The observations in (i)–(ii) were also presented in that work. 8. In some structures, under some conditions, old information can follow the verb in the linear order, e.g. in the so called (pseudo-)cleft constructions. 9. Restrictive and exclusive are equivalent terms for this concept. Contrastive focus can also be restrictive/exclusive as in the following pair: Onde está o meu casaco? O teu casaco, a Joana levou. ‘Where is my coat? Your coat Joana took away’. Our analysis covers also this case (see Ambar (1997). 10. To my ear there is no marked accent on a tarte in (13) as in prosodic focus (cf. (2–4)). The two points following the focused constituent are intended to mark the peculiar prosody of these structures. 11. A more accurate discussion is addressed in the extended version of this paper (cf. Ambar 1997). 12. As in Ambar (1988), I will assume that the so-called free inversion is an instance of a presentational focus structure in the context of a question/answer pair. 13. For commodity of exposition this node is henceforth labeled IP, independently of the discussion about the c-commanding node of sentence structure — AgrSP or TP — and about the existence or not of an Agr node. Cf. Pollock (1989), Belletti (1990) and Chomsky (1995) for qualifications. 14. The presence of an R-expression in this context would be excluded just like in the context of relative structures. As for the choice of CP as the landing site for the null operator and the verb, motivation was drawn from extraction in this context. These structures are incompatible with


47

wh-movement. We will assume, however, that CP, very plausibly, coexists with the projection where focus is licensed (among other projections) and that the opacity effects manifested in focus wh- contexts have to be derived in another way. 15. The definition of topic-like element will become clearer below. As we will see, a tarte in (7), the lexical counterpart of OP in (19), is not a true topic, but just the subject of focus. Very plausibly languages will differ with respect to the nature of TopicFocusP. Hungarian, for instance, seems not to allow a topic-like element in Spec,TopicFocusP (only a focus constituent can occur there), this correlating with the impossibility of sentences like (7a)–(7b) in this language, which, however is a null subject language (cf. Ambar 1997 and fn.2). 16. For a more detailed treatment of these aspects see the extended version of this work (Ambar 1997). 17. The parametrisation responsible for the differences concerning word order in focus constructions across languages can then be attributed to the choice languages make with respect to the projection for focus licensing. In English and French, for instance, the projection has not what I will call topic features i.e. in these languages what is projected is FocusP. I cannot pursue this here for space reasons. 18. For the relation between focus and Event see also Zubizarreta (1993), where a focus phrase has to end up under the scope of AspectP at the latest at LF in order to be Event-related. I also claim that event is involved in focus licensing, although I do not consider the existence of an Aspect projection — instead I have suggested that a T(ense)O(bject)P is a necessary projection for establishing the relevant relations between Aktionsart, object determination/quantification and tense morphology. In my proposals it is claimed that Aspect is an epiphenomenon (cf. Ambar (1996), (1997) for qualifications). 19. Or by an empty Event operator, as in answers to What happened? In this case the entire IP is focussed. (cf. Ambar (1997)). We will see that the TopicFocus head can be lexicalized, under given conditions, by inserting the only verb that lets through those intrinsic properties of lexical verbs: BE. 20. Exept in cases were the features are checked by Event operators (cf. fn 19 and Ambar (1997)). 21. For the reasons related to Event already pointed out. In fact, it is the verb that introduces events. Note that the verb capacity for checking focus corresponds to an intrinsic property of verbs, not to an optional choice. Very probably, the prominent feature for TopicP is nominal in nature; the one of FocusP is verbal. 22. Italian needs a clitic to identify pro and to link it to its antecedent. Note that this pro is reminiscent of Cinque’s (1984) and Obenauer’s (1984) pro- which occurs in long whextraction structures, avoiding islands effects. If, as we are observing, the two options (overt vs. empty) coexist, the so-called Avoid Pronoun Principle (cf. Chomsky 1981) remains mysterious. Further research is necessary for an understanding of this phenomenon, very plausibly in the direction put forward by Montalbetti (1984). Notice that the same speakers who prefer sentences (ia) to (ib) below as an answer to the question Quem vestiu o meu vestido? (“Who put my dress on?”): (i)

a. b.

O teu vestido,, A MARIA… vestiu-o … (não sei quem mais vestiu….) Your dress,, MARY… put it on … (I don’t know who else put it on …) O teu vestido,, A MARIA… vestiu … (não sei quem mais vestiu…)

48

MANUELA AMBAR do not accept (iia) as an answer to Quem comeu a tarte? (“Who ate the pie?”); in this case they prefer, (iib): (ii)

a. *?A MARIA … comeu-a … (Os outros não sei…) Mary … ate it … (about the others, I don’t know …) b. A MARIA,, comeu … (Os outros não sei…)

The oddity of (iia) is expected under our analysis: these structures are associated to a nonexclusive/non-restrictive contrastive focus reading, i.e. the speaker assumes that other entities may have eaten the pie; consequently “the pie” cannot have been totally eaten by the entity the focused phrase denotes, only a part of it can. It follows that only a pronominal  clitic would be able to refer such a ‘part’ of the entity “pie”, which has this ‘divisibility’ property. Portuguese pronominal system does not include such a kind of partitive clitic pronoun, the result being that the choice left is to use the empty counterpart of such a non-available overt partitive clitic. In (i) this problem does not arise, since o vestido ‘the dress’ cannot be dressed in a ‘partitive’ way; consequently the definite object clitic pronoun can occur and the structures where it occurs are improved for those speakers who prefer the overt resumptive strategy to the empty one. For an accurate analysis of topicalization in Portuguese see Duarte (1987, 1996). Agreeing with Duarte (1987), Raposo (1986) also shows that, with respect to these structures, Portuguese systematically differs from other Romance languages. In Raposo’s proposal the parameter opposing Spanish (and other Romance languages) to Portuguese is derived from the assumption that in the latter, but not in the former type of languages, the class of determiners includes a null element — plausibly a null expletive D, which also heads a DP with a null pro complement. The behavior of Portuguese and its differences with respect to other Romance languages are derived from licensing requirements of pro in the structure [D pro]. For our purposes here, it is sufficient to assume that the main difference underlying structures where the clitic does not occur vs. structures where it appears concerns the different status of movement and, consequently, of the gap it produces: an operator-variable relation vs. a topicpro one, in the spirit of Duarte (1987, 1996). As we will see, the position responsible for an operator-variable relation (therefore, absence of clitic) is Spec,TopicFocusP; the one responsible for the presence of the resumptive clitic is Spec,TopicP. 23. Although we are not concerned with the semantics of these constructions here, non-restrictive contrastive focus could, plausibly, be semantically analyzed in terms of l abstraction, thus providing an interesting analysis of its syntax / semantics interface. 24. Lack of space precludes the presentation of other focus contexts in which the proposal outlined makes good predictions, such as cases of contrastive focus with a restrictive/exclusive reading, where arguably the verb raises to the head of TopicFocusP and ends up in a spec-head relation with the contrastive focused element (cf. Ambar 1997, the extended version of this paper). 25. For a more detailed discussion, see Ambar (1997). 26. In sentences of the type Comeu a tarte, a Joana ‘Ate the pie Joana’ or Ofereceu as flores à Joana o Pedro ‘Offered the flowers to Joana Peter’, I assume that the entire VP moves to Spec, TopicFocusP, an option available whenever the moved elements form a constituent, but not otherwise (cf. Ambar 1997 for qualifications).


49

27. If the copula appears in the position occupied by the verbal trace, it would not precede any phonetically realized material (a condition for having focus). 28. See Ambar (1997) for detailed discussion. 29. In this case the moved elements form a constituent. Therefore the subject/object asymmetry manifested in these structures is derived from the fact that the entire VP can be moved (as in focused subject structures), but the subject + verb cannot, since they do not form a constituent (as in focused object structures). 30. In fact we do not have an algorithm to evaluate the cost of a derivation. 31. Note that sentence represented in (35), where the subject is in TopicP is more natural than the ones where TopicP is filled by the object as in (33–34), even with the clitic. Besides the question of cost of the derivation, pro in subject position is also more natural than in object position in Portuguese. The contrast then follows. 32. Zubizarreta also proposes that presentational focus constituents have to be under AspectP. In our analysis, however, we do not claim that the focused element is under AspectP (or TobjectP); it can be in different positions inside IP (presentational) or in Spec,TopicFocusP (contrastive). The effects of (36) are derived in some environments by Tense raising at LF. Note that condition (36) concerns full focused constituents — the ones where the exclusive/restrictive interpretation obtains. In fact, we do not want non-exclusive focused elements like (8) or (23) to have an exclusive interpretation — if Tense would raise in those structures at LF this interpretation would be available. Recall that non-exclusive focus is not in a sense a true focus, since it only gives a partial answer to the addressed question; note further that contrastive focused elements also have topicf features; consequently, once (36) is a focus licensing requirement, it would be undesirable to have it applying on topic-like elements. 33. For a treatment of these quantifiers in Portuguese see Peres (1987). Discussing properties of quantifiers, Peres (1987) assumes that in the context of predicates of the same type, quantifiers like muitos are always distributive. We will adopt Peres’s proposal — and, consequently, Rouveret’s analysis of proclisis vs. enclisis, against Martins’s 1995 proposal. Whatever the analysis of this type of quantification turns out to be, it seems to me that Peres is right concerning the distributive reading of these expressions, independently of the occurrence of enclisis or proclisis. But quantifiers like muitos differ from other quantifiers, e.g. todos, in that the former but not the latter can have an adjectival status. Note, for instance, that, in contrast with todos, muitos allows degree variation: muitíssimos vs. *todíssimos. 34. In work in progress, we show that adverbs, being verbal, move to this position in some contexts. Note that only elements that bear an evaluative feature move: if the verb is not assigned such feature it does not, this correlating with the different available interpretations. It is why in these sentences inversion can be dispensed with, as in Muitos livros EU lhe ofereci. I am proposing that these sentences are possible because the verb does not bear that evaluative feature, it has however a focus feature and, consequently, moves to the head of TopicFocusP; EU — which in this sentence receives a constrastive-type stress moves to Spec,TopicFocus P. This sentence receives then a contrastive focus interpretation that is not available in (47). For a different proposal for these constructions see Raposo (1995), who proposes that the two different positions for the subjects in these constructions are Spec,IP and Spec,VP (then the different informational value of these sentences cannot be derived). For the author proclisis

50

MANUELA AMBAR obtains whenever Spec,FP (which also covers Focus) is filled; enclisis when it is empty. Our description of the facts has shown that in contrastive focus, for instance, the focused element clearly is in Spec, and still enclisis is obligatory.

35. Following work by Duarte & Matos (1995).

References Ambar, Manuela. 1988. Para uma Sintaxe da Inversão Sujeito-Verbo em Português. PhD Diss., Univ. de Lisboa. Published by Ed. Colibri, Lisboa, 1992. Ambar, Manuela. 1994. “Aux-to-Comp and Lexical Restrictions Verb Movement”, in G. Cinque, J.Koster, J.-Y. Pollock, L. Rizzi and R. Zanuttini (eds.), Paths Towards Universal Grammar: Studies in honour of Richard Kayne. Washington D.C.: Georgetown University Press. Ambar, Manuela. 1996. “Inflected Infinitives Revisited — Genericity and Single Event”. Canadian Review of Linguistics. Ambar, Manuela. 1997. “The Syntax of Focus — a unified approach”, ms. Univ. of Lisbon. Ambar, Manuela, and Vasconcelos, Manuela. 1983. “O Lugar da Gramática no Ensino de uma Língua”. Actas do Congresso sobre a Situação Actual da Língua Portuguesa no Mundo, Vol. II, Lisbon: Instituto de Cultura Portuguesa, 666–679. Babby, L. 1980. Existential Sentences and Negation in Russian. Ann Arbor, Mich.: Karoma. Barwise, J., and Cooper, R. 1981. “Generalized Quantifiers and Natural Language”. Linguistics and Philosophy 1(1).145–150. Belletti, Adriana. 1990. Generalized Verb Movement. Torino: Rosenberg and Sellier. Brody, Michael. 1990. “Some Remarks on the Focus Field in Hungarian”. University College London Working Papers in Linguitics 2.201–225. (Published as “Remarks on the order of elements in the Hungarian focus field”. Approaches to Hungarian, Vol. 3, JATE.) Calabrese, Andrea. 1985. “Focus and Logical Structures in Italian”. Ms., MIT. Calabrese, Andrea. 1990. “Some Informal Remarks on Focus and Logical Structures in Italian”. Ms., Harvard University, Cambridge, Mass. Casteleiro, João. 1977. “Sintaxe e Semântica das Construções Enfáticas com É QUE (1ª e 2ª partes. Boletim de Filologia XXV. 14 (1979)97–166.)


51

Casteleiro, João. 1981. Sintaxe Transformacional do Adjectivo. Lisbon: INIC. Chomsky, Noam. 1977. “On wh-Movement”. In P. Culicover, T. Wasow & A. Akmajian (eds.), Formal Syntax. New York: Academic Press, 71–132. Chomsky, Noam. 1986. Barriers. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Cinque, Guiglielmo. 1984. “A′-Bound pro vs. Variable”. Ms., University of Venice. Cinque, Guiglielmo. 1990. Types of A′ Dependencies. Cambridge, Mass.: MIT Press. Cinque, Guiglielmo. 1993. “A Null Theory of Phrase and Compound Stress”. Linguistic Inquiry 24. 239–297. Culicover, Peter, and Rochemont, Michael. 1990. English Focus Constructions and the Theory of Grammar. Cambridge: CUP. Duarte, Inês. 1983. “Variação Paramétrica e Ordem dos Clíticos”. Revista da Faculdade de Letras 158–178. Duarte, Inês. 1987. A Construção de Topicalização na Gramática do Português: Regência, Ligação e Condições sobre o Movimento. PhD Dissertation, University of Lisbon. Duarte, Inês. 1996. “Topicalização em Português Europeu: Uma análise comparativa”. In Inês Duarte and Isabel Leiria (eds.), Actas do Congresso Internacional sobre o Português, Vol. I. Lisbon: Associação Portuguesa de Linguística and Ed. Colibri. Duarte, Inês, and Matos, Gabriela. 1995. “Romance Clitics and the Minimalist Program”. Ms., University of Lisbon. Giorgio, Alessandra, and Pianesi, F. 1992. “For a Syntax of Tense”. Ms., University of Catania and IRST, Povo (Trento). Greenberg, Joseph. 1963. “Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements”. In J. Greenberg (ed.), Universals of Language. Cambridge, Mass.: MIT Press. Guéron, Jacqueline. 1980. “On the Syntax and Semantics of PP Extraposition”. Linguistic Inquiry 11. 637–678. Halliday, M. 1967. “Notes on Transititivity and Theme in English”. Journal of Linguistics 3.199–244. Heim, Irene. 1982. The Semantics of Definites and Indefinites Noun Phrases. PhD Dissertation, University of Massachussetts, Amherst. Hockett, C. 1958. A Course in Modern Linguistics. London: MacMillan,.

52

MANUELA AMBAR

Horvath, Julia. 1986. Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris,. Horvath, Julia. 1995. “Structural Focus, Structural Case and the Notion of Feature-Assignment”. In Kiss (ed.). Kayne, Richard. 1993. “Toward a Modular Theory of Auxiliary Selection”. Studia Linguistica 47.3–31. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press. Kiss, Katalin É. (ed.) 1995. Discourse Configurational Languages, Oxford: Oxford University Press. Kuno, S. 1975. “Conditions for Verb-Phrase Deletion”. Foundations of Language 13.161–175. Kuroda, Yuki. 1972. “The Categorial and the Thetic Judgment: Evidence from Japanese syntax”. Foundations of Language 9.153–185. Kuroda, Yuki. 1992. Japanese Syntax and Semantics. Dordrecht: Kluwer,. Laka, Itziar. 1990. Negation in Syntax: On the nature of functional categories and Projections. PhD Dissertation., MIT. Martins, Ana Maria. 1994. Clíticos na História do Português, PhD Dissertation, University. of Lisbon. Montalbetti, Mario. 1984. After Binding: On the Interpretation of Pronouns. PhD Dissertation, MIT. Obenauer, Hans-Georg. 1984. “On the Identification of Empty Categories”. The Linguistic Review 4.2. Ortiz de Urbina, Jon. 1989. Parametrs in the Grammar of Basque. Dordrecht: Foris. Peres, João. 1987. Para uma Semântica Formal da Quantificação Nominal NãoMassiva. PhD Dissertation, University of Lisbon. Pesetsky, David. 1982. “Complementizer-Trace Phenomena and the Nominative Island Condition” The Linguistic Review 1(3).297–343. Pollock, Jean-Yves. 1989. “Verb Movement, Universal Grammar and the Structure of IP”. Linguistic Inquiry 20.365–424 Raposo, Eduardo. 1986. “On the Null Object in European Portuguese”. In O. Jaeggli & C. Silva-Corvalán (eds.), Studies in Romance Linguistics. Dordrecht: Foris. Raposo, Eduardo. 1995. “Clitic Position and Verb Movement in European Portuguese”. Ms., University of California, Santa Barbara.


53

Raposo, Eduardo. 1997. “Some Definite/Zero Alternations in Portuguese: Consequences for the analysis of Topic Constructions”. Ms., Univ. of California, Santa Barbara. Rizzi, Luigi. 1995. The Fine Structure of the Left Periphery. Ms., University of Geneva. Rochemont, Michael. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins. Ross, John. 1970. “Gapping and the Order of Constituents”. M. Bierwisch & K. E. Heidolph (eds.), Progress in Linguistics. The Hague: Mouton. Rouveret, Alain. 1992. “Clitics, Morphological Checking and the Wachernagel Position in European Portuguese”. Ms., University of Paris VIII. Rouveret, Alain. 1996. “Clitics, Subjects and Tense in European Portuguese”. Ms., University of Paris VIII. Suñer, Margarita. 1982. Syntax and Semantics of Spanish Presentational SentenceTypes. Washington D.C.: Georgetown University Press. Tuller, Laurice. 1992. “The Syntax of Postverbal Focus Constructions in Chadic”. Natural Language and Linguistic Theory 10(2).303–334. Uriagereka, Juan. 1992. “Focus in Iberian Languages”. Paper presented at the 15th GLOW Colloquium, Lisbon, GLOW Newsletter 28. (Published as “An F position in Western Romance”, in Kiss (ed.) 1995.) Zubizarreta, Maria-Luisa. 1993. “Grammatical Representation of Topic and Focus: Implications for the structure of the clause”. Cuadernos de Linguistica del I. U. Ortega y Gasset.

Bound Focus or How can Association with Focus be Achieved without Going Semantically Astray?* Josef Bayer Friedrich-Schiller-Universität Jena

Abstract By the notion “bound focus”, we mean phonological prominence and semantic affectedness that is dependent on the presence of a focus-demanding lexical element such as only or even. This paper tries to demonstrate that modern conceptions of syntactic theory such as the Minimalist Program explain in a natural way the various word orders that can be observed. Crucial use is made of the idea that focus particles are either base-generated in a fixed functional position or brought into this position by covert raising or by reconstruction. In deviation from the operation of feature movement (“Move-F”), arguments are provided which suggest that overt movement may indeed include the movement of phrases.

1.

Goal

The goal of this article is to show that all cases in which a quantificational focusing particle, such as only, even and perhaps others associates semantically with a focused element in the clause, can be reduced to a canonical constellation in which the particle is the head of a “particle phrase” in which it binds the focus associate. In this case, the particle occupies an inalterable scope position. There are four cases that can be distinguished. The associative relation between particle and focus is achieved (i) directly by base-generation such that the particle is in its final scope position from where it binds its associate; (ii) by covert movement of a phrase such as only John to a position where the particle

56

JOSEF BAYER

is able to take (clausal) scope; (iii) by reconstruction of a phrase such as only John to a position where the particle is able to take (clausal) scope; (iv) by reconstruction of the focus associate to a position where the particle can bind it. The treatment, which makes crucial use of the theoretical machinery developed in the Minimalist Program (Chomsky 1995) leads to a considerable reduction of the complexity that has troubled much research on bound focus in the past. It can especially be shown that intuitions about phrase structural constituency can rather easily be reconciled with the desired semantics of focusing particles. Although the main points can be made on the basis of English, German data will be drawn into the discussion because this language offers certain test cases that may enhance our understanding of bound focus in general.

2.

Four Types of Constructions

I will use the category label PRT for “particle” and PrtP for “particle phrase”, i.e. a functional projection that is headed by the semantically relevant functional category PRT. The intended focus associate of PRT is indicated by capitals. Consider the example in (1a) with approximately the phonetic form given. It is ambiguous between the readings connected to the different foci in (1b) and (1c): (1)

a. b. c.

[3%n wgd ognlI InvaIt Ám7rI] John would [PrtP only [VP INVITE MARY]] John would [PrtP only [VP invite MARY]]

In (1b) only associates with the whole VP, while in (1c) it associates only with part of the VP. The semantics is roughly as in (2): (2)

a.

b.

For the set P of (contextually relevant) properties {phone Susan, visit Sarah, kiss Mathilde, …, invite Mary} that John would have, P is exhausted by invite Mary For the set P of (contextually relevant) properties {invite (Ann), invite (Barbara), invite (Carol), …, invite (Zeldah)} that John would have, P is exhausted by invite (Mary)

Due to focus spreading, the two readings correspond to one and the same phonetic form. This is the reason for the ambiguity. PRT is a functional head which occupies a pre-VP operator position. Its complement — the VP — is a C F C (CFC) in the sense of Chomsky (1986). I

BOUND FOCUS

57

assume that it contains a trace of the subject that has been moved out of VP for reasons of Case-licensing. If we follow the standard assumption that focus presupposes a set of entities against whose other members the focused element is contrasted, we achieve a split between foreground and background which determines the set that will be affected by the operator. In (1b) it is the set of all of John’s contextually relevant properties, while in (1c) it is the set of all of John’s contextually relevant properties of inviting someone. Thus, the focused element corresponds to a variable. Simplifying matters somewhat, only according to Rooth (1985) translates into lPlx [∀Q [[Q{x}] → Q = P]]. (1b,c) are readily converted to LFs that can be semantically interpreted: If P corresponds to the entire VP, the semantic translation that is yielded is [∀Q [[Q{John}] → Q = invite Mary]]; if P corresponds to the VP [invite y], the semantic translation that is yielded is [∀Q, Q = invite y [[Q{John}] → Q = invite Mary]]. This provides the core cases, where the LF is essentially read off the syntactic string directly. Consider now the following example where this is clearly not possible: (3)

John would invite [? only [DP MARY]]

If only is part of DP — something like a “modifier” of DP, it cannot head PrtP. PRT does not c-command anything like a CFC in this case. Thus, it is prima facie unclear how PRT can have propositional scope. Nevertheless, the meaning of (3) appears to be more or less the same as the meaning of (1c). Assume now that PRT is not evaluated in situ but rather undergoes covert raising to the standard operator position. In this case, there are two options: Either PRT itself moves to the head position of PrtP from where it can bind the focused element Mary, or the entire phrase only Mary moves to the specifier of PrtP (SpecPrtP) whose head it “identifies”. In the latter case, the carried-along focused element Mary has to be reconstructed into its original position. In each of these cases, an LF is created that can be interpreted as outlined above in connection with (1c). Before we move on to a more thorough investigation of this constellation, let us look at yet two other possibilities. Consider first the situation in which PRT is part of a DP as in (3), but where this DP is higher in the phrase marker than the purported pre-VP head position. In English, this situation holds whenever PRT is part of a DP in SpecIP; in German, whenever PRT is part of a DP (or any other phrase) in SpecCP: (4)

a. b.

[? only [DP MARY]] would invite us to her home [? nur [DP MARIA]] würde uns nach Hause einladen

58

JOSEF BAYER

If PrtP were headed as in (1b,c), these examples would not conform to the standard pattern, because in that case PRT should be able to associate also with material that is lower in the phrase marker. As Jackendoff (1972) has already observed and as shown by (5), this is impossible though: (5)

a. *Only Mary would invite US to her home / *… to her HOME b. *Nur Maria würde UNS nach Hause einladen / *… nach HAUSE einladen

This seems to be clear indication that PRT and DP form a single constituent in (5), and that PRT is not in a proper scope position yet. The situation in German is even clearer. Due to the V-S (V) C, the position in front of würde in (5b) must be a single constituent. In this case, it is expected that nur is part of the DP in SpecCP and will not c-command anything but Maria. Notice now that if PrtP is as in (1b,c), only Mary/nur Maria has to undergo lowering in such a way that it fills SpecPrtP and identifies the head of PrtP. Then, the focus associate Mary has to be lowered further to SpecVP. In this case it can be bound by PRT which is now in proper scope position. As will be shown below, this situation is naturally accounted for in the Minimalist framework. Consider finally a situation where PRT does occupy the propositional scope position, but where its focus associate has been raised higher such that it cannot be bound by PRT. With respect to the particle only, modern English shows a restriction here which had been noticed at least as early as Jackendoff (1972), but which seems to have been absent in earlier stages of the language. This is shown in (6): (6)

a. b.

ANNA could [PrtP even / ?*only [VP escape from the prison]] The eldest son shall only inherit his father (18th century) (Taglicht 1984: 97, n35)

While in (6a) the focused phrase Anna can easily associate with even, for many speakers association with only seems to be much harder or totally impossible. As the example in (6b) shows, however, this must be a more recent development of the English language. The only analysis of (6b) that achieves a natural interpretation is such that only associates with eldest.1 The German example from Primus (1992) in (7) shows that there is no restriction on nur as compared to sogar (and other particles) in modern German. (7)

ANNA entkam1 [PrtP sogar/nur [VP dem Gefängnis e1]] Anna escaped even/only the-DAT prison

BOUND FOCUS

59

Perhaps the reason lies in the fact that earlier stages of English and German have the V2–property while modern English has only residues of V2.2 If we ignore the problem with English only, the situation seems to be clear enough though: PRT is in situ, and the associate that has been raised to a higher position must be lowered to a position where it can be bound by PRT. Then, the only difference between (6)/(7) and (4) is that in the former PRT overtly occupies the propositional scope position in PrtP while in the latter it has to somehow reconstruct into this position. This concludes my overview of the four situations in which PRT can bind associated focus material: (i) PRT is the head of PrtP and binds a focus in its ccommand domain; (ii) PRT and some XP form a constituent, and either PRT must undergo covert raising to a proper scope position, or [PRT XP] move together to SpecPrtP covertly; (iii) [PRT XP] is “too high” in the phrase marker for PRT to be in a proper scope position, and must therefore be reconstructed into a lower position; (iv) PRT is in proper scope position while its focus associate is “too high”; thus, the associate must reconstruct into a lower position where it can be bound by PRT. I take (i) to be representative of the situation in which PRT occupies its ultimate scope position and c-commands its focus associate. The interpretation of bound focus is straightforward in this case as has been shown by Rooth (1985), Kratzer (1991) and others. I take it that at the level of LF the other constructions can be interpreted analogously because they all involve an abstract syntactic structure which reflects this elementary configuration.

3.

Characteristics of bound focus

In this section, I want to discuss characteristics of bound focus, some of which may be novel in research on this phenomenon. 3.1 PRT requires a focus constituent PRT always requires a focus associate. As the examples in (8) and (9) show, if it cannot bind any focus associate, PRT is not licensed: (8) (9)

a. b. a. b.

Sally likes only HIM / *it Sabine liebt nur IHN / *es Sally likes even HIM / *it Sabine liebt sogar IHN / *es

60

JOSEF BAYER

The pronoun him is stressed, but since it and es are notoriously unstressed, the sentences become deviant as soon as these weak or clitic pronouns are used. This property of bound focus follows from the semantics of PRT. PRT always requires a non-trivial set such that there is an  to the focused element. Such an alternative exists, for example, for full pronouns on different dimensions (gender: {he, she}, person {I, you, …}, deixis: {this, that}, etc.), but not for clitics etc.3 With respect to only, Tancredi (1990) proposes the P  L A (PLA):4 (10)

P  L A An operator like only must be associated with a lexical constituent in its c-command domain.

The PLA accounts for the ‘?*’ in (6a), but it is a lexical stipulation that does not capture the data on foci which may be bound by even. It would also fail to account for (6b) and for (7). Despite my present ignorance concerning the restrictions on only in modern English, I would favor the following generalized version of the PLA: (11)

P  L A ₍₎ At LF PRT must be associated with a lexical constituent Lm or a trace of Lm in its c-command domain such that Lm is a member of the set {L1, L2, …, Ln} where every Lm-1 is a discourse alternative to Lm.

Since (11) requires association only at the level of LF, it fails to capture the restriction on only. If some lexical constituent L is related to the c-command domain of PRT by a trace, this trace — which I will argue is an LF- of L — can be bound by PRT. In making a statement about the whole class of PRTs, (11) is, of course, more general than (10). It is also more explanatory because it makes a statement as to what “associate” means. As the discussion of only in section 2 has shown, its meaning requires a non-trivial set of alternatives. As far as I can see, this is also true for the additive particles even, also, too and a number of elements such as at least, at most etc. as well as for most of those German particles that have been studied extensively by Altmann (1976; 1978). 3.2 PRT attaches only to a potential Xmax Strings such as those in (12) and (13) give the impression that PRT may attach to any category that satisfies the revised version of PLA in (11).

BOUND FOCUS

61

(12)

a. b.

Some students smoke even IN the classroom Only LITTLE boys are permitted to use the ladies’ restroom

(13)

a. b.

Einige Studenten rauchen sogar IN dem Hörsaal Nur KLEINE Buben dürfen die Damentoilette benutzen

Since in English even and in German both sogar and nur can optionally be postposed, it is easy to test this impression. As (14) and (15) show, it is wrong: (14) *Some students smoke [IN even] the classroom (15)

a. *Einige Studenten rauchen [IN sogar] dem Hörsaal b. *[KLEINE nur] Buben dürfen die Damentoilette benutzen

A necessary though not always sufficient condition is that PRT is postposed behind the smallest Xmax that contains its focus associate. This yields grammatical results. (16)

Some students smoke [[IN the classroom] even]

(17)

a. b.

Einige Studenten rauchen [[IN dem Hörsaal] sogar] [[KLEINE Buben] nur] dürfen die Damentoilette benutzen

This suggests that PRT is a head that takes an XP as its complement. Additional data from German at first sight seem to contradict this generalization. In (18), nur seems to form a constituent with the category V0: (18)

a. b.

Peter hat das Buch nur DURCHGEBLÄTTERT Peter has the book only skimmed-through Nur DURCHGEBLÄTTERT hat Peter das Buch

As den Besten and Webelhuth (1987), Müller (1998) and others have shown, however, there is an analysis that does not force us to the conclusion that nur attaches to V0 here. German has the option of scrambling material out of VP. Thus, in (18) PRT may well be attached to VP, but this VP contains a scrambling trace of the direct object instead of the lexical object itself. This analysis is sketched in (19): (19)

a. b.

… das Buchi … [nur [VP ti durchgeblättert]] [nur [VP ti durchgeblättert]] … das Buchi

I conclude that PRT is indeed a head that follows X-bar theory closely enough to permit only an XP-complement. We will shortly turn to the question of how from structures like [PRT [… FOCUS …]] or [[… FOCUS …] PRT] a standard

62

JOSEF BAYER

interpretation can be derived which has PRT in proper (propositional) scope position from where it can bind the focus associate. 3.3 PRT is a “minor” functional head In which sense is PRT a regular syntactic (lexical or functional) head? Notice that functional heads usually select only one type of complement: C selects IP, I selects VP, D selects NP to mention the most popular cases.5 Looking at the list in (20), we must conclude instead that PRT is totally promiscuous: (20)

a. b. c. d. e.

only Bill only to London only poor only go to London only that he goes to London

nur Willi nur nach London nur arm nur nach London fahren nur daß er nach London fährt

DP PP AP VP CP

School grammar classifies particles as . Categorial grammar in the Montague tradition introduced them like other functional elements such as negation as  expressions.6 Both traditions are right in the sense that particles are non-inflecting minor categories, that they  their target rather than determine it, and that they nevertheless make an important contribution to semantic composition. But how should PRT fit into more restrictive versions of X-bar theory? As I have already done in Bayer (1996: ch.1), I want to follow a suggestion by Rothstein (1991) who introduces the category M F H (MFH). MFHs are heads which lack a q-grid, and which do not project categorial features. Thus, if an MFH takes XP as a complement, the resulting category will nevertheless be of syntactic type XP. This achieves exactly what syncategorematic introduction in Montague grammar achieves, but now on the basis of a deficient matrix of syntactic features such that PRT becomes amenable to the parsimonious principles of phrase structure composition. Following Rothstein, I assume for the cases in (20) the phrase structure representations [DP only [DP Bill]], [PP only [PP to London]], …, [CP only [CP that he goes to London]]. Is this in conflict with our earlier assumption of a phrase PrtP? It is not, if we recognize that PrtP is essentially a  defined category. The difference between PrtP and, say, only Bill is that in the former but not in the latter PRT is in an operator position.7 The generalization in (21) tries to capture this state of affairs.8

BOUND FOCUS

(21)

63

PRT being a M F H cannot project syntactic categorial features unless it occupies an operator position; in this case, PRT heads the functionally defined phrase PP.

The semantics of PRT is such that it cannot be evaluated without a restrictive phrase and a scope in the sense of Heim (1982). This makes it comparable to standard quantificational expressions like every student. In (3), which I repeat now with the proper phrase structure in (22), (22)

[IP John [I′ would [VP invite [DP only [DP MARY]]]]]

Mary is the restrictive part and the CFC corresponding to the open proposition lx invite (John, x) is the scope. Given the P  F I (Chomsky, 1986: 98ff; 1995), only would count as an uninterpreted item at LF and therefore cause derivational crash, if it could not reach a scope position. In Bayer (1996), I discuss a host of cases which show that derivations involving PRT crash exactly where no syntactically licit derivation can be found in which [PRT XP] moves to a virtual scope position. No such obstacle is present in (22). Thus, there is a possible derivation in abstract syntax as shown in (23) where [PRT XP] has moved to SpecPrtP: (23)

[IP John [I′ would [PrtP [DP only [DP MARY]]1 [Prt′ [Prt ] [VP invite t1]]]]]

According to standard assumptions (see Rizzi, 1990; 1991), there is an agreement relation between Spec and the head. Thus, features present in SpecPrtP appear also in the related abstract head position. If this is true, we are entitled to rewrite (23) as (24): (24)

[IP John [I′ would [PrtP [DP only [DP MARY]]1 [Prt′ [Prt only] [VP invite t1]]]]]

If the trace is an LF-copy of the moved phrase, and if PRT is an operator that is licensed in propositional scope position, then we can apply Chomsky’s (1995) copying-and-deletion mechanism by which the restrictive phrase — here the focus associate — is deleted in the operator position, while it appears as a copy in the position of the trace. Since PRT now occupies the head of PrtP, PRT can be deleted in SpecPrtP as well. This leads to the LF-representation in (25). (25)

[IP John [I′ would [PrtP [DP only —— [DP MARY ———]] [Prt′ [Prt only] [VP invite [DP only —— [DP MARY]]]]]]]

64

JOSEF BAYER

(25) is semantically equivalent to the base-generated structure in (1c) and can be straightforwardly connected with a standard semantic interpretation as in Rooth (1985).9 This gives a sufficiently clear picture of the analysis of examples such as (3). 3.4 Association with focus is not focus movement toward PRT The theory of bound focus outlined so far assumes sentences of type (1) as basic in the sense that they can be interpreted more or less directly, while sentences of type (3) require abstract movement. This view is not shared by everybody, and it seems to me that some syntacticians still adhere to the idea that PRT must associate with a focus constituent by a process of movement. We know that sentences with PRT in operator position may have a reading that is indistinguishable from the meaning of sentences in which PRT is a co-constituent of some XP. The long-lived idea is that in a case like (1c), John would [PrtP only [VP invite MARY]], where Mary is the sole carrier of focus, the focus associate moves to PRT. It has, however, been observed as early as Anderson (1972) that such movement would violate all sorts of syntactic constraints. Consider the following: (26)

a. *Who1 do you dislike [the idea [that t1 is tall for a Watusi]]? b. John even has [the idea [that HE is tall for a Watusi]]

While overt movement of who in (26a) would violate both the E C P (ECP) and the C-NP-C (CNPC), focus association with even in (26b) does not seem to violate any such constraint. Discrepancies of this kind have often led to major complications of the organization of grammar. LF was said to be less restrictive than S-structure.10 In the face of the sharp grammaticality contrast in (26), it seems preferable to conclude that the focus does not move to PRT at all. Notice in addition to this, that a single PRT may bind more than one focus.11 In English and German it is, however, not possible to move different phrases overtly to one and the same target position. Consider now the following English and German examples of multiple focus binding: (27)

a.

b.

I have only suggested that DOMINGO should sing “Tristan” in VIENNA (but not that PAVAROTTI should sing it in SALZBURG) Ich habe nur vorgeschlagen, daß DOMINGO den “Tristan” in WIEN singen sollte (aber nicht daß PAVAROTTI ihn in SALZBURG singen sollte)

BOUND FOCUS

65

Domingo and Vienna are not part of any simplex constituent that could undergo movement to only. The focus-movement analysis then forces us to recursive adjunction to SpecPrtP, an operation that may be permitted, but for which there is hardly any evidence in English and German. Taken together, the problems connected with (26) and (27) suggest that there is no focus movement to PRT at all, and that in both cases PRT has propositional scope over a CFC which is an appropriate semantic object for association with focus. A set of discourse alternatives as required by (11) is easily invoked. For (26b) it would be {have the idea that he is tall for a Watusi, have the idea that she is tall for a Watusi, have the idea that that one is tall for a Watusi, …}; for (27a) it would be {suggest that Domingo should sing “Tristan” in Vienna, suggest that Pavarotti should sing “Tristan” in Salzburg, suggest that Kollo should sing “Tristan” in Berlin, suggest that Aschenbach should sing “Tristan” in Weimar, …}. Since according to the present theory PRT is in canonical operator position and binds at least one focus such that this focus invokes a set of alternatives, there is no reason whatsoever to move the focus toward PRT. In this case, no island is violated, and multiple movement into a single XP-position is not required. A desirable side effect of this result is that we can derive the fact that once PRT is in scope position its scope is fixed, whereas a scope position has to be targeted in case PRT heads a non-scopal XP. To take an example, (26b) means something else than John has [the idea [that even HE is tall for a Watusi]]. The scope of even is confined to the IP in which it occurs. As can be expected from the effect of the ECP and the CNPC, (covert) raising of [PRT XP] to the matrix clause is out of the question. Given the well-formedness of (26b) and (27a,b), however, the conclusion must be that as long as PRT c-commands a focus, this focus may be arbitrarily far away.12 This account also provides a way to deal to deal with those examples in which PRT is a MFH of a constituent which properly contains the focus. As examples (12) through (17) have shown, PRT can associate in [PRT XP] with any subpart of XP as long as this subpart corresponds to a variable that ranges over a non-trivial set as specified in (11). Take example (12a) — Some students smoke even IN the classroom. The previous discussion has made it clear that PRT does not form a constituent with P. Thus, the proper syntactic structure must be (28): (28)

Some students smoke [PP even [PP IN [DP the classroom]]]

66

JOSEF BAYER

The procedure of assigning PRT to a propositional scope position from where it can bind the focused P is the same as in (22). The only difference here is that for reasons of syntactic constituency more than the focus associate must be raised to SpecPrtP. After raising, head identification and deletion, the LF of (28) will be as in (29): (29)

Some students [PrtP [PP even —— [PP IN —— [DP ——————]]] the classroom [Prt′ even [VP smoke [PP even —— [PP IN [DP the classroom]]]]]]

The relevant set for this LF is {smoke in the classroom, smoke in front of the classroom, …}. Given the standard semantics of even developed by Karttunen and Peters (1979), interpretation is again straightforward. Notice, however, that under the assumption that nothing but PRT must be raised for semantic reasons, (25) and (29) are instances of LF-Pied-Piping. I will turn to this aspect of my analysis in Section 6.

4.

Lowering

We can so far account for two manifestations of bound focus: (i) Base-generation of PRT as the head of PrtP such that PRT c-commands a focus associate; (ii) movement of [PRT XP] to SpecPrtP such that the head of the virtual phrase PrtP is identified. Consider now those cases in which either PRT is higher than its purported canonical scope position, or the focus associate is higher than PRT. Relevant examples appeared in (4) — Only Mary would invite us to her home — and in (6) — ANNA could even/?*only escape from the prison — respectively. Turning first to (4), we notice that PRT and its focus associate form a single phrase, and that this phrase is in a non-operator surface position. For proper semantic evaluation, it has to undergo reconstruction such that PRT ends up in a head position. A solution is readily achieved, if we follow the proposal that the subject is generated as a specifier of the verb which then moves to SpecIP for Case reasons.13 According to the minimalist implementation of trace theory, [PRT DP] exists after Spell-Out as a copy in SpecVP. Let us imagine now that the copy can undergo LF-movement to SpecPrtP. This presupposes the following representation at Spell-Out: (30)

—————— MARY] invite us to her home] Only MARY would [VP [only

BOUND FOCUS

67

The PF-deleted part would then undergo LF-raising to the hypothesized pre-VP position and activate PRT in the required operator position. Although this would yield the desired LF, we must be suspicious. The reason is that the derivation amounts to the raising of a trace, an operation that has explicitly been banned for reasons I cannot go into here.14 Let us therefore pursue a different solution. Assume that the phrase only Mary has not moved to SpecIP directly but rather that it has first moved through SpecPrtP. In this case, there is a copy in SpecPrtP which could activate the head PRT in operator position. Then the actual representation of (4) is not as in (30) but as in (31a); given that the filled Specposition can abstractly activate the head position of PrtP, and deletion applies according to the principles of the LF-side of the grammar, the relevant parts of the LF of (4) would be as in (31b): (31)

a. b.

Only MARY would [PrtP [only —————— MARY] [Prt′ [VP [only —————— MARY] invite us to her home]]] ( S-O) Only —————— MARY would [PrtP [only ——————] MARY [Prt′ only [VP [only —— MARY] invite us to her home]]] ( S-O)

The above-mentioned problem in connection with the raising of trace dissolves. Of course, the question remains whether movement to SpecIP requires access of the intermediate SpecPrtP-position. I will give an argument in favor of this solution directly which is not fully conclusive but nevertheless suggestive. Let us now turn to the last type of bound focus, namely to the constellation in which PRT binds a focus that is not c-commanded but rather c-commands PRT itself. This situation is exemplified in (6) — ANNA could even/?*only escape from the prison. The proper minimalist description of this case is, of course, that the focus associate has been raised to a higher position, and that this movement has left a copy behind which at LF serves as the bindee required by the version of PLA given in (11). The question is from which position the focus associate raises. There are two options: Either it raises from a DP such as even ANNA, or it raises from a position which is already bound by PRT in its final operator position. It is easy to see that only the second option is available. Consider the examples in (32): (32)

a. SALLY1 I guess t1 was even arrested t1 b. *SALLY1 I guess t1 was arrested [DP even t1] c. *SALLY1 I guess [DP even t1] was arrested t1

68

JOSEF BAYER

These data show that PRT is in its ultimate scope position when the focus associate undergoes raising. It cannot be the case that the focus leaves a DP of which PRT is an MFH as would be expected if a DP of type [DP1 PRT DP2] were initially generated from which the focus phrase DP2 is moved to SpecIP and then to SpecCP. I will refrain from a discussion of movement from DP, but it seems plausible that such movement could only occur under very special circumstances.15 Given the facts displayed in (32), it can be assumed that there is also a PrtP in (4) although its head is only abstractly represented. I take this as, at least, indirect support for my suggestion that topicalizations of this sort have abstractly activated SpecPrtP. The proper LF-representations of (6) and (32a) would then be as in (33a) and (33b) respectively: (33)

a. b.

——— ANNA could [PrtP even [VP [even —— ANNA] escape from the prison]] SALLY was [PrtP even [VP arrested SALLY]] ———— I guess ———— SALLY

If tenable, this shows that all the word order variation that is observed in connection with bound focus can be reduced to a single format — a format that satisfies the revised PLA.

5.

A Bare Phrase Structure Implementation

In the Minimalist Program of linguistic theory as envisaged in Chomsky (1995), X-bar syntax is abandoned in favor of the basic operations M and M. Merge is reduced to the combination of lexical and categorial features. Move is split up into essentially two different operations that correspond to the derivations before and after Spell-Out. If Move applies before Spell-Out, the process is subject to generalized Pied-Piping; if it applies after Spell-Out, it is reduced to the movement of features (M-F). Features are not categories. The movement of categories is, thus, triggered by restrictions on morphophonological convergence. According to the minimalist philosophy, it is to be avoided, unless it is necessitated by morphophonological output constraints. How should the syntax and semantics of bound focus be implemented in this framework? Assume that for an example like (1c) — John would only invite MARY — the VP is formed according to the theory of bare phrase structure. The item invite and the item MARY are merged. Since the verb turns out to be a head that requires an argument, the resulting category is a projection of this head. The

BOUND FOCUS

69

result is formally {invite, {invite, MARY}}. Merge can then be applied to only and the structure which may now be called “VP”. This yields {only, {only, VP}}. According to (21), PRT does not have a syntactic category unless it occupies an operator position. In {only, {only, VP}}, only is able to take scope over VP, a CFC. Thus, {only, {only, VP}} amounts to {PRT, {PRT, VP}}. In this structure, PRT can bind either the entire VP or a focus associate that is a proper part of VP. Consider now the example only Mary as it appears in (3). The bare phrase structure is {only, {only, DP}}, but now PRT cannot project syntactic categorial features. For only to be licensed, it has to move to a position where it is. According to the theory developed in Chomsky (1995: ch.4), covert movement cannot be movement of the entire phrase. It must be movement of the relevant quantificational features of PRT.16 This would derive the proper LF of examples such as (3). Consider now (6) — ANNA could even/?*only escape from the prison. As we have shown, PRT occupies a scope position here. Then the focus associate must have been raised higher because of some feature that needs to be checked before Spell-Out. The exact nature of this feature is not relevant here. Assume it to be [top]. Since movement has left a copy of the focus associate behind, this copy will be bound by PRT. Consider finally the examples in (4) — Only Mary would invite us to her home and its German equivalent which is a V2–clause. Given the plausible assumption that at LF PRT must occupy a scope position from which it binds a focus associate, how can the Move-F theory deal with this case? There are two options: (i) Instead of raising, the features of PRT will undergo . (ii) PRT raises from the LF-copy that the phrase only Mary has left behind. Option (i) should be disfavored for the simple reason that lowering operations have mostly proved to be either untenable or unnecessary. Option (ii) seems to conflict at first sight with the verdict against the raising of trace. At closer inspection, however, one can see that it does not amount to the raising of the trace/copy but rather to the raising of a -label of the trace/copy. My sketch seems to lead to a minimalist implementation of focus association in the sense of Chomsky’s proposal that Move-F is to be preferred over Move-a as long as output constraints remain unaffected. The rest of this article deals with a set of problems which nevertheless emerge from Move-F, and which can be avoided under the more conventional theory of covert Move-a.

70 6.

JOSEF BAYER

The Derivation of Wide Scope for PRT

Let us ask the question whether the scope of PRT can be determined by Move-F where F is the semantically relevant feature of PRT. In the following, I will draw attention to some problems that are likely to be solved more straightforwardly by covert Move-a than by Move-F. 6.1 Minimal Links According to Chomsky (1995: 271), raising of a pure feature F is adjunction of F to a head which projects. F is supposed to make the shortest possible move. Any violation of the M L C (MLC) is said to be illegitimate (p. 267f). It has been found, however, that the scope of PRT is not necessarily local in the sense that PRT takes the closest possible domain as the scope domain. Taglicht (1984) has observed sentences in which only may, under certain circumstances, take scope wider than the clause in which it appears overtly. This can be seen in the following examples: (34)

a. b.

The students in the GDR were required to learn only RUSSIAN The GDR education ministry demanded that the students learn only RUSSIAN

(34a) can mean that the GDR students were required to learn no other language than Russian. This is factually false because it means that it was forbidden for students in the GDR to study foreign languages such as Chinese, English or Spanish. (Mind control was not that rigid after all!) (34a) has another reading, however, which is factually true. It means that of all foreign languages, Russian was compulsory. Nothing is said as to the freedom to learn Chinese, English, Spanish etc. The same kind of ambiguity arises in (34b) where we find a finite complement. The more “realistic” reading can be achieved if only takes scope in the matrix clause rather than in the clause where it arises. But this move is not readily compatible with the MLC. 6.2 Attraction by [prt-]? To make it compatible, we are forced to hypothesize a silent affixal feature that attracts the semantic feature corresponding to PRT. Let us call the affixal feature for the sake of the example [-]; [prt-] may enter the numeration at different

BOUND FOCUS

71

stages of the derivation. For the false reading where only remains in the scope of require, [prt-] is an affix to the verb learn; for the true reading where require is in the scope of only, [prt-] is an affix to the verb require. So far so good. The problem with this solution is that the attractor [prt-] is a device which adds too much power to the system of UG. What is the evidence for this conclusion? Consider the German versions of (34) given in (35): (35)

a. b.

Die Studenten in der DDR wurden gezwungen nur RUSSISCH zu lernen Das Bildungsministerium der DDR verlangte, daß die Studenten nur RUSSISCH lernten

These sentences are not ambiguous at all. They permit only the false interpretation, i.e. nur takes narrow scope. This is unexpected because the true reading is readily achieved, if nur is overtly inserted in the matrix clause: (36)

a. b.

Die Studenten in der DDR wurden nur gezwungen RUSSISCH zu lernen Das Bildungsministerium der DDR verlangte nur, daß die Studenten RUSSISCH lernten

Before I continue discussing this effect, a caveat is necessary to which my attention has been drawn by Büring and Hartmann (1996). We know from the work of Taglicht and Rooth that the scope of PRT is fixed as soon as PRT is in the pre-VP operator position. The English sentence The students in the GDR were required to only learn RUSSIAN ceases to be ambiguous. Büring and Hartmann are right in arguing that this might be the only analysis for sentences like (35). The head-final nature of the German VP does not allow us to see unambiguous constituency. Thus, given the fact that an operator does not undergo raising from an operator position, it would not be surprising that the scope of nur cannot be extended into the matrix clause. Recall, however, that in German nur can also be postposed.17 In that case, it must form a constituent with the preceding focus associate. The only plausible syntactic structure would be [DP [DP RUSSISCH] nur]. The question is now whether the examples in (34) turn out to be ambiguous if we change them as in (37): (37)

a. b.

Die Studenten in der DDR wurden gezwungen [RUSSISCH nur] zu lernen Das Bildungsministerium der DDR verlangte, daß die Studenten [RUSSISCH nur] lernten

72

JOSEF BAYER

They remain unambiguous. This is unexpected, if [prt-] could have been inserted in the matrix clause, as it must be assumed for English according to the theory of pure feature movement. I conclude from this and a host of further evidence which I cannot include here that “semantic attractors” such as [prt-] are not allowed in any derivation. The alternative is that, where semantically necessary, PRT must raise “by itself”. This throws us back to the question why there is this difference in LF scope options between English and German.18 As far as I can see, the theory of pure feature movement does not give an answer. 6.3 Island Constraints Assume alternatively that it is not the pure feature corresponding to PRT that undergoes covert movement, but rather the entire phrase. This phrase can target whatever landing site is there to fulfill the needs of PRT. According to our assumptions, its primary semantic need is to take sentential scope. This can be achieved in the lower or in the higher clause. As long as a syntactic derivation in terms of successive cyclicity is available, it is a matter of free choice which scope is taken. The empirical consequence is, of course, that we predict island effects. The following examples show that island constraints are active. (38)

a. b.

(39)

a. b.

The GDR education ministry made the suggestion to learn only RUSSIAN The GDR education ministry made the suggestion that the students learn only RUSSIAN The conformist student asked the ministry [where to study [only RUSSIAN]] The conformist student asked the ministry [where he could study [only RUSSIAN]]

All of these sentences are unambiguous. They permit only a narrow scope interpretation of PRT. This shows that at least the CNPC and the wh-IC must be active, contrary to the popular idea that LF-derivations are not constrained by subjacency. The question is what blocks transclausal scope of PRT in German. An answer was given in Bayer (1996) in terms of the typological difference that sets the SOV-language German aside from the SVOlanguage English.19 According to the theory developed there, scoping out of complements which appear on the non-canonical, right side of V is only possible if the matrix clause hosts an  operator. In English and other VO-languages

BOUND FOCUS

73

where the complement is on the canonical side of V, covert wide scope can be achieved, if there is a possible derivation which respects subjacency. The activity of island constraints as shown in (38) and (39) suggests that it is SpecCP that must be accessible. This implies that the covert operation is Move-a. For MoveF it is said instead that “the computation ‘looks at’ only F and a sublabel of [the target of movement, J.B.] K”. In this case, it should not matter how deeply F is embedded in the phrase in which it occurs.20 This cannot be appropriate for those cases, however, in which PRT has to be assigned sentential scope. I would like to mention only two examples. Consider first German PPs. German does not allow P-stranding. Thus, PP should be an island for the covert extraction of a DP that is headed by PRT, but not for a feature that corresponds to PRT. The data in (40) address this point: (40)

a.

Sie haben nur an ANNA gedacht the have only at Anna thought ‘They thought only about Anna’ b. *Sie haben an nur ANNA gedacht

Despite the fact that nur can form a constituent with DP, nur+DP is bad inside PP. This is not expected if PRT can raise as an abstract feature. It is expected, however, if the primary step in the LF-derivation is covert movement of the phrase nur ANNA. Consider next an example from English. As has been shown by (16) — Some students smoke [[IN the classroom] even] — English permits postposed even. If the assignment of scope to even is not affected by the degree of embedding in, say, a DP, we would not expect the following difference: (41)

a. [[ANNA’s father] even] was arrested b. *[[ANNA even]’s father] was arrested

The feature corresponding to PRT should extract from the complex DP in (41b), but contrary to this prediction it does not. If we assume, however, that it is the entire phrase ANNA even that has to undergo movement, embeddedness matters. (41b) is, in this case, ruled out as a violation of the L-B-C (LBC). The question remains, of course, how the occurrence of Wh in the same environments is licensed: (42)

a.

b.

Wer hat [an wen] gedacht? who has [at whom thought ‘Who has thought about whom?’ [Who’s father] was arrested?

74

JOSEF BAYER

I cannot try to give an answer here, but it is not obvious that it should be sought under the assumptions of the theory of pure feature movement. If we try to do so, we must draw a distinction between different types of operator features and explain why some respect islands and others don’t.21 6.4 A morphological reflex of movement Let me finally point to another property of sentences which allow a wide scope interpretation of PRT. As has been pointed out by Longobardi (1991: 187, note 8) and Richard Kayne (p.c.), neither Italian nor English seem to allow the relevant wide scope interpretation, if the finite complement which hosts PRT happens to be in the indicative mood. Thus, (43) — as opposed to (34b) — would not permit the wide scope interpretation: (43)

The GDR education ministry demanded that Michael learns only RUSSIAN

Similarly, the Italian examples of transclausal scope of solo, soltanto or solamente which are adduced in Longobardi (1991) always show finite complements in the subjunctive mood. (44)

a. b.

E’ veramente necessario che io parl -i soltanto con Gianni is really necessary that I speak - only with Gianni E’ davvero indispensabile che lui cred -a che io is truly indispensable that he believe - that I desider-i vedere soltanto Gianni wish- see only Gianni Longobardi (1991: 153)

(44a) is ambiguous between an interpretation of soltanto in the embedded clause or in the matrix clause. (44b) is three-way ambiguous with soltanto taking scope in either of the three clauses. In his discussion of differences between overt and covert movement, Chomsky (1995: 267) ascribes the presence of a morphological reflex of this kind to successive-cyclic wh-movement. The fact that such a reflex is a precondition of wide scope interpretation of PRT could then lead to two different conclusions. Either there is indeed overt movement of some kind of zero operator as Watanabe (1992) has argued, or covert movement is largely the same as overt movement, i.e. it is Move-a and not Move-F. Since I see no evidence for overt movement of a silent operator in the syntax of PRT-scope, I tend to keep to the

BOUND FOCUS

75

older theory according to which UG has the option of moving syntactic categories covertly.22 6.5 A solution without Move-F It was shown in Section 5 that by and large the minimalist conception of grammar offers important insights into the syntax of bound focus. It should not be overlooked, however, that the adoption of the subtheory of pure feature movement (“Move-F”) leads to predictions and expectations which are hardly compatible with the scope facts that are revealed by the grammar of PRT. Since Move-F is nothing like the movement of categories, it remains to be seen whether it has properties that are distinct from those of Move-a. The conceptual question that remains is how desirable a theory of grammar is in which LF and PF are “maximally different”, as suggested by this theory.23 For the time being, I would rather stick to the theory of abstract Spec-head agreement that underlies the analyses presented in Section 3 and 4. If we do so, we can show how the different interpretations of ambiguous examples like (34a) — The students in the GDR were required to learn only RUSSIAN — can be derived. Assume that the semantics of PRT forces it to move to a position where it can be interpreted unless PRT occupies a proper scope position before SpellOut. Since only, being part of the DP only RUSSIAN, does not occupy such a position, the phrase must undergo movement. This movement is under the control of syntactic constraints on movement. The first landing site that this DP can target is the VP in which it occurs. Let DP adjoin to VP. Does this move create an interpretable semantic object? It clearly does not. The adjoined DP cannot be evaluated according to the semantics we have adopted from Rooth (1985). The process of adjunction can, however, be seen somewhat differently. Assume that DP has moved to the specifier of a  phrase PrtP that comes into existence by agreement with the specifier. In terms of bare phrase structure, nothing is required but the usual assumption that the target projects. If the target VP projects according to the semantic features associated with PRT, the result will be an abstractly represented PrtP. The rest is straightforward. The DP only RUSSIAN has “identified” the head of the virtual PrtP. Since its trace is a copy of the original, deletion can now apply and reduce the structure in such a way that the final result is a structure with PRT in head position that binds its focus associate in the “reconstructed” position. The result appears in (45). Its interpretation derives the meaning according to which it was required that apart from Russian no other language could be learned.

76

JOSEF BAYER

(45)

The students in the GDR were required [IP PRO to [PrtP [only —— RUSSIAN [[Prt only] [VP learn [only —————] —— RUSSIAN]]]]]

According to the actual world, (45) achieves the truth-value 0. How is the wide scope reading derived? If PRT were attracted by the minimal scope domain that can satisfy its semantic needs, and if the feature PRT would undergo checking, no ambiguity could be expected. Notice, however, that according to the present proposal PRT is neither attracted nor checked. If the semantics of PRT can be evaluated in an arbitrary position that can satisfy its scopal needs, and if this position can be targeted without any violation of movement constraints, the ambiguity is predicted. Imagine that indicative mood corresponds to a feature which arises in I and is copied onto C, and that it erects a barrier that is absent if the mood feature chosen is subjunctive. In this case, the DP only RUSSIAN may move successive-cyclically through SpecCP to the matrix-VP which is headed by the verb require. The resulting structure is given in (46). (46)

The students in the GDR were [PrtP [only ———————] RUSSIAN [[Prt only] [VP required [CP ——————— only RUSSIAN [IP PRO to [VP learn [only —— RUSSIAN]]]]]]]

(46) represents the wide scope reading of PRT. According to the actual world, it achieves the truth-value 1. The question may arise as to why the grammar of English does not allow overt transclausal movement of this sort. Notice, however, that it does not even allow intra-clausal movement to SpecPrtP. We can stick to the minimalist conjecture that the Procrastinate principle rules out overt derivations that can equally well be achieved covertly.

7.

Conclusion

The syntax and semantics of bound focus has been an enormous challenge to linguistic theory. Much useful semantic work has been done in the past which has set standards which had to be met by syntactic analysis. Unfortunately, the syntactic side of these accounts was often sketchy and less illuminating. It seems to me that the minimalist theory of grammar has opened up interesting possibilities of bridging the gap between syntax and semantics that has always been visible in this domain as well as in a number of related phenomena such as negation.24 The reason for this lies in what I see as the greatest virtue of this theory, namely the strictly uniform treatment of overt and covert operations in

BOUND FOCUS

77

grammar. The only point where my account of bound focus diverges from the theory as outlined in Chapter 4 of Chomsky (1995) concerns the sub-theory of pure feature movement, which was rejected, and this is exactly the point where overt and covert operations diverge from each other.

Notes * I want to thank the audience of the Table ronde internationale sur la grammaire du focus where this material was presented, especially Michael Brody, Nomi Erteschik-Shir and Laurie Tuller. Special thanks to Noam Chomsky, who was kind enough to respond with a detailed letter to my questions concerning feature movement. 1. Nomi Erteschik-Shir (p.c.) suspects that the discrepancy between only and even in modern English has something to do with the fact that the former but not the latter is an eliminative operator. See also Erteschik-Shir (forthcoming: ch.3). I cannot exclude the possibility that dialectal and stylistic factors play a role as well. Laurie Tuller (p.c.) informs me that for her postposed only is in fact possible, and that such constructions occur regularly in poetry and song. See also note 2. 2. Thanks to Nomi Erteschik-Shir for suggesting this possibility. The difference between only and even can also be seen in phrases like even Mary and Mary even where both orders are possible, while with respect to only the order *Mary only is excluded. Although in German the order Maria nur is somewhat stilted, it is not ungrammatical. In the last act of Richard Wagner’s “Parsifal” we hear (i)

die Wunde schließt der Speer nur, der sie schlug the wound closes the spear only that it cut ‘Only the spear that caused the wound may close it again’

At the moment I have nothing to say about the stylistic factors that are affiliated with postposed nur/only. 3. That the phenomenon is independent of focusing particles is shown by topicalization data from German. As Travis (1984) has observed, a clitic (or weak pronoun) es cannot be topicalized. (i)

*Esi hat der Hund ti gefressen it has the dog eaten

Likewise, separable prepositional prefixes to verbs can only be topicalized if the prefixed verb P–V has a lexical alternative P′–V in which the meaning of V remains constant. Thus, there is the pair auf-machen (‘open’) ~ zu-machen (‘close’), but there is no alternative to the prefix verb auf-hören (‘stop’) or ab-liefern (‘deliver’). If topicalization requires a non-trivial set of alternatives, the following contrasts are explained: (ii)

a.

AUFi hat er die Tür ti gemacht (und nicht ZU) open has he the door made (and not closed ‘He OPENED the door (but did not CLOSE it)’

78

JOSEF BAYER b. c.

*AUFi hat er nicht ti gehört intended: “He did not stop” *ABi hat er die Bücher nicht ti geliefert intended: ‘He did not deliver the books’

4. See also Aoun and Li (1993) for discussion. 5. See van Riemsdijk (1996) for the role of functional heads in the extension of lexical projections. 6. See Karttunen and Peters (1979) and Jacobs (1983). 7. For reasons of space, I leave out one complicating aspect: The eliminative particle only can also operate on sets that are hierarchically ordered, i.e. on so-called . Consider the following examples: (i) (ii)

a. b. a. b.

The scientist found the dangerous substance in only three organisms Some only weakly gifted students applied for the job Der nur mit einer Badehose bekleidete Gangster konnte entfliehen the only with a swim-suit dressed gangster could escape Der Journalist hat die Mütter von nur zehn Jugendlichen befragt the journalist has the mothers of only ten teenagers interviewed

In none of these examples is it the case that only/nur takes scope over more than the respective scales “number of organisms”, “degree of giftedness”, “amount/value of garment”, “number of teenagers”; only/nur is in an operator position here, but it does not have propositional scope. See Bayer (1996) for extensive discussion. 8. The same seems to hold for the relation between term negation and sentential negation. I assume that (ii) involves a NegP, while (i) is a negative statement only on a more abstract level of semantic interpretation. (i)

Not John, but Fred, was here

(ii)

John was not here, but Fred (was)

9. While Rooth also considers treatments in terms of type raising, this is not necessary here because in the present system [PRT XP] phrases reduced to the standard case by LF-movement. 10. See Huang (1981) and much work that took up the conclusion that LF is not controlled by subjacency. 11. This has been denied by Tancredi (1990), who proposed that apparent cases are to be derived by V-raising from a focused VP; as I have pointed out in Bayer (1996), however, there are other cases which cannot be captured in this way. Thus the assumption of multiple bound foci is unavoidable. 12. See also Kratzer (1991) and Bayer (1996) for further discussion. I want to thank Michael Brody for a controversial discussion of this point. 13. See Sportiche (1988), Koopman and Sportiche (1991) among many others. 14. See Chomsky (1995: 304) where the following is proposed: (i)

Trace is immobile

(ii)

Only the head of a chain CH enters into the operation Attract/Move

Thanks to Peter Suchsland for discussion of this point.

BOUND FOCUS

79

15. I think of the NP/DP-split phenomenon that is likely to hold in cases of “floated quantifiers” and similar constructions where DP or NP appears to have left a more complex phrase that is headed by a quantifier or some other material. Although much insightful work has been devoted to this area over the years, the linguistic descriptions remain controversial. 16. Chomsky is very brief about similar cases, but nevertheless explicit. See Chomsky (1995: 337) where a feature [quant] is assumed that raises to a potential host. The target of movement — in Chomsky’s proposal T or v — can have an affixal feature [quant-] which may be chosen in a numeration or not. 17. Examples like (17b) — [[KLEINE Buben] nur] dürfen die Damentoilette benutzen — are to my ears completely natural. See also (i) from note 2. 18. Ortiz de Urbina (this volume) reports that in Basque neither Wh nor focus may undergo abstract movement from the (extraposed) complement to the matrix clause, while, according to Tsimpli (1995), focus in Greek seems to have matrix scope, even if the focus phrase appears in the complement. According to the theory developed in Bayer (1996), this result is expected because Basque has a head-final VP; thus a clausal complement to the right of V appears in noncanonical position. Greek has a head-initial VP; thus a clausal complement to the right of V appears in canonical position. See also my remarks in 6.3. 19. For reasons of space, I cannot repeat my argumentation here. The reader is referred to Bayer (1996). 20. See Chomsky (1995: 269) where the example Pictures of whose mother did you think were on the mantelpieces is adduced to show that the wh-feature can be deeply embedded in the DP. As we know from other languages, however, this is the exception rather than the rule. 21. A first attempt has been made in Longobardi (1991) and in Bayer (1996: 113ff), and further developed in Bayer (1995; 1998). 22. Independent evidence in favor of the same conclusion has been given in Wilder (1997). See also Guéron and May (1984) and subsequent work which shows that certain LF operations must involve Pied-Piping in order to void Principle C effects. Compare (i) and (ii): (i)

*I told heri that the concert was attended by many people last year who made the sopranoi quite nervous (ii) I told heri that the concert was attended by so many people last year that the sopranoi became quite nervous. Raising of so or its features is not sufficient. We can undo Principle C only if the entire phrase whitch contains the soprano leaves the c-command domain of her. 23. This question was raised in Brody (1996). While the expectable reaction would be to switch back to the classical LF-theory in which LF and PF are maximally similar by virtue of the fact that both involve movements of genuine syntactic entities, Noam Chomsky (p.c.) would draw a different conclusion. He pursues the idea that overt movement is also Move-F, and that all checking takes place within extended lexical items. Successive-cyclic overt movement would then be Move-F, followed by pied-piping. In this case, the purported differences between overt and covert movement would equally have to vanish. See Chomsky (1998) for an elaboration of this idea. 24. See Moritz and Valois (1994) for an interesting account which is very much in the spirit of what I had to say about PRT.

80

JOSEF BAYER

References Altmann, Hans. 1976. Die Gradpartikeln im Deutschen. Untersuchungen zu ihrer Syntax, Semantik und Pragmatik. Tübingen: Niemeyer. Altmann, Hans. 1978. Gradpartikel-Probleme. Zur Beschreibung von gerade, genau, eben, ausgerechnet, vor allem, insbesondere, zumindest, wenigstens. Tübingen: Narr. Anderson, Stephen. 1972. “How to get even,” Language 48.893–906. Aoun, Joseph, and Yen-hui Audrey Li. 1993. “wh-elements in situ: syntax or LF?”. Linguistic Inquiry 24.199–238. Bayer, Josef. 1995. “Quantification, agreement and Pied-Piping”. Talk presented at the Hungarian Academy of Science, Budapest, 03.10.95. Bayer, Josef. 1996. Directionality and Logical Form: On the Scope of Focussing Particles and wh-in-situ. Dordrecht: Kluwer. Bayer, Josef. 1998. “Two types of covert movement.” Talk presented at the workshop on Acquisition and Variation in Syntax and Semantics, SISSAL, Trieste, 3.–6.9.98. Besten, Hans den and Gert Webelhuth. 1987. “Remnant topicalization and the constituent structure of VP in the Germanic SOV-languages,” GLOW Newsletter 18.15–16. (Talk presented at the 1987 GLOW-conference, Venice). Brody, Michael. 1996. “Focus in perfect syntax”. Talk presented at the Table ronde internationale sur la grammaire du focus, Paris 22.-23.February 1996. Büring, Daniel, and Katharina Hartmann. 1996. “Bemerkungen zur Syntax und Semantik einiger fokussensitiver Partikeln im Deutschen unter besonderer Berücksichtigung sententialer Komplementation”. Talk presented at the GGS meeting, Berlin. Chomsky, Noam. 1986. Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Massachusetts: MIT Press. Chomsky, Noam. 1998. “Minimalist Inquiries: the Framework.” MIT Occasional Papers in Liguistics: 15. Erteschik-Shir, Naomi. Forthcoming. The Dynamics of Focus Structure. Ms. Ben Gurion University, Beer Sheva. Guéron, Jaqueline and Robert May. 1984. “Extraposition and Logical Form”. Linguistic Inquiry 15.1–31.

BOUND FOCUS

81

Heim, Irene. 1982. Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation, University of Massachusetts, Amherst. Huang, C.-T. James. 1982. Logical Relations in Chinese and the Theory of Grammar. Ph.D. dissertation, MIT. Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge, Mass.: MIT Press. Jacobs, Joachim. 1983. Fokus und Skalen: Zur Syntax und Semantik der Gradpartikeln im Deutschen. Tübingen: Niemeyer. Karttunen, Lauri, and Stanley Peters. 1979. “Conventional implicature”. In: Choon-Kyu Oh and David A. Dinneen (eds.) Presupposition, Syntax and Semantics, vol. 11. New York: Academic Press. Koopman, Hilda and Dominique Sportiche. 1991. “The position of subjects”. Lingua 85.211–258. Kratzer, Angelika. 1991. “The representation of focus,”. In: Stechow, Arnim von and Dieter Wunderlich (eds.), Semantics: An International Handbook of Contemporary Research. Berlin: de Gruyter. Longobardi, Giuseppe. 1991. “In defense of the correspondence hypothesis: island effects and parasitic constructions in Logical Form.” In C.-T. James Huang and Robert May (eds.), Logical Structure and Linguistic Structure. Dordrecht: Kluwer. Moritz, Luc, and Daniel Valois. 1994. “Pied-piping and specifier-head agreement”. Linguistic Inquiry 25.667–707. Müller, Gereon. 1998. Incomplete Category Fronting. A Derivational Approach to Remnant Movement in German. Dordrecht: Kluwer. Ortiz de Urbina, Jon. This volume. “Focus in Basque”. Primus, Beatrice. 1992. “Selbst — variants of a scalar adverb in German”. In Joachim Jacobs (ed.): 1992. Informationsstruktur und Grammatik. Linguistische Berichte, Sonderheft 4. Opladen: Westdeutscher Verlag. Riemsdijk, Henk van. 1996. “The extension of projections”. Ms. Tilburg University/CLS. (To appear in the Tsuru University Papers in Linguistics.) Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Rizzi, Luigi. 1991. “Residual verb second and the wh-criterion”. Technical Reports in Formal and Computational Linguistics 2. University of Geneva. Rooth, Mats 1985. Association with Focus. Ph.D. dissertation, University of Massachusetts, Amherst.

82

JOSEF BAYER

Rothstein, Susan 1991. “Heads, projections, and category determination”. In: Katherine Leffel and Denis Bouchard (eds.), Views on Phrase Structure. Dordrecht: Kluwer. Sportiche, Dominique. 1988. “A theory of floating quantifiers and its corollaries for constituent structure”. Linguistic Inquiry 19.425–449. Taglicht, Joseph 1984. Message and Emphasis: On Focus and Scope in English. London: Longman. Tancredi, Christopher. 1990. “Not only even but even only”. Ms., MIT. Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. Ph.D. dissertation, MIT Cambridge, Massachusetts. Tsimpli, Ianthi 1995. “Focussing in modern Greek”. In: Katalin É. Kiss (ed.), Discourse Configurational Languages. Oxford: Oxford University Press. Watanabe, Akira. 1992. “WH-in-situ, subjacency and chain formation”. MIT Occasional Papers in Linguistics 2. Wilder, Chris. 1997. “Phrasal movement in LF”. Talk presented at the International Conference on Pied-Piping, Jena 29–31 May 1997.

Are There Cleft Sentences in French?* Anne Clech-Darbon, Georges Rebuschi & Annie Rialland Sorbonne Nouvelle (Univ. Paris III)

Abstract We endeavor to show that the traditional label “cleft sentence” does not correspond to any specific “construction” in the usual sense of the word, at least as far as French c’est … que/qui contrastive sequences are concerned. (i) There are other semantic and pragmatic interpretations of such utterances; (ii) traditional generative analyses of clefts as such are shown to suffer from various shortcomings; (iii) the prosody of contrastive C’est … que/qui sentences has nothing specific to it; (iv) other strategies exist that convey contrastive effects. As a result, we are led to posit that the post-focal relative clause is simply right-adjoined to a an identificational IP: the phonetic properties of clefts directly follow from this configuration, and so do their logico-semantic properties, once the would-be expletive subject ce is allowed to contain a predicate variable in its translation — the coda, trivially interpreted as a predicate, simply binds that variable.

Introduction After showing that there are at least four distinct types of C’est … que/qui… sequences in French, we review various traditional analyses of clefts proper, and show that they suffer from serious shortcomings. We then turn to interface considerations and examine prosodic data, and next some semantic and pragmatic data, so as to finally propose a new representation of s–s or Spell-out which is maximally simple with respect to them: the post-focal clause is base-generated as a relative clause right-adjoined to a copular or identificational IP specified by

84

ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND

ce. Prosodically, the duplicated terminal intonation that characterizes clefts (as against other C’est … que/qui … sentences) is interpreted as a form of   which follows from the fact that the focused XP carries a terminal boundary tone, whilst the adjoined CP has no independent status. At the other interface, the reason why such a CP can be thus adjoined without violating Full Interpretation lies in the hypothesis that the relative clause is interpretable: it binds a   associated with the translation of the would “expletive” subject ce. As a result, we are led to the conclusion that there are no cleft sentences as such in French: bare output conditions simply happen to licence the merging of an identificational sentence beginning in ce and a relative clause, sans plus.

1.

Four types of C’est … que/qui… sequences in French

Consider the following examples: (1)

a. b.

(2)

a. b.

(3)

a. b.

(4)

a. b.

(— Ta fille est tombée dans l’escalier?) (‘Did your daughter fall down the stairs?’) — Non, c’est le petit qui est tombé dans l’escalier. ‘No, it’s the young one [+masc.] that fell down the stairs.’ (— Qui c’est, ce gamin?) (‘Who’s that kid?’) — C’est le petit qui est tombé dans l’escalier. ‘It’s the young one [+masc.] that fell down the stairs.’ (— Tu sembles inquiète. Qu’est-ce qui se passe?) (‘You look worried. What happened?’) — C’est le petit qui est tombé dans l’escalier. ‘The young one [+masc.] fell down the stairs.’ (— Papa a acheté trois gâteaux.) (‘Dad’s bought three cakes.’) — C’est le petit qui va être content! ‘The young one’s going to be happy!’

Since syntactic forms on the one hand, and communicative and/or pragmatic functions on the other, do not generally correlate directly, we cannot simply say that the (b) sentences above are different just because their contexts are. There

ARE THERE CLEFT SENTENCES IN FRENCH?

85

are, however, several tests that show that they must be distinguished. (i) Only type (3b) sentences allow a paraphrase in which the post-copular constituent may  the complementizer: (5)

— C’est que le petit est tombé dans l’escalier.

Moreover, (5) is altogether unfelicitous as a reply to (1a) or (2a), just as (6) could never constitute a normal comment on (4a). (6)

— C’est que le petit va être content!

We can therefore identify type (3b) sentences as belonging to a special type of sentences, which display  - . (ii) Only type (2b) sentences allow the presence of a strong, left-dislocated “copy” of the light, or non-tonic, pronominal subject ce: (7)

— Ça, c’est le petit qui est tombé dans l’escalier.

This sort of structure would be unacceptable in replies to either (1a) or (3a), just as (8) would be unacceptable as a reply to (4a):1 (8)

— Ça, c’est le petit qui va être content.

Type (2b) sentences can therefore safely be identified as pure identificational, or   , sentences. (iii) Turning to (4b), it must first be noted that it is heavily constrained lexically (the predicate must indicate the subject’s astonishment and satisfaction) and from a temporo-modal point of view (reference to the future is preferred).2 Thus, (9c) is at best awkward, and (9d) and (e) are out — (9a,b) repeat (4a,b): (9)

a. b. c.

d. e.

(— Papa a acheté trois gâteaux.) (‘Dad’s bought three cakes.’) — C’est le petit qui va être content! ‘The young one’s going to be happy!’ ?— C’est le petit qui va tomber à la renverse! ‘The young one’s going to be astounded.’ [. ‘The young one’s going to fall backward!’] *— C’est le petit qui va tomber dans l’escalier! ‘The young one’s going to fall down the stairs!’ *— C’est le petit qui a été/est content! ‘The young one was/is happy!’

86


Note further that the sequence Papa a acheté trois gâteaux; c’est le petit qui va être content!, if uttered by the same speaker, can be inverted, with c’est possibly replaced by (Il) y a: (10)

‘Y a/C’est le petit qui va être content: papa a acheté trois gâteaux’ . ‘There’s the young one that’s going to be happy…’

Such an inversion is totally impossible in the other three cases;3 sentences like (4b) can then be given their own label; we suggest   for those. Of course, it is the distinction between type 1 and type 2 which is most relevant. It is well-known that, out of context, sentences like (11) below can be interpreted as having either a broad (or presentational) focus, or a narrow (or contrastive) focus: (11)

C’est le garçon qui parle russe.4 ‘It is the boy that speaks Russian.’

There are, however, many more syntactic facts than the one illustrated by (7) above that show that two distinct structures must underlie those two interpretations. We shall only consider a few of them here. (i) First, as soon as the cleft XP is not linked to the subject or direct object in the post-focus sequence, French drastically distinguishes between the two structures, as shown in (12) and (13):5 (12)

a. b.

(13)

a. b.

C’est le garçon à qui/auquel j’ai parlé. ‘It’s the boy to whom I spoke.’ C’est au garçon que j’ai parlé. ‘It’s to the boy that I spoke.’

[] []

C’est la maison où/dans laquelle j’ai dormi.[] ‘It’s the house where I slept.’ C’est dans la maison que j’ai dormi. [] ‘It’s in the house that I slept.’

(12a) and (13a) illustrate the presentational pattern, where a relative pronoun appears, which is itself either governed by a preposition, or somehow incorporates it. On the contrary, (12b) and (13b) exhibit the contrastive pattern, where no explicit relative pronoun is allowed — only the complementizer que surfaces. For that reason, we shall avoid the label “relative clause” to refer to the postfocus material, and will refer to it as the  or c-clause.6,7 (ii) Second, in the case of presentational sentences, proper nouns are excluded if


87

followed by a relative clause, unless they are preceded by a definite article, whereas names are just fine in real cleft sentences: (14)

a. b.

C’est *(le) Michel qui parle russe.8 C’est Michel qui parle russe.

[] []

(15)

a. b.

C’est *(le) Michel à qui j’ai parlé. C’est à Michel que j’ai parlé.

[] []

(iii) A third test is provided by the possibility, in the case of broad focus, to substitute Voici ‘Here is’ for C’est ‘it is’, a manipulation forbidden in the case of narrow focusing: (16)

a.

Voici le garçon qui parle russe. [√⁄*] b. Voici le garçon à qui/auquel j’ai parlé. [id.] c. *Voici au garçon que j’ai parlé.

It follows that just as in (16), the material to the right of the inflected verb in the (a) cases of (12) through (15) is just one (complex) NP or DP which contains an ordinary restrictive relative clause, whereas a clefted XP need not be an NP or DP, but can be a PP, as illustrated in the corresponding (b) examples. (iv) This conclusion is independently corroborated by the following fact. French, contrary to English, does not allow the stacking of relative clauses, as shown by (17a) — vs. the coordinated variant (b), which is fine. However, as (18) shows, there is  case in which the linear juxtaposition of two apparent relative clauses is possible; but as the translation indicates, this case is the one provided by clefting. (17)

a. *[L’article que Chomsky a écrit l’an dernier que j’ai lu hier] m’a intéressé. b. [L’article que Chomsky a écrit l’an dernier  que j’ai lu hier] m’a intéressé. ‘[The paper that Chomsky wrote last year (and) that I read yesterday] interested me.’

(18)

C’est l’article que Chomsky a écrit l’an dernier que j’ai lu hier. ‘It’s the paper that Chomsky wrote last year that I read yesterday.’

(v) Finally, although infinitival relatives can paraphrase ordinary tensed (restrictive) relatives, it is not the case when c-clauses are considered:

88


(19)

2.

a.

Voici [le livre [qu’on doit lire]]. ‘Here is the book that we/one must read.’ b. Voici [le livre [à lire]]. ‘Here is the book to read.’ c. C’est le livre qu’on doit lire, pas la revue. ‘It’s the book that we/one must read, not the journal.’ d. *C’est le livre à lire, pas la revue. L.: ‘It’s the book to read, not the journal.’

Some traditional hypotheses concerning the structure of cleft sentences

Simplifying somewhat, the structures for cleft sentences generally proposed in the literature fall into two groups: those according to which the focused phrase (henceforth FXP) and the c-clause form one constituent, and those in which the FXP is structurally independent from the coda. Let us now examine them in turn. 2.1 Analyses in which the FXP forms a constituent with the c-clause In the first group, even setting aside those treatments that do not distinguish between presentational focus sentences and real clefts (see above), there still are several possibilities to consider, which all have in common the hypothesis that, at d-structure, the copula is simply followed by one single clause. When transformations were not constrained too much, and the “structural change” was given linearly, the FXP was simply extracted from the clause and placed between the copula and the coda — see Gross (1968) and Ruwet (1975) for such “clefting” in French. Chomsky (1977) next proposed to move the FXP into a TOP[ic] position, sister to S′ under S″ (at least when the FXP was an NP or a PP).9 In the Barriers framework, this analysis has recently been taken up e.g. by Manfredi (1993), among others, where it is proposed that the FXP raises from its argumental position to Spec,CP (see also Kayne 1994);10 the head C0 then carries a feature [+F] which must be either shared, checked or eliminated. The corresponding representation would look like (20): (20)

C’est [CP le petiti [C0 qui [ti est tombé]]]11

Against (20), we must mention an asymmetry between clefted XPs and (ordinary) wh-movement: extracting a wh-phrase from inside the object NP is


89

perfectly natural, whether it is a interrogative pronoun or a relative one, as shown by the well-formedness of (21a,b): (21)

a.

b.

[De qui]i a-t-il épousé [la fille ei]? ‘Whose daughter did he marry?’ . ‘Of-whom did he marry the daughter e?’ Le banquier dont/de qui Pierre a épousé la fille e a fait faillite. ‘The banker whose daughter Pierre married has gone bankrupt.’ . ‘The banker of-whom Pierre married the daughter…’

On the other hand, even when the context should facilitate the construction of a parallel cleft, such as (22b), that cleft will be out; compare (22c), where the whole object NP is pied-piped. (22)

a.

(— Michel a épousé la fille du général!) (‘Michel has married the general’s daughter!’) b. *— Non, c’est [du banquier]i qu’il a épousé la fille ei. c. — Non, c’est [la fille du banquier]i qu’il a épousée ei.

It therefore seems that under a movement analysis, a restriction like Subjacency is at work here, indicating that if the FXP [du banquier] is moved from inside a clause like [Il a épousé [la fille …]], the landing site must in fact be external to the CP that immediately contains that clause. Note furthermore that this very movement towards the Specifier of the lower CP could be utilized to derive the paraphrastic relation we noted in Section 1 between (3b) and (5): other things being equal, (5) would represent the underlying structure to both (the kernel sentence, to use an old-fashioned term), and (3b) would be the marked sentence. Now if this is on the right track, we can understand why the same movement cannot simultaneously serve to produce a real cleft: there is just no semantic difference between (3b) and (5), whereas (1b) conveys presuppositions and implicatures that are radically different. Let us now consider what sort of semantic interpretation could be given for (20). The only natural one, it appears, would consist in dealing with focusing by stressing, by clefting and by movement in the same way: the LFs for (1b)/(20), repeated here as (23a), for (23b), and for (23c) would uniformly be (23c) itself:12 just as, according to Chomsky (1986), the typology of wh-movement can be reduced to pre- or post-s–s raising, the typology of (non-explicitly quantified) focusing would be reduced to the same type of parameter. (23)

a.

C’est [CP le petiti [C′ qui [IP ti est tombé dans l’escalier]]].

90


b. c.

[IP Le  est tombé dans l’escalier]. Le i [IP ti est tombé dans l’escalier].

However, this means that not only the uninterpretable que≈qui under C0, but also the formatives ce and être of (23a), and, in fact, the entirety of the root IP/CP (minus the lower one, of course), should be deleted.13 To justify this, one would have to show that those elements are not interpretable either, but that they nonetheless are somehow required by the syntax of French to be present at some level other than LF to licence the raising of the FXP. Although we cannot show that such a tack cannot be followed, the optimal transparency requirement between syntactic structure and semantic interpretation clearly would not be respected (see Section 4).14 Another argument against (20) is provided by the following data. Suppose that (24a) were derived from (24b) by Raising. This would entail that the inner IP in (24b) is itself grammatical; now this is factually false, as shown by the irredeemable status of (24c). On the other hand, as the examples (24d,e) shows, only the structure that is possible as an underlying sequence for (24a) is grammatical (owing to the selection of the auxiliary, and the necessity for luimême to be licenced by the reflexive clitic se) — (24d) cannot be taken as input to the hypothetical Raising operation, since it would lead to (24e), which is even more hopeless than (24c) is.15 (24)

C’est lui-même que Jean-Pierre  vu. ‘It’s himself that J.-P. saw.’ b. C’est que Jean-Pierre  vu lui-même. c. *Jean-Pierre  vu lui-même. d. Jean-Pierre  vu lui-même. e. **C’est lui-même que Jean-Pierre  vu.

a.

Let us now consider an   variant of this approach, whereby the FXP would be base-generated either as the subject of a “small clause” — which would in fact be a full CP, or as a second specifier: (25)

a. b.

C’est [CP* le petiti [CP Opi qui [IP ti est tombé]]] C’est [CP [Spec le petiti] [[Spec Opi] [C′ qui [IP ti est tombé]]]]

Concerning the (a) sentence, it seems difficult to distinguish the analysis that underlies it from the predicational analysis that will be associated with (36) (see 2.2 below): the choice between these options really is a matter of preference for small clauses vs. predication à la Williams (1980) or vice versa. Of course, one


91

might argue in favour of (25a) that a “Small Clause” analysis allows a uniform treatment of the coda in clefts and of the relative clauses of the “third type” which appear in the complement of verbs of perception, as illustrated in (26): (26)

a. b. c.

J’ai vu le petit pleurer. ‘I saw the young one cry/crying’ J’ai vu que le petit pleurait. ‘I saw that the young one was crying.’ J’ai vu le petit qui pleurait. [same as (a) or (b)]

However, when the clefted XP is a PP, the parallelism is lost anyway: (27) *J’ai vu [PP à l’hôpital] [que Jean-Pierre a rencontré Marie e] L. ‘I saw at the hospital that J.-P. met M.’ As for (25b), this structure suffers from at least one drawback (as long as the prosodic data to be studied in the next section are not taken into account), even if it has independently been suggested that French possibly has two CP specifiers by Rizzi & Roberts (1989); their hypothesis was meant to explain French Complex Inversion phenomena. Note, however, that many languages have cleft sentences which do not simultaneously allow Complex Inversion. Moreover, the two specifiers were supposed to be linked to the two types of features carried by C0 as such on the one hand, and by the raised I0 on the other — in other words, if their analysis is correct, it can only help justify the clefting of NPs that correspond to the  position in the coda, but certainly not the clefting of either complement NPs or PPs. Finally, this hypothesis supposes that the whword will appear in the higher specifier position; but in the case of cleft sentences, as is apparent from languages like English, where real relative pronouns are allowed, the FXP will be higher than the pronoun. Admittedly, we could also imagine a structure in which FP, a phrase projected from a functional head F0, would take the CP as its complement, as in (28): (28)

C’est [FP [Spec le petit] [F′ F0 [CP [Spec Opi] [C′ qui [IP ti est tombé]]]]]

The clefted phrase or FXP would then have to be base-generated in the Spec,FP position. It is to be wondered, though, if there is any independent reason for postulating that F0 heads may subcategorize for CPs, rather than C0 heads for FPs. In any case, we are reaching here something that looks very much like the predicational cases represented by (25a), or (36) below. We shall consider them

92


after we have dealt with the Extraposition hypothesis. 2.2 Analyses in which the FXP is independent from the coda There are three basic variants in the second series of accounts. In the oldest version (cf. Akmajian 1970; Chomsky 1970; Bolinger 1972, 1977; Gundell 1977), cleft sentences are taken to derive from pseudo-clefts:16 a c-clause is analyzed as a relative clause extraposed from the subject NP, typically a free or headless relative.17 Of course, twenty or thirty years ago, scholars were not particularly bothered by the landing site of the extraposed clause. In a more recent framework, it would have to be adjoined to the initial IP, so as to ccommand its trace: (29)

a. b.

[IP [NP Ce(lui) [CP qui est tombé]] [I′ est le petit]] ⇒ [IP [IP [C(e)-Ø ti] [I′ est le petit]] [CP qui est tombé]i]

Even in English, deriving clefts from pseudo-clefts requires a motivated theory enabling us to distinguish between generic and specific free relatives, since “before Extraposition”, a free relative is ambiguous, whereas “after Extraposition”, only the specific reading is possible: (30)

a. b. c.

What you see is what you’ll get. [/] It’s what you’ll get that you see. [only ] It’s what you see that you’ll get. (id.)

Turning to the French counterparts of (30), they typically  the dislocation of whatever appears to be the subject phrase in English: (31)

a.

Ce que tu vois, c’est ce qui sortira. ., ‘What you see, it’s what will come out.’ b. *Ce que tu vois est ce qui sortira.

(32)

a.

Ce que Marie aime, c’est le riz. ., ‘What Mary likes, it’s rice.’ b. *Ce que Marie aime est le riz.

Note the sharp contrast with real predicational sentences, which do not require dislocation: (33)

a.

Ce que je vois est laid. ‘What I see is ugly.’


b.

93

Ce que Marie aime est bon. ‘What Marie likes is good.’

There also are morpho-syntactic problems concerning the form of the head of the free relative, since lui cannot be repeated in a c-clause, cf. (34c,d) — needless to say, the situation is even worse with clefted PPs, as shown in (35d). (34)

a. b. c. d.

(35)

Celui que Marie aime, c’est celui que j’ai vu. ., ‘The one who Mary loves, it’s the one who I saw.’ Celui que j’ai vu, c’est celui que Marie aime. ., ‘The one who I saw, it’s the one who Mary loves.’ C’est celui que j’ai vu [Ø/*lui que Marie aime]. C’est celui que Marie aime [Ø/*lui que j’ai vu].

a.

On l’enterrera là où il mourra. ‘He’ll be buried where he dies.’ [., ‘…there where…’] b. Là où il mourra, *(c’)est là qu’on l’enterrera. ‘Where(ever) he dies is where he’ll be buried.’ c. C’est là où il mourra Ø qu’on l’enterrera. ‘It’s where he dies that he’ll be buried.’ d. *C’est là où/qu’il mourra (là) où on l’enterrera.

The second variant consists in an   analysis. As proposed by Higgins (1973), Chomsky (1974) (cited in Chomsky 1977), Delahunty (1984), Nølke (1984), van Riemsdijk & Williams (1986) or Bergvall (1987), among others, the coda can be seen as a CP base-generated as as sister to the clefted NP under the VP node. In the notation adopted here, (36) would represent this approach: (36)

[IP [NP Ce] [I′ esti [VP ti [NP le petit] [CP Øj qui tj est tombé]]]]

Most supporters of this predicational analysis have suggested that it simply follows from Jackendoff’s (1977) basic rules, in particular the one that allows a V to be subcategorized for an NP plus an S′ (what was to become a CP). The difficulty is that such subcategorization is typical of control structures (where the embedded clause is “final”) or of complement clauses (where the content of the embedded sentence is precisely a proposition).18 Of course, the S′/CP could also be interpreted as a secondary predicate. But this would mean that its content is rhematic, and its prosody, fairly salient, two propositions that the next sections will show do not hold. Moreover, as Chomsky himself has so often noted, focusing (either by contrastive stress of by clefting) means  —

94


something that requires that the CP be (at least)  as an NP, or a PP, or an Adverbial Phrase. As we find it difficult to imagine how an  CP could undergo such a shift in type, we must turn to the final hypothesis, according to which the VP would have a structure [copula XP XP], as in (37):19 (37)

C’estj [VP tj [NP le petiti] [NP Ø [CP Opi qui [IP ti est tombé]]]]

Here again, être is interpretable as a marker of identification between the contents of two NPs/DPs, since the fundamental idea is that the two constituents to the right of the copula in cleft sentences necessarily belong to the same syntactic category. But when what is focused is a PP, as in (13b), repeated as (38a), one must  that the phrase to its right is also a PP, as in (38b) — unless both are NPs/DPs, as in (38c): (38)

a. b. c.

C’est dans la maison que j’ai dormi. ‘It’s in the house that I slept.’ C’estj [VP tj [PP dans la maison] [PP Ø [NP Ø [CP Opi que [IP j’ai dormi ti]]]]]20 C’estj [VP tj [NP Ø [PP dans la maison]] [NP Ø [CP Opi que [IP j’ai dormi ti]]]]

The difficulty is to syntactically justify the empty P0 and N′ in (38b), and the empty N′’s in both post-copular phrases in (c). Besides, it should be obvious that the system proposed by Bresnan & Grimshaw (1978), which consisted in adjoining a CP to either an NP or a PP cannot be maintained under standard GB assumptions, and cannot be revised and imported into the treatment of cleft sentences, since (39) below is totally   (there is no independent evidence, as far as we know, for restrictive relatives taking a P′ or a P0 as their antecedent): (39)

C’estj [VP tj [PP dans la maison] [PP [P′ [P′/P0 Ø] [CP Opi que [IP j’ai dormi ti]]]]]

To summarize, the traditional analyses of cleft sentences, without being necessarily altogether impossible, all raise difficult questions. We shall now investigate the prosodic properties of this sentence type.


3.

95

Some prosodic data

A wide variety of intonation patterns can be associated with sentences beginning in C’est … qui/que …, whether declarative, interrogative, or exclamatory. As already noted, some of these sentences contain a narrow focus and others contain a broad focus (cp. (1b) vs. (2b) or (3b) above). But even cleft sentences proper offer quite a range of distinct intonation patterns in which both the focused and the postfocused parts may have various contours: rising, falling, flat, etc. Such diversity may seem puzzling at first sight, all the more as        characterizes either the focused part or the coda as such. The purpose of this section is then to show that French cleft sentences have an intonative pattern that distinguishes them from broad focus C’est … qui/que … sentence types: the former are characterized by a  of what is otherwise identified as typical “terminal intonation” (it is this very duplication that we call “intonational agreement”), whilst the latter only have  such occurrence of the terminal intonation type.21 3.1 Declarative cleft sentences Narrow-focus intonation patterns in clefts have been studied by many authors (see Rossi 1985; Touati 1987; and Di Cristo 1996, among others) but only in assertive utterances. Here, we will investigate a wider range of patterns and utterances, including interrogative ones. Consider first the “classical” declarative pattern illustrated in (40) below. (40)

[C’est Jean-Pierre] [qui est sorti]. | | | | H L% L L% ‘It’s J.-P. who’s gone out.’

(see Figure 1 on p.101)22

The analysis of the melody of even such a simple example is far from straightforward. As far as the notation is concerned, we could have chosen to use contours or slopes instead of the level tones H (high) and L (low). One main reason for choosing tones is the need to decompose pitch contours into smaller units, an analytical decision which will be justified in the course of our discussion. In fact, the appropriateness of using tones for transcribing intonation has been much discussed recently (see e.g. Pierrehumbert (1981) and Thorsen (1985) for different points of view). In French, the use of a tonal notation has become

96


common (see Mertens 1987; Di Cristo 1996; as well as Rossi 1993, who considers tones to be the melodic realizations of “intonemes”). In intonational studies of French, there is also much current agreement on the idea that the melody is basically determined by the tones attached to two main points of the “accentual group” (or “rhythmic group”): the last syllable, and the first or second syllable, of the first lexical word — among which auxiliaries and certain prepositions (see Mertens 1987; Pasdeloup 1990; and Sun-Ha & Fougeron 1995, among others). Both of these syllables are often considered to be stressed, with “primary stress” falling on the second of them and “secondary stress” on the first. These two syllables are the anchors of the relevant tones, whilst the other syllables have either predictable L tones or no tone at all. Thus, monosyllabic grammatical words such as articles or pronouns at the beginning of the accentual group and the second syllable from its end (if the group has more than two syllables) are often analyzed as bearing L tones, while all syllables between the secondarystressed and the penultimate ones can be analyzed as toneless (Sun-Ha & Fougeron 1995). Here, we will only note tones on the “secondary” and “main” stressed positions, as in Mertens (1987), while referring occasionally to the overall “melody” of the whole accentual group; we will also introduce the symbol % from Pierrehumbert (1981) to refer to a class of boundary tones such as the final L% of assertions and the final H% of questions, which are syntagmatically linked to the last syllable of the group, whilst paradigmatically functioning at higher levels of organization (such as the utterance, in the case of the terminal intonation pattern). Let us therefore consider example (40) in the terms of this system. Our notation actually incorporates a large part of our analysis. In the first group (the focus), the falling contour on Jean-Pierre is broken down into two tones, a H tone on Jean- (the secondary stressed syllable) and a L% boundary tone on Pierre (the main stressed syllable). The symbol % means that we recognize this L tone as the same “declarative” L tone as that which occurs at the end of declarative utterances. Such an analysis has been proposed earlier by Rossi (1985) and Di Cristo (1996), who, however, used a different terminology. In the second group, the coda, a L tone is associated to secondary-stressed est. This syllable has no prominence, a fact which could be interpreted in terms of destressing, as has been done for other languages such as English (e.g. Pierrehumbert 1981) or German (e.g. Féry 1993). Finally, consider the L% tone associated with sorti. We propose to analyze it as the  of the L%


97

declarative marker of the focus. The motivation for this analysis will become clearer when we consider interrogative sentences later on. If we did not look at further data, the intonation contour of (40) would be open to many other analyses. For example, the falling contour of the first group could be considered an undecomposable marker of focus and the flat low contour of the coda an equally undecomposable marker of postfocus. In order to motivate our analysis in terms of tones we must compare the intonation contour of (40) with that of other focused utterances. Let us then consider a second very common intonation contour of focused utterances: (41)

[C’est Jean-Pierre] [qui est sorti]. \/ | | HL% L L%

(see Figure 2)

Here, the primary stressed syllable of the focused group bears a falling tone, which we suggest is the same as the H and L% tones of (40), both now associated with the unique syllable Pierre. Many authors have discussed this type of pattern (e.g. Di Cristo 1996), often at some length (see especially Touati 1987; who calls it “contrastive intonation”). The only difference between this pattern and the one shown in (40) is the position of the H tone, which now also falls on the primary stressed syllable. In fact, it is well-known that the pitch prominence (here, the H tone) which marks the focus may be on the first, second, or last syllable of the phrase containing the focus (Di Cristo 1996). However, we must underscore the fact that          ,    . Note further that the same duplication that we saw in (40) between the L% boundary tone at the end of both the focused phrase and the postfocus phrase is to be found here. The contour shown in (42) combines the features of the preceding patterns: (42)

[C’est Jean-Pierre] [qui est sorti]. | /\ | | H HL% L L%

(see Figure 3)

Here, the H tones marking the focus are linked to  syllables of Jean-Pierre, forming a high pitch plateau over both syllables of the word. The final L% tone of the first group is reduplicated as before. This intonation pattern is interpreted as more emphatic than the preceding ones.23 The duplication of the L% declarative tone may also involve other features

98


typical of the end of utterances, such as final lengthening. Thus in (43), which represents a teasing contour, not only the boundary tone of the first group, but also its lengthening, are repeated in the c-clause. (43) [C’est Jean-Pierre] [qui est sorti] /\ | | HL% L L% [+long] [+long]

(see Figure 4)

In this variant of the basic declarative intonation pattern shown in (40), the global pitch range is reduced and a breathy voice quality accompanies the whole utterance. 3.2 Interrogative clefts Let us now turn to interrogative utterances. Consider first the familiar pattern shown in (44). (44)

[C’est Jean-Pierre] [qui est sorti]? | | | | L H% L H%

(see Figure 5)

In this type of example, there is a sort of prosodic “double interrogation”: the interrogative H% tone occurs twice, at the end of the FXP and at the end of the coda. In our system, these H% tones belong to the same paradigm of contrasts as the L% tones seen earlier, all constituting boundary tones which occur utterance-finally. Apart from the interrogative H% tone, no H tones mark the focus. The interrogative H% excludes the possibility of another H tone occurring at the beginning of the accentual group, at least in short accentual groups such as the one examined here. We suggest that the interrogative H% boundary tone does double duty as a H focus marker, and propose that these two tones have  in utterances of this type. In the postfocal part, we have posited a L tone on est, as in the declarative pattern, although its interpretation is not entirely straightforward. We will maintain the same analysis as in the corresponding declarative form, but whatever its analysis, a H tone preceding the H% is excluded. The term “double interrogative” which we have used to describe this utterance seems appropriate from the point of view of its prosodic organization. However, semantically, there is, of course, only one interrogation. In C’est JeanPierre qui est sorti?, the question bears on the identity of the person who has left (Jean-Pierre) and not on the action of leaving itself. Thus, a single interrogative


99

morpheme has two identical intonational markers — a phenomenon of  , which will be shown to be independenty attested in 3.3. There exist interesting variants of this interrogative pattern. Our first example, (45), illustrates a case of duplication involving a modification of the pitch range between the two parts of the overall contour. (45)

[C’est Jean-Pierre] [qui est sorti]? | | | | L H% L H%

(see Figure 6)

The second part is realized on a higher pitch, but it can also be realized on a lower pitch.24 Such shifts in pitch do not affect the recognition of the “double interrogation” pattern in any respect. The interrogative pattern shown in the next example is more complex than the preceding one. A nuance of incredulity is added, whose realization is the falling contour following the rising interrogative contour. (46)

[C’est Jean-Pierre] [qui est sorti]? | | | | L HM% L HM%

(see Figure 7)

Here, we must posit two successive tones, H% and M%, forming a “contour” or non-level boundary tone, which seems independently necessary for a precise description of French.25 Again, the whole contour, constituting a single complex unit, is duplicated in the postfocal phrase. Due to space limitations we cannot consider more patterns in support of our reduplication analysis. One important remark must however be made. Up to this point, we have only considered the intonation of focus/postfocus sequences, which might have suggested that tonal reduplication was characteristic of the prosodic organization of cleft sentences. However, we must underline that the postposed theme (or right-dislocated topic, as in Il est sorti, Jean-Pierre) has the same intonative behaviour as the postfocal coda, since it also receives a     .26,27 There is, then, nothing specific to cleft sentences, although they represent a  combination of the c’est que/qui… sequence,  the tonal agreement pattern. 3.3 Non-cleft C’est … qui/que sentences As was mentioned in Section 1, utterances beginning with C’est … qui/que but

100 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND without a narrowly focused phrase can be analyzed as signalling a broad focus — in other words, as having a totally rhematic VP: they refer to an individual or an event as a whole, and can therefore constitute full answers to questions, cf. (2) and (3); moreover, the phrase or sequence to the left of que/qui does not contrast with another possible term.28 As noted in 3.1 above, such utterances have a very distinct intonation from that of clefts. Thus, in (47), which is to be interpreted like (3b), there is only one single terminal intonation: (47)

[C’est le petit] [qui n’est pas rentré]. | | H[cont]% L%

(see Figure 8)

In other words, it is a simple declarative sentence containing two accentual groups: [C’est le petit], and [qui n’est pas rentré]. The declarative L% that we have met many times previously is present — but only at the  of the entire utterance: at the end of the first accentual group, we have a  H% tone (following the traditional terminology) which differs from the interrogative H% tone mainly by its lower pitch (see Di Cristo 1996): the continuative H% tone constitutes the typical  boundary tone. Many other intonation patterns (exclamation, doubt…) could be associated to the same broad-focused sentences, but their prosodic organization would always be different from the intonation pattern of the narrow-focused ones, as they never display the intonational “agreement” pattern. To summarize, we have first shown the existence of what we have called   between the focused phrase and the coda. The terminal boundary tones and the other prosodic characteristics of such sentences are present twice, once at the end of the focused phrase and again at the end of the c-clause. Second, we have shown that this agreement is   to cleft sentences, since the same default partial copying is displayed in right-dislocated topics.29 Finally, we have drawn a clear divide between the intonation patterns of C’est… qui/que utterances with narrow, and those with broad, focus, since the latter only have one terminal boundary tone.


101

102 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND 4.

On the semantics and pragmatics of French cleft sentences

Besides cleft sentences, French exhibits two more (narrow-)focusing strategies, one by sheer stressing of the focused word or phrase, the other by explicit quantification. Cleft structures are repeated as the (a) sentences in the following examples: (48)

a. b. c.

(49)

C’est M qui est sorti. ‘It’s Michel that went out.’ M est sorti. Seul M est sorti.

a.

C’est M que j’ai vue. ‘It’s Marie that I saw.’ b. J’ai vu M. c. J’ai seulement vu M.30 c.’ J’ai vu seulement M.

We saw in the preceding sections what differentiated French clefts from other C’est que/qui… structures, both syntactically and prosodically. For our description to be complete, we also need to see the degree to which these three constructions differ from each other, semantically and/or pragmatically. The (c) variants of (48) and (49) have specific properties. First, in a context like (50a), only the seul(ement) variant is possible (the symbol ‘#’ signals when the utterance is unfelicitous): (50)

a.

(— Tous les garçons sont sortis./Tous les garçons sont sortis?) (‘All the boys went out.’/‘Did all the boys go out?’) b. — Non, seul M est sorti. ‘No, only M. went out.’ # c. — Non, M est sorti. d. #— Non, c’est M qui est sorti.31

More generally, the explicitation of contrastive material requires the addition of such expressions as par exemple ‘for instance’ in the explicitly quantified structure, but is unfelicitous in the other two: (51)

a. b. c.

C’est M qui est sorti, pas J-P (#par exemple). M est sorti, pas J-P (#par exemple). Seul M est sorti, pas J-P #(par exemple).

(52)

a.

C’est M que j’ai vue, pas J (#par exemple).


b. c.

103

J’ai vu M, pas J (#par exemple). J’ai seulement vu M, pas J #(par exemple).

According to the foregoing tests, the explicitly quantified focus sentences should therefore be drastically distinguished from the other two. However, cleft sentences share the exhaustive listing property with explicitly quantified ones — those in seul(ement) — because the truth values of (54a,b) are just as contradictory or incompatible as are those of (53a,b), two couples of examples inspired by Szabolcsi (1981): (53)

a. b.

Seuls M et J-P sont sortis. Seul M est sorti.

(54)

a. b.

Ce sont M et J-P qui sont sortis. C’est M qui est sorti.

On the other hand, as is well-known, there is no necessary contradiction between (55a) and (b) — in fact, the (a) sentence  the (b) sentence: (55)

a. b.

M et J-P sont sortis. M est sorti.

In other words, both clefts and explicitly quantified focused sentences somehow convey the information that, given a set U provided by the context, what is predicated (as true) of the referent of the focused XP (a member or a subset A of U — see below) just  true of  other individual belonging to U (i.e. to the complement set A′ = U–A). That is, of course, no news to anybody, and the simplest way to account for the foregoing apparently contradictory data is to rely on the classic distinction32 between narrowly defined semantics (concerned with truth values), and pragmatics, taken to refer to whatever else the words ‘meaning of a sentence’ refer to. Let’s be (slightly) more specific: just as the lexical item seul(ement) contributes to the meaning of the sentence that the term it associates with is somehow unique, we believe that the cleft structure supplies this sort of semantic information too. We shall see in the next section how that can be done. Meanwhile, in order to distinguish between the two “meanings”, we propose that the contextual set U referred to above can be either closed and “small”, or not. In the latter case, the explicitly quantified structure will be preferred (hence the felicity of (50b) in its context, and the necessity to acknowledge the size and/or openness of U by par exemple in (51c) or (52c)); in the former case, on

104 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND the contrary, the cleft sentence strategy will be preferred, and the felicity judgements, reversed. But since the size of the set U does not affect the truth conditions, and since it may well change from one context of utterance to another, we are quite content to assign its assessment to the pragmatic dimension of the analysis. What is it then that triggers the parallelism noted in (50)–(52) beween cleft sentences and free focus sentences? We would like to suggest that just as clefts are implicitly quantified with respect to a closed and small contextual set U, so do (narrow) free focus sentences “evoke” a small, closed set of alternatives. In other words, we would like to claim that in a language like French, as far as 33 focus is concerned, Rooth’s (1992, 1996) “alternative semantics” is a sufficient tool to describe the phenomena — all the more so as the “evocation of alternatives” is not their rejection, as the possibility of adding au moins ‘at least’ after the subject NP Michel in (48b) shows. The results of this section can be summarized as in Table (56):34 (56)

T     F Type of sentence Clefts Association with seul Free focus

5.

quantification ≈ uniqueness + + –

Properties of the contextual set closed/small open/large closed/small

Towards an analysis

How can we interrelate the prosodic and semantic properties of French cleft sentences (as described in the foregoing sections) as simply and directly as possible? Given null assumptions governing the relationship between syntax and prosody, the results obtained in Section 3 require that we reject most of the analyses reviewed in Section 2 — except possibly the one that relies on Extraposition, simply because it is the only one that guarantees that the coda will be      IP :35 all the other analyses reviewed here, in addition to the problems raised in Section 2, also have to  that a focus feature be attached to the XP right-adjacent to the copula, so as to account for the destressing of the cclause. Yet another advantage (29) has over some of the rival hypotheses is the direct semantic interpretation which it provides for the copula (see below).


105

However, we saw in Section 2 that the Extraposition analysis faces some non-trivial problems: (i) it is difficult to see how such a mechanism could either be governed by a requirement that the pseudo-cleft out of which the rightadjoined CP is extracted have a definite, rather than a universal or generic, reading — or how the extraposition transformation itself could likewise contribute a definite or specific, rather than universal, reading of the subject expression it is extracted from; (ii) the very idea of deriving clefts from pseudo-clefts in French — as opposed to some other natural languages — also meets more empirical difficulties, among which the fact that in most cases the morphology of whatever material precedes the complementizer changes, and, more generally, that pseudo-clefts in French require Left Dislocation, thereby duplicating the “demonstrative” element that only appears once (in subject position) in the cleft sentence proper; (iii) when we extend the analysis to FXPs where X is a preposition (or an adverb), the Extraposition approach seems to fail completely. A fourth problem must now be added to our list: we must also take into account the fact that Extraposition usually signals the  of the extraposed phrase,36 whether a DP, a PP or, more crucially, a relative CP, as in (57), whose prosodic pattern is like that of ex. (47), and where the last lexical word is (at least) normally stressed, in utter contrast with the prosody of c-clauses in clefts: (57)

a. b.

[IP [IP [DP Un homme ti] est venu] [CP qui voulait te parler]i] [IP [IP [DP A man ti] came] [CP who wanted to talk to you]i]

It seems to us that most, if not all, the foregoing shortcomings of the Extraposition hypothesis can be overcome if   CP  -  -     , as in (58b) vs. (58a): (58)

a. b.

[IP C’estv [VP tv [NP le petit]]] [IP [IP C’estv [VP tv [NP le petit]]] [CP Opi [C′ qui [IP ti est tombé]]]]

Such a hypothesis guarantees maximal transparence as far as the syntax-phonology interface is concerned: since the adjoined CP has no argumental, or even predicational, status, it has no autonomous prosodic properties, and the terminal intonation which le petit takes on in both (a) and (b) simply derives from its being the last (in fact, the only) canonical — and, for that matter, rhematic — XP in both structures; the intonational “agreement” of the coda then follows as the application of a default rule.37

106 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND What about the syntax-semantics interface now? If our hypothesis is correct, the only problem is to identify the property that salvages the whole of an expression like (58b): if (58a) is interpretable, by Full Interpretation, the adjoined CP should not be interpretable, and (58b) should crash at the LF interface, but it does not. Consider now the semantic parallelism between (59a) and (b) — which is naturally responsible for the Extraposition account which we have rejected on syntactic grounds:38 (59)

a. b.

[DP Ce [CP Op que j’ai acheté]](, c’)est un/le dictionnaire ‘What I’ve bought is a/the dictionary’ [[IP C’est un/le dictionnaire] [CP Op que j’ai acheté t]] ‘It’s a/the dictionary that I’ve bought’

In the (a) case, leaving aside the dislocation facts, which are not semantically relevant (see Section 2 for their syntactic import), the interpretation of the sentence is straightforward. In particular, the interpretation of the free relative — a DP — will be compositionally computed from the semantics of ce, taken to be a definite determiner, and the natural translation of the relative into a property (lw[acheter(je,w)]): (60)

a. b.

[D ce] ⇒ lP[lQ[∃x[∀y[P(y) ↔ x = y]&Q(x)]]]39 [DP ce [CP Op que j’ai acheté]] ⇒ lP[lQ[∃x[∀y[P(y) ↔ x = y]&Q(x)]]](lw[acheter(je,w)])

After two l-conversions, we get (61c): (61)

a. b. c.

lP[lQ[∃x[∀y[P(y) ↔ x = y]&Q(x)]]](lw[acheter(je,w)]) [lQ[∃x[∀y[[lw[acheter(je,w)](y)] ↔ x = y]&Q(x)]]] [lQ[∃x[∀y[[acheter(je,y)] ↔ x = y]&Q(x)]]]

Next the full sentence has its meaning compositionally calculated from the translation (61c) of the free relative as a generalized quantifier and the predicate (lz[être-un/le-dictionnaire(z)]) applied to it:40 (62)

a. b. c.

[lQ[∃x[∀y[[acheter(je,y)] ↔ x = y]&Q(x)]]] (lz[être-un/le-dictionnaire(z)]) [∃x[∀y[[acheter(je,y)] ↔ x = y]& [lz[être-un/le-dictionnaire(z)]](x)]]] [∃x[∀y[[acheter(je,y)] ↔ x = y]& [être-un/le-dictionnaire(x)]]]


107

Suppose now that when it instantiates a whole DP, ce has the lexical property of letting its restriction (i.e. predicate variable P) be free, as in (63), the crucial variant of (60a): (63)

[DP ce] ⇒ [lQ[∃x[∀y[P(y) ⇔ x = y]&Q(x)]]]

(63) will then combine with the explicit predicate (lz[être-un/le-dictionnaire(z)]), as in (64)a, successively yielding (64b) and (c): (64)

a. b. c.

[lQ[∃x[∀y[P(y)× ↔ x = y]&Q(x)]]] (lz[être-un/le-dictionnaire(z)]) [∃x[∀y[P(y) ↔ x = y]& [lz[être-un/le-dictionnaire(z)]](x)]] [∃x[∀y[P(y) ↔ x = y]&[être-un/le-dictionnaire(x)]]]

We now want to suggest that (64c) is the translation of the sentence (65) — besides being equivalent to (66), a fact which clearly shows where the implicit quantificational import of cleft sentences lies: (65)

C’est un/le dictionnaire

(66)

[lx[être-un/le-dictionnaire(x)]] (ıy[P(y)])41

with P either contextually (i.e. extra-sententially) bound, or referring deictically. Suppose now that a relative like the one in (59b) is adjoinded to the sentence (65); being a relative clause, it translates again into a property; but it has no “antecedent” whose domain it could restrict; so the natural way for it to be licenced at the syntax-semantics interface is to bind a property variable. Now that is exactly what (63) or (64c) can do for that relative — and, more generally, for the analysis of how French cleft sentences work and are interpreted (of course, the same reasoning applies to [+human] cases like (58a,b)).42 The situation is, however, slightly more complicated than appears above, because coordinated XPs and plural XPs can appear in the post-copular construction. This can be dealt with easily by adopting Jacobson’s (1995) suggestion that the iota operator (materialized by ce itself in our analysis) range over    (MPIs) rather than mere atomic individuals, an idea vindicated by the fact that the subject form ce has no plural, even though the copula itself may (and, in formal style, must), agree in number with the focused phrase, as shown in (67). (63) should then be replaced by (68), where the capital letters X and Y are meant to refer to such MPIs:43

108 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND (67)

a. b.

(68)

6.

Ce sont les enfants qui sont tombés . ‘it  the children that fell down’ C’est les enfants qui sont tombés

[] []

[DP ce] ⇒ [lQ[∃X[∀Y[P(Y) ↔ X = Y]&Q(X)]]]

Conclusions

To summarize, we have shown that in French cleft sentences, the post-focal “coda” does not occupy a subcategorized position, but is rather adjoined to a regular identificational IP, and is interpreted as a predicate that binds a predicate variable associated with the translation of the [Spec,IP] ce, thereby incorporating the semantic intuition that underlies the Extraposition hypothesis — but also making allowance for the fact that from the point of view of truth conditions, clefts and focus associated with seul(ement) share a common quantificational meaning, as was noted when pointing out the equivalence between (64c) and (66). Moreover, the copula is literally interpreted as a (binary) predicator of identification, this identification being pragmatically the most prominent dimension of the whole sentence (as opposed to seul(ement) sentences, where it is the quantification that is foregrounded); finally, the prosodic properties of cleft sentences have been shown to directly derive from their syntactic structure, since the intonation of the first group, from c’est till the end of the FXP, is exactly that of an identificational sentence built up with the same words, and the prosodic contour of the c-clause is characterized by a sequence of non-terminal L tones followed by a copy of the terminal tone of the FXP, i.e. by “intonational agreement”, a phenomenon independently attested. Our analysis therefore corroborates one of the most basic discoveries of the Principles and Parameters research program, namely, that “constructions” are a descriptive notion, with little, if any, theoretical content:44 in the case under discussion,       . What we do find (independently) is that French can build identificational sentences like (65), and relative clauses like the CP in (59b). Combining (or Merging) such sequences will then be submitted to (bare) output conditions, and we’ve just seen how these may apply to both the Articulatory-Perceptual and the Conceptual-Intentional sides of the expression thus created. This analysis is thus maximally simple as far as the syntax is concerned: no other movement is required than the raising of the abstract relative operator (or


109

of a wh feature) in the coda, which is necessary anyway. In particular, since the c-clause is higher than the surface pseudo-expletive subject ce, there is no question of adjoining the former to the latter: the solution proposed here is therefore more economical than any analysis that would involve overt or covert movement of anything; hence, it is more economical than covertly adjoining the focused XP to the pronominal subject, as suggested by Ouhalla (this vol.) to deal with clefts in Arabic, and perhaps in English too. In any case, it appears that in the case investigated here, the semantic interpretation can be directly read off the s-structure, thereby weakening the view that Spell-out is not a pertinent level of representation (see Groat & O’Neil (1996) or Kitahara (1996) for independent arguments that Spellout may well directly contribute more to the semantic interpretation proper than was thought previously). One might complain that everything’s too good to be true; however, we must acknowledge that French cleft sentences are comparatively special; the development of the Qui est-ce …que/qui…? question type as unmarked is a clear an indication of the fact. Another argument is, of course, provided by crosslinguistic variation: compare the properties of English clefts (which admit explicit relative pronouns in the coda), Wolof clefts (Kihm, this vol.), and more perspicuously still, Arabic clefts (Ouhalla, this vol.), which have the form [DP–Copular Pronoun–DP]; here, the utterance-final constituent clearly is a free relative, distributionally almost equivalent to an ordinary DP, and headed by an explicit relative marker — two properties that are significantly absent in the case of French cleft codas. We are consequently ready to entertain the view that the common “cleftness” of French and Arabic cleft sentences may well be a superficial matter, even though distinct semantic analyses applied to distinct syntactic structures may nonetheless lead to an identical semantic or denotational core.

Notes * Thanks to Ileana Comorovski, Denis Creissels, François Lonchamp and Laurie Tuller for their remarks on an earlier draft. Needless to say, the usual disclaimers apply. 1. (8) would be acceptable if Ça were given a purely exclamative value (cp. Ça alors!), rather than a deictic one; but such an interpretation is, of course, irrelevant. 2. See however Robert (1993) for a tentative reduction of such sentences to clefts on a purely pragmatic basis. 3. This fact may be due to independent causes: (3b) is causal with respect to (3a), whereas (4b)

110 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND is consecutive/resultative with respect to (4a). 4. When the predicate is dynamic, yet a third interpretation obtains — see ex. (3b); we shall not repeat this proviso when the context is clear. 5. This is not the case in English; see Chomsky (1981: 280), where It is John to whom I spoke is analyzed on a par with It’s the book that was on the table (p. 245:(15iii)); see also Radford (1988: 523); but French is by no means exceptional: according to Nølke (1984), explicit relative pronouns governed by prepositions are blocked in Danish clefts just as they are in French. 6. We borrow “c-clause” from Nølke (1984), but the intended meaning there was “cleft-clause”. 7. See Creissels & Rebuschi (in prep.) for some data drawn from a totally unrelated language, Tswana, which exhibits the same sort of distinction between restrictive relative clauses proper, and c-clauses in cleft sentences. What is interesting about Tswana is that the language has no que≈qui alternation rule (as might be expected) but rather has relative pronouns even when it is the subject or direct object slot which is relativized; consequently, Tswana has two unambiguous counterparts to (11) in the text, one that corresponds to the restrictive relative interpretation, and has both a relative pronoun and the suffix -ng that marks the finite verb form as “relative”, and the other that only displays the latter suffix, and corresponds to c-clauses. In any case, the issue is not whether the c-clause has the internal structure of a relative clause (Schachter already showed that in 1973), but whether it also entertains the usual  relationship a relative bears to the rest of the sentence, and whether it might have the  of a relative clause. 8. A sentence like C’est le Michel qui parle russe presupposes the existence of (at least) two individuals whose name is Michel, only one of which knows Russian; the restrictive character of the coda is therefore undisputable. 9. Chomsky (1977), after Pinkham & Hankamer (1975), chose to propose different derivations for different clefted phrases (see also Bresnan & Grimshaw 1978). We do not wish to follow such a tack. Chomsky’s (1977) model for clefted NPs (and PPs) was adopted by Obenauer (1981) in his treatment of French clefts. 10. For greater clarity, we will reword all the proposals in a Barriers or (immediately) post-Barriers framework, hoping that this will not mar the subsequent discussion. 11. Qui must of course be understood as the allomorph of que when the latter is followed by an empty subject position (see Kayne 1974). Note also that we will stick to a fairly traditional version of X-bar theory, more specifically the one exposed in Radford (1988). A modernized version would not affect the argumentation. 12. See Ouhalla (this volume) for Arabic clefts. It may be relevant to note that Horvath (1986), among others, translates Hungarian sentences in which focusing is realized by overt fronting, by both a stressed counterpart and a clefted counterpart. See Section 4 for some differences between clefts and the “association” of FXPs and seul(ement) ‘only’ in French. 13. For an early defense of English it and French ce as semantically non-empty entities, cf. Bolinger (1972, 1977) or Gundel (1977), and Pollock (1981) respectively. 14. Admittedly, many languages that overtly raise XPs to the left periphery do not have cleft sentences — e.g. Roumanian, or Standard Basque. However, this is by no means universal:


111

Standard Arabic (Ouhalla, this vol.), Lekeitio Basque (Hualde et al. 1994) and Malayalam (Jayaseelan 1996) do use both strategies; now, in those languages, the availability of FXP Raising should block the derivation using the less economical clefting strategy; the corresponding LFs must therefore be distinct, at least in such languages. 15. Such data could perhaps be used to justify Raising (so as to evade the requirement that luimême be licensed by se when in direct object position); but we do not see how this sort of explanation could carry over to simpler cases like (1b)≈(20), or to examples where lui-même is governed by a preposition: (a)

C’est (à propos) de lui-même que Michel parle comme cela. ‘It’s about himself that M. talks like that.’

(b)

Michel parle comme cela (à propos) de lui-même. ‘M. talks like that about himself.’

16. See Kihm (this volume) for an up-to-date analysis in those terms. 17. There are subvariants. Thus Chomsky (1970) mentions two distinct possibilities: (i)

[the one who writes poetry] is John

(ii)

[it–one writes poetry] is John

Harries-Delisle (1978) also suggests that the relative might have a head, which would be subsequently deleted. 18. Later on, some scholars even hoped to derive the very form of XPs in argumental position from their q-roles (cf. Chomsky 1986); hence, Goal could have been realized by an infinitival clause, Proposition by a (that) tensed clause, etc. 19. As far as we know, such an analysis has never been explicitly formulated for cleft sentences proper; but it suggests itself as a natural extension of Bresnan & Grimshaw’s (1978) paper on free relatives (note that Muraki (1970) already proposed that the same identificational [S Infl NP NP] d-structure be taken as underlying both clefts and pseudo-clefts.) 20. This sort of structure is more easily accessible in English than in French, as the comparative (in)acceptability of the following examples shows: (i)

Under the bed is where you should sleep.

(ii) *Sous le lit est (là) où tu devrais dormir. 21. In this paper we only present a few pitch curves to support our reduplication thesis. However, the analysis is based on our daily experience with spoken French, and it has been confirmed by the recording of 67 tokens of the sentence C’est Jean-Pierre qui est sorti, which all display the intonative agreement pattern. The recordings were obtained from four different native speakers who were asked to imagine any kind of possible situation for such an utterance. Other sentences were recorded in a more controlled fashion, using dialogue scenarios that triggered the desired intonation patterns. Moreover, most of those patterns are well-known and well documented from previous studies (which will be mentioned in due course). 22. The spontaneous character of our recordings is highlighted by the “qu’” realization of qui (whether it is due to a contraction of the latter, or the non-applcation of the ≈ Rule is irrelevant).

112 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND 23. It is more common in emphatic style, such as used in political speeches for instance. 24. See figures (5) and (6) for a clearer idea of the differences between (44) and (45). 25. In fact, the analysis of French intonation requires still other boundary contour tones, such as the LHL pattern for the well-known implication countour; such boundary tones have been proposed by other authors (Mertens 1987; Di Cristo 1996), using other terms. 26. For example, consider the assertion (i.a) and the question (i.b): (i)

a.

b.

[Il est sorti,] [Jean-Pierre.] | | L L . ‘He’s gone out, J.-P.’ [Il est sorti,] [Jean-Pierre?] | | H H ‘Has he gone out, J.-P.?’

27. The intonations found of the coda and the potsposed theme have been called “parenthetic” by some authors, beginning with Delattre (1966), who posited two variants: a low level contour in the assertive utterances and a high level contour in the interrogative utterances, depending upon the melodic end-point of the main clause (low in the assertion, high in the interrogation). Actually, besides or instead of the high level contour following a rising interrogative contour at the end of the main clause, other scholars have found a rising contour (Di Cristo, 1997) similar to the one borne by the postfocus illustrated in Figure 6. Delattre used the concept of “variants”, which is different from the notion of “duplication”, but the idea of a copy was expressed in the description of the “high parenthesis” which he called “echo”. Later on, many studies followed the same line of thought recognizing a basic parenthetic contour (Rossi 1975; Ashby 1994, among others). Wunderli (1979) remained in Delattre’s tradition but reinterpreted the “echo” as a sort of copy and came close to retaining the hypothesis of an intonational copy between the main clause and the various right extraposed XPs which he treated as a group. But he rejected this hypothesis on the basis of a few counterexamples, all very long appositions in a oralized written corpus, which we consider too litterary to be relevant, since they are very difficult to find in a spontaneous speech. More recently, in his survey of French intonation, Di Cristo (1996) has provided a short analysis of the main clause/postponed theme contours with two identical terminal intonations, which closely approximates our own interpretation of the data. To summarize, if the duplication of the terminal intonation has often been mentioned under various guises, no systematic investigation of this process has yet been done, and more research will have to be done in at least two directions: (a) on its phonological nature, more precisely on the determination of the contours (the boundary tones in our terms) and possibly other features which might be copied (such as length, see fig. 4); and (b) on its syntactic determination, which is not limited to postfocal codas and postponed topics, but also concerns other types of “parentheticals” or “right extrapositions” which will have to be more precisely defined. 28. It is even probable that sequences consisting of an “antecedent” plus a rel. clause do not correspond to syntactic phrases, since restrictive CP are adjoined to N′ or NP (depending on one’s preferred analysis of nominal expressions as NPs or DPs), thereby leaving the Det out.


113

29. The same is true of utterances like (4b), whose syntax and semantics are quite distinct from those of clefts proper. 30. As the scope of seulement would be ambiguous here, we represent the prosodic prominence which corresponds to the intended meaning; to render the data more readily comparable, we also use small capitals to indicate prosodic prominence in the other cases, where it is predictable. Besides, we must confess we have no explanation for the fact that seul(e) (as opposed to the adverb seulement) is restricted to the subject NP/DP. 31. (50c) improves if mais pas les autres ‘but not the others’ is added, but nothing of the kind salvages the clefted variant. 32. Explicitly rejected for instance by Erteschik-Shir (this vol.). 33. Things are no doubt different when explicitly quantified focusing are considered; see Krifka (1991, 1992) or Pulman (1997) for a critique of Rooth’s position  , and the intuitive discussion above. 34. Broad focus could thus be regarded as corresponding to the fourth possibility (no uniqueness implied, with the contextual set U large or open). Thus, as a reply to (i), (ii) contains an object NP which arguably corresponds to this notion: (i)

— Qui as-tu vu à la manif? ‘Who did you see at the demo?’

(ii)

— J’ai vu Jean Dupont. ‘I saw J.D.’

35. At least in the form we have given it — since for instance Akmajian (1970) and those who followed him in the seventies allowed the extraposed clause to be dominated by the the same S node that directly dominates the pronominal “impersonal” subject. 36. See Guéron & May (1984), Rochemont & Culicover (1990), among others. 37. If the post-copular XP is longer, as in (18), then it is the last lexical item it contains, dernier, which will carry the typical HL% contour. 38. To avoid complications with the allomorph celui used when the domain is restricted to [+human] entities, we temporarily choose an example with an [−human] feature. 39. For simplicty’s sake, we ignore the restrictions over the contextual set mentioned in Section 4. 40. Owing to the specific morpho-syntactic of free relatives in French, it is unnecessary to analyze them as forcing the type shifting operation proposed by Heycock & Kroch (1996) to acount for the English counterpart of (59a). 41. Cf. Löbner’s (1990) translation of the following German sentence, where the subject is focused: (ii)

[F Anna] hat mir ein Bild geschenkt

(iii)

lx[x = Anna] (ix[POS(x gave me a picture)])

where POS is the positive operator ‘it is true that’ (cited in von Stechow 1991). Here again, we find a possible explicitation of Chomsky’s paraphrases of certain sentences containing a focused element (e.g. Chomsky 1972). Recall, however, that we doubt that our analysis carries over to free focus (cf. Section 4).

114 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND 42. The idea of letting the coda bind a predicate variable which is itself part of the semantic content of a pronoun has been inspired by Cooper’s (1979) work on correlative constructions (which also involve the adjunction of a CP to an otherwise well-formed sentence). Srivastav (1991) suggests that correlative pronouns are variables which can be bound by the (translation of the) correlative clause into a property containg an iota operator — or “remain free and refer deictically” when there is no correlative CP. That would, however, entail that ce be given two totally unrelated translations, an unnecessary move at least in the case we are dealing with here — or even in a case like (i), under one possible analysis, whereby it is now the protasis as such that would provide the binder of P. (i)

Si quelqu’un fait cela, est un criminel ‘If someone does that, he [lit. ] is a criminal.’

43. We also follow Jacobson in considering atomic individuals to be a subset of plural individuals. 44. See also Chomsky (1995: 170): “The notion of grammatical construction is eliminated, and, with it, construction-particular rules. Constructions such as verb phrase, relative clause, and passive remain only as taxonomic artifacts, collections of phenomena explained through the interaction of the principles of UG, with the values of parameters fixed.”

References Abraham, W.; Epstein, S. D.; Thráinsson, H. & Zwart, J.-W. (eds.). 1996. Minimal Ideas. Amsterdam: John Benjamins. Akmajian, A. 1970. “On Deriving Cleft Sentences from Pseudo-cleft Sentences”. Linguistic Inquiry 1.2:149–168. Ashby, W. 1994. “An Acoustic Profile of Right-dislocations in French”. Journal of French Language Studies 4.2: 127–145. Bergvall, V. 1987. Focus in Kikuyu and Universal Grammar. Doctoral dissertation, Harvard University (distrib. by UMI Dissertation Services, Ann Arbor, Michigan). Bolinger, D. 1972. “A Look at Equations and Cleft Sentences”. In E. S. Firchow et al. (eds.). Studies for Einar Haugen, Presented by Friends and Colleagues. The Hague: Mouton, 96–114. Bolinger, D. 1977. “It”. In D. L. Bolinger, Meaning and Form. London: Longman, 66–89. Bresnan, J. & Grimshaw, J. 1978. “The Syntax of Free Relatives in English”. Linguistic Inquiry 9.3:331–391. Chierchia, G. 1995. Dynamics of Meaning. Chicago: The University of Chicago Press.


115

Chomsky, N. 1970. “Deep Structure, Surface Structure, and Semantic Interpretation”. In R. Jakobson & S. Kawamoto (eds.). Studies in General and Oriental Linguistics Presented to Shiro Hattori […]. Tokyo: TEC, 52–91. Chomsky, N. 1974. The Amherst Lectures. Paris: U. de Paris 7 [mimeo.]. Chomsky, N. 1975. “Questions of Form and Interpretation”. Linguistic Analysis 1.1, 75–109. Chomsky, N. 1976. “Conditions on Rules of Grammar”. Linguistic Analysis 2.4, 303–351. Chomsky, N. 1977. “On wh-movement”. In P. Culicover, T. Wasow & A. Akmajian (eds.), Formal Syntax. New York: Academic Press, 71–132. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge (Mass.), MIT Press. Chomsky, N. 1986. Knowledge of Language. Its Nature, Origin, and Use. New York: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge (Mass.): MIT Press. Cooper, R. 1979. “The Interpretation of Pronouns”. In F. Heny & H. Schnelle (eds.), Selections from the Third Groningen Round Table. New York: Academic Press, Syntax & Semantics 10, 61–92. Creissels, D., & Rebuschi, G. In prep. “Focus in Tswana”. Delahunty, G. P. 1984. “The Analysis of English Cleft Sentences”. Linguistic Analysis 13.1:63–113. Delattre, P. 1966. “Les dix intonations de base du français”. The French Review 40.1:1–14. Di Cristo, A. 1996. “Intonation in French”. In D. Hirst & A. Di Cristo (eds.), Intonation Systems: A Survey of Twenty Languages. Cambridge: CUP. Féry, C. 1993. German Intonational Patterns. Tübingen: Max Niemeyer. Groat, E. & O’Neil, J. 1996. “Spell-Out at the LF Interface”. In W. Abraham et al. (eds.), 113–139. Gross, M. 1968. Grammaire transformationnelle du français. Syntaxe du verbe. Paris: Larousse. Guéron, J. & May, R. 1984. “Extraposition and Logical Form”. Linguistic Inquiry 15.1:1–31. Gundel, K. 1977. “Where do Cleft Sentences Come from?”. Language 53.3: 543–559.

116 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND Harries-Delisle, H. 1978. “Contrastive Emphasis and Cleft Sentences” In J. Greenberg (ed.), Universals of Human Languages, 4: Syntax. Stanford, CA: Stanford UP, 419–486. Heycock, C. & Kroch, A. 1996. “Pseudocleft Connectivity: Implications “. Ms., U. of Edinburgh & U. of Pennsylvania. Higgins, F.R. 1973. “On J. Emonds’ Analysis of Extraposition”. In J. Kimball (ed.), Syntax and Semantics 2. New York: Academic Press, 149–195. Horvath, J. 1981. Aspects of Hungarian Syntax and the Theory of Grammar. Doctoral dissertation, UCLA. Horvath, J. 1986.  in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Hualde, J.I., Elordieta, G., & Elordieta, A. 1994. The Basque Dialect of Lekeitio. San-Sébastián–Donostia: Gipuzkoako Foru Aldundia (ASJU Supplements 34). Jackendoff, R. 1972. Semantic Interpretation in Generative Grammar. Cambridge (Mass.): MIT Press. Jackendoff, R. 1977. X-bar Syntax. A Study of Phrase-structure. Cambridge (Mass.): MIT Press. Jacobson, P. 1995. “On the Quantificational Force of English Free Relatives”. In E. Bach, et al. (eds.), Quantification in Natural Languages. Dordrecht: Kluwer, 451–486. Jayaseelan, K.A. 1996. “Question-word Movement to Focus and Scrambling in Malayalam”. Linguistic Analysis 26.1–2:63–83. Kayne, R. 1974. “French Relative que”. Recherches linguistiques 2–3. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge (Mass.): MIT Press. Kihm, A. “Focus in Wolof”. This volume. Kiss, K.É. 1981. “Structural Relations in Hungarian, a ‘Free’ Word Order Language”. Linguistic Inquiry 12.2:185–213. Kitahara, H. 1996. “Raising Quantifiers without Quantifier Raising”. In W. Abraham et al., (eds.), 189–198. Krifka, M. 1991. “A Compositional Semantics for Multiple Focus Constructions”. Linguistische Berichte Sonderheft 4, 17–54. Krifka, M. 1992. “A Framework for Focus-Sensitive Quantification”. In C. Barker & D. Dowty (eds.), Proceedings of the Second Conference on Semantics and Linguistic Theory. Working Papers in Linguistics 40 (Ohio State University), 213–236.


117

Löbner, S. 1990. Wahr neben Falsch. Duale Operatoren als die Quantoren natürlicher Sprache. Tübingen: Niemeyer. Manfredi, V. 1993. “Verb Focus in the Typology of Kwa/Kru and Haitian”. In F. Byrne & D. Windford (eds.), Focus and Grammatical Relations in Creole Languages. Amsterdam: Benjamins, 3–51. Mertens, P. 1987. L’intonation du français. Doctoral Dissertation, Katholieke Universiteit Leuven. Muraki, M. 1970. “Presupposition and Pseudo-Clefting”. Papers from the Sixth Regional Meeting. Chicago: CLS, 390–399. Nølke, H. 1984. “Clefting in Danish?”. Nydanske Studier & Almen Kommunikationsteori 14:71–111. Obenauer, H.-G. 1981. “Le principe des catégories vides et la syntaxe des interrogatives complexes”. Langue française 52:100–118. Ouhalla, J. (this volume) “Focus and Arabic Clefts”. Pasdeloup. V. 1990. Modèles de règles rythmiques du français appliqués à la synthèse de la parole. Doctoral dissertation, Université d’Aix-en-Provence. Pierrehumbert, J. 1980. The Phonology and Phonetics of English Intonation. Doctoral dissertation, MIT. Pinkham, J. & Hankamer, J. 1975. “Deep and Shallow Clefts”. Papers from the Eleventh Regional Meeting. Chicago: CLS, 429–460. Pollock, J.-Y. 1981. “On Case and Impersonal Constructions”. In R. May & J. Koster (eds.), Levels of Syntactic Representation. Dordrecht: Kluwer, 219–252. Pulman, S. 1997. “Higher Order Unification and the Interpretation of Focus”. Linguistics & Philosophy 20.1:73–115. Radford, A. 1988. Transformational Grammar. Cambridge: CUP. Riemsdijk, H. van & Williams, E. 1986. Introduction to the Theory of Grammar. Cambridge (Mass.): MIT Press. Rizzi, L. & Roberts, I. 1989. “Complex Inversion in French”. Probus 1. Robert, S. 1993. “Structure et sémantique de la focalisation”. Bulletin de la société de linguistique 88.1:25–47. Rochemont, M. & Culicover, P. 1990. English Focus Constructions and the Theory of Grammar. Cambridge: CUP. Rooth, M. 1992. “A Theory of Focus Interpretation”. Natural Language Semantics 1.1:75–116. Rooth, M. 1996. “Focus”. In S. Lappin (ed.), The Handbook of Contemporary Semantic Theory. Oxford: Blackwell, 271–297.

118 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND Rossi, M. 1985. “L’intonation et l’organisation de l’énoncé”. Phonetica 42: 135–153. Rossi, M. 1993. “A model for predicting the prosody of spontaneous speech (PPSS model)”. Speech Communication 13:87–107. Ruwet, N. 1975. “Les phrases copulatives en français”. Recherches linguistiques 3:143–191. Schachter, P. 1973. “Focus and Relativization”. Language 49:19–46. Srivastav, Vaneeta. 1991. “The Syntax and Semantics of Correlatives”. Natural Language and Linguistic Theory 9.637–686. von Stechow, A. 1991. “Current Issues in the Theory of Focus”. In A. von Stechow & D. Wunderlich (eds.), Semantik/Semantics. Berlin: Walter de Gruyter, 804–825. Sun-Ha, J., & Fougeron, C. 1995. “The Accentual Phrase and the Prosodic Structure of French”. ICPhS 95 (Stockholm Session 38.3, vol. 2), 722–725. Szabolcsi, A. 1981. “The Semantics of Topic-Focus Articulation”. In J. Groenendijk, T. Janssen, & M. Stockhof (eds.), Formal Methods in the Study of Language 2. Amsterdam: Matematisch Centrum, TRAC 136, 513–540. Thorsen, N.G. 1985. “Intonation and Text in Standard Danish”. Journal of the Acoustical Society of America 77:1205–1216. Touati, P. 1987. “Structures prosodiques du français et du suédois”. Travaux de l’Institut de Phonétique de Lund [Lund University Press]. Williams, E. 1980. “Predication”. Linguistic Inquiry 11.1:203–238. Wunderli, P. 1979. “Au sujet de l’intonation du français: la parenthèse en position finale”. Travaux de linguistique 6:83–111 [Gent].

Focus Structure and Scope* Nomi Erteschik-Shir Ben-Gurion University of the Negev

Abstract Focus Structure (f-structure), an annotated structural description in which Topic and Focus constituents are marked, is viewed as the interface between syntax and semantics. The f-structure theoretical approach to interpretation is dynamic in that it presupposes a discourse theory which defines the state of a file both before and after the utterance of the sentence. The state of the file before the utterance of the sentence determines the potential f-structures licensed for it. The resulting state of the file determines the application of the semantic rule of predication which assigns a truth value to the sentence by evaluating the predicate with respect to the topic. It follows from the rule of predication that quantified phrases which function as topics must have wide scope. F-structures are thus scopally transparent.

Introduction F-structure theory is a model of informational structure. It provides a natural account of a wide range of linguistic phenomena, usually accounted for by syntactic constraints and rules of LF, including quantifier scope, interpretations of wh-questions, anaphora, and extraction. A model of grammar necessarily includes f-structure since it determines intonation. F-structure theory allows for a large part of syntax, including LF syntax, to be handled by f-structure principles. The model I propose is therefore maximally simple in that it enables interpretation of f-structures directly without the mediation of LF. F-structure is an annotated structural description in which Topic and Focus

120

NOMI ERTESCHIK-SHIR

constituents are marked. F-structure feeds both PF and semantics and is sensitive to lexical information. It feeds PF since this level provides the explicit phonetic spell-out including intonation. I argue that f-structure and not LF is the input to a semantic rule of Predication. Under this view, the model of grammar takes the following shape: (1)

SYNTAX SD f-structure / \ PF semantics

A variety of results follow from the view that f-structure mediates between syntax and semantics. This paper concentrates on issues related to scope and demonstrates that f-structures are scopally disambiguated. One important result is that the topic by definition takes scope over the rest of the sentence. Further, the existence of f-structures with implicit stage topics allows for unscoped interpretations. The idea that F-structure affects truth conditions is not new. Rooth (1985) among others has shown that Focus assignment may determine truth conditions. Chierchia (1992) shows that Topic is what forms the restrictor on adverbs of quantification. Partee (1992) discusses the idea that the restriction of tripartite discourse representations is akin to Topic and that nuclear scope is the Focus. Following Reinhart (1981), I adopt the Strawsonian view that the topic is the pivot for assessment and show that f-structures involving both main and subordinate assignments of topic and focus are required for interpretation. In particular, I show that quantifier scope is determined by f-structure and that it makes unnecessary the need for an LF component. F-structure theory thus feeds semantics. It is a property of dynamic semantic theories that the borderline between pragmatics (involving circumstance of use) and semantics is blurred. Fstructure theory blurs this line further:

1.

F-structure

The f-structure theoretical approach to interpretation is dynamic in that it assumes a theory of discourse which defines the state of the common ground both before and after the utterance of a sentence. I adopt and modify Reinhart’s

FOCUS STRUCTURE AND SCOPE

121

(1981) file metaphor for the organization of the common ground. The common ground is viewed as having the following structure: It consists of a set of file cards with indexed headings which represent existing discourse referents. The common ground propositions form entries on these file cards. Only propositions which are interpretable as properties of a particular discourse referent are entered on the file card for that discourse referent. Common ground information is thus ordered according to the topics defined by each discourse referent. The stack of cards is partitioned into prominent and nonprominent cards: the top-of-the-file is where “prominent” cards are to be found. The cards on top-ofthe-file are licensed as potential topics of an utterance. Thus, the state of the file before the utterance of the sentence determines the potential f-structures licensed for it. An NP contained in a focus constituent is considered to be focused. Focusing a NP in the sentence results in positioning a card on top of the file. Focusing thus triggers a new state of the file.1 The following f-structure rules apply to f-structures (SDs annotated for Topic and Focus): -  I

TOPIC instructs the hearer to locate on the top of his file an existing card (or an existing set of cards) with the relevant heading and index.

II

FOCUS instructs the hearer to either i) open a new card and put it on the top of the file. Assign it a heading and a new index (in the case of an indefinite) or ii) locate an existing card and put it on the top of the file (in the case of a definite)

III PREDICATION instructs the hearer to evaluate the predicate with respect to the topic where the predicate is taken to be the complement of the topic. If the result of the evaluation is TRUE the UPDATE rule applies: IV UPDATE instructs the hearer to enter the predicate on the topic card and then to copy all entries to all cards activated by the focus rule. F-structure theory is a pragmatic theory which is concerned with felicity conditions on the relation between sentences and context. Thus TOPIC can be

122

NOMI ERTESCHIK-SHIR

assigned only to constituents for which file cards are available on top of the file, i.e., cards which have been positioned there by the application of the f-structure rules to previous utterances. The constraints on f-structure (f-structure rules I and II) are pragmatic, i.e., context determines whether a particular f-structure can be assigned to a sentence. Yet, f-structure determines interpretation (f-structure rule III) and therefore the view of grammar proposed here requires a semantic theory which takes fstructure, rather than LF, as its input. In Erteschik-Shir (1997) I also claim that certain syntactic constraints must be defined on f-structure. The model of grammar proposed here is therefore distinct from most other proposed models in that it incorporates pragmatics (in the form of f-structure).2 1.1 Illustration Rules I and II apply to the constituents marked TOP and FOC respectively. Rules III and IV are interpretive rules which apply to all f-structures to which TOP and FOC have been freely assigned. The latter two rules are informally sketched here and clearly need semantic formalization. Assume the following interaction: A is speaking, B is listening. The cards for the speakers are available on top of the file: A1 is licensed as the topic of (2): A says: (2)

I [have a dog]. [It] is brown. FOC TOP

B’s update: 1. Select the card for A1 (first person) from the top of the file. (TOPIC rule) 2. Evaluate “A1 has a dog” with respect to A1. (PREDICATION) 3. If 2 yields TRUE, enter “e has a dog” on A’s card. (UPDATE) 4. Open a new card, label it dog2. Put it on top of the file (FOCUS rule i.) 5. Enter “A1 has e” on this card. (UPDATE) The following cards are now on top of the hearer’s file and are available as future topics: card #1

card #2

A1

= heading

dog2

e has dog2

= entry

A1 has e


123

The card index is assigned by a function which maps the set of cards onto the set of discourse referents. The heading, however, is an attribute of the card (such as dog). When a card is first introduced the constituent which introduces it determines the form of this heading. In the new card for dog2, the heading allows future definite references to this dog such as the dog. Once entries are added to the card, any other attribute can take the place of this heading, deriving new headings such as the dog you have (your dog) or the brown dog. Entries can in this way be viewed as restrictions on the heading. The entry for the second sentence in (2) can now be made by B. The pronoun is interpretable only if it is matched with an available card from the top of the file, i.e., the referent must have been introduced and a card for this referent must have been positioned on top of the file. Card 2 represents such a card since the features of its heading in match those of the pronoun. The entry can therefore be licensed on this card. The following steps are now taken by B: 1. Select card 2 from the top of the file. (TOPIC rule) 2. Evaluate “e is brown” with respect to dog2. (PREDICATION) 3. If 2. yields TRUE, enter “e is brown” on card 2. (UPDATE) Note that a sentence may have a variety of f-structures depending on context. (2), for example, could be a response to questions such as (3)

a. b.

Do you have a pet? Which pets do you have?

In that case the f-structure of (2) would have only the object in focus as in: (4)

ITOP have [a dog]FOC

Here the focus and the predicate are not coextensive as in the previous case. The questions themselves trigger a manipulation of the file. They generate cards with variables in them. The card generated by the question in (3b) roughly looks as follows:3 A1 e has x Presuppositions are assumed to be existing entries on cards. It follows that the entry on the card is presupposed by the answer.4 The answer simply replaces the variable with the entry a dog, resulting in a card identical to the one in card #1. This analysis accords with the traditional view that the constituent which answers

124

NOMI ERTESCHIK-SHIR

a wh-question is the focus. If the object were a potential topic, the following f-structure could be assigned: (5)

TOPi [John has [the dog]i]FOC

Here the topic is the object. Again, a card with the heading the dog must be available on the top of the file. And if the sentence is assessed as true with respect to the topic the dog, the new entry will be added: dog2 John has e The f-structure in (5) has an initial TOP followed by a focus constituent. The fstructures associated with syntactic topicalization structures are isomorphic to the parallel f-structures without topicalization. This f-structure thus exhibits a notational reflection of syntactic topicalization or can alternatively be viewed as a f-structure theoretical variant of LF topic movement.5 The following f-structure is ruled out by the premise that a focus must constitute a syntactic constituent: (6)

[John has]FOC [the dog]TOP

From the perspective of discourse theory (6) is, however, merely a notational variant of (5). F-structures are freely assigned to SDs. F-structures are, however, constrained by context. If, for example, Topic is assigned to a constituent which is not available on top of the file, the f-structure rules cannot apply and no interpretation will be derived. Further constraints on f-structure are the Topic Constraint and the Subject Constraint discussed below. 1.2 Stage Topics It follows from the rule of Predication that all sentences must have topics, since the assignment of truth values depends on their presence. Out-of-the-blue sentences (answers to What happened?) have traditionally been considered to be all focus and to have no topic. I argue here that such sentences, in fact, have what I refer to as a  topic. A stage topic (sTOPt) defines the spatiotemporal parameters of the utterance. Stage topics may be overt (‘this afternoon’, ‘on Park Avenue’) or discoursally implied (the here-and-now). The truth value of a sentence with such a topic is determined by examining a card with a spatio-


125

temporal heading. Out-of-the-blue sentences can be uttered because a card which signifies the ‘here-and-now’ of the discourse situation is always located on top of the file providing an implicit stage topic: (7)

sTOPt [It is raining]FOC

sTOP1 is raining (at) e (7) is assessed by examining the implicit stage topic (the here-and-now) to see if it is raining there. In this case the whole sentence is taken as a focused event predicated of a stage. 1.3 Subordinate f-structure Individual level predicates are characterized by requiring individual topics.6 In view of the fact that indefinites do not represent existing cards, indefinites cannot function as topics. The following examples of sentences with individual level predicates are therefore ruled out (unless given a contrastive or generic interpretation): (8)

a. b.

#A #A

friend is intelligent. student likes linguistics.

Specific indefinites can, however, be topics: (9)

a. b. c.

A friend of mine is intelligent. A certain student likes linguistics. A student who I know likes linguistics.

Since topics must be either generic or specific (i.e., a card with the relevant heading must exist), it is not surprising that specific indefinites are allowed as topics. It is, however, rather mysterious what it is that makes these indefinites specific. In each case the indefinite is modified in some way. It is often recognized that specific indefinites are specific only to the speaker and that the hearer is not assumed to have an available referent in these cases. It follows from fstructure theory that complex constituents must be assigned an f-structure to get interpreted. F-structures assigned to subconstituents of the utterance are called  f-structures and do not involve predication (assessment) since only full sentences are assessed. The subordinate f-structure assigned, is however

126

NOMI ERTESCHIK-SHIR

interpreted by subordinate update, the process that associates subordinate topics with their foci. Subordinate update involves the application of the Topic, Focus and Update rules to a subordinate f-structure and differs from main f-structures only in that the rule of predication does not apply. Its role is to rearrange the cards according to the discourse rules and to make the relevant entries. The file manipulation for (9c), for example, is equivalent to that of the following sequence of sentences: (10)

ITOP [know a student]FOC. HeTOP [likes linguistics]FOC.

In the first sentence of this sequence, ‘a student’ is contained in the focus and hence a new card is introduced (by focus rule I) allowing for the introduction on top of the file of a new indexed card for ‘a student’: I

student2

e known student2

I know e

Once this card is opened, the indefinite is (speaker) referential and qualifies for topic status. The fact that the relative clause introduces a ‘subordinate’ application of the rules is also what explains the speaker-perspective of the reference. When a speaker introduces a new referent by means of a relative clause s/he instructs the hearer to open a new card, position it on the top of the file, and then go on to the sentence (without the relative clause). Only the speaker, not the hearer, is assumed to already have a card for student2 explaining speakerspecificity. Subordinate update must occur prior to the file manipulation triggered by the main f-structure, otherwise the card for the main topic would not be available, making the sentence uninterpretable. Subordinate f-structure and its associated update thus feeds the main application of the rules. Note that relative clauses are necessarily presupposed in this system because they represent existing entries on cards. (9a) works the same way: a friend of mine receives the following (subordinate) f-structure: (11)

Topi [a friend of minei]FOC

The subordinate topic is the available card for the speaker (first person). Again, a friend is in the subordinate focus domain which triggers the opening of a new card in the hearer’s file. This card is placed on top of the file ready for the relevant entry. After the rules have been applied to (9a), the hearer’s card looks


127

as follows (Ai = the speaker):7 friendj e is a friend of Ai e is intelligent Finally, adjectives such as certain and specific provide the same service, i.e., they trigger a subordinate f-structure and a new card is opened by subordinate update as before. These adjectives simply mean that the speaker ‘has a reference’ for the following NP. (9a) renders the following card: studentj Ai has reference for e e likes linguistics Subordinate f-structures do not involve assessment for truth value, but only a rearrangement of the cards. A subordinate f-structure is assigned whenever a card is located on top of the file for the subordinate topic which represents a phrase which is not analyzed as the  topic of the sentence. The subordinate focus functions to place cards on the top of the file. This is how the main topic, with respect to which the assessment of the sentence is performed, was derived in the case just examined. Subordinate f-structures are constrained primarily by the availability of subordinate topic cards on top of the file at the point at which the sentence is uttered. 1.4 Restrictive sets The focus constituent may define a  of cards, the students in my class, for example, forms such a set. Cards which contain a set of cards are called  cards, the set they define is a  set and the focus which introduces them is a  focus. A subordinate f-structure can be assigned to such a restrictive set in a partitive: (12)

[[Two]FOC-sub of [the students]TOP-sub]TOP [are intelligent]FOC

The subordinate f-structure is formed around a card on top of the file which represents a discoursally available set. A constituent which defines a subset of

128

NOMI ERTESCHIK-SHIR

this topic set is focused triggering the partitioning of this set. The new subset card is now available as the main topic. In sum, Topic and Focus are defined within a theory of discourse. Both individuals and stages provide topics. A rule of Predication is defined which assigns truth values. This rule is viewed as a relation between the topic of a sentence and its predicate. The truth value of a statement is determined by assessing it as putative information about its topic. The f-structure Topic and Focus rules apply to subordinate f-structures as well, but the rule of Predication applies only to the main f-structure. Since the topic is the pivot for the assignment of truth value, it follows that every sentence must have a topic.8

2.

R-dependencies

A distinction is made here between R-dependencies (relevant to the speaker’s file) and I-dependencies (relevant to the hearer’s file). R-dependencies (certain scopal relations, for example) are subject to the Topic Constraint, an f-structure theoretical constraint which determines a markedness hierarchy on f-structure. I-dependencies (e.g. anaphora, bound anaphora, negation and the scope of negation, wh-movement) are sensitive to the Subject Constraint, a syntactic constraint on f-structure. Here I briefly discuss R-dependencies. For a discussion of I-dependencies see Erteschik-Shir (1997). I use the term R-dependence when reference is defined by f-structure update, as in the sentences in (13): (13)

a. b. c. d.

John talked to a boy. He criticized her. Two girls arrested three boys. Everyone talked to someone.

In (13a) the indefinite object in the focus triggers the construction of a new card. The content of this card is determined solely by the entry made by update. All we know about a boy is that he is someone John talked to. He does not have an “identity” independently of the utterance at hand. This is what I call R-dependence. If an R-dependency involves introducing a new card into the hearer’s file, the card is assigned its reference merely in terms of the predication in which it occurs, it may be specific to the speaker, but not to the hearer. In this sense R-dependencies are   and the only requirements on such depen-


129

dencies are the ones applying to f-structure. R-dependencies thus encompass those dependencies that follow from the f-structure determined update. Coreferential pronouns, which are interpreted by being entered on the cards with the relevant headings, are R-dependent in this sense. R-dependencies are defined by f-structure update rules. Any NP in a Rdependency which receives a specific interpretation, must therefore represent an existing card in the file. Importantly, sentences with R-dependencies are constrained only by f-structure and the Topic Constraint, a pragmatic constraint which constrains f-structures and is part and parcel of f-structure theory. An fstructure in which the object is the topic renders a marked f-structure. Canonical or unmarked f-structures are those in which the topic precedes the focus: The Topic Constraint on R-dependencies (> = less marked): SUBJECTTOP […]FOC > TOPi[SUBJECT […NPi…]]FOC sTOPt […]FOC The Topic Constraint says that if both the subject and an other argument qualify as topics, i.e., represent cards on top of the file, the subject must be picked as the topic of the main f-structure. The constraint is also intended to reflect the intuition that the unmarked f-structure is one in which either the subject or a stage is the topic. (14) is therefore the unmarked f-structure for the sentence (13b), for example: (14)

TOPi-sub [HeTOP [criticized heri]FOC]FOC-sub

Both pronouns are referentially dependent on cards discoursally placed on top of the file and the main f-structure takes the subject as its topic. The object pronoun is also interpreted through entry on a card with the relevant heading (due to the subordinate f-structure). If the hearer does not have a card on top of the file for both pronouns before the utterance, the sentence will not be interpretable.9 The f-structure in (14) takes the subject pronoun as the main topic and the object pronoun as the subordinate topic. The notation used here may be confusing in that it looks like the main f-structure is nested within the subordinate one. This is merely a result of the notation and has no significance with respect to the application of the f-structure rules. If both subject and object qualify as potential topics, i.e., they are definite and therefore represent existing cards, the subject will be chosen unless the discourse pushes heavily for the other f-structure. Although f-structures are available in which the object is interpreted as a topic, the unmarked interpretation

130

NOMI ERTESCHIK-SHIR

is still the one in which the subject is taken to be the topic. This can be seen in the following interchange which fixes John as the topic: (15)

Tell me about John: a. He is in love with Mary. b. ??Mary is in love with him.

F-structures with stage topics are also unmarked. (16a), in which the object is interpreted as the topic, is less natural than (16b) with the stage topic: (16)

a. b.

TOPi [A girl talked to Johni]FOC (Answers: Tell me about John) sTOPt [A girl talked to John]FOC (Answers: What happened next?)

The Topic Constraint, which applies to R-dependencies, thus provides an explanation for the well known fact that certain scopal interpretations are much harder to get than others. (13c), for example, illustrates an example with two weak quantifiers. Weak quantifiers, as will be shown in the next section, can be interpreted as topics only if they are interpreted partitively. Again the unmarked case will be the one in which the subject is taken as the topic, but if a set of boys is available in the context and a set of girls is not, the object and not the subject will function as the main topic. Since the topic has wide scope, context determines scopal relations in sentences. Similarly for (13d).10 The topic constraint is based on the intuition that processing is facilitated when the topic (‘what we are talking about’) precedes the predicate (‘what we say about the topic’). It can be verified only empirically. The topic constraint is here viewed as a pragmatic constraint on f-structures.

3.

The Scopal Properties of Topics

It is a property of topics that they take wide scope. A topic provides a link to the preceding discourse in which it is introduced by a focus constituent which may or may not be restrictive. In the former case, its quantificational nature follows from the application of the rule of predication which requires assessment for every single member of this restricted set. A simple definite topic such as the man can be viewed as quantificational in that predication ranges over the ‘single’ member of the topic-set. The link of the topic to the discourse also makes the


131

contextual restriction of quantifiers fall out automatically. So, for example, the topic everyone must be interpreted as ‘everyone we are talking about’, it is a context specified set represented by a card with the heading everyonei.11 3.1 Weak Determiners Let us now examine weakly quantified topics: (17)

a. b.

TWO students are inTELLigent. Two STUdents are inTELLigent.

(18)

a. b.

SOME students are inTELLigent. Some STUdents are inTELLigent.

A possible context for (17a) is one in which a set of students is supplied by the context: (19)

[I have a class of six students.] TWO (students) are intelligent, THREE (students) are mediocre, and ONE (student) is a total idiot.

In (19) the set of students focally introduced in the bracketed sentence is fully partitioned. The subject of (17a) is therefore interpreted partitively if the necessary restrictive set has been introduced in the preceding discourse. The same analysis applies to (18a) The intonation of (17a) and (18a) supports this analysis. In both cases a subordinate f-structure is indicated in which the unstressed noun ‘students’ is the topic and the stressed quantifier is the focus: (20)

[TWOFOC-sub studentsTOP-sub]TOP …

(17b) and (18b) are acceptable under a contrastive reading or else under a specific reading. The focused (stressed) nouns enable the necessary subordinate f-structure. Weak determiners allow for either specific or partitive readings by means of subordinate f-structures. These readings have been referred to as “presuppositional” readings in Diesing (1992), and as “strong” readings (of weak determiners) by de Hoop (1992). I will use the latter term. It is also possible to verify (as was done for the partitives) that assessment of these topics must be distributive. Distributive readings range over the sets defined by these strong NPs. A “weak” topic is a contradiction in terms involving quantifying over undefined sets.

132

NOMI ERTESCHIK-SHIR

3.2 Strong Determiners What it means for an NP to be strong or definite is that a card representing a discourse restricted set (of students, here) is assumed to be available in the hearer’s file. The set interpretation is necessary for the distributive assessment of the sentence. Examine the following examples: (21)

a. b. c.

Every student is intelligent. Every student in my class is intelligent. Everyone is intelligent.

To determine the truth value of (21a) each of the individual students in the set must be examined. This set is represented by a restrictive set of cards on top of the file which receives a unified heading. The individual cards which constitute this set must be examined in order for a truth value to be assigned. The restrictive set can be introduced by a subordinate f-structure as in (21b); I assume no partitioning of this set since every ranges over all the members of the set. How the set is to be partitioned is a property of the particular quantifier in question. The evaluation process for most is more complex than the one suggested for universal quantifiers. In the following (22)

Most students in my class are intelligent.

a topic set encompassing the students in my class is available and this set is partitioned by the quantifier. As before the derived set provides the main topic for assessment. How do we effect such a partitioning with most? Clearly there is no single set that most students could represent. As mentioned by an anonymous reviewer, when our model contains students a, b, c, d, then the sets {a,b,c}, {a,c,d}, {b,c,d} and {a,b,c,d} provide possible partitions for which the evaluation of the sentence comes out true. If most is interpreted roughly as ‘more than half’, then the result is achieved if any partitioning of the set of students which results in more than half of them is licensed. The result of partitioning should therefore not be viewed as defining a necessarily unique new subset. For most, many, etc. it suffices that a subset of the appropriate size is definable. Note the interpretation of the pronoun in the following: (23)

Most of the students came to the party. They had a good time.

The pronoun refers to whichever set of students (counting as most students) actually came to the party.


133

Whether the quantifier is strong or weak, if it quantifies a topic, it must receive a strong reading, i.e., there must be a card, or set of cards available on top of the hearer’s file.12 I have shown that topic sets generate a quantificational reading when they consist of more than one member. This reading is obtained because the rule of predication must distribute over the individual members of this set. In (24)

The boy is intelligent.

The set {boyi} (which happens to have only one member) is assessed in the same manner. The quantificational reading is therefore a direct outcome of the fact that topics may have set readings and that sets may have more than one member.

4.

Quantifier Scope

I will now argue that f-structures are scopally transparent and can be interpreted directly. LF processes such as Quantifier Raising (QR) are rendered superfluous. I will also argue that suggestions such as May’s (1977) that focus constituents are raised by QR in LF are not only superfluous but give bad results. For the moment, I limit the discussion to scope interactions in simple transitive sentences. At least the following three f-structures are available for sentences with quantifiers in subject and object position (Q1 = subject, Q2 = object): (25)

a. b. c.

[Q1]TOP [V Q2]FOC TOP2 [Q1 V Q2]FOC sTOPt [Q1 V Q2]FOC

Two important results follow from the interpretation of f-structures argued for in the previous sections: 1. Topic quantifiers take wide scope over any other quantifier 2. There are non-scoped f-structures I will start by discussing Topic-scope: A topic has been defined as a card on top of the file. The existence of such a card presupposes the existence in the discourse of the referent of the card ‘heading’. Predication takes the focus constituent and assesses its truth value with respect to the topic. It follows that the topic necessarily has wider scope than any constituent contained within the focus. Thus, in (25a) Q1 is the topic, Q2 is contained within the focus. Predication applies as follows: For each individual contained in Q1, the truth of the focus

134

NOMI ERTESCHIK-SHIR

constituent containing Q2 is assessed, i.e., Q1 has scope over Q2. If we follow the same reasoning for (25b), we get the opposite scope relation. Two scoped readings thus follow without further ado from applying predication to these f-structures. In (25c) the topic is a stage topic, hence neither of the quantified NPs is a topic and a nonscoped reading results. In the next section I show, following Landman (in prog.), that there are actually 4 nonscoped readings, all of which have the same focus structure. Under this view, for each case in which a quantified NP is analyzed as the topic, there are two readings to be derived. 4.1 Cardinal Scope According to Landman a sentence such as (26) has the eight readings listed in (27) and (28). (27) represents four unscoped readings in which each of the cardinals enables a collective and a distributive reading. (28) represents the four scoped readings which Landman derives by a special scope rule. [C = collective, D=distributive, subscript s=subject, subscript o=object, scope=parenthesis]: (26)

Two girls arrested three boys.

(27)

a. b.

c. d.

(28)

a. b.

c.

d.

C C {+(a,b)} → {+(1,2,3)} C D {+(a,b)} → 1 →2 →3 DC a → {+(1,2,3)} b → DD a →1 b →2 →3 a → [1,2,3] b → [4,5,6] Ds (Do) a → 1 b → 4 →2 →5 →3 →6 →1 Do (Cs) [a,b] [c,d] →2 [e,f] →3 Do (Ds) a → 1 c → 2 e → 3 b→ d→ f →

Ds (Co)


135

The f-structure of all the cases listed in (27) is (25c) which has a stage topic. For each plural NP, there are two possible interpretations: collective and distributive. (27a) gets collective readings for both NPs, i.e., the group, ‘2 girls’ arrested the group ‘3 boys’. (27b) says that the group ‘2 girls’ arrested 3 individual boys. (27c) involves 2 individual girls who arrest a group of 3 boys. (27d) is the double distributive reading in which 2 individual girls arrest 3 individual boys, i.e., there is some pairing between the 2 girls and the 3 boys such that for each girl there is one or more boys (of the three) that she arrests, and no boy gets left unarrested. (Since the reading leaves the number of pairings unspecified with a minimum of three and a maximum of six, the number of ‘arrows’ which indicates the number of ‘pairings’ has been left open. This reading includes the ‘all-all’ reading in which each of the two girls arrests each of the three boys. This reading, according to Landman, is the borderline case of (27d). All of these are assessed with respect to the discourse here-and-now, i.e., only  event is involved for each reading (and only two girls and three boys are involved in each case). A distinction is thus made between the number of pairings, in this case the number of arrests (which ranges between 1 and 3), and the number of events. Imagine, for example, the following spatio-temporal parameters: Today, between 6–7 p.m., in Beer Sheva. These parameters define the stage upon which all the arrests take place. No individual stages for each separate pairing is made available. This is a requirement of the stage topic reading evidenced by the presence of an overt stage topic: (29)

a. b. c.

Today, two girls arrested three boys. At 6 o’clock, two girls arrested three boys. On the corner, two girls arrested three boys.

The two scoped readings (28a) and (28b) are derived from the f-structure in (25a) by allowing the object to be either distributive or collective. Similarly, for (28c) and (28d) which are derived from the f-structure (25b) These readings in which the object NP is interpreted as having wide scope are highly marked. According to the Topic Constraint, if both the subject and the object are candidates for topichood, i.e., both represent available cards on top of the file, then the subject will take precedence as the main topic. In the cases discussed here, both the subject and object are cardinal. Pending contextualization both are therefore equally likely candidates for topichood. Only a context that enhances the objecttopic reading will make such a reading possible. This is the reason it takes brute force (i.e., ample contextualization) to convince speakers of the readings in

136

NOMI ERTESCHIK-SHIR

which the object has scope over the subject. The various readings in (27) and (28) are intonationally distinguishable. The stress rule associated with the theory of f-structure presented here is simply that a focus is stressed.13 The following intonation patterns are predicted: (30)

a. b. c. d.

Two GIRLS arrested three BOYS. (sTOPt C C) Two GIRLS arrested THREE boys. (sTOPt C D),(TOPo C) TWO girls arrested three BOYS. (sTOPt D C),(TOPs C) TWO girls arrested THREE boys. (sTOPt D D),(TOPs D),(TOPo D)

I predict that the most popular intonation pattern for sentences of this sort will be (30d) since three different readings are associated with it. It is therefore not surprising that speakers find it hard to supply the variety of readings available for each intonation pattern. (30a) is also not impossible to get, but the other two are not only tongue twisters, but also brain twisters. The three f-structures in (25), together with Landman’s analysis of plurals as either distributive or collective, renders the eight readings listed above. Landman’s scope rule becomes superfluous if predication is read off f-structures, since this is what forces a distributive reading of topics. The unscoped readings are also a direct outcome of the view that f-structures allow for stage topics. Since stage topics are ruled out for individual level predicates, it is predicted that the four unscoped readings are not available for them. Only the readings (28) and none of the readings in (27), should be available for (31): (31)

Two students know three languages.

Note, however, that the f-structure (25a), with an additional subordinate fstructure in the VP, in which the object quantifier is the topic, gives a symmetric interpretation mimicking (27d) The intonation involves stressing the verb to the exclusion of the NPs: (32)

TOPi [Two Chinese girls]TOP [ADORE [three American boys]i]FOC

Such a reading is not excluded in the present system. Certain (individual level) predicates block a reading in which the object is interpreted as the topic.14 For such sentences only the readings predicted by the f-structure (22a) (in which the subject is the topic) are possible. This is the case in the following examples: (33)

a. Two boxes contain three books. b. Two lakes are close to three houses.


137

c. Two women have three husbands. d. Two dogs require three caretakers. The only readings we get for the sentences in (33) are (34)

a. Ds (Co)

a → [1,2,3] b → [4,5,6]

b. Ds (Do)

a→1 →2 →3

b→4 →5 →6

Neither of the other two scoped readings, nor the four unscoped readings are possible for these predicates. I take this fact as strong evidence in favor of my position that f-structures are disambiguated with respect to scope. F-structure thus replaces LF in deriving structures which directly feed the semantic component. The LF rule of quantifier raising (cf. May 1977, 1985) will not be necessary. May (1989) argues that LF does not render structures disambiguated with respect to scope. The readings are rather semantically derived. F-structure, under my view, is disambiguated and feeds the semantic rule of predication to render the correct readings. These disambiguated f-structures interact with the semantics of the various quantifiers as will be argued presently. In the next sections I argue that all cases of quantifier scope boil down to f-structure predication. Under this view, there is no need for semantic scope rules since scope is read off f-structure. The theory of f-structure and predication also leaves obsolete any type of syntactic quantifier raising rules. If quantifier raising (QR) is rendered obsolete, LF loses at least some of its raison d’être.15 4.2 Some-Every Scopes Let us test whether these predictions can be verified with other quantifiers: (35)

a. b.

Someone arrested everyone. Everyone arrested someone.

In view of the fact that someone is weak, it cannot be a main topic (unless it is contrastive). The two readings in which someone would be a topic are thus excluded from the sentences in (35). Further, someone is singular, therefore the distributive/collective distinction is irrelevant. That cuts another 3 readings and we

138

NOMI ERTESCHIK-SHIR

are left with 3 readings for each sentence represented by the following f-structures: (36)

a. b. c.

sTOPt [everyoneD arrested someone]FOC sTOPt [everyoneC arrested someone]FOC EveryoneTOP [arrested someone]FOC

(37)

a. b. c.

sTOPt [someone arrested everyoneD]FOC sTOPt [someone arrested everyoneC]FOC TOPi [someone arrested everyonei]FOC

(36a) involves one event in which one person gets arrested by each of the members of the set defined by everyone. (36b) differs only in everyone being viewed collectively, i.e., the arrest is performed as a group action.16 In (36c) everyone is the topic of the sentence. Everyone must therefore be a discourse specified set represented by an indexed card on top of the file. Since predication involves assessment for each individual member of this set, a distributive reading is achieved. This reading takes an ‘undefined’ someone for each of the members of the set and gives us the interpretation: (38)

∀x, ∀y (x arrested y)

The Topic Constraint picks (36c) as the least marked f-structure for the sentence. It follows that (38) is the most natural reading. The readings derived from the fstructures with stage topics are very highly marked. This is because strongly quantified NPs necessarily presuppose a contextually defined set, i.e., they represent an existing card. In principle, this card, as any other definite, need not be positioned on top of the file forcing a topic reading. However, for the pronominal form everyone (as opposed to every teacher, say), only few contexts would allow unambiguous reference to such a card if it were not already to be found on top of the file. It follows that the most likely use of everyone is as a topic or at least as a subordinate topic. If a context with a stage topic is contextually forced, the following f-structure with a subordinate f-structure is most plausible: (39)

sTOPt [everyoneTOP-sub [arrested someone]FOC-sub]FOC

The subordinate f-structure again derives a ‘scoped’ reading in which everyone has wide scope. The readings resulting from (36a & b), in which a ‘single’ someone is arrested, are therefore almost impossible to get, unless of course an  triggered subordinate f-structure on someone is provided as in: (40)

Everyone arrested someone, namely his best friend.


139

The story is the same for the f-structures in (37a–b) except that the ‘arresting’ is reversed. Here again, properties of the strong quantifier force a subordinate fstructure as follows: (41)

sTOPt [TOPi-sub [someone arrested everyonei]FOC-sub]FOC

The f-structure (37c) is almost impossible to contextualize at least with the predicate ‘arrest’. Destressing everyone together with an appropriate context generally works. Note that, according to the Topic Constraint, since everyone is a strong quantifier and someone is not (unless it is used partitively), the odds are already skewed in favor of everyone being the topic.17 Each only allows a distributive reading in which the individual members of a set are scanned, i.e., no collective reading is available independently of whether Each quantifies a topic or not. This is the reason Each always seems to take wide scope. Vendler (1967) argues this point by noting the following differences between Each and every: (42)

a. b.

each of them vs. *every of them one of them vs. every one of them

??each

(42b), according to Vendler, is redundant for Each since it already implies one. A, on the other hand, can get a collective reading, even when it functions as a topic:18 (43)

a. b.

All the items in the store cost $500. Every item in the store costs $500.

(43a) allows a collective reading in which the total value of the items in the store is $500. This reading is excluded in (43b). Note that the italicized NP necessitates a referential card. All may be used similarly to any to emphasize that no item is excluded, i.e., the NP is viewed as a plural individual and all is used as an emphatic marker, rather than as an operator in this case.19 Partee (1994) addresses the question of whether focus affects pragmatic or semantic aspects of interpretation. In particular she examines the Prague school view that “theme-rheme structure is always considered an aspect of linguistic meaning, whether it has any truth-conditional effects or not.” In particular, Partee examines the relation between the restrictor and the notion “theme” (=topic) and the relation between the Nuclear Scope and the “rheme” (=focus). Since the f-structure theoretical approach presented here concurs with this view, Partee’s discussion is particularly pertinent.

140

NOMI ERTESCHIK-SHIR

Partee offers two examples of D-quantifiers, one in which the tripartite structure is syntactically determined and one which, according to Partee, is pragmatically or contextually determined: (44)

A: B:

Most logicians like linguistics. Most NICE logicians like linguistics.

(45)

Most ships pass through the lock at NIGHT. a. MOST (ships)(pass through the lock at night) b. MOST (ships that pass through the lock)(pass through the lock at night)

In Section 1.3. Individual level predicates were seen to be restricted to non-stage topics. The topic of (44) is therefore most likely to be the syntactic subject — in this case the generics ‘logicians’ and ‘nice logicians’, which provide the domain of quantification. In (45), however, a stage topic is available rendering the b. reading as follows: The stage topic is defined by the nonfocused spatio-temporal parameters of the sentence. In this case ‘passing through the lock’ provides the spatial parameter and the present tense gives a generic reading. The stage topic thus provides the correct restrictor here as well.20 The following example further strengthens my claim that quantifier scope is predictable from f-structure: (45)

Many Scandinavians have won the Noble prize in LITERATURE. One. [Many Scandinavians]TOP [have won the Noble prize in LITERATURE]FOC Two. Many Scandinavians have won [the Noble prize]TOP in [LITERATURE]FOC

The f-structure in a. would indeed predict an unlikely interpretation for the sentence, namely the one that says that out of all Scandinavians, many are Nobel prize winners. However, this f-structure is an unlikely one for the sentence, it assumes that Scandinavians are under discussion. A more likely context is the one offered in (b) in which Noble prizes are under discussion, and  is contrasted, rendering an f-structure which predicts a more likely interpretation.21 My purpose here is not to list the properties of each individual quantifier, but rather to argue that the full range of interpretations of sentences with (two) quantifiers is predicted by the three f-structures in (25). For strong quantifiers in subject position, the topic constraint predicts that unscoped readings will be very highly marked. For strong quantifiers in object position, a subordinate topic


141

reading is the most natural, making unscoped readings for strong quantifiers very highly marked in general. If the quantifiers involved each allow both distributive and collective readings, a total of eight readings are available. The specific properties of individual quantifiers interact with the three f-structures to either limit (as in the case of each) or extend () the actual number of readings. 4.3 Scoped Stages In the preceding sections I showed that topics provide a restrictive set over which quantifiers range. Stage topics are no different in this respect. They too function as the restriction on quantifiers. Relevant quantifiers are sometimes, everywhere, always, etc. These quantifiers cannot be used as stage topics with individual level predicates, neither can they be used with stage level predicates unless they are assigned a f-structure with a stage topic. Examine (46) and its fstructure (47): (46)

Sometimes a boy meets a girl.

(47)

sometimessTOPt [a boy meets a girl]FOC

Here, a set of times {ti, …tn} are derived exactly as some students was derived in Section 2.2. The focussed sentence is then assessed with respect to each of these times. (The stage topic also includes a location with respect to which the sentence is assessed. Again this aspect of the stage topic is supplied by the context.) What I have argued so far is that any individual NP topic will have wide scope with respect to any quantifier in the focus constituent. This follows from the predication rule, in which the focus is predicated of the topic. I have also argued for a class of unscoped readings. These are the cases in which the whole sentence is predicated of a stage topic. It follows that if the stage topic is overt, any quantifier phrase included in it will also take wide scope: (48)

a. b. c. d.

In In In In

every city, John loves someone. every city, someone loves you. some city, John loves everyone. some city, everyone loves you.

None of these are ambiguous as predicted. As noted by one an anonymous reviewer, the following example  ambiguous:

142

NOMI ERTESCHIK-SHIR

(49)

In every city that he visited John met someone he went to high school with.

The ambiguity here results from two subordinate f-structures that can be assigned to the main object: (50)

a. b.

someoneFOC [he went to high school with]TOP someoneTOP [he went to high school with]FOC

The f-structure in (50a) renders the reading in which the stage topic has wide scope as expected and for each city in the topic-set a different ‘someone’ is derived. The f-structure in (50b), in which someone is a subordinate topic, is also possible. In this case reference is made to an existing card, i.e., a particular ‘someone’ and no scopal interaction transpires, rendering the seeming wide scope reading of the object. It is therefore important, in making scopal predictions according to f-structure, to examine not only the main f-structures but also the potential subordinate ones. 4.4 Scope and f-structure, other views The “tendency” for topics to take wide scope has been noted before (see, for example, Ioup (1975), Kuno (1982), Reinhart (1983), and others). The opposite view has, however, also been prevalent, namely, that focused constituents are the ones to take wide scope. May (1977) proposes that the rule of quantifier raising apply to focused constituents in order to raise them and give them wide scope. This view is also to be found in Williams (1988, 143) who claims that heavily stressed objects get wide scope as in Williams’ (26): (51)

Someone loves EVERYONE.

It seems to me that this is factually wrong. It is indeed the case that contrastively focused constituents get a wide scope interpretation and that this follows from the fact that these constituents form a subordinate f-structure in which the contrast set provides the topic, and the overt stressed constituent, the (contrastive) focus. Noncontrastive focused constituents necessarily take narrow scope, however. It has also often been noted that it is easier to get the reading in which the subject has scope over the object, the unmarked case under the current view. Reinhart (1979) offers a syntactic view of this situation. She argues that the ccommanding quantifier in surface structure necessarily has wide scope. Excep-


143

tions to this generalization involve f-structures with stage topics or cases in which the subject is not a candidate for topichood: Chierchia and McConnelGinet (1990, 117) mention the following exceptions to the c-command constraint: (52)

a. b. c.

There was a name tag near every plate. A flag was hanging in front of every window. A student guide took every visitor to two museums.

The subjects are not possible candidates for topichood. I have underlined the unmarked topics in (52a) and b. in which the topics are stages. In (52c) either the object or the PP can be interpreted as the topic and take scope over the subject as well as the remaining quantifier constituent. What looks like a structural constraint is in effect a consequence of the structural relationship between subjects and objects and the fact that subjects are unmarked topics, i.e., the Topic Constraint. A recent analysis of scope in German and its interaction with focus is offered in Krifka (1994). Krifka discusses scope inversion under a rise-fall pattern as in the following illustrations (his (4a, 12a (rise is marked by “/” and fall by “\”)): (53)

a.

(∀∃ only)

b.

(∃∀, ∀∃)

[CPfast jeder Jungei [CHAT nearly every. boy has [ti[mindestens einen Roman [gelesen]]]]] at.least one. novel read Fast/JEDer Junge hat mindestens \EINen Roman gelesen.

Whereas the subject necessarily has wide scope in (53a), (53b), with the rise-fall intonation, allows object wide scope as well. Krifka’s explanation involves details of German syntax and focus assignment that would take us too far afield.22 An explanation within the current framework is however also available. In German, topics are fronted and word order therefore largely determines fstructure. (53a) must therefore be interpreted with the subject as topic fixing the scope as indicated. According to Krifka, the rise-fall intonation in (53b) indicates multiple foci, as in the context of a multiple wh-question. If this is the case, we can conclude that the two focal stresses in the rise-fall cases signal subordinate f-structures since this is the only way, in the current framework, that more than one focus can be generated. Krifka also suggests that the stressed quantifiers determine the alternatives to the expression in focus (p. 144). Although such

144

NOMI ERTESCHIK-SHIR

topic sets can be made contextually available for both the NPs, the word order should still allow only the subject to be interpreted as the main topic. The fact that a ∃∀ reading is also available can be explained, however, if we assume a subordinate contrastive f-structure on the object. The following context could work. Assume that an ordered list of novels were assigned as summer reading to a set of boys and that they were supposed to read as many novels as possible starting from the top. In this context a contrast is formed between (at least) “one” novel and “no” novels, and the interpretation is derived in which one particular novel, namely the one figuring on top of the list was read by all the boys. This type of reading can be made contextually available without interfering with the main f-structure which is syntactically determined in German. The current framework may also afford an explanation for the fact that scope inversion is blocked or very hard to get with stage topics (Krifka’s (49a)): (54)

(∃∀, ??∀∃)

Heute hat mindestens /EIN Junge fast \JEDen Today has at least /one boy almost \every Roman gelesen. novel read

This sentence must be interpreted with a stage topic which leaves only an unscoped reading, i.e., only (at least) one boy is involved. I assume that the presence of a stage topic makes a context which triggers scope reversal extremely hard to conjure up. Unfortunately, Krifka does not indicate the necessary context for scope reversal. Krifka’s explanation necessitates focus marking in d-structure. According to Krifka discontinuous foci occur in s-structure in German requiring focus marking in d-structure independently. Although it seems that scope inversion receives a natural explanation in f-structure theory, the other case presented by Krifka to argue for his position must be examined here as well (p. 145 (43)): (55)

Peter gab das Verbrechen sofort \ZU. Peter admitted the crime immediately \AFFIX

Krifka proposes the following derivation: (56) D-structure: Focus Assignment: C0 movement: Spec-CP movement:

[CPe[C′e[IPPeter[das Verbrechen[sofort[zu[gab]]]]]]] [CPe[C′e[IPPeter[das Verbrechen[sofort[zu[gab]]F]]]]] [CPe[C′gabi[IPPeter[das Verbrechen[sofort[zu[ti]]F]]]]] [CPPeterj[C′gabi[IPtj[das Verbrechen[sofort[zu[ti]]F]]]]]


145

The verb has moved out of the surface focus constituent leaving a trace. If focus is marked on s-structure, the rules that interpret f-structure will involve reconstruction to trace position. Stress assignment will also correctly derive stress on the affix left in the focus constituent. I therefore see no reason why the focus should not be marked on s-structure in these cases as well.

5.

Conclusion

F-structure is a basic component of the language faculty which interacts essentially with Phonology, Syntax and Semantics as well as Pragmatics. F-structure marks the organization of sentences into focus and topic constituents. In Erteschik-Shir and Lappin (1987) we argue that all modes of perception are organized into foreground and background constituents. Focusing is viewed as a single task-specific mechanism which identifies the foregrounded constituent in representations of all modular systems. Focusing is therefore a nonmodular process which provides the interface between the modular system and the central cognitive mechanisms. It follows that although focusing plays a central role in grammar, it is not unique to the language faculty. The identification of a topic is, however, characteristic of linguistic structure. In vision, for example, foregrounding is necessary for the formation of a visual report, but this report is not “about” anything, i.e., it does not have a topic. Whereas the capacity to identify a topic is a distinctive property of the human brain and of the language faculty in particular, focusing is a basic property of all cognitive systems. The linguistic level of F-structure in which both topic and focus are identified is therefore a fundamental part of Universal Grammar (UG) which determines the class of possible languages. The rule of predication takes fstructures as its input producing a well-formed discourse representation. Fstructure can therefore be viewed as an interface level between the grammar and the conceptual — intentional system replacing LF. F-structure also mediates grammar and PF (Phonetic Form), itself an interface level with the articulatory — perceptual system. This is necessary for the derivation of intonation. Replacing LF with F-structure, as I suggest, is a natural consequence of its function as an interface to conceptual-intentional structure since focusing is an innate reflex, essentially involved in perception as a whole, and topic identification is an innate and essential part of conceptual-intentional structure itself.

146

NOMI ERTESCHIK-SHIR

Notes * This research was supported by grant No. 90–00267 from the United States-Israel Binational Science Foundation (BSF), Jerusalem, Israel. I wish to thank the audiences at the Workshop on Focus, Paris (1996), the Conference in honor of R. Jakobson, Prague (1996), and in particular, Josef Bayer for helpful comments. 1. New cards are also constructed for constituents other than NP, allowing for pronominal reference. Here I limit discussion of the application of f-structure rules to focused NPs and NPs contained in focused constituents. If the focused constituent does not contain any NPs, the Focus rule does not apply and no cards are positioned on top of the file. The appropriate entry on the topic card, is, however, made by the Update rule. 2. For a discussion of how this framework relates to Kamp/Heimian Discourse Representation Theory, see Erteschik-Shir (1997: Chapter 1). 3. Note that new cards are not introduced in questions. 4. Topics are also presupposed, but are distinguished in the file-system as existing cards (positioned on top of the file). 5. The topicalized version would have the following form: [the dogi]TOP [John has ti]FOC 6. See Kratzer (1989) for arguments for the view that individual level predicates do not take a spatio-temporal argument. Within f-structure theory this is tantamount to claiming that individual level predicates cannot take a stage topic. 7. Permanently available cards on top-of-the-file are the cards for the speaker, the hearer as well as the current stage. 8. Vallduví (1992) presents a model of update semantics, Information Packaging, which is in many ways similar to the one defined here in that it takes as its aim to account for syntactic phenomena and intonation in terms of discourse structure. There are two major differences between Vallduví’s framework and mine. The first one is that he claims no connection between the assignment of truth values and update semantics. I show in the following that this is an integral aspect of f-structure theory with consequences for, among other issues, the account of quantifier scope. The second difference is that Vallduví does not distinguish the top-of-the-file as I do. One of the consequences of distinguishing those cards that are available as future topics is that it predicts possible sequences of sentences in discourse. Moreover, in view of the fact that topics are not represented in Vallduví’s framework unless they are overt new topics (Links), stage topics, which are often covert, cannot in any obvious way be incorporated in his system. 9. I thus claim that pronouns must be either main or subordinate topics. For a discussion of contrastive pronouns in this context see Erteschik-Shir (1997). 10. The view that an unmarked f-structure is one in which the syntactic structure and the f-structure are isomorphic, i.e., subject is the topic and the VP is the focus, has been recognized for other languages as well. Schwartz (1976) argues that what he refers to as the Ilocano focus, which is actually the topic as defined here, is necessarily the subject. Anderson (1991) argues for Dinka, a major Western Nilotic language, that the preverbal noun phrase is a topic. Li and Thompson (1976: 484) argue that Subjects are essentially grammaticalized topics. This “is why many of the


147

topic properties are shared by subjects in a number of languages.” Reinhart (1981: 87) also claims that the syntactic subject is the preferred topic for the following reason: “…it is easier to interpret the sentence as being about its subject, than, say, about its object, since in the logical form, something is predicated directly of the subject’s interpretation.” More recently Lambrecht (1994) argues that the unmarked information-structure sequence is topic followed by focus and that the subject is the unmarked topic. I make no assumptions here with respect to fstructure markedness in a topic-prominent language. 11. See also Bühring (1999) for an analysis of scope in terms of topics. 12. Note that the topic is the contextual set (of students, in this case), not the quantified phrase as a whole. It follows that topic tests identify this topic only (as pointed out to me by Josef Bayer, p.c.): (i) *Every student in my class, he is intelligent. (ii) As far as the students in my class are concerned, every one of them is intelligent. 13. In Erteschik-Shir (1997: Chapter IV) I show that stress on a complex constituent is actualized by stressing each of the major constituents contained therein. In fast speech middle stresses are reduced. Subordinate foci are stressed by the same stress rule. 14. This is evidenced by the impossibility of (i)

*A box contains the book.

(ii) *A lake is close to the house. 15. LF movement has been motivated by other linguistic phenomena such as crossover and superiority. In Erteschik-Shir (1997: Chapter VI) I offer a f-structure theoretical account for these phenomena as well. Note that I-dependencies, which hold in cases of wh-quantifier interactions, multiple wh-questions, bound anaphora, negation and its scope, only and its scope are restricted by a syntactic constraint on f-structure which limits I-dependencies to canonical f-structures. The cases of quantifier scope discussed here exemplify R-dependencies for which noncanonical f-structures are discoursally marked but not ruled out. 16. I argue below that the most likely reading of everyone is as a topic (main or subordinate). A collective reading is therefore highly marked and feasibly completely blocked. When the quantifier ‘every’ is combined with an NP as in every teacher, only a distributive reading may be construed. 17. Stressing someone (rendering the partitive reading in which ‘someone’ is selected out of a context defined set) allows for a reading in which someone is the topic. 18. For more details on the differences between each, every, any and all see Vendler (1967). 19. Szabolcsi (1995) argues that the syntactic position of a quantifier in Hungarian defines its scope. In her system a TOPIC position c-commands all other positions and it therefore follows that topics necessarily have wide scope. Other syntactic positions are Quantifier, Focus/ Predicate Operator and Negation, which precede the verb in that order. It is possible that a careful analysis of the properties of the elements that can go into these slots together with the f-structure theoretical approach can be made to work for Hungarian as well, but it is not immediately obvious how the structural properties can be made to follow. This must await an analysis of scrambling in terms of f-structure. 20. For an analysis in terms of event-induced measures of quantification see Krifka (1990a) who is

148

NOMI ERTESCHIK-SHIR the source of the example. For further discussion of this approach see Eckardt (1994).

21. The indicated f-structure could be the main or the subordinate one. 22. Krifka notes that (53a) from Frey (1993) uses accent on the finite verb to exclude the effect of focus. This renders the so-called verum focus, which focuses on the truth polarity of the sentence. In the current framework, such a (metalinguistic) accent masks the intonational marking of the nonmetalinguistic f-structure but is not predicted to interfere with the scopal interpretation (cf. Erteschik-Shir 1997).

References Anderson, T. 1991. “Subject and Topic in Dinka”. Studies in Language 15(2). 265–294. Büring, D. 1999. “Topic”. Submitted to P. Bosch & R. van der Sandt (eds), Focus. Cambridge: CUP. Bayer, J. 1996. Directionality and Logical Form: On the Scope of Focusing Particles and wh-in-situ. Dordrecht: Kluwer Academic Publishers. Bosch, P. and R. van der Sandt (eds.). 1994. Working Papers of the Institute for Logic & Linguistics 7. Vol. 1: Intonation and Syntax. Heidelberg: IBM Deutschland Informationssysteme, GMBH Scientific Center. Chierchia, G. 1992. “Anaphora and Dynamic Binding”. Linguistics and Philosophy 15.111–183. Chierchia, G. and S. McConnel-Ginet 1990. Meaning and Grammar: An Introduction to Semantics. Cambridge, Mass.: MIT Press. Diesing, M. 1992. Indefinites. Cambridge, Mass.: MIT Press. Eckardt, R. 1994. “Focus with Nominal Quantifiers”. In Bosch, P. and R. van der Sandt (eds.), Working Papers of the Institute for Logic & Linguistics. Heidelberg: IBM Deutschland Informationssysteme GMBH Scientific Centre. Erteschik-Shir, N. 1997. The Dynamics of Focus Structure. Cambridge: CUP. Erteschik-Shir, N. and S. Lappin 1987. “Dominance and modularity”. Linguistics 25.671–685. Frey, W. 1993. Syntaktische Bedingungen für die semantische Repräsentation. Über Bindung, implizite Argumente und Skopus. (Studia Grammatica XXXV). Berlin: Akademie Verlag. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation, University of Mass., Amherst.


149

de Hoop, H. 1992. Case Configuration and Noun Phrase Interpretation. Ph.D. Diss, Univ. of Groningen. Ioup, G. 1975. “Some Universals for Quantifier Scope”. In J.P. Kimball (ed), Syntax and Semantics. Vol.4. New York: Academic Press. Kratzer, A. 1989. “Stage-Level and Individual-Level Predicates”. In Papers on Quantification. Department of Linguistics, University of Mass., Amherst. [appeared in: Carlson, G. and Pelletier, F.J. (eds.), 1995. The Generic Book. Chicago: University of Chicago Press, 125–175.] Krifka, M. 1990. “4000 Ships Passed through the Lock: Object induced Measure Functions on Events”. Linguistics and Philosophy 13. Krifka, M. 1994. “Focus and Operator Scope in German”. in Bosch, P. and R. van der Sandt (eds.), 133–152. Kuno, S. 1982. “The Focus of the Question and the Focus of the Answer”. In Papers from the Parasession on Nondeclarative Sentences from the Regional Meeting of the Chicago Linguistics Society. Lambrecht, K. 1994. Information structure and sentence form. A theory of topic, focus, and the mental representations of discourse referents. Cambridge: CUP. Landman, F. In progress. “Events and Plurality. The Jerusalem Lectures”. Ms., Tel Aviv University. Li, C. and S. Thompson 1976. “Subject and Topic: A New Typology”. In C. Li, (ed.), Subject and Topic. New York: Academic Press. May, R. 1977. The Grammar of Quantification. MIT. Ph.D. dissertation. May, R. 1989. “Interpreting Logical Form”. R. May (ed) Studies on Logical Form and Semantic Interpretation. Linguistics and Philosophy. 12,4. 387– 435. Partee, B., H. 1992. “Topic, Focus and Quantification”. in Proceedings of the 1991 SALT Conference. Department of Modern Languages and Literatures. Cornell University. Partee, B., H. 1994. “Focus, Quantification, and Semantics-Pragmatics Issues”. Preliminary Version in Bosch, P. and R. van der Sandt (eds.). Reinhart, T. 1979. “Syntactic Domains for Semantic Rules.” In F. Guenthner and S.J. Schmidt (eds.), Formal Semantics and Pragmatics for Natural Language. Dordrecht: Reidel. Reinhart, T. 1981. “Pragmatics and Linguistics: An analysis of sentence topics”. Philosophica 27. Reinhart, T. 1983. Anaphora and Semantic Interpretation. London: Croom Helm.

150

NOMI ERTESCHIK-SHIR

Rooth, M. 1985. Association with Focus. Ph.D. dissertation, University of Mass., Amherst. [Reproduced by GSLA.] Schwartz, A. 1976. “On the Universality of Subjects: The Ilocano Case”. In Charles N. Li (ed.), Subject and Topic. New York: Academic Press. Szabolcsi, A. 1995. “Strategies for Scope Taking”. Working Papers in the Theory of Grammar 2,1. Budapest: Research Institute for Linguistics. Hungarian Academy of Sciences. Vallduví, E. 1992. The Informational Component. New York: Garland. Vendler, Z. 1967. Linguistics in Philosophy. Ithaca, New York: Cornell University Press. Williams, E. 1988. “Is LF Distinct from S-Structure? A Reply to May”. Linguistic Inquiry 19(1).135–146

The Interaction between Focus and Tone in Bantu* Larry M. Hyman University of California, Berkeley

Abstract Both African languages in general and Bantu languages in particular are known for their “focus prominence”: As in (all?) other languages, focus can determine the syntactic structures that will be used in a given context, e.g. the word order of major constituents, the choice of clause type (main, relative, cleft) etc. What is particularly striking about focus in African languages is its effect on other parts of the grammar, specifically on the morphology and the phonology. Within the morphology, these focus effects may determine verb marking, as when a tense is marked differently according to whether the verb is included within the focus or not. Similar distinctions can be made on the morphological marking of NP complements (in- vs. out-of-focus, as in the celebrated case of Aghem). In this paper I take a close look at the effects of focus on tone in Bantu. Although the literature abounds with examples where surface tones appear to vary according to what is in focus, I argue here that the relationship is not a direct one: The semantics of focus does not directly affect tone in Bantu. Instead there is always mediation by the grammatical system such that tone-focus correlations are imperfect at best. Examples come from a number of Bantu languages which are selected to show the range of focus-tone interactions that may be found within this language group.

1.

Introduction

Given that almost all of the approximately 500 Bantu languages are tonal, it will perhaps come as no surprise that their tone systems are frequently sensitive to considerations of focus. Based on their familiarity with English, linguists find it

152

LARRY M. HYMAN

quite natural that the realization of stress should be dependent on what is in vs. out of focus. Pitch is of course a (or the) major phonetic correlate of stress. Since tone = pitch, tone might at least sometimes be expected to act as English stress. Such an expectation is fulfilled in many of the Eastern Bantu languages. In these languages there is an underlying opposition between marked H(igh) tone-bearing units vs. unmarked toneless ones. Whether these H tones come to the surface appears to depend in part on focus. This has considerable consequences for the linguist interested in studying either phenomenon. The phonologist interested in accounting for the properties of a Bantu tone system must undertake a thorough-going study of the grammar of focus in the language in question. On the other hand, the linguist interested in the syntax or semantics of focus may find enlightenment in considering the phonology of tone in these languages. In other words, tone can be seen as providing an interesting window on the nature of focus in general. While the intersection of these two domains is quite complex and interesting, I will argue in this paper that the relationship is not a direct one: The semantics of focus does not directly affect tone in Bantu. Instead there is always mediation by the grammatical system such that tone-focus correlations are imperfect at best. There are unmistakeable correlations such that focus may be associated with a syntactic position (or construction), a morphological spell-out, or a phonological process. In all cases that I know, however, the construction, morphological exponent or phonological process may also characterize elements not semantically in focus; or they may fail to characterize constituents which clearly are focused. Perhaps this is true in all languages that mark focus formally. To account for this imperfect alignment of semantic focus and linguistic form, it is thus necessary to evoke the Grammar as a mediator. The goal of the present paper is to illustrate some of the ways in which tone and focus interact in Bantu. We begin in Section 2 by examining how [+focus] may correlate with tone followed in Section 3 with cases where tone marking correlates with the absence of focus. In Section 4 we consider the interaction between focus, morphology and tone. In Section 5 the major point is reaffirmed: that the tonal features that appear to be focus-conditioned are instead conditioned by certain grammatical configurations which in turn only imperfectly correlate with the expression of focus.

2.

Tonal marking of [+focus]

The interaction between focus and tone in Bantu is profitably typologized

THE INTERACTION BETWEEN FOCUS AND TONE IN BANTU

153

according to which of the two values of the feature [±focus] is “tonally active”.1 By tonally active is meant a tonal alternation or modification that is conditioned by focus. In some languages it is [+focus] that occasions a tonal modification, while in others it is [−focus].2 When it is [+focus] that is tonally active, the tones of the focused constituent generally undergo processes that are typical of what I refer to as “tonal finality”, i.e. tonal properties that characterize the end of a phrasal domain. In the opposite situation, [−focus] is tonally active. In this case the attested processes are quite different. Instead of tonal finality, [−focus] modifications may best be seen as marking a “tonal integration”, i.e. as tonal properties that are associated with phrase-internal position. These two potential focus-tone interactions are summarized in (1). (1)

Parameters of focus-tone interactions in Bantu a. What is tonal active? i. [+focus] ii. [−focus] b. How is this interaction marked? i. [+focus] → tonal finality = end demarcation ii. [−focus] → tonal integration = reduction

As a final introductory point, either active value of [±focus] can be syntactically defined or may be morphologized in ways to be discussed below. In this section we consider the tonal marking of [+focus]. The tonal marking of [−focus] is then treated in Section 3. We begin by considering the end-marking of assertive focus in lu-Haya, a Bantu language spoken in Tanzania.3 The nouns in (2a,b) show a surface opposition between HL and H tones in phrase-penultimate position: (2)

lu-Haya: H, L and HL a. o-mu-kâzi ‘woman’ b. e-m-bûzi ‘goat’ c. o-mu-kázi a-ka-gw-a d. e-m-búzi e-ka-gw-a ‘a woman/goat fell’ e. o-mu-kâzi % Káto f. e-m-bûzi % Káto ‘a ⁄a , Kato’

e-m-púnu ‘pig’ o-mú-ti ‘tree’ e-m-punú e-ka-gw-a o-mu-tí gú-ka-gw-a ‘a pig/tree fell’ e-m-púnú % Káto o-mú-tí % Káto ‘a /a , Kato’

In each case two prefixes precede the noun stem: a noun class prefix (mu-, m-)

154

LARRY M. HYMAN

and a prefixal vowel (o-, e-) referred to in Bantu as the “augment”. In (2c,d) we see that these same nouns are realized differently as subject of the verb ‘fall’: The nouns that have a penultimate falling tone in (2a) have a penultimate H tone in (2c), while those that have a penultimate H tone in (2b) have a final H tone in (2d). A similar, but slightly different situation is observed in (2e,f). In these examples the nouns appear before the right-dislocated proper noun ‘Kato’ used vocatively, e.g. in answer to Kato’s question: ‘What do you see?’. Answer: ‘ , Kato’, etc. The right-dislocated vocative is marked by the % symbol. As seen, the nouns in (2e) again have a penultimate HL tone. This time, however, the nouns in (2f) end in a H-H sequence. These facts are summarized below in (3). (3)

Underlying a. b.

…H-Ø …Ø-H

Internal (2c,d) …H-L …L-H

Before % (2e,f) …HL-L …H-H

Before Pause (2a,b) …HL-L …H-L

As is common in Bantu, we propose that vowels in lu-Haya are either underlyingly H or are toneless. Noun stems such as /-kázi/ and /-búzi/ are set up with an underlying penultimate H, while stems such as /-punú/ and /-tí/ have an underlying final H. To derive their internal realization as subject in (2c,d) it suffices to assign a default L tone to any vowel that does not have an underlying H. To derive the % phrase-final forms, we introduce the right-to-left tone spreading rules in (4a,b), which follow default L spelling:

(4)

a. σ σ H L %

b. σ σ = L H %

c. σ = H

L] pause

In (4a) a phrase-final L spreadings to the left to create a HL falling tone. In (4b) a phrase-final H similarly spreads to the left. Since lu-Haya does not permit LH rising tones, the L automatically delinks, creating a phrase-final H-H sequence. The rules in (4a,b) also apply before pause, which necessarily involves a % phrase edge as well. However, lu-Haya does not permit a H to surface before pause. In the rule in (4c) we posit a L boundary tone before pause which, as shown, modifies the H-H derived from (4b) into H-L. What then is the nature of this % phrase marker? We contend that it marks assertive focus, i.e. the end of an assertion. To see this, consider the sentences in (5).


(5)

a.

b.

c.

155

a-ba-kázi ni-ba-bal-íl-a ó-mw-ána é-m-bûzi women -.-count-for child goats ‘the women (cl.2) count the goats for the child’ a-ba-kázi ó-mw-ána é-m-búzi ni-ba-zi-mu-bal-îl-a women child goats -.-.-.-count-for ‘the women, the child, the goats, they have counted them for him’ ni-ba-zi-mu-bal-îl-a á-ba-kâzi ó-mw-âna é-m-bûzi -.-.-.-count-for women child goats ‘they have counted them for him, the women, the child, the goats’

The input sentence in (5a) shows the subject noun a-ba-kázi ‘women’ preceding the verb ni-ba-bal-íl-a, which in turn is followed by the two objects ó-mw-ána ‘child’ and é-m-bûzi ‘goat’.4 Each noun as well as the verb has an underlying penultimate H which, in phrase-final position, would become HL. In (5a) only the noun é-m-bûzi ‘goats’ occurs phrase-finally, and hence only it has a surface penultimate HL. In (5b) we have left-dislocated the two objects, which now require coreferential object markers prefixed to the verb stem. In this case it is the verb ni-ba-zi-mu-bal-îl-a ‘they have counted them for him’ whose penultimate H becomes HL. Now compare this result with the corresponding rightdislocations in (5c). Again, object markers are required on the verb which agree with the dislocated objects. In this case we obtain four instances of HL tone: on the verb, on the right-dislocated subject, and on the two dislocated object nouns. What this means is that there is an asymmetry between left- and right-dislocation in this language. Why should this be? As shown in (6), the answer has to do with assertive focus: (6)

a.

b.

c.

[ LD1 [ LD2 [ LD3 [ Assertion ] ] ] ] ………… % % % % (LD = left-dislocation) [ [ [ [ Assertion ] RD1 ] RD2 ] RD3 ] …………% …………………% …………………………% …………………………………% (RD = right-dislocation) [ ni-ba-zi-mu-bal-îl’ á:-ba-kâzy’ ó:-mw-â:n’ é:-m-bûzi ] la%á zi%ó na%é

156

LARRY M. HYMAN

If we assume that % marks the end of an assertion, then the left-dislocations in (6a) will not assign a % until the entire proposition is completed. By contrast, as seen in (6b), an assertion (complete proposition) precedes each right-dislocation. Thus multiple %’s are assigned in (6b), one at the end of each of the four completed assertions. Although the change of H to HL is an instance of tonal finality, note in (6c) that there are no pauses before (internal) %’s. In fact, as the apostrophes indicate, there is obligatory vowel coalescence: each word final vowel either deletes (if non-high) or glides (if high), conditioning compensatory lengthening on the following vowel. Examples of such tonal finality are easy to duplicate elsewhere in Bantu. As shown by Bresnan and Mchombo (1987) and Kanerva (1989), for instance, ciCewa marks tonal finality in a way similar to that seen in lu-Haya. Thus, the penultimate H on the verb stem /mény-a/ ‘hit’ in the two examples in (7a) becomes a HL fall in focus-phrase-final position: (7)

Examples of tonal finality in ci-Cewa a. tí-náa-mê:ny-a ‘we hit’ [distant past] tí-náa-mu-mê:ny-a ‘we hit him’ b. tí-náa-mény-á m-khuˇ:ngú

c.

H tí-náa-mu-mê:ny-a % m-khuˇ:ngú ‘we hit him, the thief’

As also seen in these examples, a focus-phrase-penultimate vowel is lengthened by a rule which can be informally stated as in (8). (8)

a.

Ø → V / ___ C V %

The first example in (7a) has the simple verb with no object, while the second example has the object marker -mu-. In either case the penultimate H becomes HL. In (7b), however, where the verb is followed by an overt object noun, the verb ends in a H-H sequence. This is because of a general rule that spreads a H tone one vowel to the right, as shown by the dotted association line in (7b).5 In (7c), where the object noun is right-dislocated and there is consequently a coreferential object marker, the verb again exhibits a penultimate HL tone.


157

Except for the application of H tone spreading in (7b), the effects of being final in a focus phrase on a penultimate H are the same as in lu-Haya. As seen in (9a), (9) a.

b.

c.

Cewa-Ntcheu and Cewa-Nkhotakota VVCV% H VVCV% | H Cewa-Nkhotakota s H

L ]pause

the HL fall is automatically created by the creation of penultimate vowel length by rule (8). Now what about a form with a final H tone? The tonal realization of the noun /m-khungú/ in (7b,c) is typical. As shown in (9b) the final H spreads leftward onto the second half of the lengthened penultimate vowel, thereby creating a LH rising tone. The resulting final LH-H sequence is widely attested in ci-Cewa, e.g. in the Ntcheu dialect spoken by Al Mtenje. In the Nkhotakota dialect spoken by Sam Mchombo, however, such forms surface as LH-L before pause. This is the result of the linking of a L boundary tone, as in (9c), a process which is highly reminiscent of what we saw in lu-Haya in (4c). The processes that are summarized in (4) and (9) are some of those typically associated with phrase-final phonology: attraction of a tone to the penultimate syllable and reduction of a final H tone. What is likely is that the attraction to penultimate position originally took place before pause and only later generalized to utterance-internal positions. It is this generalization (or “boundary narrowing”) that causes one to note the association between focus phrase and final tonality. Not every Bantu language with a focus-tone interaction can be accounted for in such a straightforward manner. Creissels (1996) and Chebanne, Creissels and Nkhwa (1996) document a similar phenomenon in seTswana. They offer the minimal pair in (10). (10)

Tonal effect of % in se-Tswana ≠ tonal finality a. bátSáàb´7r´7ká % líb`fn´7 ‘they1 will work, they1 too’

158

LARRY M. HYMAN

b.

bátSáàb´7r´7kà líb`fn´7 ‘they1 will work with them2’

In (10a) the form líbfn ` 7´ is separated from the verb by % and means ‘they too’. In (10b) líbfn ` 7´ means ‘with them’ and occurs in the same focus phrase as the verb. However, the tonal difference on the final vowel (H vs. L) does not represent the phonologization of final tonality: There is no general reason why the final vowel should be H before % in (10a) vs. L internally in (10b). In fact, as Creissels shows, the expression of the opposition in (10) is realized quite differently according to the “tiroir verbal”, i.e. the combinations of tense, aspect, polarity etc. that define the different morphological forms of the verb. The conclusion is therefore that the tonal difference seen in (10) necessarily represents a  fact, not the result of the application vs. non-application of phrase-final phonology.6 We have just seen that tonal effects related to focus need not be derived by general phonological rules. In addition, these effects need not be solely tonal, nor need they be restricted to occuring at the end of a “complete assertion”. A relevant illustration comes from the ci-Bemba in (11), cited from Sharman (1956): (11)

The tonal effect of % in ci-Bemba ≠ the end of a complete assertion a. bushé mu-la-peepa (Present, [+focus]) ‘do you smoke?’ b. ee tu-peepa sekelééti (Present, [−focus]) ‘yes, we smoke cigarettes’ (Present, [+focus]) c. bámó bá-la-lyá ínsoka ‘some people actually eat snakes’

In the question in (11a) the verb form is marked by the [+focus] present tense marker -la- which is present because the verb falls under the scope of focus. In the answer in (11b), on the other, where the verb is presupposed by the preceding question, -la- does not appear. Instead the [−focus] form of the present tense is characterized by the lack of any tense prefix. The sentence in (11c), which requires no previous set-up (other than perhaps something like an introductory “Did you know that…”), shows that the use of -la- is quite subtle. Here, as seen from the translation, the verb is included as part of the focus/assertion. If -lahad been left out the verb would unambiguously have been presupposed. A similar observation can be made with respect to Sharman’s pair of examples in (12), where there is also a relevant tonal difference:


(12)

a. b.

159

[+focus] nga mw-aa-tóbá úmutóndó, bálééisaafúlwá ‘if you break the pot they will get angry’ nga mw-aa-tóba úmutóndó, tsáákuláatápíla múnsupa [−focus] ‘if you break the pot we will have to use a calabash to draw water’

In (12a) the consequent clause responds to the whole proposition in the first clause: one will get angry because of your breaking the pot. Put slightly differently, focus in the protasis is on the whole action represented by the proposition rather than on any subconstituent within it. The verb is therefore “in focus,” marked by the spreading of the H tone to the final vowel. In (12b) on the other hand, attention is on what will be used if there is no pot. Focus is thus on the post-verbal object and the verb is “out of focus,” and there is no H tone spreading onto the final vowel. The kinds of differences just illustrated in se-Tswana and ci-Bemba have been observed in other Bantu languages as well and have been referred to the relationship between the verb and what follows it in a variety of ways. Thus the (a) clauses in (10)–(12) have been referred to as “open context”, “weak link”, “disjoint” and “verb-focused” by different scholars, while the (b) clauses have been referred to as “close context”, “strong link”, “conjoint” and “post-verb focused” (Carter 1962; Meeussen 1959, 1963; Givón 1971; Hyman and Watters 1974). Meeussen (1959) sets up a thorough-going “conjoint”/”disjoint” opposition in Kirundi, as summarized in (13). (13)

“Conjoint” vs. “Disjoint” tenses in ki-Rundi C [−focus] D [+focus] [−TR] P0 -Ø- … -a/-ye [+TR] -ra- … -a/-ye P1 -a- … -a/-ye [+TR] -aa- … -a/-ye [−TR] [−TR] P2 -á- … -a/-ye [+TR] -ára- … -a/-ye

The tenses in question are here identified as P0, P1 and P2, roughly the immediate, general and distant past, although the semantics vary. Meeussen points out that the system divides up into a binary opposition: his “conjoint” corresponds to our [−focus], while his “disjoint” corresponds to our [+focus].7 As seen the prefixal marking of these tenses is fuller in the [+focus] column. This is a common feature in Bantu languages (cf. Sharman 1955 and Givón 1971 for ciBemba; Carter 1963 for ci-Tonga etc.). The [+TR] feature indicates that there is tonal reduction of the H tone of a H verb root in the conjoint form. Such tone reduction does not take place in the disjoint forms. As we shall see in the next section, this too is a common feature of focus-tone interactions in Bantu.

160 3.

LARRY M. HYMAN

Tonal marking of [−focus]

The prefix forms in (13) clearly indicate that it is [+focus] that is morphologically marked in ki-Rundi. At the same time the tonal reduction rule applies only to the verb when it is out of focus. In contradistinction to the lu-Haya and ci-Cewa examples discussed in Section 2, where [+focus] triggered a tonal process, we can refer to TR in ki-Rundi as tonal marking of [−focus]. A similar phenomenon actually occurs in lu-Haya as well. The sentences in (14) illustrate the only trace of the old conjoint/disjoint opposition in the prefix system of luHaya: (14)

One last trace of the conjoint/disjoint opposition in lu-Haya a. y-a-kom-a káto ‘he tied Kato’ (-a- = “conjoint” P1) b. y-áá-mu-kôm-a ‘he tied him’ (-mu- ‘him’; -áa- = “disjoint” P1)

As seen, the conjoint form in (14a) of the today past (P1) tense is marked by the prefix -a-, while the disjoint form in (14b) is marked by -áa-. In lu-Haya the conjoint is used whenever anything follows the verb in the same clause, e.g. the object noun káto in (14a). The disjoint form is used in case the verb is final within its clause, as in (14b). In both sentences the verb root -kóm- ‘tie’ has an underlying H tone. In the conjoint form in (14a), however, this H undergoes a tone reduction (TR) similar to that characterizing ki-Rundi. A comparison of the disjoint forms in (15a) with the corresponding conjoint forms in (15b) shows the effects of TR in a number of tenses in lu-Haya: (15)

Tonal reduction (TR) in lu-Haya a. ‘they tie’ etc. Present habitual ba-kóm-a Past1 bá-á-kôm-a Past2 ba-kom-íl-e Past habitual ba-a-kóm-ag-a ba-laa-kôm-a Future1 ba-li-kóm-a Future2

b.

‘they tie Káto’ etc. ba-kom-a káto ba-a-kom-a káto ba-kom-il-e káto ba-a-kom-ag-a káto ba-laa-kom-a káto ba-li-kom-a káto

As seen, all H tones in these tenses under TR when the verb is not final in its clause, i.e. when it is followed by an NP, adverb, or whatever. However, as seen in (16),


161

(16)

No tonal reduction in the corresponding negative forms a. ‘they don’t tie’ etc. b. ‘they don’t tie Kato’ etc. Present habitual ti-bá-kom-a ti-bá-kom-a káto Past1 ti-bá-á-kom-a ti-bá-á-kom-a káto ti-ba-kom-íl-e ti-ba-kom-il-é káto Past2 Past Habitual ti-bá-á-kom-ag-a ti-bá-á-kom-ag-a káto Future1 ti-ba-a-kôm-e ti-ba-a-kóm-e káto Future2 ti-bá-li-kom-a ti-bá-li-kom-a káto TR does not occur in the corresponding negative forms of these tenses. In addition, the examples in (17) show that TR does not apply when an affirmative verb form incorporates a “marked” tense, aspect and mood (TAM): (17) No tone reduction in “marked” tenses, aspects and moods a. ‘they are tying’ etc. b. ‘they are tying Kato’ etc. Progressive ni-ba-kóm-a ni-ba-kom-á káto Perfect (“already”) bá-á-kóm-il-e bá-á-kóm-il-e káto ba-lá-kom-íl-e ba-lá-kom-il-e káto Experiential8 Persistive (“still”) ba-kyáá-kôm-a ba-kyáá-kóm-a káto Subjunctive ba-kóm-e ba-kom-é káto Imperative kóm-a kom-á káto Past3/Perf/Consec9 bá-ka-kôm-a bá-ka-kóm-a káto The idea is that such TAM’s have an intrinsic morphosyntactic focus, [+F], which derives from their marked semantic status (Hyman and Watter 1984). Thus, negation is the marked polarity, subjunctive and imperative are marked moods, progressive and persistive are marked aspects, and the perfect is a marked tense. It is this [+F] which exempts the verb forms in (17) from TR, as well as their corresponding negatives in (18). (18) No tone reduction in negative marked TAM’s a. ‘they are not tying’ etc. b. ‘they are not tying Kato’ etc. Progressive ti-bá-li-ku-kôm-a ti-bá-li-ku-kóm-a káto Perfect ti-bá-ka-kóm-il-e ti-bá-ka-kóm-il-e káto Experiential ti-bá-ka-kóm-ag-a ti-bá-ka-kóm-ag-a káto Persistive ti-bá-kyáá-kôm-a ti-bá-kyáá-kóm-a káto Subjunctive ba-tá-kom-a ba-tá-kom-a káto Imperative o-tá-kom-a o-tá-kom-a káto ti-bá-á-kom-il-e ti-bá-á-kom-il-e káto Past3

162

LARRY M. HYMAN

By comparison, the affirmative forms in (15) which do undergo TR do not contain any of the marked TAM’s: they simply incorporate simple (i.e. nonintrinsically focused) tense distinctions. Note in this regard the relation between the present habitual in (15) and the progressive in (17). As indicated in (19), the latter is clearly derived from the former by the addition of a initial prefix ni-. (19)

Progressive = ni + Present habitual cf.: ní káto ‘it’s Kato’

As the example in (19) illustrates, this ni- is none other than the focus marker in lu-Haya. It is thus as if the progressive consists of the present habitual (or “zero” tense) preceded by a focus marker: ‘it is they tie’. As has been noted, the [+F] progressive does not undergo TR, while the [−F] present habitual does. In the above account we assume, first, that [+F] represents a “secondary focus” and, second, that it is this secondary focus that “protects” the verb from tone reduction. It is a secondary focus that has become morphologized and which, therefore, only imperfectly corresponds to the semantics that motivates it. We see this particularly clearly in verb forms that are marked by the prefix -ka-, as in (20). (20)

Verb forms marked by -ka- are redundantly [+F] a. bá-ka-kóm-a káto ‘they tied Kato’ (Past3) b. ti-bá-ka-kóm-il-e káto ‘they have already tied Kato’ (Perf) ti-bá-ka-kóm-ag-a káto ‘they have tied up Kato before’ (Exp) c. ti-bá-á-gend-a bá-ka-kóm-a káto ‘they didn’t go and tie Kato’ (Consec) a-bá-á-gend-a bá-ka-kóm-a káto ‘they who went and tied Kato’ (Consec)

In (20a) -ka- marks the distant past (Past3) in main clause affirmatives. That it alone is responsible for the [+F] specification is seen in the corresponding relative clause affirmative in (21a). (21)

The same [±F] marking carries over to relative clauses a. a-ba-a-kóm-il-e → a-ba-a-kom-il-e káto ‘they who tied Kato’ (Past3) b. a-bá-á-kóm-il-e → a-bá-á-kóm-il-e káto ‘they who have tied Kato’ (Perf)


163

The corresponding relative clause Past3 form has the prefix -a- instead of -ka(and, in addition, ends in the perfective sequence -il-e instead of the final vowel -a). Since -ka- is not present, and since Past3 is not otherwise [+F], the H of the verb root -kóm- is reduced in the output of (21a). This contrasts with the tonally distinct perfect form in (21b), which is [+F] and therefore does not undergo TR even in a relative clause.9 In the case of -ka- we see therefore that there is some arbitrariness: The [+F] property of -ka- in the Past3 and consecutive does not derive from the semantics (as it does in the other cases), but rather from the fact that *-ka- reconstructs as a perfect (Mould 1979). Although occurring in this function only in negative verb forms, its effect on tone provides additional evidence of its prior historical status. We have thus established that the (largely) semantically-determined [±F] feature protects the H’s of verbs from TR, an example of the “tonal integration” marking referred to in Section 2. Although TR also applies within noun phrases, the situation is quite different. Tonal integration in this case depends solely on grammatical factors. Starting with the nouns in (22a), (22)

Tonal integration in the noun phrase in lu-Haya a. /é-ki-kómbe/ /ó-mu-tí/ [e-ki-kô:mbe] [o-mú-ti] b. e-ki-ko:mbe kyáitu o-mu-ti gwáitu ‘our cup’ ‘our tree’ e-ki-ko:mbe kya:= káto o-mu-ti gwa:= káto ‘Kato’s cup’ ‘Kato’s tree’ e-ki-ko:mbe ki-lú:ngi o-mu-ti mu-lú:ngi ‘good cup’ ‘good tree’ c. e-ki-kó:mbe kya:nge o-mu-tí gwa:nge ‘my cup’ ‘my tree’ d. e-ki-kó:mbe kî-li o-mu-tí gû-li ‘that cup’ ‘that tree’ e-ki-kó:mbe kî-mo o-mu-tí gû-mo ‘one cup’ ‘one tree’

we observe first in (22b) that their stem H tones are reduced when there is a following possessive pronoun, genitive noun or adjective.10 The examples in (22c) show that TR does not apply when the following word lacks a H tone, while those in (22d) show that TR is not conditioned by either a demonstrative or a numeral. It seems inappropriate to introduce a [+F] feature in (22d). We

164

LARRY M. HYMAN

assume rather that TR applies more generally to verbs than with nouns. If nonfinal in its clause, a verb will undergo TR unless it is [+F]. By contrast, TR will apply to a noun only if it is non-final within its NP and if the word that follows has a H tone.11 For this to work we assume that demonstratives and numerals fall outside the NP, as they would if we assume that the NP appears in turn within a determiner phrase (DP).

4.

Focus, morphology and tone

Besides tone, the [+F] feature may have a major effect on the morphology (and even syntax) of a language. A case in point is the out-of-focus (OF) determiner f´ in Aghem, a Grassfields Bantu language spoken in Cameroon. In this language the immediate after verb (IAV) position is used to mark focus (Watters 1979). The word order in (23a) potentially expresses an utterance with neutral or “even” focus or possibly one where the object tí-bvú ‘dogs’ is the focus: (23)

The out-of-focus (OF) determiner -´f in Aghem — !n´7 a. ò mf` tà] tí—-bvú he  count dogs today ‘he counted dogs today’ ! bvú — !t´f b. ò mf` tà] n´7 he 1 count today dogs  ‘he counted dogs today’ c. ò kà tá] bvú — !t´f n´7 he P1– count dogs  today ‘he didn’t count dogs today’ d. tá] bvú— !t´f n´7 count dogs OF today ‘count dogs today!’ — !t´f n´7 e. wizí—n wì—l à ò m`f tà] bvú woman   she P1 count dogs OF today ‘the woman who counted dogs today’

In (23b) contrastive focus is marked on the adverb n7´ ‘today’ by placing it in the IAV position. In this case the following direct object noun bv} ¢ !tf´ appears with 12 the OF determiner. As shown by Hyman (1985) a non-empty determiner is required in Aghem if an NP is not properly governed, i.e. if it does not appear


165

adjacent to its c-commanding lexical head. The semantically empty (or “dummy”) OF determiner -f´ appears iff a demonstrative does not appear in the determiner slot of a non-properly governed NP. If the NP appears directly after its head, it may appear without any determiner — and must not in this case appear with -f. ´ However this generalization appears to be violated in (23c–e), where the object noun appears as bvú!tf´ even though it follows its verb head in each case. The solution proposed in Hyman (1985) is to propose a structure where the verb does not c-command the following object, e.g. by having the verb join the auxiliary morphemes under INFL. It will be observed that (23c) involves a negative and (23d) an imperative, i.e. marked polarity and marked mood. As in lu-Haya above (and lu-Ganda below), negation and marked TAM’s have a secondary focus which requires the direct object noun to have an overt determiner. We can thus say that their [+F] attracts the verb to INFL, leaving behind a trace as head of VP which cannot properly govern the immediately following NP. In the case of (23e) the OF determiner is required on the object noun because its verb head is in a relative (i.e. backgrounded) clause.13 Not only can [+F] affect the morphology, but also the syntax. In many languages there is no negative imperative: instead, there is a paraphrastic form, such that ‘don’t tie Kato’ is built on ‘refrain from tying Kato!’. Could this be evidence of a desire to avoid two [+F]’s in one clause, one from the negative, one from the imperative mood, as one frequently avoids two [+focus] constituents? In addition, Takizaka (1973) points out an interesting situation where negative marking is not allowed in relative clauses. Instead, as seen in (24), a paraphrastic structure is used: (24)

kit ki a-khoon-in Kipes ku-suum chair  -fail- Kipes to-buy ‘the chair that Kipese didn’t buy’

(Kihung’an)

The literal translation of the negative clause in (24) is ‘the chair that Kipes failed to buy’. Here it appears that the [+F] of negation cannot occur in an out-offocused (backgrounded) clause. In other words, the morphosyntactic [+F] conflicts with the [−focus] of the clause. In more complex situations [+F] interacts with the morphology, the syntax, and the tone system. A remarkable case of this is found in lu-Ganda.14 As indicated in (25a),

166

LARRY M. HYMAN

(25)

The tone group (TG) in lu-Ganda X a.

Y X Z [........................]TG b.

c.

e-bi-kópò e-bi-kópò e-bi-kópò e-bi-kópó HØ

e-bi-nénè bi-sátu bi-rî by-áá=wálúsìmbì HL L

‘big cups’ ‘three cups’ ‘those cups’ ‘cups of Walusimbi’

a tone group (TG) is syntactically defined as consisting of the X head of an X” plus the first word (Z) of a c-commanded Y”. The TG in turn serves as the domain in which a rule of L tone deletion (LTD) applies. As indicated in (25b), the noun e-bi-kópò ‘cups’ has a H to L pitch fall on its bisyllabic stem. This is seen when it is followed, respectively, by an adjective, a numeral and a demonstrative. In (25c), on the other hand, there is no pitch drop on the syllable [po]. The L’s of a H-L pitch drop of a noun are deleted by LTD whenever it is followed by a genitive complement. Since the latter must contain a H tone, the rule is actually one by which a H-Ln-H sequence becomes H-Hn-H. While (25a) generally schematizes the syntactic configuration that must obtain in order to form a TG, the (non-) application of LTD depends to a large extent on the nature of the “X” and of the “Z” in the formulation. As seen in (26), this X may also be a verb: (26)

Nature of X: LTD can apply in most affirmative verb tenses a. tú-làb-a tú-láb-á wálúsììmbì HL HØ HLL L ‘we see Walusimbi’ [] b. tú-làb-y-e tú-láb-y-é wálúsììmbì H L LL H Ø ØØ H LL L ‘we have seen Walusimbi’ [] c. tw-áá-láb-y-ê tw-áá-láb-y-é wálúsììmbì HØ Ø ]HL HØØ ]H H LL L ‘we saw Walusimbi’ []


d.

e.

f.

g.

h.

i.

167

tw-áá-làb-à tw-áá-láb-á wálúsììmbì H L L H ØØ HLL L ‘we saw Walusimbi’ [] tú-náá-láb-à tú-náá-láb-á wálúsììmbì H Ø H L H Ø HØ H LL L ‘we will see Walusimbi’ [] tú-lì-làb-a tú-lí-láb-á wálúsììmbì HL L H ØØ H LL L ‘we will see Walusimbi’ [] ne tú-làb-à ne tú-láb-á wálúsììmbì HL L HØ Ø H LL L ‘and we saw Walusimbi’ [] tu-lab-ê tu-lab-é wálúsììmbì HL HØ HLL L ‘let’s see Walusimbi’ [] tw-áándí-láb-y-ê tw-áándí-láb-y-é wálúsììmbì H ØØ ]HL H Ø Ø ]HØ H LL L ‘we would see Walusimbi.’ []

Since it is beyond the scope of this paper to develop a full discussion of verb tones in lu-Ganda, note only in (26) that the final L(‘s) of the forms in the first column are deleted (marked by Ø) when followed by the proper noun Walúsììmbì. This contrasts with the corresponding negative tenses in (27) where LTD fails to apply: (27)

LTD does not apply in negative verb tenses a. te-tú-làb-à walúsììmbì H L L H LL L ‘we do not see Walusimbi’ [ ] b. te-tú-láb-y-ê walúsììmbì H Ø ]HL H LL L ‘we haven’t seen Walusimbi’ [ ] c. te-tw-áá-làb-y-è walúsììmbì H LL L HLL L ‘we didn’t see Walusimbi’ [ ] d. te-tw-áá-làb-à walúsììmbì H L L HLL L ‘we didn’t see Walusimbi’ [ ]

168

LARRY M. HYMAN

e.

f.

g.

h.

te-tú-ú-láb-è walúsììmbì H HL HLL L ‘we will not see Walusimbi’ te-tú-lì-làb-a walúsììmbì H LL HLL L ‘we will not see Walusimbi’ ne tú-tà-làb-à walúsììmbì H L LL HLL L ‘and we didn’t see Walusimbi’ te-tw-áándì-làb-y-è walúsììmbì H H L L L HLL L ‘we wouldn’t have seen Walusimbi’

[ ] [ ] [ ]

[ ]

This of course is analogous to the situation described in lu-Haya in Section 3 where TR failed to apply to negative verb forms. As in lu-Haya there are some intrinsically [+F] affirmative tenses which block LTD in lu-Ganda: (28) a.

b.

c.

d.

LTD does not apply in certain affirmative verb tenses, which are [+F] lab-írìr-à walúsììmbì H L L H LL L ‘look after Walusimbi!’ [] tú-kyáá-láb-à walúsììmbì H Ø H l HLL L ‘we still see Walusimbi’ [] tw-aaka-láb-à walúsììmbì HL HLL L ‘we have just seen Walusimbi’ [] o-ku-láb-à walúsììmbì H L HLL L ‘to see Walusimbi’ []

Thus, although the processes are different in the two languages (TR in lu-Haya vs. LTD in lu-Ganda), there is an unmistakeable resemblance. Unlike TR in lu-Haya, however, what follows the verb can determine whether LTD will apply. The examples in (29) show that the Z in (25) can bear any grammatical relation to the verb, as long as it appears within the same clause:


(29)

169

Z can bear any relation to the verb X within X” a. tw-áá-géénd-á tútùtu H ØØ Ø H L ‘we went slowly’ b. tw-áá-géénd-á lúlî H ØØ Ø HL ‘we went the day before yesterday’ c. tw-áá-géénd-á ná=wálúsììmbì H ØØ Ø H LL L ‘we went with Walusimbi’ d. tw-áá-láb-w-á wálúsììmbì HØ Ø Ø H LL L ‘we were seen by Walusimbi’ e. tw-áá-mù-làb-à wálúsììmbì H L L L H LL L ‘we saw him, Walusimbi’ (-mu- him’) f. *tw-áá-mú-láb-á wálúsììmbì H Ø Ø Ø H LL L

In (29a–d) the verb is followed, respectively, by a manner adverb, a temporal adverb a prepositional phrase, and the agent of the passive (which is Ø-marked in lu-Ganda). In all four cases LTD applies. In (29e), however, where Walúsììmbì has been rightdislocated with a coreferential object marker -mu- appearing within the verb complex, LTD cannot apply. As a consequence (29f) is ungrammatical. We conclude that Z can be any element as long as it appears within the same clause. There is one important exception to this last statement: As seen in (30), when Z begins with the “augment” morpheme, LTD is impossible: (30)

LTD does not apply if Z begins with an “augment” a. tú-làb-a e-bi-kópò ‘we see cups HL HL a′. tú-láb-á bí-kópò ‘we see cups’ HØ HL b. tú-làb-a o-mu-kázì ‘we see a woman’ HL HL b′. tú-láb-á mú-kázì ‘we see a woman’ HØ HL

170

LARRY M. HYMAN

c. c′.

tú-làb-a a-ma-fútà ‘we see oil’ HL HL tú-láb-á má-fútà ‘we see oil’ HØ HL

The augment morpheme, also known as the “initial vowel” or as the preprefix, takes the shape e-, o- or a-, depending on the shape of the following noun class prefix. It is reconstructable to Proto-Bantu and occurs in many of the daughter languages with varying distribution, function and complexity. As outlined in Hyman and Katamba (1993a), its presence vs. absence has been characterized in the various ways listed in (31). (31)

Previous characterizations of the presence vs. absence of the augment a. list of arbitrary conditions, e.g. Ashton et al. (1952) b. semantic characterization, e.g. Bokamba (1971), Givón (1971), Mould (1974) c. syntactic characterization, e.g. Meeussen (1959), Dewees (1971) d. other characterization (stylistic, phonological), e.g. Carter (1963) e. combination of factors, e.g. De Blois (1970), Hyman & Katamba (1993a)

Although Mould (1974) in particular sees the augment as a [+definite] marking, there are at least three arguments, indicated in (32), that militate against this interpretation: (32)

Three arguments against the semantic characterization in lu-Ganda a. The augment can appear on a noun which is [−definite, -specific] ànáákólá=kí, bwè ànáálábà è-njóvú ‘what will he do if he sees an elephant?’ b. The augment can appear on an adverbial (etc.) which is [−referential] y-a-yíímbá bú-lúúngì (*o-bu-lúùngí) HØ Ø H L ‘he sang well’ b′. e-y-a-yíìmbà o-bu-lúùngì (*bu-lúúngì) HL L HL L ‘the one who sang well’


c.

171

The augment can sometimes be absent on a [+definite] noun te-y-a-láb-à bi-kópò bì-nò (*e-bi-kópò) HL HL L L ‘he did not see these cups’

Although the reader is referred to Hyman and Katamba (1993a) for the full story, the solution is outlined in (33). (33)

The solution a. [−A] must be licensed by being in the scope of one of two licensers:  (negative) or  (focus). b. [+A] cannot be so licensed, i.e. cannot fall within the scope of  or  (and is interpreted as the “elsewhere case”).

The application or not of LTD and the presence or absence of the augment logically define four possibilities. However, as seen in (34), only three of these actually exist. (34)

Three of the four possibilities: [−A]

[+A]

[−LTD]

te-yalábà bikópò

yalábà e-bikópò

[+LTD]

yalábá bíkópò

*

The sentence te-yalábà bi-kópo ‘he didn’t see cups’ shows that a noun must not have an augment after a negative verb (because it would be in the scope of the operator ). As indicated, in no case can LTD apply before an augmentmarked nominal. Important for our discussion, the difference between the two remaining sentences in (34), glossed in (35), appears to be one of focus. (35)

a.

b.

y-à-gúl-à e-bi-kópò (bì-nó) HL H L ‘he bought these cups’ [“even” focus] y-à-gúl-á bí-kópò (bì-nó) HØ HL ‘he bought (these) cups’ [postverbal focus]

With the augment present in (35a), focus is “even”, i.e. with the verb included within the focus. One context for (35a) could be in answer to a question, ‘What happened?’. In (35b), on the other hand, focus is on the object ‘(these) cups’ as

172

LARRY M. HYMAN

it would be in answer to the question, ‘What did he buy?’. As also observed, there is no LTD in (35a), since e-bi-kópò ‘cups’ begins with an augment, while LTD does apply in (36b), where bi-kópò has no augment. That it is not focus, however, that is responsible for LTD is seen in forms such as in (36). (36)

a.

b.

y-à-láb-á walúsììmbì H Ø H LL L ‘he saw Walusimbi/Walusimbi’ e-y-à-láb-á walúsììmbì H Ø H LL L ‘the one who saw Walusimbi’

Since proper nouns such as Walusimbi do not take an augment, LTD takes place in (36a). As seen in the gloss, however, the rule applies whether (36a) is construed as even focus or as post-verbal focus. The corresponding form in (36b) also shows LTD applying, even though it is impossible to mark postverbal focus in relative clauses (Hyman and Katamba 1993a). We conclude that it cannot be focus that is directly conditioning LTD in these examples. It must therefore be the augment that somehow blocks the application of LTD in (35a). The solution comes from two sources: the formulation of the LTD rule and the tonal representation of the augment. First, as demonstrated by Hyman (1988), LTD is a “domain juncture rule” in the sense of Selkirk (1980) and Nespor and Vogel (1986), formulated as in (37). (37)

LTD:

Ln → Ø/[TG … [PW … H ___ ] [PW H … ]… ]

A L tone or tones will be targeted at the end of a prosodic word (PW) when the following prosodic word begins with a H tone. Turning to the underlying representation of the augment, evidence such as in (38) establishes that an underlying H-L drop is involved: (38)

a.

b.

ku-láb-à byàà=mùlímí H L ‘to see those of the farmers’ o-ku-láb-a è-byáá=ó-mùlímí H L H L ‘to see those of the farmers’

In (38a) there is no augment and no H other than in the H-L pitch drop on the infinitive ku-láb-à. In (38b) the augment is present on the infinitive, the genitive


173

proclitic, and the possessor noun. As seen, there is a second H-L pitch drop on the noun: the H occurs on the augment and the L on the following prefix mu-. This means that the augment must be underlyingly “tonic”, with its H-L pitch drop preserved only if preceded by a proclitic.15 Following Hyman and Katamba (1993b), we assume that underlying “tonicity” is represented by a HL contour tone. The augment in (38b) is thus underlyingly /ô-/. As indicated in (39a), (39)

a.

b.

[CG V | H → Ø / ____ [ tw-áá-làb-à ] [ è-bi-kópò ] H L L L HL ‘we saw cups’

the H of a HL contour is lost on a vowel that is initial in its clitic group (CG). Otherwise put, a CG cannot begin on a H tone vowel (Hyman & Katamba 1990). As a result, as seen in (39b), the noun è-bi-kópò begins with a L tone at the relevant stage of the derivation. Since the second constituent does not begin with a H tone, as per the requirement in (37), LTD cannot apply. This, then, explains the tonal behavior of the augment without necessary recourse to focus.

5.

Conclusion

In the preceding sections we have investigated the pervasive relationship that exists between focus and tone in Bantu languages. These tonal effects have been characterized in one of two ways. On the one hand, languages may extend domain-final effects on tone to [+focus] domain-internal forms. On the other hand, domain-internal effects on tone may become associated with [−focus]. In some of the cases we have seen focus has been morphologized as [+F]. In these, negation and certain “marked” TAM’s resist [−focus] tonal effects. In no case however have we seen what can be called a “direct mapping” from focus to tone. That is, I am unaware of a “pure” example where semantic focus (and only semantic focus) unambiguously conditions a [+focus] tonal effect, or where the absence of semantic focus (and only its absence) conditions as [−focus] tonal effect.16 In each case the grammar mediates between semantic focus and tone. For example, in lu-Ganda, focus plays an imperfect rôle in determining whether the augment will be present. The augment in turn has a tonal effect (blocking

174

LARRY M. HYMAN

LTD). One cannot say, however, that the blocking of LTD is a property of “even focus”, with which the presence of the augment is sometimes identified. Similarly, we have seen that [+neg] blocks the application of TR in lu-Haya. That this is not due to focus directly is seen from the fact that this effect is observed even when the negation is presupposed (Hyman & Watters 1984). The lesson to be learned from these complex interactions is obvious: Whether one is a syntactician or semanticist wishing to study focus or whether one is a phonologist wishing to study tone, one must consider all aspects of the grammatical system of a Bantu language. As should be clear form the examples surveyed in this paper, to not do so would be to risk drawing the tempting — but wrong — conclusion that there is a direct link between semantic focus and pitch in these languages.

Notes * This paper was presented while I was on sabbatical leave in residence as a Chercheur Associé in the Laboratoire “Dynamique du Langage” (Université Lumière Lyon 2/C.N.R.S.). 1. Throughout this study we shall use the feature [±focus] as a convenience. Later, also for convenience, I shall distinguish between the syntactic feature [+focus] vs. its morphologized analogue [+F]. 2. A third logical possibility is a language where both [+focus] and [−focus] are tonally active. 3. This section is based on Byarushengo, Hyman & Tenenbaum (1976). In these and other examples an acute accent (á) marks H(igh) tone, while either a grave accent (à) or no accent (a) marks L(ow) tone. A circumflex (â) marks a HL falling tone. For more information on lu-Haya tonology, see Hyman & Byarushengo (1984) and Section 3. 4. The augment prefixal vowel is underlyingly H, but is realized L after pause. 5. However, as shown by Kanerva 1989, H not cannot be spread onto either the penultimate or final vowel of a focus phrase, the term he gives to this phrasal domain. As a result, a phrasepenultimate H may not spread either to the final vowel or to V inserted by rule (8). 6. Creissels et al. refer to the verb form in (10a) as “disjoint” and that in (10b) as “conjoint”, an opposition and terminology introduced by Meeussen (1959) for ki-Rundi (cf. (13) below). 7. In (13), the final -a/-ye distinction refers to imperfective vs. perfective, respectively. Meeussen’s analysis is a formal rather than semantic one. He thus points out that the imperfective Po has present tense meaning if [−focus], but immediate future meaning if [+focus]. In addition, the P1 imperfective is translated as a present conditional if [−focus] and a past conditional if [+focus]. 8. The experiential refers to having had an experience at least once in the past, i.e. ‘they have tied before’. In the last example the -ka- prefix is used to mark distant past in the main clause affirmative, the perfect in negative forms, and a consecutive tense after negative or non-main antecedent clauses. In this case it appears to be the morph -ka- which exempts the verb form from TR rather than its actual semantics, which vary quite a bit from language to language —


175

and, as we see in lu-Haya, even within the same language (cf. Mould 1979; Botne, in press; Nurse 1996). 9. In languages that have the conjoint/disjoint distinction, relative clauses typically show the properties of the conjoint, i.e. [−focus] form of the verb. This makes sense since relative clauses are by definition “backgrounded” or “out of focus” with respect to the asserted proposition of the main clause. In lu-Haya, where the conjoint/disjoint distinction has been replaced by the morphologized [+F]/[−F] one, the negative and “marked TAM’s” are [+F] independent of the clause type in which they are found. The one exception is the special behavior of -ka- in the Past3. 10. The H on the augment prefix /é-/ and /ó-/ is not affected by TR in lu-Haya. Instead, its H tone is always reduced to L after pause, as in all of the examples in (22); see note 4. 11. Although the exact realization involves other complexities, TR does apply to a verb that is followed by a word lacking H tone (see Hyman & Byarushengo 1984). 12. It also loses its prefix tí- by a general rule of prefix deletion whenever a noun is followed by an agreeing modifier other than a numeral (see Hyman 1979). 13. In Hyman (1985) this is accounted for by requiring a proper-government chain from the affected determiner “slot” up to the top node of the sentence. 14. This section represents a summary of parts of several detailed studies on the interaction of tone with syntax and with the phonological and grammatical properties of the augment in lu-Ganda. See especially Hyman, Katamba & Walusimbi (1987), Hyman (1988), Hyman and Katamba (1990) and Hyman and Katamba (1993b). 15. In fact, the augment reconstructs as *H in Proto-Bantu. Cf. notes 5 and 11. 16. Odden (1991) suggests that a case of direct semantic conditioning of tone exists in Kikongo, though a syntactic interpretation appears possible here too.

References Anderson, S.C. 1979. “Verb structure”. In L. M. Hyman (ed.), Aghem Grammatical Structure, 73–136. SCOPIL 7, 73–136. Los Angeles: USC. Ashton, E. O. et al. 1954. A Luganda Grammar. London: Longmans, Green & Co. Bokamba, G. D. 1971. “Specificity and definiteness in Dzamba”. St. in Afr. Ling. 2.217–237. Botne, R. 1999. “Future and distal -ka-’s: Proto-Bantu or nascent form(s)?” In J. M. Hombert & L. M. Hyman (eds), Bantu Historical Linguistics: Theoretical and Empirical Perspectives. Stanford: C.S.L.I. Bresnan, J., and S. A. Mchombo. 1987. “Topic, pronoun and agreement in Chichewa”. Language 63.741–782.

176

LARRY M. HYMAN

Byarushengo, E. R., L. M. Hyman, and S. Tenenbaum. 1976. “Tone, Accent and Assertion in Haya.” In L. M. Hyman (ed.), Studies in Bantu Tonology, 183–206. SCOPL 3. Los Angeles: USC. Carter, H. 1962. Notes on the tonal system of Northern Rhodesian Plateau Tonga. London. Carter, H. 1963. “Coding, style and the initial vowel in Northern Rhodesian Tonga: A psycholinguistic study”. Afr. Lang. Stud. 4.1–42. Chebanne, A.M., D. Creissels et H.W. Nkhwa. 1996. Tonal Morphology of the Setswana Verb. Grenoble: Université Stendahl. Creissels, D. 1996. “La tonalité des finales verbales et la distinction entre formes verbales conjointes et formes verbales disjointes en tswana”. Africana Linguistica XI, 27–47. Tervuren: Musée Royal de l’Afrique Centrale. De Blois, K. F. 1970. “The augment in the Bantu languages”. Africana Linguistica IV, 85–165. Tervuren: Musée Royal de l’Afrique Centrale. Dewees, J. W. 1971. The Role of Syntax in the Occurrence of the Initial Vowel in Luganda and some other Bantu Languages. Ph.D. dissertation. Univ. of Wisconsin. Givón, T. 1971. “Studies in ChiBemba and Bantu Grammar.” St. in Afr. Lin., Supplement 3. Hyman, L. M. 1979. “Phonology and noun structure”. In L. M. Hyman (ed.), Aghem Grammatical Structure, 1–72. SCOPIL 7. Los Angeles: USC. Hyman, L. M. 1985. “Dependency relations in syntax: the mysterious case of the empty determiner in Aghem”. St. in Afr. Ling., Supplement 9, 151–156. Hyman, L. M. 1988. “Direct vs. indirect syntactic conditioning of phonological rules”. Proceedings of E.S.C.O.L. 1987. Ohio State Univ. 147–163. Hyman, L. M. and E.R. Byarushengo. 1984. “A model of Haya tonology”. In G.N. Clements and J. Goldsmith (eds), Autosegmental Studies in Bantu Tone, 53–103. Dordrecht: Foris. Hyman, L. M. and F.X. Katamba. 1990. “The augment in Luganda tonology”. J. of Afr. Lang. & Ling. 12.1–45. Hyman, L. M. and F. X. Katamba. 1993a. “The augment in Luganda: syntax or pragmatics?”. In S. Mchombo (ed.), Theoretical aspects of Bantu Grammar, 209–256. Stanford: C.S.L.I. Hyman, L. M. and F. X. Katamba. 1993b. “A new approach to tone in Luganda”. Language 69.34–67. Hyman, L. M., F.X. Katamba and L. Walusimbi. 1987. “Luganda and the strict layer hypothesis”. Phonology Yearbook 4.87–108.


177

Hyman, L. M. et J. Watters. 1984. “Auxiliary focus.” St. in Afr. Ling. 15. 233–273. Kanerva, J. 1989. Focus and phrasing in Chichewa Phonology. Ph.D. Dissertation, Stanford Univ. Meeussen, A.E. 1959. Essai de Grammaire Rundi. Tervuren: Annales du Musée Royal du Congo Belge. Meeussen, A. E. 1963. “Morphophonology of the Tonga verb”. J. of Afr. lang, 2.72–92. Mould, M. 1974. “The syntax and semantics of the initial vowel in Luganda”. In E. Voeltz (ed.), Proceedings of the 3rd Annual Conference on African Linguistics, 223–229. Bloomington: Indiana University Press. Mould, M. 1979. “Reconstructing Bantu tense systems”. Paper presented at the Tenth Annual Conference on African Linguistics, University of Illinois, Urbana, April 5–7, 1979. Nespor, M. & I. Vogel. 1986. Prosodic Phonology. Dordrecht: Foris. Nurse, D. In press. “Tense and aspect in Lacustrine Bantu languages”. In J. M. Hombert & L. M. Hyman (eds), Recent Advances in Bantu Historical Linguistics. Stanford: C.S.L.I. Odden, D. 1991. “The intersection of syntax, semantics and phonology in Kikongo”. In K. Hubbard (ed.), Special Session on African Languages, 188–199. Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society. Selkirk, E.O. 1980. “Prosodic domains in phonology: Sanskrit revisited”. In M. Aronoff and M-L. Kean (eds), Juncture, 107–129. Saratoga: Anma Libri. Sharman, J.C. 1955. “The tabulation of tenses in a Bantu language (Bemba: Northern Rhodesia)”. Africa 25.393–404. Takizala, A. 1973. “Focus and relativization: the case of Kihung’an”. In J.P. Kimball (ed.), Syntax and Semantics 2, 123–148. New York: Academic Press. Watters, J.R. 1979. “Focus in Aghem: a study of its formal correlates and typology”. In L.M. Hyman (ed.), Aghem Grammatical Structure, 137–197. SCOPL 7. Los Angeles: USC.

The Syntax of the P-Focus Position in Turkish* Sarah D. Kennelly Utrecht Institute of Linguistics OTS

Abstract In Turkish (an SOV language) the discourse function of an element is mapped onto a linear representation; hence we see the emergence of a discourse–function interface. Focused elements are analyzed here as derived predicates (drawing on Herburger 1995) which require contiguity with the verb, motivating their immediately preverbal position. A sub-class of Focused elements, call it P-Focus, only have a cumulative reading with respect to the subject, that is no scope distinction whatsoever. This natural class minimally includes nonspecific object DPs, defined here as the existential quantification of a free-choice partial function, and interrogative wh-exps, a nonspecific plus a Q feature. P-Focus elements are then existential quantifiers and therefore undergo movement to an adjoined position, in Turkish in the overt structure. Call that position ‘F’. It is shown in this paper that in Turkish the verb is in I0 and that the subject remains in Spec,VP such that ‘F’ can be analyzed as rightmost adjoined to VP, maintaining an m-command relation with the subject, resulting in no scope effects. Thus the discourse interface (here Focus) and principles of interpretation (quantification) interact with the derivational economy principles in mapping sentential elements onto the syntactic structure.

Introduction Turkish is a discourse configurational language where the discourse function of an element is mapped onto a linear representation; thus we see the emergence of a discourse function interface. This paper will show that the discourse interface (here Focus) and principles of interpretation (interrogative and quantification)

180

SARAH D. KENNELLY

interact with the derivational economy principles in mapping elements onto the syntactic structure. Turkish is an SOV head final language1 where a Focused element, roughly ‘new information’, appears immediately preverbally (Erkü 1982). Drawing on Herburger (1995), Focused elements are analyzed here as derived predicates, which require contiguity with the verb. Though the subject normally takes wide scope over arguments to its right, a sub-class of Focused elements, call it P-Focus, show an absence of scope construals with respect to the subject. This class minimally includes nonspecific object DPs (NODP),2 defined here as the existential quantification of a free-choice partial function whose domain is the set described by the NP, and interrogative wh-exps (wh-Q),3 a nonspecific plus a Q feature. The NODP is morphologically explicit in Turkish in the absence of the usual Accusative Case marker, indicated by . P-Focus elements are then existential quantifiers and hence undergo movement to an adjoined position, in Turkish in the overt structure. In order to derive their predicate role P-Focus elements must also be contiguous with the verb, while the verb is in I0. Call the position(s) that host(s) these elements ‘F’. The standard form of adjunction under current linguistic theory is Quantifier Raising (QR) at the level of Logical Form (LF). Under QR, quantified NPs raise to adjoin to IP or VP (May 1985) to be interpreted. Thus movement to adjunction is an alternative in Universal Grammar to the feature-checking system of Chomsky (1995: 377). Since the subject remains in Spec,VP then ‘F’ can be analyzed as rightmost adjoined to VP, sketched in (1). Assuming May’s (1985) analysis of segments, ‘F’ and the subject mutually m-command (Chomsky 1986: 8) each other, accounting for the absence of scope. (1)

IP Spec

I’ VP

VP Spec Subject

I° Verb j ‘F’

V°tj

THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH

181

Section 1 considers the scope construals and distribution of wh-Qs and NODPs, motivating ‘F’ and P-Focus (‘P’ for presentational). Then the linear component in terms of verb contiguity is considered in Section 2. ‘F’ as adjunction to VP is supported in Section 3 while the VP internal subject is motivated in Section 4. Section 5 examines the adjunction analysis in the iteration of ‘F’, the scope construals of quantified VP adverbs, and the inability to extract from ‘F’. Section 6 addresses the problematic order of P-Focus elements, discounting Case Licensing and Incorporation for the adjacency constraint on NODPs. Section 7 summarizes the issues.

1.

Motivation for Position ‘F’ and P-Focus

In Turkish there is no covert QR such that the overt structure is the LF for quantification.4 (2)

Genç bir doktor her hastayı tedavi etti.5 young a doctor every patient- treated ‘A young doctor treated everyone.’ [adapted from Göksel 1995] [there is one young doctor such that s/he treated each of the individuals] *[for each of the individuals, there is a young doctor such that s/he treated that individual]

In (2) there is unambiguously only one doctor and the immediately preverbal argument has no scope with respect to the subject. That is the universal quantifier does not have a quantificational interpretation with a distributed reading of the preceding argument. A discussion of (2) in terms of its role as a Focus structure lies beyond the scope of this paper. In sharp contrast, wh-Qs and NODPs obligatorily appear left-verb adjacent and demonstrate no scope construals wrt the subject, indicating that they reside in a mutual m-command relation, which I propose results from their position in ‘F’. (3)

a.

b.

(Bu) Üç çocuk kimleri gördü?6 this 3 child who-- saw ‘Who did the(se) 3 children see?’ [who are the individuals such that 3 children saw them] [for each of 3 children who did they see] Deniz’i ve Ufuk’u Deniz- and Ufuk-Acc [as an answer to (3a)]

182

SARAH D. KENNELLY

c.

Deniz Ufuk’u, Ug˘ur Ümit’i ve Remzi’de Ahmet’i D. Ufuk- U. Ümit- and Remzi-also Ahmet- gördü. saw ‘Deniz saw Ufuk, Ug˘ur saw Ümit and Remzi saw Ahmet.’ [as an answer to (3a)]

Scope construals of interrogatives are disambiguated in their possible answers. In response to (3a) either (3b) or the pair-list response in (3c) is a felicitous response, indicating an absence of scope between the subject and the wh-Q in Focus position. The preverbal wh-exp that is not in the Focus position is translated as a cleft construction in (4a) and is strongly marked in Turkish, even ungrammatical for some speakers. (4)

a. ?*Kimi üç çocuk gördü? who- 3 child saw [a simple request for information] ‘Who is it that the 3 children saw?’ b. *Üç çocuk gördü kimi? [totally unacceptable]

The immediately preverbal wh-Q may take prosodic prominence or marked contrastive stress, whereas in other positions it obligatorily takes marked contrastive stress. Cheng (1991: 19,58) discusses cross-linguistic data that support an analysis of base-generated sentence initial wh-exps which result in a wh-cleft construction as well as the scrambling or ‘Topicalization’ of wh-exps to an initial position. These are distinct from wh-fronting to obtain an interrogative interpretation. Cheng’s analysis is in keeping with Horvath’s (1986: 118) proposal for UG that an interrogative interpretation can be derived only if the wh-Q Op bears the feature Focus, which in Turkish is obtained in the left-verb adjacent position. The wh-exp in the postverbal position in (4b), reserved for backgrounded information which remains unchanged as long as the context is constant, is infelicitous.7 Thus it is clear that in Turkish the linear position of an element plays an integral role in its interpretation and that there is a left-verb adjacency constraint on straightforward wh-Qs. Turkish is also a ‘scrambling’ language (Kural 1992) in that the arguments may appear in any order, though any order other than SOV is marked.8 Consider the hypothesis that the wh-Q in (3) is in its base generated position. Then we wouldn’t expect to find adverbials base generated in the same position. In contrast with argument wh-Qs in English, adverbial wh-Qs are ungrammatical in situ.


(5)

183

a. *Who arrived why? b. Why did who arrive?

The conclusion is that they obligatorily undergo overt wh-mvt. to a position that takes scope over the sentence. In Turkish they obligatorily appear in the immediately preverbal position. (6)

a.

Murat nereye gitti? to.where went ‘Where did Murat go?’ b. *Nereye Murat gitti? to.where went ‘Where did Murat go?’

[Erguvanlı 1984: 35 (92)]

Horvath’s proposal together with (3) and (6) support the hypothesis that a wh-Q in ‘F’ a) undergoes movement and b) is in a mutual m-command relation with the subject. NODPs, like wh-Qs, occur left-verb adjacent where they may show wide scope effects wrt the subject: (7)

Üç çocuk/Çocuklardan üçü yeni bir araba-@ almıs¸.9 3 child/of.the.children 3 new a car-@ bought ‘Three (of the) children bought a new car.’ [a new car is such that 3 (of the) children bought it] [*each of 3 (of the) children bought a new car] b. *Yeni bir araba-@ üç çocuk/çocuklardan üçü almıs¸. c. #Üç çocuk/Çocuklardan üçü taze bir çekirdek-@ yemis¸. 3 child/of.the.children 3 fresh a sunflower.seed-@ ate ‘Three (of the) children ate a fresh sunflower seed.’ [a fresh sunflower seed is such that 3 (of the) children ate it] [*each of 3 (of the) children ate a fresh sunflower seed]

a.

Unlike its English counterpart, (7a) is unambiguous: there is only one car, indicating a wide scope construal of the NODP. (3c) has demonstrated that the subject ‘3 NP’ does not have an intrinsically collective interpretation. However if the subject cannot be treated collectively, as in (7c) where it is impossible for 3 (of the) kids to eat a unique sunflower seed, the sentence is anomalous. (7b) indicates that NODPs are obligatorily left-verb adjacent in the written language, patterning with wh-Qs. Furthermore, the absence of ambiguity in (7) supports the proposal that there is no covert QR at LF in Turkish.

184

SARAH D. KENNELLY

The wide scope construal in (7) is in fact a cumulative interpretation (Scha 1984) of both arguments, resulting from an absence of scope between the subject and the NODP, clarified in (8): (8)

Üç kiz üç sepet-@ kaldFrmis¸. 3 girl 3 basket-@ lifted. ‘3 girls lifted 3 (nonspecific) baskets.’ [Cumulative reading: : x lifts y, |x| = 3, |y| = 3]

The only available reading in (8) is cumulative whereby it is not at all clear how many of the girls lifted which of the baskets. Informants offered that one possible interpretation is that there is a one-to-one relation between the three girls and the three baskets, positing that cumulativity subsumes a constrained form of distribution, i.e. that ‘det NP’ cannot be regarded as a lattice sum of individuals under cardinality. The scope facts are not ambiguous: the distributed reading which would result in either 9 baskets or 9 girls does not obtain. In conversation NODPs may occur postverbally, where verb adjacency isn’t required but a determiner is. (9)

Bu adamlardan kaçı görmüs¸ dün *(bir) film-@? these men.of how.many saw yesterday *(a film-@ ‘Of these men how many saw a film yesterday?’ [Göksel 1995: 21 #27a; restricted to conversation]

The backgrounded NODP10 is highly marked in Turkish, while the left-verb adjacent Focus position is unmarked, as in (7)/(8). (9) clearly demonstrates that Case Licensing is irrelevant to the left-verb adjacency constraint for NODPs, thereby supporting de Hoop’s (1992) analysis of Weak Case for NODPs in Turkish. The data in (2)–(9) have prompted the proposal of a position ‘F’ which is rightmost adjoined to VP, as shown in (1), and which minimally hosts wh-Qs and NODPs. ‘F’ hosts the Accusative wh-Q-exp (3a), the Dative wh-Q-exp (6a) and the ‘Weak’ Case for NODPs (7)/(8). Surely all these elements are not base generated in the same position; thus there must be movement to ‘F’. The absence of scope of the Focused wh-Q and NODP wrt the subject stems from the interaction of linearity and hierarchical structure in relation to quantificational interpretation.


2.

185

The Linear Component

In Turkish there is evidence that the LF for quantification is the overt structure. Since it is a head final language, the verb moves to the right under head movement to check the strong inflectional morphology. If one issue for ‘F’ is contiguity with the verb, then in order to move to an adjoined position, the Focused element would also have to move to the right. So it is crucial to ascertain if verb adjacency is at stake. Cross-linguistically the Wh/Focus position is consistently verb adjacent: Hungarian (Szabolcsi 1981: 143; Horvath 1986; Brody 1990), Aghem (Rochemont 1986: 19), Georgian (Nash 1995) and Basque (Rebuschi 1983; Ortiz de Urbina this vol). According to Chomsky (1971: 199) “…the Focus is the predicate of the dominant proposition of the deep structure”. In the following discussion it is assumed that the verb (predicate adjective/noun/PP)11 is the natural predicate of an assertion, and hence, in Chomsky’s terms, that it is the natural or ‘neutral’ Focus. When an element other than the verb is Focused, Chomsky doesn’t mention how it becomes the predicate; what mechanism is at work. Nor does he elaborate on what the status of the verb is when an argument is Focused, i.e. is the verb then a ‘secondary’ predicate, is it relieved of its status as the predicate, or is there a verb-Focus predicate complex? I propose that the motivation for (i) the verb adjacency constraint and (ii) the adjoined position ‘F’ are distinct: (i) the Focused element needs to derive the role of predicate from the verb to be interpreted as ‘Focus’; call it ‘predicate derivation’. That is, given that the natural predicate of a sentence is the verb, if another element is to become the predicate it must derive that role from the verb; in Turkish the crucial factor is contiguity. (ii) P-Focus elements are existential quantifiers which require movement to an adjoined position for interpretation. Consider the adjacency constraint. A sentence can be taken as a description of an event such that the verb is not a direct predicate of the individuals denoted by the subject but rather it is a predicate of an event while the subject is a participant in that event (Davidson 1967; Parsons 1990). Then the arguments have a relation to the event through their theta-roles, as exemplified in (10). (10)

∃(e)[Lee(e,agent) & hugged(e) & Kim(e,theme)] ‘There was a hugging by Lee of Kim.’

Herburger (1995) notes that this representation of the structure of a sentence may include the restriction of the existential quantification over events, and conse-

186

SARAH D. KENNELLY

quently the Restrictor of the Event Operator is distinguished from the Matrix. I am assuming that in a ‘neutral’ sentence the Matrix of the Event Operator is the verb, the natural predicate. (11)

∃(e)[Restrictor[Lee(e,agent) & Kim(e,theme)]Matrix[hugged(e)]]

Herburger then proposes that Focus structures the quantification over events such that the Focused element is the Matrix of the Event Operator. If we link her proposal with Chomsky’s then we have (12). (12)

P: The predicate of a sentence is the Matrix of the Event Operator.

If the theme is a NODP, which is P-Focused unless marked as backgrounded, we get a representation of the structure of the sentence where the Matrix of the Event Operator is now the theme and the verb constitutes part of the restrictor: (13)

∃(e)[R[Lee(e,agent) & hugged(e)]M[a child(e,theme)]] [first attempt]

Perhaps a more precise way of analyzing the step from (11) to (13) (D. Lebeaux p.c.) is to identify it as a form of raising of the Focus element, leaving the original Matrix as the matrix of the Restrictor, shown in (14) and (14′). This is what occurs in the syntax under adjunction to obtain interpretation by the PFocus element(s). Note that this is not a one-to-one mapping onto the syntactic structure since the adjunction site there remains verb internal. (14)

∃(e)[R[R[Lee(e,agent)]M[hugged(e)]]M[a child(e,theme)]] [final form]

(14′)

New Restrictor

original restrictor

New Matrix

original matrix


187

It is the re-organization of the structure of the sentence that occurs under predicate derivation. That is, assuming that the natural Matrix of the Event Operator is the verb, the restructuring of the sentence to a Focus structure where the Focused element is the Matrix requires a process which I have termed ‘Predicate Derivation’. (15)

P D: Predicate Derivation restructures the sentence such that the role of Matrix of the Event Operator is shifted from the original predicate to another element.

Focus is then a constraint on predicate derivation whereby the Focus element is the derived predicate while the verb, in the restrictor, is the event variable in the sentence, in Parson’s terms the predicate of the event. In Turkish, as in many languages, there is a syntactic constraint on predicate derivation that requires contiguity at Spell-Out. (16)

C  P D  T Predicate Derivation occurs from a predicate head H to an element in position P iff a. H c-commands the element in P; b. the relevant branches are both either left or right-branching; and c. there are no intervening elements.

Notice that the stricter form of government, c-command, is used here rather than the m-command relation that is relevant for the subject/P-Focus element. The Focus elements that of themselves occupy ‘F’ are labeled P-Focus, a natural class. (17)

P-F —  N C [following Cheng 1991]: P-Focus is the set of elements that minimally contains wh-Qs and Nonspecific Object DPs.

Drawing on the work of Hamblin (1973), Karttunen (1977), Cheng (1991), Rebuschi (1994), Kratzer (1995), and Reinhart (1995), I propose that the wh-Q is the existential quantification of a free-choice partial function where the range is a subset of the domain. The domain is defined by the wh-Q itself (who operates on the domain of people) and by the context, while the range argument remains undefined due to the Q feature. Cross-linguistically, the nonspecific is frequently either homophonous with the wh-Q or has one additional morpheme

188

SARAH D. KENNELLY

(Cheng 1991: 80). This has led Cheng to propose that the wh-Q is itself without quantificational force (p.84). Another way to capture that idea is my proposal that the additional morpheme found on the nonspecific is the representation of the range argument of the free-choice function while the Q feature of the wh-Q precludes the realization of that argument. That is, there is a choice function and there is a domain but the mapping onto a range argument is not operative. The analysis of wh-Qs as predicates patterns with Pearson’s (1996) analysis of Malagasy. Furthermore, in Japanese there is an optional scope taking particle that occurs in the presence of the in situ wh-Q. Watanabe (1992: 19) has proposed that the particle is right adjoined to VP, which is consistent with this analysis of ‘F’.

3.

The Position ‘F’

Since ‘F’ is left-verb adjacent, it is crucial to establish the position of the verb. Following Chomsky (1986: 6) I assume that all adjunction is to XP rather than X′, and that all adjuncts are base generated in an adjoined position and remain in that position (Pollock 1989). Hence they are crucial in disambiguating verb movement and the position of arguments. Sentential adverbs adjoin to IP while VP adverbs adjoin to VP (Jackendoff 1972: 106). In Turkish, sentential adverbs seen in (18) occur postverbally without a pause. A pause would lead to an analysis in which the postverbal elements were extraposed. Its absence indicates the reverse. (18) then supports right-adjunction. (18)

Deniz filmi çevirdi galiba /bile /zaten. film- turned probably/even/anyway ‘Deniz probably/even made the film.’ ‘Deniz made the film anyway.’

In contrast, VP adverbs may not occur postverbally without a pause, indicating that the verb is higher than VP. (19) *Deniz filmi çevirdi hızla /aceleyle. film- turned quickly/urgently ‘Deniz made the film quickly/urgently.’ Distinct morphology for number and person as well as aspect/tense, termed ‘strong’ morphology, has been cited in cross-linguistic studies (Pollock 1989;


189

Vikner 1991: 134) as the driving element for verb movement to I0. Turkish strong inflection then provides the motivation for the head movement of the verb to I0. There is no evidence of a CP category in Turkish since there is no lexical complementizer in the language, subordination is in terms of DP (Kennelly 1990) and there is no wh-fronting in the usual sense, as seen in (4)/(6).12 Consequently it is assumed that at Spell-Out the verb resides in I0 rather than higher up. This analysis is in the spirit of Chomsky (1995) whereby a derivation is structure building such that the only structure that exists for a given sentence is that necessary for its derivation, and also in the spirit of Grimshaw’s (1997) claim that there are no useless projections. The preverbal position for the NODP is an A-bar position, which cannot act as an A-binder in a binding relation but is interpreted in situ for A-binding purposes, as seen in (20). (20) *Kendi bir hasta-@ öldürmüs¸. self a patient-@ killed ‘Self killed a patient.’ When it occurs preverbally the NODP occupies an A-bar position that is leftadjacent to the verb in I0, providing further support for the adjunction analysis for ‘F’. It is claimed here that movement occurs for interpretation. In addition, contiguity with the verb of Focused elements is imposed to obtain predicate derivation, forcing rightward movement of the P-Focused elements in a head final language. The condition on interpretation and the overtness of the logical structure feed the discourse interface in the mapping of P-Focus elements onto the syntactic structure.

4.

The Subject in Spec,VP

The distinction between final sentential adverbs and ungrammatical VP adverbs in (18)/(19) indicates that the verb has undergone head movement to the category above VP, I0, where it governs the arguments internal to VP. It is assumed that all arguments, including the subject, are base-generated internal to VP. This analysis also claims that the subject remains in its base generated position in Spec,VP where it is governed by the verb. The proposal that all arguments may remain within VP has been made by Jackendoff (1977) for English, as well as by Brody (1990) and Kiss (1994) for Hungarian; Haider (1985) and Diesing

190

SARAH D. KENNELLY

(1992) for German; de Hoop (1992) for Dutch; Costa (1996) for Portuguese and Pinto (1994) (among others) for Italian. Traditionally Nominative Case is checked in Spec,IP by Tense within a Spec-Head checking relation. Koopman and Sportiche (1988: 17) propose that while Tensed INFL assigns (checks) Nominative Case in English by agreement in a Spec-Head configuration, forcing the external argument DP to raise to Spec,IP, Tensed INFL in Arabic, Irish and Welsh can assign Case structurally, under government, permitting lexical DPs to surface in Spec,VP iff the verb is in I0. Since the Turkish verb is in I0, under the Koopman/Sportiche analysis, government licensing of the Nominative in Spec,VP is then possible. From I0 the verb governs Spec,VP. It is then predicted that a sentential subject should be transparent to extraction by relativization and this is attested. It is shown that if a quantificational determiner appears on the subject DP it then takes wide scope over the NODP, indicating a distinction between subjects in Spec,IP and those in Spec,VP. This is clearly an overt form of Quantifier Raising which follows Woolford’s (1994) Principle of Exclusion in that quantified subject DPs are excluded from a VP internal position. The position of subject DPs can best be explained in terms of quantification, as proposed by Diesing (1992: 4), with nonquantificational determiners in Spec,VP where they are governed by the verb. 4.1 Turkish subjects behave like complements rather than adjuncts In English sentential subjects behave like adjuncts in that they are islands to extraction, here relativization, while complements are not: (21) *The patienti whoi [[ti would eventually recover] is extremely doubtful] walked out of the hospital. (22)

The patienti whoi [John thought [ti would eventually recover]] walked out of the hospital.

The distinction in (21)/(22) can be attributed to the ECP which roughly states that a trace must be properly governed. P government is either (a) thetagovernment, government by a theta-marking element, or (b) antecedent government. The latter has the effect of imposing a locality condition (Chomsky 1986). Thus objects are transparent to long distance extraction while subjects are not and theta-government is the crucial difference. A subject in Spec,VP is also theta-governed and hence it behaves as an object rather than as an adjunct. The mechanism that is used to describe govern-


191

ment of a Specifier position is grounded in the idea of L- (Chomsky 1986). The effect of L- is transparency to extraction. a L-marks b iff a is a lexical category that theta-governs b and they are in a sisterhood configuration. Chomsky continues (p. 24) that if a L-marks b, then it also L-marks the Specifier of b. Although I0 cannot L-mark VP because INFL is not lexical, once the verb has moved to I0 the V+I complex can. The result is that Spec,VP is Lmarked while Spec,IP is not. Koopman and Sportiche (1988) make use of this analysis to propose that a sentential subject that is not an island to long distance extraction must then be in Spec,VP. In English the subject obligatorily moves to Spec,IP where it is an island; i.e. it behaves like an adjunct, as in (21). Sentential subjects are not islands in Turkish, as seen below. Under the above analysis the sentential subject must then be located in Spec,VP where it is L-marked by the verb in I0. (23)

Relativization out of a sentential subject: [ex. from Sezer 1982: 2]. [[ti ˙Iyiles¸eceg˘i] son-derece s¸üpheli olan] hastai hastaneden will.recover extremely doubtful being patient from.hospital yürüyerek çıktı. on.foot emerged ‘The patient, whose (will be) recovering is extremely doubtful, walked out of the hospital.’

(23′)

VP

Spec,VP DP

VP

Adv VP hastahaneden (from hosp) V° Adv Spec,VP VP VP yürüyerek (on foot) (emerged) DP V° Adv Spec V° hasta supehli olan son derece ¸ Iyilesecegi ¸ (will recover) (extremely) (doubtful being) (patient) (

VP

(23) is the Turkish equivalent of (21) and it is grammatical. It has been argued that Turkish relativization does not involve movement. If that were the case then

192

SARAH D. KENNELLY

there should also be relativization out of an adjunct clause. This is not attested, as seen in (24). (24) *[[[Ays¸e ti kırdıg˘ı] için] Ahmet’in bag˘ırdıg˘ı] bardaki… [[[A. broken for Ahmet’s yelled glass ‘the glass that Ahmet yelled because Ays¸e broke (it)…’ [Kural 1994: 5 ex. 6] Thus the relativization out of the sentential subject indicates that the subject is governed by the verb and hence that it is in Spec,VP. A similar proposal has been put forth by Haider (1985) for German where extraction is possible out of sentential subjects. He concludes that the subject in German remains within VP unless it is scrambled. 4.2 Contrast with Quantified Subjects in Spec,IP When a necessarily quantificational determiner appears on the subject DP of the matrix clause the result is a wide scope construal with respect to the NODP. (25)

Herkes yeni bir araba-@ almıs¸.13 everyone new a car-@ bought ‘Everyone bought a new car’ *[a new car is such that everyone bought it] [each one bought a new car]

In English existential sentences are distinguished by the fact that the logical subject remains internal to VP due to the fact that there occupies Spec,IP. Thus the determiners that cannot appear on the logical subject of existential sentences in English are those that cannot remain internal to VP. It is only her ‘each’ that is necessarily quantificational in subject position in Turkish where it takes a wide scope construal over the NODP, as seen in (25). It then adheres to Woolford’s (1994) Principle of Exclusion from VP, based on quantification as suggested by Diesing (1992: 4), and undergoes an overt form of Quantifier Raising in subject position. In this case it does not adjoin to IP, as in classical QR. I propose that it undergoes Spec-to-Spec movement from Spec,VP to Spec,IP under a quantificational interpretation. The distinction between the scope construals in (7)/(8) and (25) provides striking evidence that the subjects in (7)/(8) remain in Spec,VP where they are in a mutual m-command relation with the NODP, as predicted by the proposal. The analysis presented here is that raising to adjunction occurs to


193

obtain the interpretation of the P-Focus existential quantifiers. At the same time QR is analyzed here as overt Spec-to-Spec movement. Clearly the use of the adjunction strategy for rightward Focus movement does not preclude overt QR to a Spec position. (25) is also further evidence that the NODP may be distributed. To recap what has been brought to light so far: The LF and the discourse function are read off the overt structure. The requirement that Focused elements be verb adjacent is motivated by predicate derivation. The verb is in I0. P-Focus, that is the minimal set of wh-Qs and NODPs, are left-verb adjacent and have no scope relation with respect to the subject in Spec, VP. They require an adjoined position to obtain the interpretation of existential quantification as well as contiguity with the verb.

5.

Arguments for Adjunction

There are four arguments in favor of the analysis that ‘F’ is right-most adjoined to VP and one against it. In favor are: i) Multiple occupants of ‘F’ have ambiguous scope; ii) VP adverbs have ambiguous scope readings wrt elements in ‘F’; iii) Extraction out of ‘F’ is ungrammatical; iv) Movement to adjunction obviates the need for feature checking. The argument against adjunction is that it is difficult to explain the strict order of wh-Qs wrt NODPs in ‘F’ in terms of adjunction. 5.1 Iteration of ‘F’ The first argument in favor of the adjunction analysis is grounded in the theory of segments. Chomsky (1986: 7) adopted May’s (1985: 34) analysis for adjunction structures: two adjoined elements mutually c-command each other and hence govern each other iff there is no maximal projection boundary between the two, where a Xmax boundary crucially dominates all segments of XP. Consequently, adjoined elements to XP are free to take any type of relative scope relation. In Turkish, multiple wh-Qs line up adjacent to one another with the rightmost left-verb adjacent. Their order is reversible and they have ambiguous scope construals with respect to each other — that is, whatever the order of wh-Qs, the same question is being asked. This is taken to indicate that they are all in ‘F’.

194

SARAH D. KENNELLY

(26)

a.

b.

(27)

a.

b. c.

Bu s¸ehirde kim kimi arıyordu? in.this.city who- who- was.seeking ‘Who was looking for who in this city?’ Bu s¸ehirde kimi kim arıyordu? in.this.city who- who- was.seeking ‘Who was looking for who in this city?’ Can Deniz’i arıyordu; Ufuk Ümit’i arıyordu. C. Deniz- was.seeking U. Ümit- was.seeking ‘Can was looking for Deniz; Ufuk was looking for Ümit….’ [as an answer to (26a/b)] Deniz’i Can arıyordu; Ümit’i Ufuk arıyordu. [as an answer to (26a/b)] Can ve Ufuk Deniz’i arıyordu. [as an answer to (26a/b)] C. and U. Deniz- was.seeking ‘Can and Ufuk were looking for Deniz.’

(27) (a) (b) and (c) are all felicitous answers to both (26a/b), indicating that the scope construals of the wh-Qs are ambiguous and that the order of the wh-Qs may be changed with no effect on the possible readings since both (26a) and (26b) ask the same question. The only difference lies in which reading is more salient; in (26b) it is more probable that only one person is sought. ‘F’ is then a term used to indicate a series of adjoined positions. Parallel to multiple wh-Qs, a wh-Q and NODP may co-occur in ‘F’ with ambiguous scope construals. The prosody of the wh-Q in (28) patterns with the wh-Qs that are immediately preverbal, taking either prosodic prominence or strong, contrastive stress. (28)

a.

b.

Ufuk kimlere güzel bir hediye-@ vermis¸? U. to.whom-. beautiful a present-@ gave ‘Who-all did Ufuk give a beautiful present to?’ [a beautiful present x [who-pl [Ufuk gave x to who]]] [who-pl [a beautiful present x [Ufuk gave x to who]]] [Can ve Deniz]’e güzel bir hediye-@ vermis¸. C. and Deniz- beautiful a present-@ gave ‘S/he gave a beautiful present to Can and Deniz (together).’ [as an answer to (28a)]


c.

d.

195

Can’a ve Deniz’e güzel bir hediye-@ vermis¸. Can- and Deniz- beautiful a present-@ gave ‘S/he gave a beautiful present to Can and Deniz (together or separately).’ [as an answer to (28a)] Can’a ve Deniz’e birer hediye-@ vermis¸. Can- and Deniz- one.each present-@ gave ‘S/he gave a present to Can and s/he gave a present to Deniz.’ [as an answer to (28a)]

In answer to the wh-Question in (28a), which includes both a wh-Q and a NODP, the response may be either (28b) or the ambiguous (28c), as well as with the explicitly distributed (28d). That is, there may be a unique present given to all concerned collectively in (28b) or (28c), or there may be a distinct present given to each person in (28c) or (28d), indicating that a wh-Q and NODP have ambiguous scope. The ambiguity in (28), as with (25), also indicates that the NODP with bir may have a distributed interpretation; that is it is not a necessary characteristic of bir to invoke a unique interpretation. Furthermore if there were two distinct projections for the wh-Q and the NODP ambiguity would not occur. The fact that there is ambiguity supports an analysis in terms of multiple adjunction. In support of the proposal that the wh-Q and the NODP are both in ‘F’ while the other (unscrambled) arguments are not, under co-occurrence neither is distributive wrt the subject. Taken together with the previous data this is further support for a position ‘F’, whose every segment is in a mutual m-command relation with the subject, resulting in a cumulative reading for all the arguments. (29)

a.

Üç çocuk kime güzel bir hediye-@ vermis¸? 3 child who- beautiful a present-@ gave ‘Who did the three children give a beautiful present to?’ [for a beautiful present x who is the individual that 3 children gave x to] [who is the individual such that for a beautiful present x, 3 children gave x to them] *[3 children are such that who gave a beautiful present to each of them] *[who is the single person such that for 3 children they gave a beautiful present to each of them] *[for one beautiful present for each of 3 children who is the

196

SARAH D. KENNELLY

individual who gave it to each of them] Denize vermis¸ler. Deniz- they.gave ‘They gave (it) to Deniz.’ [as an answer to (29a) — Turkish is an object drop lang.]] c. #Can Deniz’e, Ug˘ur Ays¸e’ye, ve Ufuk Ümit’e C. Deniz- U. Ays¸e- and U. Ümit- birer hediye-@ vermis¸. [as an answer to (29a)] each.one present-@ gave ‘Can gave a present to Deniz, Ug˘ur to Ays¸e, and Ufuk to Ümit.’ d. #Can Deniz’e, Ug˘ur Ays¸e’ye, ve Ufuk Ümit’e bir C. Deniz- U. Ays¸e- and U. Ümit- a hediye-@ vermis¸. [as an answer to (29a)] present-@ gave ‘There was one present such that Can gave it to Deniz, Ug˘ur to Ays¸e, and Ufuk to Ümit.’ ‘Can gave one present to Deniz, Ug˘ur gave one to Ays¸e, and Ufuk one to Ümit.’ e. #Deniz’e üç hediye-@ vermis¸ler. [as an answer to (29a)] Deniz- 3 present-@ they.gave ‘They gave 3 presents to Deniz.’ b.

The only appropriate answer to (29a) is seen in (29b) with a unique recipient of a unique present. Adjunction to VP doesn’t interfere with head movement of the verb due to Relativized Minimality (Rizzi 1990): movement of X0 is blind to adjunction since that is not a potential landing site for X0. This is also the case with VP adverbs in French which presumably left adjoin to VP, yet the verb moves into I0. The segment analysis of adjunction readily accounts for the iterative property of ‘F’ with multiple occupants. 5.2 VP Adverbs — Ambiguous Scope The second argument for the proposal is the fact that VP adverbs take ambiguous scope with respect to the elements in ‘F’, and not with other arguments. Suppose there were an ‘F’P. Though the Specifier positions are to the left in Turkish, there is nothing in X-Bar theory that requires a consistent directionality. However


197

the Head Parameter does. If there were a righthand Spec position to host wh-Qs and NODPs, say Spec, ‘F’P, then multiple occupants due to movement would not govern their respective traces, resulting in an illegitimate structure. The Focus Particles, discussed in Sec. VI might be proposed as ‘F’0, but under a righthand Spec analysis they should occur to the left of the elements in Spec, ‘F’. Instead they occur to the right of an element in ‘F’. Worse still they are not associated with one position; they may occur as a clitic on any element in the sentence (e.g. the Q particle), hence it will be proposed that they are incorporated. The only possible ‘F’0 would then be an abstract head. If there were an ‘F’P, one would expect occupants of the ‘F’P projection to take scope over VP adverbs. (30)

a.

?Bu

çocuk iki kere üç kitap-@ okumus¸. this child twice three book-@ read ‘This child read three books twice.’ [this child 2 times read 3 books — so 6 books — which entails: there were 3 books such that this child read each one twice — 3 books] b. ?Üç çocuk iki kere yeni bir kitap-@ okumus¸. 3 child twice new a book-@ read ‘Three children read a new book twice.’ *[twice three children read a new book — so 6 books] *[three children twice read a new book — so 6 books] *[three children read a new book twice — so 3 books since the same book is read twice] [twice a new book such that three children read it — so 2 books] [a new book such that twice three children read it — so 1 book]

198

SARAH D. KENNELLY

IP

(30′)

I’

Spec VP VP

Spec Subject

I° Verb j ‘F’ NODPk

VP

Adv

DPtk

V°tj

The sentences in (30) were judged odd by some informants, fine for others. However all were clear that the salient interpretation of the number of books in (30a) was six, with a less salient reading of three. And for (30b) there were either two or one. Then the wide scope construal of the NODP predicted by a ‘F’P projection is not attested in (30), where the NODP takes ambiguous scope with respect to a VP adverb. The VP adverb in (30a) is to the left of the object, yet has ambiguous scope with respect to that object. The object in turn is in a mutual m-command relation with the subject, seen in (7)/(8) and (30b). If the adverb were left adjoined to VP, with the subject in Spec,IP, it would be impossible for the object to demonstrate an absence of scope wrt the subject while remaining left of the verb. The fact that the subject does not take wide scope over the adverb in (30b) can be accounted for if the adverb is right adjoined to VP, where it c-commands the subject in Spec,VP. There is a distinction between elements  by all segments of XP and those dominated by some segments,  in XP. Thus an element in Spec,XP, here the subject in Spec,VP,  by VP, would be differentiated from adjoined elements which are simply  by VP. In addition, the fact that a VP adverb has ambiguous scope with elements in ‘F’ is further support for the proposal that the right-adjacent verb is in I0. The analysis of the NODP in ‘F’ seen in (1) is shown in (30′), with a parallel analysis for wh-Qs.


199

At this point we have seen c-command as crucial in predicate derivation, XP segment analysis, and adverbial scope. Why then should m-command hold for PFocus? This is an issue that will have to remain an open question for the moment. 5.3 Extraction from ‘F’ Since adjunction creates an island to extraction the prediction is that extraction out of a NODP should be ungrammatical. Though extraction out of a specific object is fine, extraction out of a NODP is infelicitous, providing the third piece of evidence in favor of the adjunction analysis. Since Focus movement is clausebound in Turkish extraction in the form of relativization is used. In Turkish there is relativization out of a Relative Clause (RC). If the nonspecific object is in fact in an adjoined position as proposed here then the prediction is that relativization out of a RC with a nonspecific head noun should pattern with relativization out of adjuncts seen in (24). That is it should be ungrammatical. Consider RCs with a ± specific head noun in object position. (31)

a.

b.

[tj Arabayı tamir eden] kadınıj s¸imdi gördüm. [tj car- repairing woman- now I.saw ‘I just saw the woman who repaired the car.’ [tj Arabayı tamir eden] bir kadınj-@ gördüm. [tj car- repairing a woman-@ I.saw ‘I saw a woman who repaired the car.’

Extraction out of the RC with a definite, and therefore specific head noun in object position in (31a) is grammatical, seen in (32a). If the nonspecific object is adjoined, then extraction out of the same RC with a nonspecific head noun in (32b) should be ungrammatical. And it is: (32)

a.

[[ tj ti tamir eden] kadınıj s¸imdi gördüg˘üm] arabai… [[ tj ti repairing woman- now my.having.seen car ‘the car that I just saw the woman who repaired (it) …’ b. *[[ tj ti tamir-eden] bir kadınj-@ gördüg˘üm] arabai… [[ tj ti repairing a woman-@ my.having.seen car ‘the car that I saw a woman who repaired (it) …’

In (32) the head of the RC is the original subject of the RC and the object of the matrix clause, such that the final extraction is the original object of the RC.

200

SARAH D. KENNELLY

While relativization is felicitous out of a RC with a specific head noun, relativization out of a RC with a nonspecific head noun is not. A RC with a head noun that is the nonspecific object of a matrix clause patterns with the adjunct seen in (24), regardless of whether or not relativization in Turkish involves movement. Furthermore, (31b) is a counterexample to the proposal put forth by Heim (1982) and Enç (1991) that a nonspecific head of a RC is disallowed. In addition, the ‘Specificity Effects’ (Enç 1991), whereby extraction is not possible out of specific DPs while it is fine out of nonspecifics, do not hold in Turkish, indicating that the issue is syntactic rather than semantic, and that no universals are involved. 5.4 Motivation in the Interpretation Fourth, under the Minimalist Program, movement to adjunction for interpretation is allowed, which obviates the need for feature checking (Chomsky 1995: 388). The points in favor of adjunction are iteration, scope ambiguity with a VP adverb, extraction, and the need for interpretation. Since wh-Q/NODPs have no scope construals wrt the subject, and since they are immediately preverbal, which we know from the distribution of adverbs in (18)/(19) resides in I0, then I propose that ‘F’ is the rightmost adjunction(s) to VP and that the left edge of the leftmost of the P-Focus elements constitutes the left edge of ‘F’.

6.

A problem: The order of the P-Focus Elements

In contrast with the reversibility of wh-Qs in (26)/(27), the order of a wh-Q and a NODP in ‘F’ is fixed with the NODP strictly left-verb adjacent, as in (29a), all the while maintaining ambiguous scope construals seen in (28). This is difficult to explain in terms of multiple adjunction. (33)

a.

Üç çocuk 3 child ‘Who did b. *Üç çocuk

kime güzel bir hediye-@ vermis¸? [=(29a)] who- beautiful a present-@ gave the three children give a beautiful present to?’ güzel bir hediye-@ kime vermis¸?

No other element may appear between the NODP and the verb, seen in (7b), including the wh-Q (33b). It might be suggested that the adjacency of the NODP is due either to Case Licensing by the verb, or to incorporation of the noun-verb complex. However the facts indicate that neither of these explanations holds.


201

6.1 Adjacency of the NODP is not determined by Case licensing As seen in (9) the NODP may occur anywhere post-verbally in a marked construction. Adjacency is a requirement on the preverbal position, with an exception in the form of Focus clitics, but not on the postverbal. So Case Licensing of the Weak Case object cannot be a factor in the preverbal adjacency constraint. 6.2 Adjacency of the NODP is not determined by Incorporation Certainly incorporation exists in Turkish, but of the nonspecific object NP and not of the NODP. The fact that NODPs may occur postverbally in (9), while the object NP may not, is the first piece of evidence for this analysis. There is also a contrast in scope construals between that of the unmarked object NP in (34), which results in ambiguity, and the NODP in (7), which results in a cumulative reading. (34)

a.

Üç çocuk araba-@ almıs¸. 3 child car-@ bought ‘Three children bought a car.’ [a car is such that 3 children bought it] [each of 3 children bought a car] b. *Araba-@ üç çocuk almıs¸.

Under Dayal’s (1992: 49) analysis, a nonspecific bare NP in object position in Hindi has no restriction on number due to the fact that it is an incorporated noun-verb complex, that is it may be perceived either as singular or plural. If we apply her analysis to sentences of the type seen in (34a) we have a straightforward explanation for the ambiguity, without resorting to structure. In addition, the ambiguity in (34a) provides further evidence that the subject in (7) does not have an intrinsically collective interpretation, indicating that the explanation for the absence of distributivity of the NODP lies in the syntax rather than in the semantics. In support of the proposal that it is incorporated, the bare NP has a strict adjacency constraint in that it may only occur immediately preverbally, as seen in (34), with the possible intervention of a Focus Clitic. A Case marked noun without a determiner has a definite interpretation in Turkish, as seen in (18) and (31). Parallel to Dayal’s analysis of Hindi (p.43), the Case clitic encodes an Iota Operator that results in a unique interpretation of the noun. It is the equivalent of a null definite determiner. If there is a lexical determiner the uniqueness feature of the Iota Operator is overridden. Consequent-

202

SARAH D. KENNELLY

ly all Case marked nouns are DPs in Turkish. The unmarked object NP is then the only bare NP that surfaces in the language. This fact in itself is strong evidence in favor of the proposal that, like the Hindi equivalent, the bare NP is incorporated. 6.2.1 Distinct Truth Conditions There are distinct truth conditions for a sentence with a nonspecific argument and for one with an incorporated noun-verb complex which can be used to test whether the unmarked DP and/or the bare NP are incorporated or not. The progressive is used to indicate an action that continues over a period of time, plus a time adverbial is added to reinforce the idea of a drawn out time span. Such a sentence is grammatical with an incorporated noun, indicating that the incorporated noun is an integral part of the activity, such as fish-catching, but anomalous with a NODP — one cannot catch an individual fish all day, not even a nonspecific fish. (35)

a.

Deniz bütün gün balık-@ tutuyordu. D. all day fish-@ was.catching ‘Deniz was fishing all day.’ b. #Deniz bütün gün bir balık-@ tutuyordu. D. all day a fish-@ was.catching #‘Deniz was catching a fish all day.’

The truth conditions for (35) require that it is true iff one is occupied with the activity of fish-catching all day. This makes sense in Turkish if the object is a bare NP as in (35a) but it is anomalous if the object is a DP in (35b). These facts are consistent with the proposal that the bare NP is incorporated while the DP is not. Thus in the same fashion as Hindi the ambiguity of the bare NP is due to a ±plurality interpretation rather than to structure. 6.2.2 Focus Clitics It is unclear if the bare NP is a classic case of syntactic noun incorporation (Baker 1988) at the level of word in which the N0 incorporates with V0, or whether it is an example of semantic incorporation. There is a closed set of Focus particles, including the Q particle mI, that may occur between the bare NP and the verb.


(36)

203

O çocuk gaze@te- de/bile/mi okudu. @that child newspaper- also/even/interrog. read ‘That child also/even read a newspaper.’ ‘Was it a newspaper that child read?’

Under a syntactic incorporation analysis the Focus clitic is adjoined to N0 before incorporation with V0. 6.2.3 Adjectives Although an adjective may modify the incorporated noun, seen in (37), multiple adjectives are excluded with the bare NP, as is a modified adjective, while these are both fine with a NODP. (37)

Deniz yeni araba-@ almıs¸. D. new car-@ bought ‘Deniz bought (a) new car.’

(38)

a. *Güzel ve eski araba-@ almıs¸. beautiful and old car-@ s/he.bought ‘S/he bought (a) beautiful and old car.’ b. Güzel ve eski bir araba-@ almıs¸. beautiful and old a car-@ s/he.bought ‘S/he bought a beautiful and old car.’

Under the proposed analysis, a bare NP and a single adjective function as a compound, which is then incorporated into the verb. Support for this can be seen in the prosody. A compound noun has prosodic prominence at the right edge of the first element of the compound. (39)

BAS¸-bakan head-minister ‘prime minister’

An unmodified bare NP has prosodic prominence (40a). However if the bare NP is preceded by an adjective, the adjective+bare NP complex patterns with compounds (40b), with prominence at the right edge of the first element. This contrasts with the NODP which may retain the immediately preverbal prominence when modified (40c).

204

SARAH D. KENNELLY

(40)

a.

b.

c.

Deniz -@ almıs¸ D. piano-@ bought ‘Deniz bought (a) piano.’ Deniz  piyano-@ almıs¸. D. old piano-@ bought ‘Deniz bought (an) old piano.’ Deniz eski bir -@ almıs¸. D. old a piano-@ bought ‘Deniz bought an old piano.’

It is my claim that the bare NP object in Turkish is incorporated and hence the ambiguity in sentences of the type seen in (34) is due to an ambiguity in ±plurality rather than due to structure. This leaves the absence of distributivity in (7)/(8) to be explained by a structural analysis. 6.3 The Proposal Multiple wh-Qs obligatorily line up adjacent to one another with the rightmost left-verb adjacent and with ambiguous scope construals, taken to indicate that they are all in ‘F’, which must then be made up of a series of adjoined positions. A wh-Q and NODP may both occur in ‘F’, maintaining ambiguous scope construals with respect to each other, but the NODP must be left-verb adjacent. Inasmuch as it can be shown that neither Case nor Incorporation determine this adjacency, and since an analysis of two distinct projections can be ruled out by the ambiguity of scope construals in (28), I propose that wh-Qs and NODPs have distinct mechanisms responsible for their respective movement to ‘F’ which apply at distinct levels of the derivation, resulting in the word order constraint. Although wh-Qs and NODPs are analyzed as members of the same natural class and in the unmarked form they are P-Focused, class membership is clearly not a primitive since their distinct characters surface in the constraint on the order of their occurrence. P-Focus elements are existential quantifiers of choice functions. NODPs are analyzed as the existential quantification of a free choice partial function which operates on the set of individuals defined by the NP while wh-Qs are analyzed as a nonspecific with a Q feature which blocks the realization of the range argument. From the order constraint visible in (33) it can be concluded that it is the Q feature, i.e. the property of having an undefined range argument, which prompts immediate P-Focus raising of wh-Qs to right-adjunction in the


205

bottom-up derivation. The NODP, instead, undergoes a last-resort style of overt quantifier raising, reminiscent of Heim’s (1982) Existential Closure, that occurs at the closure of the VP shell, resulting in the higher position for the NODP, leftverb adjacent. The discourse interface and principles of interpretation interact with the derivational economy principles in mapping sentential elements onto the syntactic structure.

7.

Summary of the Issues

All Focused elements are analyzed as derived predicates which, in Turkish, require contiguity with the verb. P-Focus elements necessarily seek an adjoined position,‘F’, parallel to the VP adjoined Focus position proposed by Brody (1990) for Hungarian and Neeleman (1994) for Dutch. The analysis of an adjoined position as an integral part in the interpretation of its occupants provides an alternative to the feature checking system as motivation for movement, along the same lines that adjunction to IP/VP applies in English for QR. One of the most striking issues that has emerged from this study of Turkish is support for the proposal that wh-Qs and nonspecifics constitute a natural class; it is proposed that they are existential quantifiers of a free choice partial function. Consequently they require movement to an adjoined position to be interpreted, similar to QR. When they are P-Focused, which is always the case for Wh-Qs, they adjoin to VP. In Turkish this raising is further constrained in terms of the discourse interface. Since the verb undergoes rightward head movement to I0, to maintain contiguity with the verb in order to obtain their role as a derived predicate, P-Focus elements are compelled to righthand adjunction to VP. whmovement is then movement to form a predication structure and is clause-bound. That is, once the wh-Q has obtained the interrogative interpretation in ‘F’, there is no motivation for further movement. Predictions based on the adjunction analysis for ‘F’ are upheld by binding data, iteration with ambiguous scope construals, VP adverbs, and extraction. The only consistent island to extraction in terms of relativization in Turkish is an adjunct. A RC with a specific head is not an island to further relativization, while an identical RC with a nonspecific is; nonspecifics then pattern with adjuncts. Similarly, sentential subjects pattern with complements rather than adjuncts. It is then concluded that the subject is in Spec,VP and that in situ arguments are VP internal, supporting the analysis of (3)/(7)/(8) with the position ‘F’ adjoined to VP. One conclusion that can be

206

SARAH D. KENNELLY

drawn from the data is that existential quantification forces raising while other quantified NPs have the option of being interpreted  , that is collectively rather than distributively. The overt structure is the LF in Turkish, at least with regard to quantification. Consequently the properties of NODPs are transparent. It is clear from the data that the common assumption (Milsark 1977: 27; Enç 1991; Reinhart 1995) that nonspecificity can be defined in terms of narrow scope is simply false. If in language X the other arguments remain internal to VP, then the NODP will have a cumulative interpretation. The proposal that a nonspecific DP cannot be the head noun of a RC (Heim 1982; Enç 1991) is also found to be in contradiction with the empirical facts. Potentially crucial issues such as Case and Incorporation are shown to be irrelevant to the verb adjacency constraint of NODPs. Furthermore, the ‘Specificity Effects’ pointed out by Enç (1991) for English, whereby extraction is not possible out of specific DPs while it is fine out of nonspecifics, do not hold in Turkish, indicating that the issue is syntactic rather than semantic, and that no universals are involved. The study of the syntax of ‘F’, the P-Focus position in Turkish, has revealed new insights not only into Focus, but also into the potentially divers nature of wh-movement. Derivational economy principles participate with the need for Full Interpretation in feeding the mapping of the discourse interface onto syntactic structure. This study of Turkish P-Focus not only brings to light some of the primitives (here interrogatives and quantification) underlying discourse structure, but it begins to tease apart the complicated interaction between hierarchical structure and linearity, while providing counterexamples for some of the general assumptions about nonspecifics.

Notes * Earlier forms of this paper were presented at the Conference on Focus in Paris, Spring 1996; at Groningen Univ., at York Univ., at Bilkent Univ., at the VIII International Conference on Turkish Linguistics in Ankara, at CUNY, and at the 1997 GLOW Colloquium in Rabat, Morocco. The ideas presented here have benefited from discussion with those audiences as well as with D. Adger, G. Alpan, C. Balın, A. Birtürk, M. Brody, J. Costa, M. Inal, V. Déprez, C. Dobrovie-Sorin, K. Eren, R. Fiengo, A. Göksel, J. Grimshaw, M. Haverkort, J.Hoeksema, K. Johnson, R. Kayne, M. Kural, G. Kuruog˘lu, D. Lebeaux, K. Oflazer, O. Orgun, M. Özdemir, I. Pembeci, G. Rebuschi, B. Say, R. Schwarzschild, N. Shir, E. Thompsen, G. Tsoulas, M. Yıldırım, and H. Yükseker. All errors are my own.


207

1. The head of a RC is to the right and there are postpositions rather than prepositions. 2. It has been proposed there are no nonspecific subjects in Turkish (Kennelly 1997). Turkish speakers vary in their judgments wrt obliques so that discussion will have to wait for further research. 3. Drawing on Baker (1970), Cheng (1991: 19) notes a correlation between the presence of a Q particle in a language and the absence of wh-movement. However Turkish has a yes-no Q particle, mI, and wh-movement is to the Focus position, first noticed by Güliz Kuruog˘lu (p.c.). Turkish is then a counterexample to Cheng’s analysis. 4. The problem of linearity vs. hierarchy was first brought to my attention by Aslı Göksel of SOAS, London and Bog˘azici Üniv, Istanbul. The analysis that there is no covert QR in Turkish predicts that any ambiguity is due to other factors. This is supported by the work of Göksel (1995) and by the analysis of ambiguity due to Incorporation in (34). 5. Like the German ein, the unstressed form of bir is the indefinite determiner while the stressed form is the number ‘one’; only the unstressed form is considered here. The determiner interpretation is forced when an adjective appears before bir while the cardinal interpretation is forced if an adjective appears after bir. 6. When numerals are used the noun occurs in the singular and the verbal agreement is 3 sg. When the subject has the 3 plural form, as in (7), it is normal to omit the plural marker on the verb. The nominal and verbal plural morphemes are homophonous so the repetition is usually considered redundant. 7. The infelicity of sentences like (4b) is attributed by Göksel (1995) to the fact that postverbal constituents in Turkish are de-stressed, while wh-Qs may not be; i.e. that the motivation for the absence of postverbal wh-Qs is found in the prosody. I would rather suggest that there is some kind of semantic type mismatch: backgrounded information is never affected by the proposition while a wh-Q is, and consequently it is this mismatch that is reflected in the prosody. 8. Specific indefinite objects are generally infelicitous in the Focus position. This constraint can be overridden if there is a clear relation between the specific indefinite and a previously introduced DP This is seen in (i) in that the subject introduces a profession to which the specific indefinite has a relation. An unrelated specific indefinite, as in (ii) is ungrammatical. (i)

S¸u tamirci bir arabayı tamir etti. that mechanic a car- repaired ‘That mechanic repaired a (specific) car.’

(ii) */?S¸u tamirci bir elmayı yedi. that mechanic an apple- ate ‘That mechanic ate a (specific) apple.’ This fact clearly indicates a functional interpretation for Focused specific indefinites which entails a relation with an element in the discourse. 9. Though the partitive acts like a strong determiner on the object in that it requires the specificity marker in the form of Accusative Case marking, in the subject position it patterns with the weak determiners in that it is nonquantificational, thereby supporting de Hoop’s (1992) analysis that partitivity is independent of definiteness.

208

SARAH D. KENNELLY

10. According to de Hoop’s (1992) analysis, an argument with Weak Case, the NODP in Turkish, may move into any A-bar position. I will modify her analysis by claiming that it may move into an A-bar position as long as occurrence in that position is consonant with its interpretation. As brought out in Chierchia & McConnell-Ginet (1991: 282) a nonpresupposed element, that is new information, may also be backgrounded, seen in English in a nonrestrictive relative. Nonspecific arguments are taken to be new information and hence nonpresupposed. So they may occur either in the ‘new information’ Focus slot, left verb adjacent in Turkish, or backgrounded postverbally. 11. The discussion may be generalized to other lexical elements (Williams 1980: 208) that may function as the predicate. 12. There is the Persian loan word ki ‘that’ which triggers SVO word order but it is unintegrated into the language. Native speakers tell me that it is primarily used by Türks who grew up in a bi-lingual environment, such as the Türks who were born in Germany. 13. It has been proposed here that wh-Qs and NODPs both occupy the P-Focus position, rightadjoined to VP. We would then expect them to interact with the distributive her ‘each’ in subject position similarly, which is not attested. The wh-Q retains a wide scope construal over the subject. However the impossibility of quantifying into an interrogative has been widely attested cross-linguistically, with the striking exception of English, so it is not a problem specific to this analysis.

References Baker, C.L. 1970. “Notes on the Description of English Questions: The Role of An Abstract Question Morpheme.” Foundations of Language 6: 197–219. Baker, M.C. 1988. Incorporation. Chicago: Chicago Univ. Press. Brody, M. 1990. “Remarks on the Order of Elements in the Hungarian Focus Field.” In I. Kenesei, ed., Approaches to Hungarian. Szeged: JATE, 95–121. Diesing, M. 1992. Indefinites. Cambridge (Mass.): MIT Press. Cheng, L.L.-S. 1991. On the Typology of wh-Questions. Ph.D. Diss, MIT. Chierchia, G. 1992. “Anaphora and Dynamic Binding.” Linguistics and Philosophy 15: 111–183. Chierchia, G. and S. McConnell-Ginet. 1991. Meaning and Grammar: An Introduction to Semantics. Cambridge (Mass.): MIT Press. Chomsky, N. 1971. “Deep Structure, Surface Structure, and Semantic Interpretation.” In D. D. Steinberg and L. A. Jakobovits, ed., Semantics. Cambridge: CUP, 183–216. Chomsky, N. 1986. Barriers. Cambridge (Mass.): MIT Press. Chomsky, N. 1995. The Minimalist Program. Cambridge (Mass.): MIT Press.


209

Costa, J. 1996. “Positions for Subjects in European Portuguese.” WCCFL XV, Stanford, CSLI. Davidson, D. 1967. “The Logical Form of Action Sentences.” Proceedings of the Conference on the Logic of Decision and Action, Univ. of Pittsburgh Press. Dayal, V. S. 1992. “The Singular-Plural Distinction in Hindi Generics.” SALT II, Ohio State Univ. de Hoop, H. 1992. Case Configuration and Noun Phrase Interpretation. Ph.D. Diss, Univ. of Groningen. Enç, M. 1991. “Specificity.” Linguistic Inquiry 22.1: 1–25. Erguvanlı, E.E. 1984. The Function of Word Order in Turkish Grammar. Berkeley: Univ. of California Press. Fodor, J.D. and I.A. Sag. 1982. “Referential and Quantificational Indefinites.” Linguistics and Philosophy 5: 355–398. Göksel, A. 1995. Linearity, Focus and the Postverbal Position in Turkish. Ms. SOAS, Univ. of London. Grimshaw, J. 1997. “Projection, Heads, and Optimality.” Linguistic Inquiry 28.3: 373–422. Haider, H. 1985. “A Unified Account of Case and Theta-Marking: the Case of German.” Papiere zur Linguistik 32: 3–36. Hamblin, C.L. 1973. “Questions in Montague English.” Foundations of Language 10: 41–53. Heim, I.R. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. Diss, UMass at Amherst. Herburger, E. 1995. “Focus Structures Quantification Over Events.” SCIL 7, MITWPL. Hintikka, J. 1986. “The Semantics of A Certain.” Linguistic Inquiry 17.2: 331–336. Horvath, J. 1986. Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Jackendoff, R.S. 1972. Semantic Interpretation in Generative Grammar. Cambridge (Mass.): MIT Press. Jackendoff, R.S. 1977. X-Bar Syntax. Cambridge (Mass.): MIT Press. Karttunen, L. 1977. “Syntax and Semantics of Questions.” Linguistics and Philosophy 1: 33–44. Kennelly, S.D. 1990. “Theta Government in Turkish.” GLOW Workshop, SOAS London.

210

SARAH D. KENNELLY

Kennelly, S.D. 1996. “Nonspecificity = Theticity An Instantiation of Presentational Focus.” SCIL 8, MITWPL. Kennelly, S.D. 1997. “*Nonspecific External Arguments in Turkish.” Dilbilim Arastırmaları 7. Kiss, K.É. 1994. “Sentence Structure and Word Order.” In F. Kiefer and K. É. Kiss, eds., The Syntactic Structure of Hungarian. New York: Academic Press, 1–90. Koopman, H. and D. Sportiche. 1988. “Subjects”. Ms. UCLA. Kratzer, A. 1995. Scope or Pseudoscope? Are there Wide-Scope Indefinites? Ms. UMass at Amherst. Kural, M. 1992. Properties of Scrambling in Turkish. Ms. UCLA. Kural, M. 1994. Postverbal Constituents in Turkish. Ms. UCLA. May, R. 1985. Logical Form. Cambridge (Mass.): MIT Press. Milsark, G.L. 1974. Existential Sentences in English. Ph.D. Diss, MIT. Milsark, G.L. 1977. “Toward an Explanation of Certain Peculiarities of the Existential Construction in English.” Linguistic Analysis 3.1: 29. Nash, L. 1995. Portée argumentale et marquage casuel dans les langues SOV et dans les langues ergatives: l’exemple du géorgien. Ph.D. Diss, Univ. Paris 8. Neeleman, A. 1994. “Scrambing as a D-structure Phenomenon.” In N. Corver and H. van Riemsdijk (eds.). Studies on Scrambling. Berlin: Mouton de Gruyter, 387–429. Ortiz de Urbina, J. This vol. “Focus in Basque.” Parsons, T. 1990. Events in the Semantics of English: A Study in Subatomic Semantics. Cambridge (Mass.): MIT Press. Pearson, M. 1996. “Domain Phrases and Topic Arguments in Malagasy Existentials”. In M. Pearson et I. Paul. UCLA Occasional papers in Linguistics 17. Pinto, M. 1994. “Subject in Italian: Distribution and Interpretation.” In R. BokBennema and C. Cremers (eds.). Linguistics in the Netherlands. Amsterdam: John Benjamins. Pollock, J.-Y. 1989. “Verb Movement, Universal Grammar, and the Structure of IP.” Linguistic Inquiry 20.3: 365–424. Rebuschi, G. 1983. “A Note on Focalization in Basque”. Journal of Basque Studies 4.2: 29–42. Rebuschi, G. 1994. “Skolémisation et spécificité.” Faits de Langues 4: 121–128. Reinhart, T. 1995. Interface Strategies. OTS Working Papers. Utrecht Rizzi, L. 1990. Relativized Minimality. Cambridge (Mass.): MIT Press.


211

Rochemont, M.S. 1986. Focus in Generative Grammar. Philadelphia: John Benjamins. Scha, R.J.H. 1984. “Distributive, Collective and Cumulative Quantification.” In J. Groenendijk et al. (eds.). Truth, Interpretation and Information. Dordrecht: Foris, 483–512. Sezer, E. 1982. The Unmarked Sentential Subject Constraint. Ms. Harvard Univ. Szabolcsi, A. 1981. “Compositionality in Focus.” Folia Linguistica 15.1–2: 141–161. Vikner, S. 1991. Verb Movement and the Licensing of NP-Positions in the Germanic Languages. Ph.D. Diss, Universität Stuttgart. Watanabe, A. 1992. Wh-In-Situ, Subjacency, and Chain Formation. MITWPL2. Williams, E. 1980. “Predication.” Linguistic Inquiry 11.1: 203–238. Woolford, E. 1994. “Object Agreement in Palauan: Specificity, Humanness, Economy and Optimality.” Papers in Optimality Theory (UMOP), 655–700.

Word Order and Focus Positions in Universal Grammar* Ayesha Kidwai Jawaharlal Nehru University, New Delhi

Abstract This paper explores the relationship between word order and focus positions in Hindi-Urdu, Malayalam, Western Bade and Tangale. Characterizations of positional focus in terms of a Focus Phrase can neither satisfactorily capture the crucial role that non-canonical word orders play in such constructions, nor explain how positional focus languages differ from those that mark it by morphology or prosody. The paper seeks to develop a minimalist theory of [+FOCUS], by which the feature is checked under strict adjacency to the verb in the PF-component. This adjacency is affected by PF-movement rules of XPand X0-adjunction, and is driven by the conjecture that UG generates [PF[±Interpretable]] features that can be checked only in the PF-component. Structures derived as a result of PF-movement are interpreted at a level distinct from LF, named Domain Discourse, located at the edge of the PF-component. [+FOCUS] is argued to be a [PF[+Interpretable]] feature that may survive to the interface unchecked, and hence may be accessed by all three of the subcomponents inside PF, PF-movement, Morphology and Phonology. Hence, the heterogeneity of focus-marking mechanisms in natural language.

1.

Introduction

This paper explores the relationship between word order and focus positions in Hindi-Urdu, Malayalam, Western Bade and Tangale. It will be our claim that word-order variation, expressed structurally as adjunction in the PF-component,

214

AYESHA KIDWAI

is responsible for the superficial occurrence of focus positions. We argue this PFscrambling to be driven by the checking requirements of [PF[±Interpretable]] features, which are interpreted at a level distinct from LF termed D D() and located at the edge of the PF-component. The crucial role that scrambling plays in the derivation of positional focus constructions in languages like Malayalam, Western Bade, Tangale and HindiUrdu has largely been ignored in the literature. Section 2 delineates this relationship and examines some of the problems that linguistic theory faces regarding the characterization of positional focus in natural language. Section 3 presents a reanalysis of focus constructions from a minimalist perspective on the organization of UG (Chomsky 1995, 1996), and demonstrates that the proposals can account for positional focus in the four languages we consider in the paper. Section 4 concludes the paper with a discussion of some focus constructions in English. Before we proceed, a statement of the view of focus and its interpretation that we will work with is in order. Following Erteschik-Shir (1997) and Zubizarreta (1996), we define focusing as primarily (but not entirely) a strategy of indexical assertion, the means by which a speaker attempts to render an entity in the discourse salient for the hearer(s) of the utterance: T FOCUS    S   ( )     S          / () ,   S (Erteschik-Shir 1997). We therefore consider the interpretation of focused constituents to be determined in terms of truth1 by the set of the pragmatically determined presuppositions common to the speaker and the hearer at a given time in the discourse. However, even while both the terms involved in the interpretation of focus are determined by the intentions and (shared) knowledge of speakers, focusing is a product of  rule, therefore analogous to topicalization in which the grammatical rule that fronts topics serves what is essentially a discursive function. The set of pragmatically determined presuppositions as well as the indexical assertion itself may well be given different (and simultaneous) characterizations by the rules of discourse, such as topic/comment, topic/focus, theme/rheme, old/new information, etc., but these discourse-grammar distinctions derive from the presupposition-assertion structure created by grammatical rule. We also suggest that the only theoretically relevant distinction in the study of focus is the one between  and  focus. Contrastive and presentational interpretations of wh- and non-wh- foci derive from the   of shared presuppositions on the basis of which the focus is interpreted (Erteschik-

WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 215

Shir 1997; Zubizarreta 1996) — if the discourse provides a contrast set as below, the focus will receive a contrastive interpretation. Outside of context, the presentational focus interpretation may well be preferred. A: Who wants to read the paper, Ram or Sita? B:  wants to read the paper.2 The distinction between narrow and wide focus, on the other hand, is particularly relevant to the study of focus constructions in natural language, since the focus involved here is invariably narrow. Wide focus, which we shall refer to as  focus, is typically associated with utterances in out-of-the-blue contexts, and is marked by ambiguity as to which constituent is singled out as the focus of the utterance. In an out-of-the-blue utterance of {Rehman {bought {a book}}}, any of the constituents may well be intended to be the focus of the sentence (possible foci are indicated by curly braces). Neutral focus utterances are pragmatically felicitous answers to questions like What happened?. In a language like English, which marks focus by prosodic means, neutral focus utterances are those generated by the core algorithm for the assignment of stress in the language, the Nuclear Stress Rule. Narrow focus, also referred to as - , requires the hearer to partition the utterance into presupposed and asserted parts, and provides him/her unambiguous cues as to which constituent constitutes the focus of the utterance. Embedding Rehman bought a book in a discourse like the one below, identifies Rehman to be the indexical assertion of the utterance: A: Who bought a book? B:  bought a book. Narrow focus utterances are pragmatically infelicitous answers to questions like What happened? In languages like English, Zubizarreta (1996) claims, narrow focus involves a marked prosodic pattern generated by an independent algorithm.

2.

Movement and Focus Positions in UG

The two most remarkable descriptive facts about the (often typologically unrelated) languages that choose to mark focus by syntactic position are that these languages typically define the focus position in terms of proximity to the verb, and that this focus position is the position targeted by overt wh-movement

216

AYESHA KIDWAI

in these languages. A closer scrutiny of the data, however, argues for a  generalization regarding positional focusing — in languages like Malayalam (Jayaseelan 1989, 1995), Western Bade and Tangale (Tuller 1992), the positional focusing of subjects involves a -  , the canonical order being reserved for neutral focus utterances:3 (1)

Malayalam: SOV a. ninn-e 6 aziccu?4 you- who beat ‘Who beat you?’ b. ninn-e  aziccu you- Ram beat ‘RAM beat you.’ c. raman ninn-e aziccu Ram. you. beat ‘Ram beat you.’ (neutral focus)

(2)

Western Bade: SVO a. tl6mp6t6] 6 zaneenii? tore what gown-your ‘What tore your gown?’ b. zanee]aa, tl6mp6t6-g 6 gown-my, tore wood ‘my gown, WOOD tore it.’ c. Saku aa b6naa kajluwaan Saku  cook tuwo ‘Saku will cook tuwo.’ (neutral focus)

(3)

Tangale: SVO a. wad Billiri ] dooji? go Billiri who tomorrow? ‘Who will go to Billiri tomorrow?’ b. tui worom mono  ate beans my she ‘ ate my beans.’ c. ~n~g lfshfgflf ti lfwei gave fish to child-the ‘(She) gave fish to the child.’ (neutral focus)


The examples in (1)–(3) demonstrate that the focus position in Malayalam, Western Bade and Tangale can host both wh- and non-wh- subject foci. Each of these constructions entails variation from the canonical order of the language. Malayalam, an SOV language that defines its focus position as left-adjacent to the verb, requires an OSV order for the preverbal focusing of subjects, Western Bade, an SVO language whose focus position is immediately right-adjacent to the verb, requires a VSO order (or left-dislocation/ topicalization) for postverbal focusing. Normally SVO Tangale, which locates its focus position in terms of right adjacency to the direct/locative object, requires a VOS order for postverbal focusing. The fact that (1b)–(3b) are pragmatically infelicitous either as out-ofthe-blue utterances or as responses to questions like What happened?, indicates that narrow focus is involved. Moreover, these examples are also inappropriate responses to questions like What did Ram do (to you)?, What did the wood do (to your gown)?, What did she do (to your beans)? since these queries assume a discourse in which the highlighted constituent in (1a)–(3a) is part of the presupposition rather than the assertion. This infelicity indicates that a narrow focus interpretation in these examples accrues to the constituent that occupies the position occupied by the wh-phrases in (1a)–(3a). In the canonical order of these languages (1c)–(3c), however, narrow focus on the pre/post-verbal element does not obtain. These examples are entirely appropriate out-of-the-blue utterances or responses to questions like What happened? The focus position thus surfaces only in the non-canonical word order, and it will be our claim that these noncanonical orders are at least partly the product of XP and X0 scrambling rules in these languages. Hindi-Urdu represents the converse case, where the role of  in inducing word order variation has been largely ignored. Generative research on Hindi-Urdu scrambling (Mahajan 1990), which has explored the syntactic constraints on such movement has been confined to the reference and coreference relationships in scrambled constructions. The fact that scrambling in Hindi-Urdu always entails focus on the immediately preverbal constituent, and that whphrases are preferentially placed in this position (except in discourse-initial questions) has gone unnoticed. (4)

Hindi-Urdu: SOV a. kitaab f laaye-gaa book who bring- ‘Who will bring the book?’

218

AYESHA KIDWAI

b.

c. d. e.

kitaab  laaye-gaa book Ram bring- ‘It is Ram who will bring the book.’ What did Ram bring? What did Ram do to the book? raam kitaab laaye-gaa Ram book bring- ‘Ram will bring the book.’ (neutral focus)

(4b) is an appropriate response only to (4a). Neither (4c), which requires narrow focus on the DO kitaab ‘the book’ in the response, nor (4d), which requires narrow focus on the verb laaye-gaa, ‘will bring’ in the response, can elicit (4b) as a pragmatically felicitous response. (4b) also cannot answer a neutral focus question like What happened? — that question can only be answered by (4e), which is in the canonical word order of the language. Thus, constructions that involve variation from the default word order in Hindi-Urdu, like those in Malayalam, Western Bade and Tangale, require the hearer to partition the utterance into presupposed and asserted parts and to interpret the constituent in the focus position as the focus of the utterance. This link between movement and positional focusing is not limited to leftward clause-internal argument scrambling alone.5 Clause-bound leftward X0-scrambling and rightward argument scrambling within the clause also yields the same results: (5)

a.

b.

(6)

a.

b.

kal 7˜ aau˜u˜]-gaa tomorrow I come- ‘I will come tomorrow.’  aau˜u˜]-gaa 7˜ tomorrow come- I ‘I will come tomorrow.’6 anjum-ko  dii nuur-ne Anjum() book() gave Noor() ‘It was a book that Noor gave Anjum.’ anjum-ko - dii kitaab Anjum() Noor() gave book() ‘It was Noor who gave Anjum a book.’

Rightward scrambling is not, however, the preferred strategy for preverbal


focusing in Hindi-Urdu. In syntactic terms as well, it does not share the properties exhibited by leftward scrambling — neither are rightward scrambled DPs interpreted as specific, nor can they serve as potential antecedents for coreference. We return to the issue later in the paper; for the moment it is sufficient to note the generality of the link between scrambling and preverbal focusing in Hindi-Urdu. This in turn suggests that the driving force behind the scrambling operation in Hindi-Urdu is  itself, i.e., scrambling in Hindi-Urdu is the word order variation we have found to be an important variable in the derivation of focus constructions in Malayalam, Western Bade and Tangale. Confronted with a set of four languages that mark narrow focus by position, where this positional focusing always involves a non-canonical linear order, the central question is how this data is to be used to arrive at a predictive theory of positional focus in natural language and to isolate the set of UG-specified properties that enables some languages to employ this option, and others to (apparently) eschew it altogether. Even a cursory look at the data reveals that the answers to these questions cannot be easily obtained. For example, take the question of the distribution of focus positions in natural language and the related (somewhat loose)7 --- requirement observed in positional focus languages. It appears that the crosslinguistic distribution of focus positions is at least partially constrained by the settings of the directionality (Head and feature-checking) parameters in UG (Horvath 1986), since the distribution of focus positions in the four languages we look at here appears to depend on the way that the Head Parameter is set by the language in question — preverbal focus positions surface in SOV languages and postverbal ones in SVO languages. However, a language such as Hungarian, that is SVO and yet has preverbal focus positions, question the universality of the generalization. The facts of languages like Hungarian could, however, be made to follow from the speculation that [+FOCUS] is a syntactic feature analogous to [+CASE] or [+OPERATOR], i.e. a feature that must be licensed by the UG mechanisms of feature-checking in local domains. Suppose then that the local domain in which [+FOCUS] feature-checking takes place is defined in terms of strict adjacency to the verb, with the physical position of the verb at Spell-Out counting as a factor in calculating where [+FOCUS] will be assigned. Languages whose focus positions do not appear in the direction set by the Head Parameter could then differ from languages that do so (e.g. Hindi-Urdu and Western Bade) on two counts: one, the former have overt verb-raising to a position higher than the canonical subject position, and two, these languages set the feature-checking

220

AYESHA KIDWAI

parameter in a direction opposite to that of the setting of the Head Parameter. Preverbal focusing in SVO Hungarian would then follow from the fact that the verb raises to a position higher than the canonical subject position (say C0), and since [+FOCUS] can be checked only under left-adjacency to the verb in the language, preverbal focusing (in [Spec,CP]) is the only option permitted by the grammar of the language. This solution to the problems of Hungarian is not satisfactory in crosslinguistic terms though, as it strengthens the proximity-to-the-verb requirement beyond the empirical facts themselves. Languages like Tangale do not appear to respect this requirement in the way formulated here, since focused elements are disallowed in a position right-adjacent to the verb — the focus position must in fact appear after the DO/Locative in all cases of postverbal focus. Also note that the literal nature of this physical proximity-to-the-verb requirement in positional focusing prohibits a situation in which the verb raises to a position higher than that of the focused element in the overt syntax. Yet, this is exactly the case that obtains in Hindi-Urdu, where the physical adjacency of the verb and the focused element is only illusory, as the verb must raise to Tense for licensing verbagreement and checking of the subject in [Spec,TP] (e.g. 4b). Given the complications of even a description of the distribution of focus positions in natural language, recent research in the area (Brody 1989, 1996; Jayaseelan 1989, 1995; Kidwai 1995) has assumed that focus positions have an ontologically prior status in terms of phrase-structure, usually formalized as a   projection immediately dominating VP. These analyses maintain the hypothesis that [+FOCUS] is a feature that has syntactic relevance in that it needs to be checked and suggest that the defining property of focus positions is not the proximity-to-the-verb requirement, since the verb may well have moved beyond F0 to a higher functional projection in the overt syntax. The fact that this FP projection is the landing site targeted by wh-movement is held to follow from the principal observation that wh-phrases are inherently [+FOCUS] elements, since they function as place-holders for indexical assertions in the response, and that if this feature is not licensed, the derivation will crash. These analyses therefore equate the positional focusing of wh-phrases with that of non-wh elements, both in terms of identity of position as well as degree of “obligatoriness”. This latter conclusion is not, however, supported by the empirical facts, since it appears that at least in some languages, the positional focusing of wh-phrases is more “obligatory” than that of non-wh foci. In HindiUrdu, for example, non-wh foci need not be preverbally positioned if recourse is


taken to a morphological strategy of -hii cliticization, an emphatic particle that marks narrow focus in the language. wh-foci cannot, however, take recourse to this strategy, which we term as in situ focusing, and must be preverbally positioned: (7)

-hii kitaab laaye-gaa Ram– book bring– ‘RAM will bring the book.’ b. *f-hii kitaab laaye-gaa who– book bring– ‘Who will bring the book?’

a.

Another problem with an analysis of positional focusing in terms of a functional FP projection is the status of the [Spec,FP] position that such analyses are forced to assume. Besides the fact that wh-phrases target this position, the fact that English focused elements pattern with wh- and quantifier operators in exhibiting W C (WCO) violations has usually been taken to suggest that focused elements are operators, undergoing QR to an A-bar position at LF. It is therefore expected that focused pronominals in a language like Hindi-Urdu cannot corefer with an antecedent in the same clause, but as is well known, scrambling in Hindi-Urdu actually overrides WCO violations, and (somewhat more marginally) licenses monomorphemic reflexives: (8)

a.

uskiii b7h7n-ne kis??i-ko dekhaa his sister– who– saw ‘Whoi did his??i sister see?’ b. apniii b7hen-ne mohan*i-ko maaraa self’s sister– Mohan– hit ‘Selfi’s sister hit Mohan*i’ c. ?kisi-ko uskiii b7h7n-ne ti dekhaa who– his sister– saw ‘Whoi did hisi sister see?’ d. ??mohani-ko apniii b7h7n-ne ti maaraa Mohan– self’s sister– hit ‘Selfi’s sister hit Mohani.’

The examples in (8a–b) demonstrate that Hindi-Urdu neutral focus constructions exhibit WCO effects and require monomorphemic reflexives to be locally bound. In the non-neutral focus constructions in (8c–d), however, where the possessive pronoun/reflexive is preverbally focused, these requirements appear to be

222

AYESHA KIDWAI

suspended. By the assumptions outlined above, this focused phrase should QR at LF to a position outside the domain of its antecedent, yielding the expectation that both the examples will be ungrammatical — an expectation belied by the facts, which demonstrate that focusing actually  the conditions for referential dependencies within a clause. This conjecture is confirmed when we consider coreference in in situ focus constructions, where the morphological focus marking of a constituent equally improves the possibility of coreference with an antecedent within the clause that contains it: (9)

a.

b.

c.

d.

uskiii biiwii kis??i-ko bfhft maartii h7 his wife who- a lot hits . ‘Whoi does his??i wife hit a lot (habitually)?’ apniii b7h7n-ne mohan*i-ko maaraa self’s sister- Mohan- hit ‘Selfi’s sister hit Mohan*i.’ uskiii-hii biiwii kis??i-ko bfhft maartii h7 his- wife who- a lot hits . ‘Whoi did his??i wife hit a lot (habitually)?’ apniii-hii b7h7n-ne mohan??i-ko maaraa self’s- sister- Mohan- hit ‘Selfi’s sister hit Mohan*i.’

The greatest difficulty faced by the functional projection analysis of positional focus is its inability to build in the crucial role that non-canonical linear orderings play in the derivation of positional focus constructions themselves. In Western Bade and Tangale, this non-canonical ordering requires verb-raising, whereas in Hindi-Urdu (and Malayalam),8 the orders (represented here) require argument-scrambling. While an X0-adjunction analysis for verb-raising is uncontroversial, the claim that XP-scrambling in Hindi-Urdu involves XPadjunction is novel. Since Mahajan (1990), it has been commonly held that clause-internal argument scrambling in Hindi-Urdu is wrought by Case-driven movement. In Kidwai (1995), we review Mahajan’s arguments for this claim in some detail and demonstrate that his results neither follow from, nor are confirmed by, the empirical facts of Hindi-Urdu Case and verb agreement, since these are determined completely independent of whether XP-scrambling takes place or not. In any case, for Mahajan, the singular fact that argues for a Casedriven analysis for Hindi-Urdu clause-internal argument scrambling is that it can override WCO effects as in (8a–b), however, we have seen that this fact may


not necessarily follow from the A-status of the binder, but could originate from the  state of the bindee. Now, given that positional focus constructions in Hindi-Urdu, Western Bade, Malayalam and Tangale can all be argued to involve adjunction, the question is how an FP analysis can deal with this putative factor in the derivation of positional focus constructions. The standard approach (Jayaseelan 1989, 1995; Brody 1989, 1996) has been to describe the superficial non-canonical orderings that accompany positional focusing as the fortuitous result of a conspiracy of independently motivated movements, such as movement for Case and verb agreement checking, left-dislocation, topicalization, etc. This independence of the movement to [Spec,FP] from the other movements observed in positional focusing predicts that positional focus (specially of non-wh) elements may take place in the canonical order, i.e. that neutral focus utterances should not be attested. Again, this prediction is not confirmed by the data. Finally, analyses of positional focus in terms of raising to a [Spec,FP] position can only postpone, but not answer, the central question regarding the occurrence of positional focus constructions in natural language, the question as to why only some languages choose to employ this option.9 The postulation of an FP projection is ultimately descriptive, since it does not originate from a wellarticulated theory of focus and focus phenomena in UG. FP analyses have no real UG explanation for the fact that the tasks served by a FP projection in the overt syntax in positional focus languages are spread over more than one phrasal projection and component in languages that employ other means of focusing — for example, English realizes wh-focus in [Spec,CP], and marks non-neutral focus on non-wh by either prosody or by the means of syntactic focus constructions like PP-extraposition, Directional/Locative Inversion, etc. Moreover, even positional focus languages may use a heterogeneity of focus-marking mechanisms. Hindi-Urdu, for example, has three strategies for realizing non-neutral focus: a syntactic strategy of preverbal positioning, a morphological strategy of in situ focus via -hii-cliticization, and a prosodic strategy of heavy (contrastive) stress.10 Furthermore, neither of these strategies are in complementary distribution with each other, and as demonstrated by (10), all three may be used simultaneously in a single utterance. (10)

kitaab -hii laayegaa (siitaa nahii) book Ram- bring- Sita not ‘RAM will bring the book, not Sita.’

224

AYESHA KIDWAI

Languages like Aghem, an SVO language with a postverbal focus position, also pattern with Hindi-Urdu in employing multiple strategies to indicate nonneutral focus (Watters 1976; Rochemont 1986). Aghem makes extensive use of focus-related morphology, both in the derivation of positional focus constructions as well as for in situ focus. Hungarian represents an interesting case in which, even while there are prosodic patterns typical of non-neutral focus constructions (a rule of Stress Reduction), this prosodic pattern does not uniquely pick out the focused phrase as the prosodically most prominent constituent of the phrase, and treating it in certain contexts at par with presupposed constituents (Kenesei & Vogel 1989). These facts demonstrate a complexity of interaction between the various and multiple strategies employed by non-neutral focus, and suggest that an account of positional focusing in natural language can only originate from a  of focus in UG, a theory which expresses the intuition that the heterogeneity of (non-neutral) focus-marking mechanisms attested in natural language — positional, prosodic and morphological — differ only in terms of superficial realization of an identical feature. The minimalist theory of positional focusing we develop in the next section attempts to capture this intuition.

3.

A Minimalist Theory of [FOCUS]

The discussion in the previous section demonstrated the distribution of focus positions in natural language to be contingent upon (1) XP- and/or X0-adjunction (2) a weak proximity-to-the-verb requirement, and (3) the grammatical nature of the feature [+FOCUS] in that it needs licensing (possibly in multiple ways) by the time phonetic output is reached. We propose that these three factors interact in the following manner to give rise to the superficial occurrence of focus positions: (11)

[+FOCUS] is a feature in UG that requires licensing under adjacency to a verbal projection. XP-/X0-adjunction is driven by this adjacency requirement, since this scrambling results in an adjacency between the focused phrase and the licensing verbal projection that would not have otherwise obtained.

We argue that current minimalist assumptions require that we locate this licensing of [+FOCUS] within the PF-component, and that this licensing is driven by discourserelated, rather than LF-relevant, considerations. In order to arrive at these results, let us consider each aspect of the proposal in (11) in some more detail.


3.1 The Status of Adjunction A first step in developing the proposals in (11) necessarily involves an investigation of the status of adjunction under current assumptions about the structure of UG. Chomsky (1995, 1996) proposes that minimalist assumptions cannot countenance overt XP-adjunction as an instance of pre-Spell-Out syntactic movement, since overt movement is interpreted as the movement of features and takes place only in the L R, as defined in (12): (12)

Move-F raises F to target K only if F enters into a checking relation with a sub-label of K.

Since Chomsky (1995) rules out checking from adjoined positions, XP-adjunction can never satisfy Last Resort, and should therefore never take place. In a similar vein, Chomsky (1996) suggests that head-movement (adjunction to X0) should also be eliminated as an option within the overt syntax, since it too lacks strong LF-effects. Chomsky (1995: 325–26) suggests that we revert to the distinction in early transformational grammar between “stylistic” rules and others, since: […] the core computational properties […] differ markedly in character from the many other operations of the language faculty, and it may be a mistake to try to integrate them within the same framework of principles. XP-adjunction […] may not really belong to the system we are discussing here, as we keep closely to Last Resort movement driven by feature-checking within the N→l computation. It is within this core component of the language faculty that we find the striking properties highlighted by minimalist guidelines. It seems increasingly reasonable to distinguish this component of the language faculty.

Assuming the theoretical validity of these proposals, the elimination of adjunction as a pre-Spell-Out movement suggests that the XP and/or X0 scrambling necessary for the derivation of positional focus constructions cannot take place in the overt syntax, with the result that positional focus constructions are given their ultimate form in the mapping from Spell-Out to phonetic output. There are, however, at least two problems with a naive interpretation of Chomsky’s proposals. The first pertains to their execution, for as yet the PFcomponent is too primitive and unstructured to accommodate fine-grained distinctions between languages that, say, have focus positions/constructions and those that do not. The second problem deals with the LF-effects of these focus constructions — relegating focus constructions and other stylistic rules to the PFcomponent would entail that, given the current shape of the grammar, they

226

AYESHA KIDWAI

should not have any discernible effect on interpretation. As is obvious, this is untrue, as not only does the feature [+FOCUS] receive an interpretation, structures involving XP-adjunction have been shown to have consequences for the binding theory. We return to the interpretative issues directly. A step in the direction of structuring the PF-component would be to assume that the PF-rules that reorder parts of the strings generated by the computational component, employ mechanisms similar to those in the overt component. That is, (at least some) PF-displacement rules too can be argued to be driven by the requirements of feature-checking, although presumably without the formation of checking domains, since the only option for Move in this component is adjunction. Rather, let us assume that although feature-checking in PF operates on hierarchical structure, this checking is effected under conditions of adjacency, linearity and the like, conditions which must derive from the requirements of the sub-component of Morphology inside PF. Cummins and Roberge (1994), following Halle and Marantz (1993) propose that inflectional affixes are available pre-syntactically only as features, and not as fully realized as bound morphemes, and that they are spelled out and linearized only by morphophonological rules in the PF-component. In this approach, the only function of the overt syntax is to ensure that these morphosyntactic features are licensed so that they are visible at the PF-interface. We suggest that this function is not exclusively the property of the overt syntax, and that feature-checking in the PF-component also shares the same function — to render morphological features visible to Morphology — a guiding intuition of the minimalist program. The question now is, what is the nature and content of features that can only be checked in the PF-component, since the standard notions of feature-checking cannot accommodate checking in this component, and it is to this question that we now turn. 3.2 Interpretability of Features Chomsky (1995: 277–79) proposes that along with strength, features need to be specified for  at the LF interface. This criterion of LFinterpretability pertains to the interpretability of the   of a linguistic item at LF. Included in the set of [+Interpretable] features of a linguistic item are its f-features and categorial features, while the prototypical [−Interpretable] feature is Case. The key difference between the two features is that the former must survive till LF, since they are needed for LF-interpretation,


and hence convergence, but the latter, which have no interpretable role at the LF interface, must delete for convergence. Consider the feature [+FOCUS]. Can it be argued to be the morphological property of a lexical item in a way analogous to Case, categorial features, etc., springing as it does from speakers’ intentions? In all likelihood not, for, as Culicover (1993: 5) puts it: Focus is orthogonal to most other aspects of interpretation, in that it can vary independently of the literal sense of individual phrases and the compositional properties of the interpretation. So the book literally means the same thing whether or not it is in focus and read the book literally means the same thing whether or not the book is in focus.

Focusing is largely a pragmatic strategy, and especially the interpretation of non-wh focus is context-dependent and constrained by a number of discourse factors. For example, in Hindi-Urdu and Malayalam, positional focus constructions are not very good discourse initiators, since they presuppose discourse. Therefore, let us assume that the feature [+FOCUS] cannot be argued to be [LF [+Interpretable]].11 Rather, we propose that the interpretation a focused constituent gets is determined in a separate component inside PF, since the notion of presupposition relevant to the interpretation of focus is contextual, and pertains to the domain of pragmatics rather than LF. It is nevertheless part of the grammar as it is the domain that evaluates the felicity of what otherwise seem to be optional rules of a particular grammar. Wiltschko (1995) terms this hypothetical component D D (originally proposed by Chomsky 1981), and assumes it to be a level at which presuppositionality, and we claim focusing and coreference, effects are interpreted. In keeping with Chomsky’s suggestions (1995: 220), this interpretation is “accessed at the interface along with PF and LF.”12 We are thus faced with a feature that is outside the [LF[±Interpretable]] distinction, but which clearly requires licensing by the PF-interface. Suppose now, we extend the notion of Interpretability to accommodate feature-specifications that are PF[±I], requiring that [PF[−Interpretable]] features must be deleted by the PF-interface, but [PF[+Interpretable]] features, can survive unchecked, and do not delete even when checked. Now consider the feature [+FOCUS]. In all likelihood, this is a [PF[+Interpretable]] feature that survives to the interface. Indeed, this seems plausible when one recalls fact that languages usually have more than one way of marking this feature. Hindi-Urdu, for example, has not one but  ways — one, by position (with scrambling), two, by heavy word-stress (especially when the focus is to be given a contrastive

228

AYESHA KIDWAI

interpretation), and three, by affixation of the emphatic focusing particle -hii: (13)

a.

b.

c.

kitaab  laayegaa, (siitaa nahii) book Ram bring- Sita not ‘RAM will bring the book, not Sita.’  kitaab laayegaa, (siitaa nahii) Ram book bring- Sita not ‘RAM will bring the book, not Sita.’ -hii kitaab laayegaa, (siitaa nahii) Ram- book bring- Sita not ‘RAM will bring the book, not Sita.’

These facts are in consonance with the characterization of the feature [+FOCUS] as [PF[+Interpretable]], since it can be licensed at any of the levels internal to PF: either via PF-movement (scrambling), Morphology (-hii cliticization) or Phonology (stress). And, just like [LF[+Interpretable]] features, it can be accessed repeatedly by the ‘PF-computation’, since all three types of focus marking can be found in a single example, as in (10) above. The problem with classifying [+FOCUS] as a [PF[+Interpretable]] comes from wh-focus constructions. In at least two of the languages we have looked at in the paper — Hindi-Urdu and Malayalam — wh-focusing in the designated pre/postverbal position seems to be more obligatory than non-wh positional focusing, which can access alternative strategies of marking focus. This means that it is a feature that  be licensed by the time the construction arrives at the PF-interface, either by the PF-movement rules or by Morphology. And since we have claimed that PF-rules themselves are driven by the need to make linguistic items visible to Morphology, feature-checking in wh-focus in these languages actually has to be satisfied by the time the derivation reaches Morphology. The difference between the two types of foci, we believe to lie in the lexicon. Chomsky (1995: 235–241) proposes that the lexical entry of a linguistic item consists of its intrinsic semantic features, such as categorial features, Caseassigning features, gender, etc., that are not predictable from other properties of the lexical entry. The operation that forms the numeration adds on the optional features that come into play in the derivation — features of number, Case, tense, etc. The difference between wh- and non-wh-foci lies in the fact that the former are intrinsically [+FOCUS] elements in the lexicon, while for non-wh-foci the [PF[+Interpretable]] [+FOCUS] feature is added on in the formation of the numeration. Let us therefore exploit this distinction, and suggest that while the


feature [+FOCUS] is [PF[+Interpretable]], it must be checked by Morphology if it is an intrinsic feature of the element in question. Additionally, this requirement may originate from the role that wh-phrases play in Domain D, functioning more or less as place-holders for the assertion. If their intrinsic feature of [+FOCUS] is not licensed, they will not function as wh-questions at all. The feature thus remains [PF[+Interpretable]] for both wh- and non-wh-foci, except that for the former, the feature is somehow [stronger] by virtue of it being intrinsic to the lexical entry of the wh-phrase.13 We are going to keep with the traditional assumption that the feature [+FOCUS] is checked by a verbal projection, though not with the view that [+FOCUS] is a feature analogous to Case. It appears that focusing pertains more to the realm of the verbal paradigm than to the nominal one, and is in fact more predicational, and rather like mood, or choice of tense and aspect, i.e. it is more like a feature of T0 rather than V0. Chomsky (1996) suggests that overt verbraising can be eliminated by assuming T0 to be a super-category of V0, with the result that verbal affixes are checked by covert-raising. If this is correct, then it is possible for us to maintain that the PF-checking of the feature [+FOCUS] is done by V0. 3.3 Deriving Adjacency Finally, let us formalize the intuition that adjacency between the licenser and licensee is a requirement imposed by Morphology. Bobaljik (1994) proposes an account of adjacency in the PF-component that derives it from a UG requirement, that bans stray affixes. The adjacency condition, which states that “in order for an affix and a stem to be combined, they must be adjacent” then reflects one of the configurations in which this UG condition is met. Bobaljik suggests that: “As affixation is a morphological condition, adjacency is defined at (an intermediate stage in) the spell-out or mapping from syntax to phonology. Adjacency is sensitive, then, only to those elements that are relevant to the mapping process. Headedness is relevant (linearization), while traces and empty projections are irrelevant (PF-deletion).” In the system we are assuming here, where inflectional morphology is available presyntactically only as features, and where not only the hard-core morphological features of a lexical item require licensing but so do intrinsic features of lexical items, both the UG requirement and the adjacency condition, would have to be phrased in terms of licensing:

230

AYESHA KIDWAI

(14)

a. b.

UG Requirement on Morphology Features must be licensed. The Adjacency Condition Features can be licensed only under adjacency.

We shall assume that the adjacency requirement in Morphology follows from the general poverty of the component itself — as it lacks the operation Move, it can only operate on strings that it can see as linear. The relevant configurations for feature-licensing, however, differ from those in Bobaljik (1994: 2). Of the possible configurations in (15), only (15b) provides an appropriate environment for adjacency, as traces are phonetically empty material. (15)

a. b. c.

…X [YP NP[overt] …X [YP trace …X [YP adverb

[Y′ …Y… [Y′ …Y… [YP [Y′ …Y…

In the overt syntax, given the locality of checking domains, (14) is always satisfied, but at PF, the adjacency condition is relevant. PF-feature checking movement of the type we are considering here is therefore motivated solely to satisfy this adjacency requirement, which may vary further in its specific formulation across particular languages. Summarizing the results of the discussion so far: Move, as adjunction, is an option in the PF-component that is driven by the need to license [PF[±Interpretable]] features. This feature-licensing is Morphology-driven. FOCUS is a [PF[+Interpretable]] feature that can survive to the interface unchecked, but only if it is not intrinsic to the lexical item; in which case, this feature must be checked by Morphology/Domain D.14 The verb is the relevant licenser for the feature [+FOCUS]. Adjacency between it and the focused category is required by the fact that Morphology can only operate on strings that it can see as linear, since it lacks the operation Move. The feature [+FOCUS] on non-wh foci is interpreted in Domain D, a level where presuppositionality and focusing effects are interpreted. The feature is an intrinsic one of wh-foci, and thus it can be accessed for interpretation both at LF as well as Domain D. 3.4 Hindi-Urdu and Malayalam Earlier in the paper we saw that Hindi-Urdu scrambling served to somehow land the focused ±wh element into the preverbal focus position. In the framework we have proposed, this scrambling is Morphology-serving PF-movement, driven by


the need to establish an adjacency relation between the focalizing verbal feature and the focused category. Since Hindi-Urdu has overt verb agreement with the direct object, it is reasonable to assume that v licenses multiple specifiers at Spell-Out. The structure of (16) that arrives in the PF-component is therefore (17): (16)

kitaab - p6qhii book() Noor() read() ‘NOOR read the book.’

(17)

[TPi nuur-ne[+F] [vP[Spec2 kitaabj [Spec1 ti [v p6qhiiv[+F] ]]] [VP ti tj tv ]]]

Now, in a construction where subject focus is involved, the preferred word order involves DO scrambling. Given that it is a [PF[+Interpretable]] feature, this must be checked under adjacency with the verb. This requirement of adjacency is, however, not fulfilled in the structure at Spell-Out, since the shifted DO intervenes. Consequently, DO scrambling, expressed as left-adjunction to TP, takes place to yield adjacency between the verb and the subject. At PF, as an input to Morphology, (17) would then have the structure of (18): (18)

[TP kitaabj [TP -i[+F] [vP[Spec2 tj [Spec1 ti [v p6qhiiv[+F] ]]] [VP ti tj tv ]]]

After PF-movement, adjacency is achieved as only phonetically empty material intervenes between the subject and the verb, and the feature of [+FOCUS] on the subject is licensed in satisfaction of (13).15 Note that the proposals do not require PF-movement to necessarily take place, since the feature [+FOCUS], realized pre-syntactically only as a feature-specification, can be spelled out in different ways by the sub-components inside the PF-component. Hence, the feature can also be licensed by Morphology in (13a) and Phonology in (13b) and by all three strategies in (10) above. (13)

a.

b.

-hii kitaab laayegaa, (siitaa nahii) Ram- book bring- Sita not ‘RAM will bring the book, not Sita.’  kitaab laayegaa, (siitaa nahii) Ram book bring- Sita not ‘RAM will bring the book, not Sita.’

Two questions remain regarding Hindi-Urdu scrambling and focus. The first queries why rightward scrambling, which is analyzed as right-adjunction to VP, should not yield the same results, since it also results in adjacency between the

232

AYESHA KIDWAI

verb and an element that would not normally appear in that environment. Indeed for some speakers, preverbal focusing is indeed possible with rightward scrambling, though not preferred. What appears to be happening is that the presuppositionality effects are more difficult to calculate in rightward scrambled constructions. For example, in an example like (6a), is it is unambiguous to interlocutors that the scrambled DP is presupposed, specific and topical, but in (6d), none of these facts seem to be quite so clear. We suggest that this preference for leftward scrambling, i.e., left-, rather than right-, adjunction to VP follows from Fukui’s (1993: 400) P V P M: (19)

A grammatical operation (Move-a, in particular) that creates a structure that is inconsistent with the value of a given parameter is costly in that language, whereas one that produces a structure consistent with the parameter is costless.

This PVP measure classifies all those movements as optional and free which proceed in the canonical direction of the language. Now, Hindi-Urdu has a head-final setting of the Head Parameter, with the result that leftward movement is free. Rightward movement can, in this system, only take place, if it is needed for convergence, and hence the preference for leftward scrambling amongst native speakers. The other problem has to do with the oft-claimed “LF-effects” of scrambling, whereby clause-internal argument scrambling can license coreferential pronouns and monomorphemic reflexives. In the discussion on (8) and (9), we demonstrated that focused XPs in Hindi-Urdu are more amenable to coreference. Following Erteschik-Shir (1997), we suggest that this coreference is favored in scrambled constructions because of the fact that scrambled XPs tend to be interpreted as topics. Kidwai (1995) shows that even the putative reflexive binding in (8b) does not follow from the Binding Theory, but rather from the fact that in Hindi-Urdu monomorphemic reflexives may, in certain discourse contexts, be used referentially. Since these coreference possibilities are heavily discourse-conditioned, we propose that coreference is determined in Domain D. There is apparently no difference between Hindi and Malayalam as far as the structures that exit PF are concerned, and therefore the analysis for HindiUrdu carry over quite straightforwardly to Malayalam. The one difference between the two pertains to the obligatoriness of wh-focusing in the preverbal position — Hindi-Urdu does not obligatorily require preverbal focusing when the question is discourse-initial. This difference is not entirely unexpected in terms


of the model of PF that we are proposing here. If wh-focusing is interpreted as a Morphology as well as a Domain D requirement, in that the intrinsic feature of [+FOCUS] of a wh-phrase must be licensed before the PF output, it is expected that these requirements will be influenced by other factors related to languageparticular Domain Ds, although determining these factors is no trivial task. We shall not attempt it here, but clearly what is crucial for the analysis is the [+Interpretable] nature of the feature. Since it must survive to the interface even when licensed, the fact that discourse-initial questions in Hindi-Urdu may fail Morphology does not yield non-convergence at the interface. 3.5 Western Bade and Tangale The most important consequence of our proposals is that the notion of a focus position turns out to be a mere taxonomic artifact. Consider Western Bade first, an SVO language when either the focus is neutral or the DO is in focus: (20)

a.

b.

Saku aa b6naa kajluwaan Saku  cook tuwo ‘Saku will cook tuwo.’ Saku aa b6naa 6 Saku  cook what ‘What will Saku cook?’

Adapting the analysis of Tuller (1992), Western Bade does not appear to have object shift. Accepting Chomsky’s proposal that verb-raising in the VSO order takes place post-Spell-Out and to a position beyond the TP projection, the structure of (20b) that enters the PF-component would then be (21): (21)

[TP sakui[+F] [vP [Spec1 ti [v aa banaav[+F] ]]] [VP ti 6 tv ]]]

In this structure, the condition for adjacency between the question-word and the licensing V-feature is satisfied, and consequently no PF-movement is required. It is therefore expected that in all instances of DO-focus, the structure at SpellOut should by default meet the licensing requirements on focused phrases, thereby precluding the necessity of PF-movement in these cases. Consider now subject-focus in Western Bade, which we argue to have the structure in (22b) at Spell-Out:

234

AYESHA KIDWAI

(22)

a.

b.

gafa-n KE viiridgwar6n caught who giant-rat ‘Who caught a giant-rat?’ [T0 [vP [Spec1 i [+F] [v′ gafanv[+F] [VP ti tv viiriidgwar6n]]]]]

By the proposals in Chomsky (1995), the more standard analysis would be to locate the subject in [Spec,TP] at Spell-Out in order to satisfy the Extended Projection Principle (EPP). PF-movement in the form of verb-raising to C0 would then take place to establish the relevant licensing domain for the wh-phrase.16 This analysis would not be able to explain why this universal requirement is overridden in Tangale, where subject-focus requires a VOS order. We therefore adopt the interpretation of the EPP proposed by Chomsky (1996), where the EPP is taken to be a UG requirement of . We suggest a fairly literal interpretation is in order, by which anything which is a focus cannot be a subject. This will entail that the subject stays in situ in [Spec1, vP] at Spell-Out, and subsequent PF verb-raising to T0 licenses the [+FOCUS] feature of the whphrase. The question is that why such PF verb-raising is required at all, since in the order at Spell-Out in (22b) no material relevant to the blocking of adjacency intervenes between the subject and the verb substituted into v. We suggest that languages are parametrized with respect to what is considered the correct direction of adjacency by the Morphology of that language, and that this parametrization is sensitive to settings of the Head Parameter, as first suggested by Horvath (1986). In an SVO VP, the canonical direction of adjacency (rightward) cannot be respected in the case of the subject, and hence the [+FOCUS] feature of the subject is not licensed in the base configuration. Tangale represents a case of further parametrization of the definition of adjacency. As mentioned earlier, Tangale does not allow focused constituents to intervene between the verb and the direct object, as shown in (23): (23)

a.

b.

c.

wad Billiri ] dooji? go Billiri who tomorrow? ‘Who will go to Billiri tomorrow?’ tui worom mono  ate beans my she ‘SHE ate my beans.’ [TP T0 [VP [Spec2 OBJ [Spec1 SU [v′ VB [VP tsu tvb tob]]] ]]]

Adapting Tuller’s insights again, it appears that Tangale has overt object shift.


Consequently, the abstract structure of the Tangale sentence at Spell-Out will be as in (23c) above. If Tangale were exactly like Western Bade, we would expect the PF verb-raising mechanisms to T0 to license subject focus to necessarily yield the order VSO, but since the language has overt object shift, the observed surface order is VOS. While in the last object-shift language we looked at the shifted object was found to be a barrier for adjacency, clearly this is not the case in Tangale. It is however, reasonable to expect that different definitions of adjacency might be operating here as well, by which Tangale does not consider a second Spec to be a barrier for adjacency, whereas Hindi-Urdu does. That is, in Tangale, the mechanism that calculates adjacency simply cannot “see” a filled multiple Spec.17 Note, however, that Tangale has the identical setting for the directionality of adjacency parameter as Western Bade, since it too requires verbraising to license subject-focus. A consequence of these proposals is that we expect DOs, IOs and adjuncts in Tangale to be focused in situ by verb-raising to T0. The predictions are born out, since in those cases, the subject is in [Spec,TP] by Spell-Out.18 To conclude, Tangale and Western Bade demonstrate that the way in which a particular setting of Morphology determines adjacency turns out to be crucial for the form of focus constructions in that particular language. In these two languages, directionality is relevant, whereas in Hindi-Urdu and Malayalam, those requirements are met by default, since they are head-final languages. The choice of the type of adjunction (XP or X0) is a language-specific choice, since at least Hindi-Urdu appears to prefer XP-movement for focusing rather than X0-raising.

4.

Generalizing the Proposals

One of the consequences of the proposals made in this paper is that we would expect all languages to have scrambling and verb-raising for focusing to some degree, since the feature [+FOCUS] is not claimed to be particular to only languages with focus positions. This prediction is being confirmed by the ‘discovery’ of scrambling by Zubizarreta (1996) for Spanish and French, a movement rule that performs the very same function as in the languages we have discussed here. There are also some instances in English where only PF-movement can serve to license [+FOCUS] elements. The first and the most obvious one is whquestions in the language, since there the [+FOCUS] element does not bear the

236

AYESHA KIDWAI

primary stress. In the analysis of wh-movement in Chomsky (1995), the structure that exits Spell-Out of a query like (24) could be either (25a) or (25b), but never (25c), since the [strong] Q-feature need only be satisfied once for convergence: (24)

[CP Q [IP you will read what]]

(25)

a. b. c.

[CP [Co will-Q [IP you read what]]] [CP what [Co Q [IP you will read]]] [CP what [Co will-Q [IP you will read]]]

While the analysis has the advantage of unifying an analysis of yes-no questions with wh-ones, it fails to delineate why when wh-phrases are involved, a yes-no interpretation is impossible, and AUX-raising to C0 in object questions is obligatory. Of the three structures in (26), the structure that cannot exit PF, (25c), is the only legitimate PF-output. The proposals we make regarding the licensing of wh-phrases in languages without wh-movement to [Spec,CP] carry over quite nicely to explain why this happens. As stated, wh-phrases are universally intrinsically specified for a [+FOCUS] feature that must be licensed by Domain D, which is where the difference between yes-no and wh-questions becomes relevant. A yes-no question, while it satisfies the Q-feature of a clause, does not have the same type of consequences for the discourse as the wh-question, since only the latter creates a presupposition-assertion configuration. In a Spell-Out structure like (25a), which has the form of a yes-no question, the [+FOCUS] wh-phrase cannot be licensed as the lexical subject intervenes between the object and the licensing auxiliary. PFmovement of the wh-phrase to adjoin to CP (substitution no longer being an option) is ruled out by the PVP-measure, leftward movement being too expensive in a head-initial language. Therefore, the wh-phrase cannot raise to be leftadjacent to the auxiliary, and hence its feature [+FOCUS] cannot be licensed. wh-raising unaccompanied by do-support or auxiliary raising do not converge for the same reasons — the raised wh-phrase is not adjacent to its licenser, the auxiliary bearing the feature [+FOCUS] without such raising. Verb-raising at PF, an option in English, is therefore necessary to license the wh-phrase, for it to be interpreted as a question. Note that directionality of adjunction does not appear to be an issue in English, at least as far as auxiliaries are concerned. This ability to check [+FOCUS] in the non-canonical direction appears to be a special property of English auxiliaries alone, since English focus constructions with main verbs usually involve structural focusing to the right, rather than the left. Consider


stylistic inversion of locatives in English, where the focused element occurs to the right of the main verb: (26)

a. b.

Into the forest nude ran   In front of her sat  

Culicover and Rochemont (1990) analyze these constructions as involving VPtopicalization, followed by V-to-I, and inversion of I around the subject in [Spec,IP]. Although their claims about VP-topicalization seem to be on the right track, the proposals regarding verb-raising do not seem to be correct. It is difficult in this analysis to explain why constructions such as the ones in (27a–b) never surface, since auxiliaries are the only verbal categories that normally undergo V-to-I in English. Nor can it explain why main verb-raising in the presence of an auxiliary is good, or at least okay, in these contexts (27a–b), but terrible in all other cases of V-to-I, viz., the extreme ungrammaticality of *Who had given you the book to?: (27)

a. *Into the forest had   run b. *In front of her had   sat c. #Into the forest had run   d. #In front of her had sat  

Suppose, then, we discount this part of their analysis, and maintain that, in fact, what is involved here is PF-movement of the verb. On analogy with our interpretation of the Extended Projection Principle for Tangale, let us assume that something which is a focus cannot simply raise to [Spec,TP] in the overt syntax. However, English is different from Tangale in that it allows VP-topicalization to satisfy the EPP. The PP raises to [Spec,TopP] and the order that exits Spell-Out in these cases is then: (28)

[TopP[VP tv into the forest nude] [TP [vP [Spec1 [+F] [v V[+F] ]][VP tVP]]]]

In this structure, the direction of feature-licensing is not compatible with the VO setting of the Head Parameter, and therefore V-raising by PF-movement must take place to establish the correct adjacency relation. In examples involving auxiliaries, then, the facts follow straightforwardly — (27a–b) are ruled out because the feature [+FOCUS] of the subject will not be licensed in the configuration. The auxiliary, in all the examples in (27), we shall assume to be in a projection higher than the subject, and hence it undergoes no movement. (27c-d)

238

AYESHA KIDWAI

are therefore better than (27a–b) because there is no unlicensed feature in the construction. The oddness of the construction is dissolved by embedding in discourse:19 (29)

… You’ve got it all wrong! At the wedding, in  of her had sat her mother, at her , her father, and to her , her brother!

If these claims are true, and main V-raising in English is a PF-movement rule for focus-licensing, then the impossibility of main verb raising in questions follows — main verb-raising will never satisfy the adjacency requirement.

5.

Conclusion

Within the pre-minimalist approaches to UG, the study of focus constructions in natural language were necessarily plagued with the apparent irreconcilable differences between the mechanisms that languages use to indicate focus, as the strategies necessarily involved more than one component of the grammar. As we have seen in this paper, current minimalist approaches open up avenues of inquiry into focus in natural language, the results of which lead to a re-appraisal of the very hypothesized shape of UG itself.

Notes * I am grateful to Anvita Abbi, Hans Bennis, Michael Brody, Noam Chomsky, Probal Dasgupta, Jacqueline Lecarme, Nomi Erteschik-Shir, Laurie Tuller and Maria-Luisa Zubizarreta for comments and suggestions on issues concerning word order and focus. Needless to say, the views held in this paper are not necessarily shared by any of them, and the mistakes and oversights are quite definitely my own. 1. Josef Bayer (personal communication) points out the necessity of spelling out the fact that the notion of presupposition relevant to truth is  presupposition rather than logical presupposition. That is, we argue that truth-conditional differences with focused alternatives arise from an evaluation in terms of a shared set of pragmatic speaker-presuppositions. This, as Zubizarreta 1996: 3–4) observes, explains why the (logically presupposed) complements of factive predicates can be focused. That is, a discourse like the one below: A: B:

I thought you realized that Mary had a husband I did! But I didn’t realize that Mary was .

Zubizarreta suggests that what is asserted by B is not the presupposition denoted by the complement, but rather that “the denial expressed by the proposition expressed by the

WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 239 complement was part of the presuppositional set of the interlocutors prior to the present.” The focus-relevant notion of presupposition is thus contextually defined. 2. Context-induced contrastive and presentational interpretations are not typical to foci alone, as the interpretations a topic may receive are identically constrained by context — if a contrast set of topics is available, the topic will be interpreted contrastively: A: Tell me about Ram and Sita. Do you like them? B: , I like (but Ram I cannot tolerate). These observations hold for morphological topics as well. Miyagawa (1989) and Kidwai (1995) trace the contrastive interpretation accorded to -wa topics in Japanese and -to topics in HindiUrdu respectively, to the role of the context-set provided by the discourse. 3. In Malayalam and Hindi-Urdu, IO focusing also involves a non-canonical linear order, where the DO is either scrambled to a VP- or IP- adjoined position: (i)

(ii)

(iii)

(iv)

m7˜ -ne kitaabi - ti dii I() book() Sita() gave ‘I gave the book to SITA.’ kitaabi m7˜ -ne - ti dii book() I() Sita-() gave ‘I gave the book to SITA.’ john oru pustakami -6 ti ko2uttu John() a book() Sita() gave ‘John gave the book to SITA.’ oru pustakami john -6 ti ko2uttu a book() John() Sita() gave ‘John gave the book to SITA.’

(Hindi-Urdu)

(Malayalam)

In both languages, DOs may be interpreted as focused in situ, either by employing heavy contrastive word-stress or by affixation of the emphatic focusing particle onto the DO. 4. The abbreviations used in the is article are:  = Subject,  = Direct Object,  = indirect Object,  = Ergative Case,  = Dative Case,  = Accusative Case,  = Future Tense,  = Emphatic particle. Focus is indicated by upper case. 5. Hindi-Urdu also exhibits the phenomenon of long-scrambling of arguments. While longscrambling out of finite complement clauses is accepted by only a minority of speakers (e.g. Mahajan 1990), speakers are unanimous about the acceptability of long-scrambling out of nonfinite complement clauses. The facts of preverbal focusing in these configurations are quite complex, but appear to center around the basic generalization that long-scrambling of an XP licenses preverbal focusing only of a constituent of the clause of origin of the scrambled XP itself. In (i) below, where long-scrambling moves the embedded DO to the matrix clause, the embedded IO is focused, while in (ii), where the whole non-finite complement clauses is longscrambled, the matrix subject is interpreted as the focus: (i)

kitaabi m7˜ -ne˜ [- ti dene-ka] vaadaa kiiyaa book I Sita- to give- promised ‘It was Sita that I had promised to give the book to.’

240

AYESHA KIDWAI (ii)

[siitaa-ko kitaab dene-ka]i - ti vaadaa kiiyaa Sita- book to give- I promised ‘It was I who had promised to give the book to Sita.’

Speakers who allow long-scrambling out of finite complement clauses also subscribe to this generalization. 6. Scrambling the verb to a sentence-initial position, as in (i), does not involve preverbal focusing of arguments: (i)

aauungaa kal m7˜ come -  tomorrow I

Interpretative judgments on these constructions vary according to context — either the utterance is interpreted as one in which the verb itself bears narrow focus or as one in which the whole clause is presupposed. We shall have little to say about these constructions in this paper. 7. Cf. Tangale, which observes the order V-DO-FOC rather than expected, but actually prohibited, order of V-FOC-DO. 8. Malayalam patterns very closely with Hindi-Urdu in that argument-scrambling can override WCO effects. We will assume that the analysis of Hindi-Urdu scrambling as XP-adjunction that we provide below carries over to Malayalam. 9. This is not, however, to be taken to imply that FP analyses ignore the question altogether. Brody (1996) suggests that positional focus languages have overt movement, triggered by the [strong] features of F0, to a universally available FP projection and that languages that apparently do not, actually exhibit covert raising to [Spec,FP] at LF. He claims English to be such a language, as association with focus is clearly subject to island constraints in the language: (i) (ii)

John only likes  = It’s only MARY that John likes John only said that to meet  would be nice ≠ It’s only MARY that John said that to meet would be nice

The fundamental assumption in Brody’s approach is that the feature checked in [Spec,FP] is the feature [CONTRASTIVE FOCUS]. This enables him to unify what are, intuitively, very distinct types of focus — the focusing involved in association with focusing particles, which we refer to as  , appears to originate in the set of grammatical conditions imposed upon an operator like only — only must be associated with a focused element in its c-command — whereas positional focusing, which we refer to as  , derives its existence from speaker intentions and the properties of wh-phrases, but not from the immediate grammatical environment. The two types of focus also differ in interpretation, as even while the truth of both types of focus is adjudicated on the basis of a shared context set, only bound focus forces the contrastive interpretation of the focused constituent, since the operator only obligatorily partitions that context set into a contrast set. In free focus, on the other hand, a contrastive focus interpretation of the focused element is not obligatory, and even when it does obtain it is not induced by a grammatical operator analogous to those like even and only. 10. Hindi-Urdu and Malayalam are syllable-timed languages that lack the Nuclear Stress Rule

WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 241 altogether. The use of word-stress for contrastive focus is however an option in the grammar. 11. It has been argued that in positional focus languages, [+FOCUS] is a [−Interpretable] feature like Case. It turns out that besides the adjacency facts and the fact that it is a feature that appears to be checked by a verbal projection, there is very little actual support for this characterization of focus — it does not seem to be the property of particular types of predicates, the morphological form of foci and verbal morphology is invariant across all GF-foci. 12. Josef Bayer (personal communication) points out that our characterization of focus as pragmatic and hence evaluated in terms of speaker-presuppositions at Domain D attributes too much to the PF-interface to really distinguish it from the LF-interface, as we conceive of it currently. While this is true, at least part of the research agenda of the approach suggested here would be to restrict interpretations at LF to core semantic properties of natural language by transferring the burden of discourse-dependent interpretations to Domain D. This is by no means a trivial task, and further raises the question as to whether there can be Domain D and LF interactions, and how these are to be captured. 13. This difference could, in fact, be the reasoning behind a claim like Chomsky’s (1992) that +wh is universally [strong], in a way that no longer identifies overt wh-movement as the basis on which issues of strength are decided. That is, the universal strength of wh- lies in its [+FOCUS] intrinsic feature, which must be licensed by Morphology inside the PF-component. An obvious problem with this proposal is the problem of wh-in-situ in languages with overt wh-movement. It appears that languages may actually differ as to whether the features of each [+wh]-element must be licensed (cf. Hungarian), or whether the licensing of the [+FOCUS] feature of one [+wh] is enough to meet Domain D requirements. In this context, note that languages like German tend to locate wh-in-situ items leftadjacent to the verb (Josef Bayer, personal communication): (i)

a. b.

wer hat bei den Hausafgaben  geholfen who has with the homework who helped ?wer hat  bei den Hausafgaben geholfen

Although this data is not foolproof, as shown by (ii), it could be argued that German licenses wh-in-situ by positional focus. The data in (ii) could be explained by considering the PP nach Hause to be part of the lexical meaning of nach Hause fahren, hence the adjacency condition is met by considering the complex predicate to be the relevant category licensing the in-situ whphrase. (ii)

a. b.

wer hat  nach Hause gefahren who has whom to home driven *wer hat nach Hause  gefahren

Bayer, in fact, explicitly cautions against this analysis, since it is also true that the PP nach Hause can be topicalized, thereby suggesting that it does not constitute a complex predicate with the verb. This conclusion, however, is hasty, as there are many languages, e.g. Hindi-Urdu, which also allow PP and NP constituents of what are clearly complex predicates to be extracted. Furthermore, the fact that a constituent can be topicalized does not necessarily entail that it may be scrambled (the operation that enables positional focusing). In fact, if topicalization involves substitution into a Spec position as suggested by Müller & Sternefeld (1993), it must be a pre-

242

AYESHA KIDWAI Spell-Out movement, and hence irrelevant to the discussion. With this in mind, consider the contrast between (i) and (ii). In (ia) what is clearly involved is PP-scrambling, i.e. the word order variation required for positional focusing. In (iia) this operation is disallowed, because nach Hause is part of the [+FOCUS] licensing predicate, and its scrambling is therefore prohibited as in (iib). The fact that the preverbal focusing requirement is not very strong in German (witness the relative acceptability of (ia)), however, then puts the question back into the domain of language-particular enforcement of Domain D requirements.

14. Noam Chomsky and Jacqueline Lecarme (personal communication) question this characterization of [+FOCUS] as a [PF[+Interpretable]] feature. Chomsky suggests that it may well be that focus interpretation takes place when various conditions like adjacency obtain, and Lecarme suggests the proper characterization of focus to be a post-Spell-Out feature adjoined in the PFcomponent. Both these suggestions, even while they have the same intuitive force insofar as they also do not characterize [+FOCUS] as a core property of lexical items, fail to explain the link between word order and positional focusing. That is, if [+FOCUS] is not a feature that requires licensing, and hence the PF-reordering in positional focus constructions, then what is the driving force behind the XP/X0 adjunction operation that takes place in scrambling positional focus languages? 15. Similar accounts of DO and adjunct focus can be given, assuming that IOs are generated as vP adjuncts in (some) languages without the dative alternation (Kidwai 1995) and with the order IO-DO — the DO must be scrambled out of the way for adjacency between the licenser and licensed. 16. See also Belletti & Shlonsky (1995) for a similar proposal regarding the link between verbraising and focus in Italian. 17. Laurie Tuller (personal communication) cautions against a possible prediction of our analysis by which a language like Tangale would allow SOV   , where the subject is in [Spec,TP], the object in [Spec2, vP], and the verb remains in situ inside the VP. This prediction would, however, follow from an absolute correlation between verb-raising and [+FOCUS] feature-checking — a thesis too strong to be maintained. A more tenable position would be one that allows PF verb-raising to be triggered in satisfaction of morphological requirements other than [+FOCUS] feature-checking as well. PF verb-raising in neutral focus utterances could then still be obligatory, but with a different morphological imperative from the one involved in non-neutral focus utterances. 18. However, there is a problem here, since Tuller, citing Kenstowicz (1985), provides evidence to show that tone sandhi processes in the language argue for a distinction between the syntactic position occupied by a focused DO versus a nonfocused one. We have no account for this difference, but it may well be possible that the rules that ensure the licensing of certain PFfeatures are not the only rules there are, and subsequent processes either physically disassociate the DO from its shifted position, or mark it as unavailable for the sandhi processes involved. 19. Significantly, (32) involves cases of paired contrastive focusing, with the topicalized VP bearing narrow focus as well. Topicalization in English has often been noticed to involve narrow focus. If auxiliaries in English also raise by PF-movement, then the fact they can somehow license the feature [+FOCUS] leftward explains the narrow focus on the topicalized VP/PP. Although the question why all topicalization in English does not require verb-raising at PF must remain unanswered.


References Bayer, Josef. 1996. This volume. Belletti, Adrianna & Ur Shlonsky. 1995. “The Order of Verbal Complements: A Comparative Study”. Natural Language & Linguistic Theory 13.489–526. Bobaljik, Jonathan. 1994. “What does Adjacency do?”. In Heidi Harley and Colin Phillips (eds.), 1–31. Brody, Michael. 1989. “Some Reflections on the Focus Field in Hungarian”. University College London Working Papers in Linguistics 2.201–225. Brody, Michael. 1996. “Focus in Perfect Syntax”. Paper presented at the Table ronde internationale sur la grammaire du focus, University of Paris II & University of Paris X, February 1996. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press:. Chomsky, Noam. 1996. “Chapter 5”. Talk given in Central Institute of English & Foreign Languages, Hyderabad. Culicover, Peter. 1993. “Focus and Grammar”. In P. Ackema and M. Schoorlemmer (eds.), Proceedings of the Workshop on the Semantic and Syntactic Analysis of Focus. Utrecht: OTS, 1–18. Culicover, Peter and Michael Rochemont. 1990. English Focus Constructions and the Theory of Grammar. Cambridge: Cambridge University Press. Cummins, Sarah and Yves Roberge. 1994. “Romance Inflectional Morphology In and Out of Syntax”. In Heidi Harley and Colin Phillips. (eds.), 53–70. Erteschik-Shir, Nomi. 1997. The Dynamics of Focus Structure. Cambridge: Cambridge University Press. Fukui, Naoki. 1993. “Parameters and Optionality”. Linguistic Inquiry 24.399– 420. Halle, Morris and Alex Marantz. 1993. “Distributed Morphology”. In Samuel Jay Keyser and Kenneth Hale (eds.), The View From Building 20. Cambridge, Mass.: MIT Press, 111–176. Harley, Heidi and Colin Phillips (eds.), The Morphology-Syntax Connection. MITWPL 22, MIT. Horvath, Julia. 1985. Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Jayaseelan, K.A. 1989, 1995. “Question-Word Movement to Focus and Scrambling in Malayalam”. Ms., Central Institute of English & Foreign Languages, Hyderabad.

244

AYESHA KIDWAI

Kenesei, István & Irene Vogel. 1989. “Prosodic Phonology in Hungarian”. Acta Linguistica Hungarica 39.149–193. Kenstowicz, Michael. 1985. “The Phonology and Syntax of wh-expressions in Tangale”. Studies in the Linguistic Sciences 15.2. Kidwai, Ayesha. 1995. Binding and Free Word Order Phenomena in Hindi and Urdu. Doctoral dissertation, Jawaharlal Nehru University, New Delhi. Müller, Geroen. & Wolfgang Sternefeld. 1993. “Improper Movement and Unambiguous Binding”. Linguistic Inquiry 24.461–507. Rochemont, Michael. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins. Tuller, Laurice. 1992. “The Syntax of Postverbal Focus Constructions in Chadic”. Natural Language and Linguistic Theory 10.303–334. Watters, John. 1979. “Focus in Aghem: A Study of Its Formal Correlates and Typology”. In Larry Hyman (ed.), Aghem Grammatical Structure. Southern California Occasional Papers in Linguistics 7.137–198. University of Southern California, Los Angeles. Wiltschko, Martina. 1995. “Syntactic Options vs. Interpretational Options: Licensing at Different Levels of Interpretation”. Talk given at the Workshop on Optionality, September 1–2, at OTS, Utrecht University. Zubizarreta, Maria-Luisa. 1996. Prosody, Focus and Word Order. Ms., University of Southern California, Los Angeles.

Focus in Wolof A Study of What Morphology May Do to Syntax Alain Kihm CNRS, Paris

Abstract This study of focus in Wolof (Niger-Congo, spoken in Senegal) is a study of how morphology may opacify syntactic structures. Focusing of arguments in Wolof can only be achieved through a construction that is shown to be similar to English clefts ‘it is x that…’. The crucial difference with English is that the ‘it is’ element of Wolof is a multicategorial item incorporating the copula and its immediate arguments. The special properties of Wolof clefts, in particular clause-straddling by the pivotal element, follow from this. Wolof facts also support the assumption that the relative clause of clefts is predicated, not of the NP it seems to follow, but of the variable spelled out as ‘it’ in English, so that clefts and pseudo-clefts are derived from the same initial structure. A copy-theory account of this derivation is proposed.

1.

Introduction

‘Focus’ is a notoriously multivalued term. In this article it is taken only in the sense of contrastive focus, as expressed in English cleft sentences such as It is Joyce who wrote Ulysses. Such utterances convey both an assertion (viz. Joyce wrote Ulysses) and an implied contrastive negation (viz. No other author but Joyce wrote Ulysses). They typically occur in debating contexts — e.g. as a rejoinder to somebody who has just claimed (or who we think would be up to claiming) that, say, Homer wrote Ulysses. From this discursive condition on the

246

ALAIN KIHM

felicity of contrastive focuses follows naturally the fact that they always impart new information in the context of the discourse or the interchange. The interest of unraveling the syntax and morphology of Wolof focus constructions is twofold.1 First of all, despite the opacifying effect of morphology, focused sentences in Wolof, it will be shown, are in fact bi-clausal clefts just like their English counterparts, and do not merely involve a so-called ‘emphatic conjugation’ as concluded in previous studies (see references in fn. 1). Note, moreover, that such a construction is the only means in the language to put an element into focus. In particular, focus stress as in English or focusing through position as in Hungarian (see Horvath 1985) are not available.2 Wolof contrastive focusing thus represents a case study in the way morphology will blur otherwise common syntactic structures, being the real locus of language variation and ‘imperfection’ (see Chomsky 1995). The facts of Wolof are also illuminating for the syntax of clefting itself. It is well known that clefts in English and perhaps generally are potentially ambiguous between a focused and what may be called a ‘presentational’ interpretation (see the studies by ClechDarbon et al., and by Ouhalla, this volume, for an analysis of clefts). For instance, It is the horse that I bought may be understood as meaning either ‘What I bought was the horse (not the cow, etc.)’ or ‘What you see here is the horse that I bought’. This suggests that the phonological form of the sentence covers two different structures: one yielding the presentational reading, where the relative clause [that I bought t] is indeed predicated of the horse as its surface position suggests it to be; the other yielding the focused reading, where the same relative clause is in fact predicated of the logical variable spelled out as it in English, and is then ‘extraposed’ to the end of the sentence. Clefts and pseudocleft can thus be shown to have a common origin. In Wolof, interestingly, the two structures do not get mixed up at PF. It is therefore a morphological peculiarity of English and other languages that a confusion occurs at this level. Wolof thus provides evidence both for the double structure of apparent clefts and, in addition, for the fact that morphology alone is the dissembling factor.3 The article is organized as follows. Section 2 below is devoted to an examination of copular sentences which are of central relevance to our main topic, since clefts take an ‘it is’ expression as their pivot. Focused constructions are then described in Section 3, including negative focusing (‘It is not X that…’), which raises special issues in Wolof, and an analysis is proposed in Section 4. In Section 5 I deal with a few complex problems of the Syntax-PF interface raised by Wolof focused constructions. In the conclusion, I return to the issue of the

FOCUS IN WOLOF

247

proper locus of language ‘imperfection’, and I briefly discuss the appropriateness of a Focus projection (FocP) in Wolof and generally (see Rizzi 1998).

2.

The structure of copular sentences

Wolof makes a clear distinction between identificational (1a) and predicational (1b,c) sentences (for the distinction, see, e.g., Pollock 1983):4 (1)

a.

b.

c.

Góor gii di sunu njiit man this be our chief ‘This man is our chief Góor gii, sunu njiit la man this our chief . ‘This man is our chief’5 Fas la horse . ‘It is a horse’

What (1a) asserts is an identity relation between the two terms ‘this man’ and ‘our chief’. Typically, (1a) can be inverted without changing the informational content of the proposition (Sunu njiit di góor gii ‘Our chief is this man’). Since Wolof is an SVO language, identificational copular sentences appear structurally similar to non-copular simple sentences.6 Predicational copular sentences (1b,c), in contrast, look as if they have a completely different structure. In (1b,c), an attribute is predicated as a property of a subject, which in English or French (see fn. 5) is a third person neuter pronoun (it, ce), but is not visible in Wolof. The issue of the syntactic category of la, glossed as a third person singular copula, is thus raised. That la is a verbal element is demonstrated by the following two examples: (2)

Fas lawoon horse .. ‘It was a horse’

(3)

Jaaykat lañu merchant . ‘They are merchants’

In (2), we see that the Past or Perfective marker /-(w)oon/ which only attaches

248

ALAIN KIHM

to [+V] lexical items can be suffixed to la.7 (3) further shows that la incorporates the person and number features of the subject of which the property ‘merchant’ is predicated, building the following paradigm:8 (4)

laa ‘I am’, nga ‘you are (sg.)’, la ‘s/he/it is’, lanu ‘we are’, ngeen ‘you are (pl.)’, lañu ‘they are’

Given the undeniably verbal character of la, therefore, two hypotheses can be made concerning (1b,c) and (2)–(3).9 One consists in considering that the overt word order of these sentences directly mirrors the underlying syntactic structure. This is the approach adopted in non-generative studies, but also in Njie (1982).10 The implication is that predicational copular sentences present the order Predicate-Subject (e.g. ‘a merchant we’ in (3)), in contradistinction to identificational copular sentences and verbal sentences generally, which all instantiate the order Subject-Predicate (see above). la’s category and syntactic role then remain a complete mystery. A more illuminating possibility (and the only one compatible with the Universal Base Hypothesis — see Kayne 1994; Zwart 1997) is that la-sentences do share the same phrase structure with other sentence types, but that this structure gets blurred by special morphological processes. A correlate of this hypothesis is that la, far from being a simple lexical item, represents the morphological compacting of a complex syntactic structure (a multicategorial word in the sense of Solà 1996). Since the issue is directly relevant to the study of focusing, as already mentioned, I will pursue it in detail. The gist of the proposed analysis is that la compacts an entire VP, so that, e.g., (1c) has the overall structure shown in (5), where fas ‘horse’ is basically topicalized in Spec,CP: (5)

[CP fas [IP… [VP la]]]

Construing la as a VP entails that (a) it must incorporate a subject in its specifier and possibly a complement; (b) its morphological make-up must somehow mirror the syntactic hierarchy of the terms (for instance in the sense of Baker 1985). How then shall we break up this item? Let us review paradigm (5). We see three recurrent elements: (i) initial /l/, except in the second person singular and plural; (ii) medial or final /a/, except in the second person plural; (iii) a final person-number morpheme clearly distinct at all persons, except the second and third singular. As we shall see when we come to the study of focused constructions themselves, there are good reasons to

FOCUS IN WOLOF

249

view /a/ as an autonomous verbal morpheme. I will therefore assume that /a/ represents the predicational copula (as distinct from the identificational copula di), to be noted a (this analysis is already found in Kobès 1869). Initial /l/, in its turn, may be identified with the noun class prefix L- of generic reference.11 Given their semantics and frequent use as agreement morphemes (see, e.g., Greenberg 1978; Creissels 1991), noun class markers seem to be best analyzed as pronominals. I therefore assume /l/ to be the subject (Spec,VP) of the a copula (see below for the second persons), i.e. a neuter pronominal unmarked for number, spelling out a logical variable and analogous to English it or, better still, to French ce in c’est/ce sont ‘it is’.12 A complementary assumption is that a takes only /l/ as a pronominal subject, for which we may bring in French again: see the unacceptability of *Il est un écrivain with the predicational interpretation (OK C’est un écrivain ‘She/He is a writer’). Moreover, no subject but /l/ is possible, due to the obligatory topicalization of all lexical arguments (see below). Finally, I assume that the postposed person-number morpheme is a clitic element representing the complement of the a copula. That is to say, laa means ‘it’s me’, nga means ‘it’s you’, etc. In the third person singular, the clitic is phonologically null. In the second persons, it absorbs the stem /l-a/, resulting in a form that is homophonous with the second person subject pronouns.13 In this way, the argument grid of the a copula is saturated, hence the necessity for the lexical item (e.g. fas ‘horse’ in (1c)) predicated of the logical variable (i.e. /l/) to be basically merged into Spec,CP, that is topicalized, this being the only free position. I will therefore assign the following syntactic structure to la, where Pro means ‘pronominal’: (6)

[VP l [V′ a [DP Pro]]]

Given the SVO typology of Wolof, no further operation can modify this basic order. Compacting /l-a-Pro/ into laa, nga, etc. is then an operation of Morphology (viewed as in Halle & Marantz 1993). A brief comparison with French may highlight the properties of Wolof copular sentences. Compare (1b) (Góor gii, sunu njiit la ‘This man is our chief’) with its French, closer translation, Cet homme, c’est notre chef. In both languages (universally, we want to assume), OUR CHIEF is the semantic attribute or property predicated of the semantic subject THIS MAN. In French, the semantic subject (cet homme) appears in topic position, while the semantic attribute (notre chef) appears in complement position. The neuter pronoun ce, in syntactic subject

250

ALAIN KIHM

position, is therefore coindexed with the semantic subject.14 In Wolof, if my analysis is correct, there is double topicalization, first of the semantic subject (góor gii ‘this man’), then of the semantic attribute (sunu njiit ‘our chief’). The syntactic pronominal subject /l/ can only be coindexed (in the broad sense alluded to in fn. 14) with the latter, since the former may be entirely absent from the structure (see (1c) Fas la ‘It is a horse’). What is the semantic subject coindexed with? The only possible candidate is the complement pronominal. This is confirmed by such examples as the following: (7)

Yow sunu njiit nga15  our chief . ‘You are our chief’

with a structure as shown in (8): (8)

[CP yowi [CP sunu njiitj [IP… [VP lj [V′ a [DP Proi]]]]]]

In (8), coindexation holds under c-command as it must. Note, however, there is a shortest link condition on the  coindexation, viz. that of the semantic attribute with /l/. Indeed, yow is not a necessary constituent of (7), which may appear as Sunu njiit nga ‘You are our chief’, with the same meaning minus some emphasis. In that case, the [Pro2SG] element of the copula is not coindexed with anything, but discursively bound. This means that, metaphorically speaking, yow (or góor gii in (1b)) is merged into the structure only after the minimal coindexation of the semantic attribute with the syntactic subject /l/ has been realized, and it is then coindexed with the remaining complement pronominal, which it c-commands in any event. The surface ordering is thus explained, as well as the fact that reversing it causes ungrammaticality (see */sunu njiit yow nga/). One question we might ask is: Why is Pro phonologically null in the third person singular while there is a clitic subject pronoun mu ‘s/he/it’? Why, in other terms, do we not find a form */lamu/?16 In fact, the unavailability of such a form suggests that the null Pro of la should be analysed as maximally underspecified in the sense of Benveniste’s (1966) ‘unperson’. That is to say, the null Pro of Fas la ‘It’s a horse’, with la analyzed as /l-a-Ø/, is discursively bound by the non-entity that would answer the question ‘What is a horse?’ where what questions it.17 Such an element may be called a pure expletive lacking phifeatures (see Chomsky 1995: 288), therefore distinct from the third person singular pronoun mu ‘s/he/it’ one finds in, e.g., mu rot ‘s/he/it fell’.18 Note that (1b) — Góor gii, sunu njiit la‘This man is our chief’ — where Ø is coindexed with

FOCUS IN WOLOF

251

góor gii according to (8) — does not run counter to this assumption. Indeed, the crucial distinction here is that between nonmatch and mismatch (see Chomsky 1995). Between an expletive lacking phi-features and a third person singular expression such as góor gii ‘this man’ negatively defined as [−1, −2, −Number] there is only nonmatch, which does not cancel the derivation. There is mismatch, in contrast, with a positively defined first, or second, or third plural expression, which explains why */yow sunu njiit la/ *‘You, it’s our chief’ is ungrammatical. As for the appropriateness of assuming a null expletive rather than nothing at all, this will become manifest in the study of focused constructions in the next section. The inner constituency of la may be summarized by saying that la includes a double orientation: towards the (semantic) subject and towards the (semantic) attribute. Now, this formulation takes us back to French and to the fact that, in this language, the copula optionally agrees in number with the attribute rather than with ce, which becomes evident when the semantic (topicalized) subject with which ce is (broadly) coindexed and the attribute differ for this feature (see, e.g., Ce groupe, ce sont nos chefs vs. Ce groupe, c’est nos chefs ‘This group are our chiefs’).19 Above, we established that ce corresponds to the Wolof complement pronominal Pro incorporated into la in that both are coindexed with the semantic subject.20 The comparison with Wolof and the possibility of (apparent) agreement of the copula with the attribute, however, suggest that ce might also include an optionally activated feature matching the attribute, and similar to Wolof /l/. If that is so, the difference between French and Wolof would then reduce to the morphological fact that subject and attribute features are fused within ce and can only be alternatively activated, whereas they are independently realized in Wolof as /l/ and Pro. This would further explain why double topicalization is impossible in French (as in *Cet homme, notre chef, c’est on the model of Góor gii, sunu njiit la), because ce would have to be doubly coindexed, which it cannot be given the alternative activation of the features it includes.21 What about the spelling out of the complex element la as one of the members of paradigm (4)? One possibility, at least, seems to be ruled out, viz. head raising and adjunction (head-to-head movement), if we adopt Kayne’s (1994) hypothesis that adjunction is obligatorily to the left. For instance, adjoining the copula a (a V0) to /l/ (probably a D0) would yield the wrong ordering with a to the left of /l/. I will therefore keep to the simplest assumption, namely that (6) is merged as such in the syntax. As already mentioned, given Wolof’s SVO typology, no further movement (e.g., of /l/ from Spec,VP to

252

ALAIN KIHM

Spec,AGRsP) can modify the basic ordering of the elements of la, i.e. [l-a-Pro]. Compacting [l-a-Pro] to one phonological word is then entirely an operation of the component Morphology. The operation may be trivial agglutination as in the case of lanu /l-a-nu/ ‘it’s us’ or la /l-a-Ø/ ‘it’s her/him/it’; or it may involve more complex morphophonological processes, as with laa /l-a-ma/ ‘it’s me’ or nga /l-a-nga/ ‘it’s you (sg.)’. In both cases, we see here an instance of fusion (see Halle & Marantz 1993: 116), implying that the members of paradigm (4) are indeed Vocabulary items.22 I will return to this issue.

3.

The structure of focused sentences described

All arguments, i.e. subjects, objects, and adjuncts, can be focused in Wolof. Only subjects and objects will be considered here, however, because syntactic adjuncts (i.e. PP) behave exactly like objects do.23 3.1 Object focusing (9)

Fas wi la jaaykat bi jënd horse the . merchant the buy ‘It is the horse (that) the merchant bought’

Just looking at the data makes it plain to what extent Wolof and English are alike here, differing only in the position of the elements la and it is. There is no doubt, then, that the la we see in this example is indeed the same as the one studied in the preceding section: a copular expression. This is confirmed by (10) and (11) below, where the subject of the predicate out of which the object is focused is a pronoun instead of a noun: (10)

Fas wi nga jënd horse the . buy ‘It is the horse (that) you bought’

(11)

Fas wi la jënd horse the . buy ‘It is the horse (that) s/he bought’

In (10), the subject of jënd ‘to buy’ is the second person singular pronoun (compare Nga jënd fas wi ‘You bought a horse’), and it fuses with la in accor-

FOCUS IN WOLOF

253

dance with paradigm (4). In (11), again in accordance with paradigm (4), the fusion of the third person singular pronominal subject of jënd ‘to buy’ with la has no phonological correlate. The pronoun surfaces, in contrast, in the corresponding unfocused sentence: (12)

Mu jënd fas wi  buy horse the ‘She/He/It bought the horse’

Without the subject pronoun, (12) would be ungrammatical (*jënd fas wi), while (11) with the pronoun would be ill-formed (*fas wi la mu jënd). I will return to this issue. Meanwhile, these data are strong and, I think, indisputable evidence that la and nga in the examples are indeed forms of the predicational copula la.24 la, then, in addition to focusing the phrase that precedes it — its attribute according to the analysis of copular sentences in Section 2 —, fuses with a pronominal subject of the predicate having the focused phrase as its basic object. It may also fuse (or rather merge in this case) with the Tense morpheme dominating the VP, under conditions illustrated in the following examples: (13)

Fas wi la jaaykat bi di jënd horse the . merchant the T buy ‘It is the horse (that) the merchant will buy’

(14)

Fas wi ngay jënd horse the .. buy ‘It is the horse (that) you (sg.) will buy’

(15)

Fas wi ngeen di jënd horse the .. buy ‘It is the horse (that) you (pl.) will buy’

The Tense morpheme di merges with the copula, thereby reducing to /y/, just in case the latter fuses with a pronominal subject, and the resulting form ends in a vowel as in (14).25 3.2 Subject focusing Subject focusing seems to involve a different construction altogether (although, as we shall see presently, the difference is only apparent):

254

ALAIN KIHM

(16)

Jaaykat bi a jënd fas wi merchant the  buy horse the ‘It is the merchant who bought the horse’

(17)

Yaa jënd fas wi . buy horse the ‘It is you (sg.) who bought the horse’

Clearly, everything hinges on the identity of the a morpheme that shows up visibly in (16) and not so visibly in (17). As implied in the glosses, I assume that a is the core of la, i.e. the predicational copula itself, and that the absence of the fused subject /l/ follows from the fact that a subject, not an object, is focused. This will be demonstrated in the next section. As a result, (16) and (17) will be seen not to be syntactically different from object-focusing constructions. In (16), a is clearly an autonomous item, even though the vowel sequence [i a] coalesces into a single segment [e:] through a general phonological process of hiatus reduction (see Ka 1994).26 I will return to this. In (17), on the other hand, the pronoun fuses with a, yielding what looks like a paradigm of so-called ‘emphatic’ pronouns: (18)

maa ‘it’s me who…’, yaa ‘it’s you (sg.) who…’, moo ‘it’s her/him/it who/that…’, noo ‘it’s us who…’, yéena ‘it’s you (pl.) who…’, ñoo ‘it’s them who…’

Given the complexity of Wolof morphophonemics, the actual derivation of these forms is far from transparent.27 This is not a problem, however, with a theory of fusion that allows us to view maa, yaa, etc. as Vocabulary items that spell out several syntactic categories. Actually, paradigm (18) is of more frequent use than one would gather from the pair of examples (16) and (17). Indeed, although it is perfectly grammatical, (16) is less common than its paraphrase (19): (19)

Jaaykat bi moo jënd fas wi merchant the . buy horse the ‘The merchant, it is him who bought the horse’

with the lexical subject topicalized and resumed by a focused subject pronoun. The Tense morpheme di behaves in the same way as with object focusing: it merges with a only if the latter undergoes fusion. We thus find the following contrast:

FOCUS IN WOLOF

(20)

Jaaykat bi a di jënd fas wi merchant the   buy horse the ‘It is the merchant who will buy the horse’

(21)

Yaay jënd fas wi .. buy horse the ‘It is you who will buy the horse’

255

3.3 Negative focusing Negative focusing, be it object or subject focusing, is noteworthy as it involves a shift from predication to identification, as shown in the following examples:28 (22)

Du fas wi la jaaykat bi jënd be. horse the . merchant the buy ‘It is not the horse that the merchant bought’

(23)

Du jaaykat bi moo jënd fas wi be. merchant the . buy horse the ‘It is not the merchant who bought the horse’

(24)

Du yow yaa jënd fas wi be.  . buy horse the ‘It is not you (sg.) who bought the horse’

Du represents the fusion of the identificational copula di (here glossed as ‘be’ — see Section 2) with the suffixal negation /-u(l)/ entailing a null third person subject.29 Determining its scope seems to be the main problem posed by these sentences. This issue will be tackled in the next section.

4.

The structure of focused sentences analyzed

The cleft character of Wolof focused sentences should by now emerge clearly from the data. It also follows from the analysis of the predicational copula la proposed in Section 2. What is in need of an explanation, however, is the precise way in which the structure is attained through the interplay of presumably universal syntax and language-specific morphology. As in the preceding section, I will first deal with object focusing, then with subject focusing, and finally with negative focusing.

256

ALAIN KIHM

4.1 Object focusing The first thing to be noticed is that the surface ambiguity of the English cleft construction is not found in Wolof. Consider the following contrast ((24) = (9)): (24)

Fas wi la jaaykat bi jënd horse the . merchant the buy ‘It is the horse (that) the merchant bought’

(25)

Fas wu jaaykat bi jënd la horse  merchant the buy . ‘It/this is the horse (that) the merchant bought’

Cleft and ‘presentational’ sentences are clearly distinct in Wolof, whereas they translate identically into English, except for the possible alternation of it and this in (25), not authorized in (24). Actually, the structure of (25) is the same as that of (1c) (Fas la ‘It’s a horse’), except that the topicalized DP resumed by /l/ includes the relative clause [wu jaaykat bi jënd t] ‘that the merchant bought’. We can therefore assign the following structure to (25), with the DP [fas wu jaaykat bi jënd t] topicalized in Spec,CP:30 (26)

[CP [DP [CP fas [C′ wu jaaykat bi jënd t]]]i [IP… [VP li [V′ a [DP Ø]]]]]

The parallel differences in interpretation of (24) and (25) suggest that the differences in the respective syntactic structures must also be parallel. This assumption entails that (24) involves relativization, even though the relative pronoun wu is not apparent in it as it is in (25).31 The latter discrepancy, in turn, is linked to the real difference between (24) and (25), in Wolof as well as in English, namely what the relative clause is predicated of. In (25), as we have just seen, the relative clause is predicated of the head fas ‘horse’. The assumption I want to make is that, in clefts, it is in fact predicated of the logical variable realized as an expletive, which is overt in English (it) and silent in Wolof (the Ø in (26)). It follows from this assumption, as we shall see, that clefts have the same syntactic and LF structure as pseudo-clefts, as seems to be required by their logico-semantic equivalence, abstracting from the different pragmatic conditions of their use (see Foley & Van Valin 1985). Let us see how this works for English, before extending the reasoning to Wolof. Informally, we want to say that both ‘What the merchant bought was the horse’ and ‘It is/was the horse that the merchant bought’ in the focused reading derive from a common syntactic and LF structure that looks like [it that the

FOCUS IN WOLOF

257

merchant bought is/was the horse], i.e. ‘the x such that the merchant bought x is/was a horse’.32 Deriving a pseudo-cleft from such a structure is straightforward, modulo the lexical specification that the Vocabulary item what spells out the syntactic sequence [it that].33 Deriving a cleft, on the other hand, involves a process that earlier models would have described as extraposition of the relative clause [that the merchant bought t] to the end of the sentence, so that it now seems to be predicated of horse. Note that, given the copular nature of the sentence, it and horse bear the same index, which makes the operation semantically possible. The homophony of presentative and cleft sentences is thus accounted for. Also accounted for is McCawley’s (1988: 427–8) observation that parenthetical expressions separating the antecedent from the relative clause are acceptable in cleft constructions (see ‘It was Fred, incidentally, who asked John for help’), while they are excluded from other contexts (see * ‘Fred was just talking to the person, incidentally, who asked John for help’). Such a contrast is expected since, according to the extraposition account of clefts, Fred is not the basic, but only the apparent (i.e. post-Spell Out) antecedent of who in the cleft sentence.34 Now, extraposition is not the kind of operation that has much status in the current framework. It is possible, however, to reinterpret it in terms of a copy theory of movement, along the lines set out in Brody (1995), Chomsky (1995), Groat & O’Neil (1996), and Solà (1996). That is to say, the focused reading of ‘It is the horse that the merchant bought’ will be given the following structure at LF: (27)

[it [that the merchant bought t] be the horse [that the merchant bought t]]

Copying the relative is necessary in order to ensure full identification of it and horse. Only one copy may be pronounced, however, and the remaining one(s) must be deleted. According to Solà (1996: 223), the pronounced copy should be the highest one (or the first one, in case they are equivalent). Such a forced choice, however, implies that the successive copies are hierarchically ordered, which is indeed the case whenever ‘pronouncing the highest copy’ means spelling out a head (and whatever goes with it) in a landing site for head movement. In (27), on the other hand, there is arguably no hierarchy between the two copies, since the subject DP [it [that the merchant bought t]] and the predicate [be the horse [that the merchant bought t]] are sister nodes under VP.35 In such a situation, therefore, we may assume there is no a priori reason for choosing one copy over another for spelling out. To put it differently, this is a

258

ALAIN KIHM

case where either derivation seems as economical. We thus get the equivalent of optional movement (see Collins 1997). Consequently, either one of the two copies may be deleted, as shown in (28) and (29), where the deleted copy is struck out: (28)

[it [that the merchant bought t] be the horse [that ———————— the merchant ————t]] bought

(29)

[it [that ————————————t] the merchant bought be the horse [that the merchant bought t]]

Deleting the second copy and pronouncing the first one as in (28) yields a pseudo-cleft, modulo replacement of /it that/ by what. Deleting the first copy and pronouncing the second one as in (29) yields a cleft. It must be emphasized again that, although the undeleted relative clause of (29) seems to be predicated of horse, in fact it is not, since it only exists qua second element of a chain. Neither horse nor it, but only the complex [it = horse] can therefore be considered the exhaustive antecedent of such a distributed relative clause. There is no chain, in contrast, in the presentational correspondent of (29): (30)

[it be the horse [that the merchant bought t]]

so that horse is indeed the sole antecedent of the single member relative clause, and it is not coreferential with it, but probably unbound. Let us now turn to Wolof. First, we need an example of a pseudo-cleft: (31)

Lu jaaykat bi jënd fas wi la what merchant the buy horse the . ‘What the merchant bought was the horse’

Lu ‘what’ consists in the generic noun class marker-pronominal /l/ and the relative determiner /u/. It is thus very similar to what = /it that/ (see above). Deriving (31) and the cleft (24) from a single LF representation does not seem to be as straightforward as it is in English, though. The difficulty stems from a morphological peculiarity of Wolof which has already been pointed out: the predicational copula a is not a free standing element like be, but it can only be inserted as part of the fused Vocabulary item la, i.e. [VP l [V′ a [DP Pro]]]. Actually, subject focusing as in (16) seems to run counter to this claim, since a appears then as an autonomous item, cliticized to the preceding element only in the phonology.36 I will return to this point shortly, as it bears directly on the crucial question raised by the data: why do we find la with object focusing, but a with subject focusing?

FOCUS IN WOLOF

259

Applying to Wolof the reasoning developed above for English leads us to the following pre-Spell Out structure for (24): (32)

[CP [DP [CP Ø [jaaykat bi jënd tØ]]] [CP fas wii [IP… [VP li [V′ a [DP [CP Ø [jaaykat bi jënd tØ]]]]]]]]

In (32), Ø is the silent pure expletive corresponding to it. This means that (32) is a copular expression similar to (27). Given Wolof rules, however, both the subject [DP [CP Ø [jaaykat bi jënd tØ]]] = [it [that the merchant bought t]] and the attribute fas wi ‘the horse’ have to be topicalized (see Section 2), and they are resumed within la, the subject by another occurrence of the silent expletive in the position of Pro, the attribute by /l/. Hence (32), translates as ‘As for the x such that the merchant bought x, the horse, it is the x such that the merchant bought x’.37 As in English, albeit in a different way, the relative clause has to be copied in order to warrant full identification of the two occurrences of the expletive. The necessity of a silent head to the right of la now becomes apparent. Since the two relative clauses constitute a copy chain, only one can be pronounced. Now, however, we are no longer in the same situation as in English: the two copies are not under sister nodes. According to Solà’s (1996) principle, therefore, only the highest, i.e. first one should be realized. But recall that topicalization is assumed to be basic in Wolof copular sentences. This means that the copies are not created by movement, but are rather a case of multiple selection in a numeration. A reasonable assumption in this case is that the copy that is retained is the one that gives rise to a possible derivation. (If both do, then we have optionality.) Suppose we retain the second copy and delete the first one in (32) as shown in (33): (33)

[CP [DP [CP Ø [jaaykat ———————tØ bi jënd — ]]] [CP fas wii [IP… [VP li [V′ a [DP [CP Ø [jaaykat bi jënd tØ]]]]]]]]

This gives rise to a possible derivation, since (33) is straightforwardly realized as (24) (Fas wi la jaaykat bi jënd ‘It’s the horse (that) the merchant bought’). What is crucial for (33) is that the expletive in the Pro position of la has syntactic (post-Spell Out) reality since it belongs to a pronominal paradigm.38 It is therefore able to function as an antecedent for the relative clause. The fact that it lacks phi-features, on the other hand, explains why no relative pronoun such as wu in (25) may appear, since relative pronouns in Wolof must agree with their antecedent. What we get, then, is an empty Co which is interpretable thanks to the interpretability of the silent antecedent.39 Things are different, however, if the second copy is deleted and the first one retained, as in (34):

260

ALAIN KIHM

(34)

[CP [DP [CP Ø [jaaykat bi jënd tØ ]]] [CP fas wii [IP… [VP li [V′ a [DP [CP Ø [jaaykat ———————tØ]]]]]]]] bi jënd

Here, the first occurrence of Ø exists only at LF. Therefore, (34) cannot be continued past Spell Out, as this would leave us with a relative clause that is not visibly predicated of anything. The only way to make something pronounceable of (34) is to change the numeration, inserting lu ‘what’ in the position of Ø, producing (31). Now, we might argue that lu has no meaning and is only there to satisfy the requirements of PF. In this way, we would still be able to maintain that Wolof is like English, except for details of the interface with the articulatory-perceptual system. I leave this as an unsettled issue, assuming for the present that (33) and (34) are indeed optional variants of (32), as both give rise to a possible derivation. The same process will account for cases where the subject of the copied relative clause is a pronoun rather than a noun. Take, for instance, (10) (Fas wi nga jënd ‘It’s the horse that you bought’). Its representation is given in (35): (35)

[CP [DP [CP Ø [nga ————tØ jënd — ]]] [CP fas wii [IP… [VP li [V′ a [DP [CP Ø [nga jënd tØ]]]]]]]]

The first copy is deleted as in (33). As just argued, the Ø in Pro position exists in post-Spell Out syntax. It is not visible in Morphology, however, since it has no phonological form. Therefore, the syntactic object ‘[VP li [V′ a [DP [CP Ø [nga’ is nondistinct from the syntactic object ‘[VP li [V′ a [DP nga’ at this level, assuming further, as is standard, that syntactic boundaries also are (or may be) invisible in Morphology. Consequently, ‘[VP li [V′ a [DP [CP Ø [nga’ is open to fusion, and is realized as the Vocabulary item nga, meaning ‘it is x (that) you (sg.)’, as well as ‘it is you (sg.)’, and ‘you (sg.)’. Deleting the second copy, on the other hand, will yield the pseudo-cleft Lu nga jënd fas la ‘What you (sg.) bought was the horse’, as explained above. 4.2 Subject focusing Applying the same assumptions, (16) (Jaaykat bi a jënd fas wi ‘It’s the merchant who bought the horse’) must be given the following pre-Spell Out structure: (36)

[CP [DP [CP Øi [ti jënd fas wi]]] [CP jaaykat bii [IP… [VP jaaykat bii [V′ a [DP [CP Øi [ ti jënd fas wi]]]]]]]]

FOCUS IN WOLOF

261

The meaning of (36) is ‘As for the x that bought the horse, the merchant, he is the x that bought the horse’. As can be seen, (36) differs from (32) on two counts: (i) there is a copy of the topicalized subject in Spec,VP instead of the resuming pronoun /l/; (ii) there is only one index. Both features are related. Indeed, what we have in (32) is a subject, ‘what the merchant bought’, call it p, resumed by Pro (Ø), and an attribute, ‘the horse’, call it q, resumed by /l/, and (32) tells us that p has the property of being q. In (36), in contrast, we have a subject, ‘who bought the horse’, but no attribute. In fact, (36) is a statement of identity over ‘who bought the horse’ and ‘the merchant’, i.e. a tautological equation p = p, where both p’s are subjects. This is why /l/ does not appear in Spec,VP, but a copy of ‘the merchant’ does, since /l/ resumes attributes according to the structure of copular sentences in Wolof. If (36) is a statement of identity, why do we not find the identificational copula di instead of the predicational copula a? The answer lies once again in the syntax-morphology interface. To the difference of a, di is a simple word that cannot include the silent expletive necessary as an antecedent of the second copy of the relative clause when it is retained to yield (16). Therefore, *Jaaykat bi di jënd fas wi cannot be derived. Proof for this is that di will appear the moment a visible antecedent is supplied as in (37): (37)

Jaaykat bi di ku jënd fas wi40 merchant the be who buy horse the ‘The merchant is the one who bought the horse’

Not surprisingly, (37) is reversible (Ku jënd fas wi di jaaykat bi ‘(The one) who bought the horse is the merchant’). In a way, then, a is used as a replacement of di in (36) because only a is morphologically equipped to fulfill the syntactic requirements of the construction. In consequence, we would expect di to be usable when the second copy of the relative clause is deleted, therefore not in need of a visible antecedent any more, and the first one is provided with a visible antecedent as in (34), thus yielding the pseudo-cleft *Ku jënd fas wi jaaykat bi di. Such a sequence is ungrammatical, however, because di must have an overt complement.41 Only the predicational equivalent Ku jënd fas wi jaaykat bi la ‘(The one) who bought the horse, it’s the merchant’, parallel to (1b) (Góor gii sunu njiit la ‘This man, it’s our chief’), is derivable. It does not seem to make a difference in (36) which occurrence of jaaykat bi ‘the merchant’ is retained, and which deleted. Consider, however, that the structural parallelism of Ku jënd fas wi jaaykat bi la ‘(The one) who bought the

262

ALAIN KIHM

horse, it’s the merchant’ and (1b) forces us to conclude that, in this case, jaaykat bi is topicalized in Spec,CP. By implication, we will also conclude that the copy in Spec,VP has to be retained, and the one in Spec,CP deleted, should (36) derive the cleft sentence (16), which will therefore receive the following representation: (38)

[CP [DP [CP Øi [t—i——————]]] jënd fas wi [CP ———— jaaykat bi—i— [IP… [VP jaaykat bii [V′ a [DP [CP Øi [ ti jënd fas wi]]]]]]]]

In sum, all topicalized material is deleted, and (38) is indeed structurally identical with the virtual identificational *Jaaykat bi di jënd fas wi, which cannot be derived purely for morphological reasons, as we saw.42 I realize that such an interplay of syntax and morphology raises far-reaching questions, both theoretical and empirical. I have neither the space nor the means to deal with them here, however. One could also ask in what sense (38) may still be said to be focused. I will try to answer this question in the conclusion. 4.3 Negative focusing The relevant examples are (22)–(24). As was already made clear, du which appears at the beginning of them represents the fusion of the identificational copula di ‘to be’ with the negative suffixal morpheme /-u(l)/ which has the property of absorbing a third person singular pronominal subject. Du jaaykat ‘He isn’t a merchant’ is thus semantically the negation, not of Jaaykat la, the closest translation of which is French C’est un marchand, but of Mu di jaaykat ‘He is a merchant’. Negation thus entails the conversion of predication into identification, a phenomenon that would be well worth pursuing. To return to our immediate concerns, it seems that (22)–(24) can be explained assuming that du ‘is not’ actually has the meaning of the logical negation ¬ ‘it is not the case that’. This interpretation is supported by such sentences as the following (Church 1981: 251): (39)

Du xam nga sama jabar? be. know . my wife Don’t you know my wife? (literally, ‘Is it not [the case that] you know my wife?’)43

In this way, (22) can be interpreted as meaning, [it is not the case [that [it is the horse that the merchant bought]]]; (23) as [it is not the case [that [the merchant]

FOCUS IN WOLOF

263

it is him who bought the horse]]]; (24) as [it is not the case [that [you] it is you who bought the horse]]]. In other words, the focused sentence is directly embedded under the negative copula. The nominal character of CP (see Webelhuth 1992) is probably what makes the presence of an overt noun head such as ‘the case’ or ‘the fact’ unnecessary.

5.

Problems of the Syntax-PF interface

Coupling the syntax with PF is by no means a simple matter in Wolof.44 This much should by now be evident. In this section, I consider a few issues which highlight the extent to which underlying syntax may be obscured in focused constructions and which support my claim that the traditional analysis in terms of ‘emphatic conjugations’ is untenable. Subject focusing appears as the least complex in this regard. As already mentioned, the copula a following a lexical subject DP such as jaaykat bi ‘the merchant’ is only absorbed into what linearly precedes it in late phonology. Yet, absorption may vary in depth. While there is no doubt that the sandhi form [be:] realizing the sequence /… bi a … / cannot be a word, things are not so clear-cut when the final segment of the lexical subject DP is a consonant: (40)

Umar a jënd fas wi Umar . buy horse the ‘It is Umar who bought the horse’

An interesting prosodic feature of Wolof focused sentences is that the copula and what precedes it cannot be separated by a pause, but have to be pronounced legato (see Sauvageot 1965: 201, and below). Because of this, the syntactic sequence /Umar a/ of (40) surfaces as a phonological word [‘umara]. Such forms would justify to some extent viewing a as an emphatic or Focus particle (see Dunigan 1994) cliticizing to (the last segment of) the phrase it modifies. It is hard to see, in contrast, what sense ‘conjugation’ may have in such a case. When the focused subject is a pronoun, on the other hand, a absorption gives rise to paradigm (18) of so-called ‘emphatic pronouns’, as we saw. Since the form these elements take is underivable by any regular process in present-day Wolof, they were analyzed as fused Vocabulary items spelling out several syntactic categories. Traditional examples of fusion (or merger) involve inflected items, e.g.

264

ALAIN KIHM

inflected verbs in English (see Halle & Marantz 1993). In this case, fusion takes two heads, say kiss and /-ed/ and fuses them into what is still a head, viz. kissed. Wolof is interesting in that it shows that the result of fusion (or merger) may not be a head (unless some reanalysis occurs), may not even be a constituent. Take, for instance, the second person singular ‘emphatic’ pronoun yaa that occurs in Yaa jënd fas wi ‘It’s you (sg.) who bought the horse’. As we saw, it consists of a pronominal stem /ya/ that is not found outside of the combination, of the predicational copula a, and of a silent expletive. The whole sequence constitutes a maximal projection VP, not a head. Moreover, since what follows yaa is a relative clause predicated of the silent expletive fused in yaa, yaa is not a constituent. Constituents are the full VP [VP ya [V′ a [DP [CP Ø [jënd fas wi]]]] or the DP [DP [CP Ø [jënd fas wi]]]. In spite of what the surface string would lead one to believe, there is thus no deep sense in which yaa can be said to be the subject of jënd fas wi. Fusion combines two heads into one, realized as one Vocabulary item; merger subsumes two heads, realized as two Vocabulary items, under one node; in either case, we get something that counts as a head and an immediate constituent. Insofar as Wolof ‘emphatic pronouns’ are neither heads nor constituents, but they still must be considered Vocabulary items because of their opacity (at least for some of them), they may instantiate a third process which I will tentatively call ‘agglutination’. Agglutinated Vocabulary items would be characterized by the fact that they are unique, like fused items and unlike merged items, while being recognizably composite, like merged items and unlike (a significant number of) fused items, and that they are blind to syntactic boundaries. Another possible example are the West Flemish ‘inflected’ complementizers in such forms as dase (zie) komt, literally ‘that-she (she) comes’ (see Bennis & Haegeman 1984; Solà 1996). Since dase does not seem to be derivable from the sequence /dat zie/ by any process that the speakers might be said to apply or the children to learn — see also dank for /dat ik/ ‘that-I’ — it must be considered a Vocabulary item. At the same time, the fact that dase agglutinates the two Vocabulary items dat and zie remains visible.45 Finally, in the surface string dase komt, dase cannot be mapped to any single position of the underlying syntactic structure, but it is spread over at least two which do not form a constituent, i.e. Co and Spec,IP.46 West Flemish ‘inflected’ complementizers thus seem similar to Wolof ‘emphatic pronouns’. Such spreading or absence of one-to-one mapping raises real interface problems. In sentences like Fas wi ngay jënd ‘It’s the horse (that) you (sg.) will

FOCUS IN WOLOF

265

buy’, the biclausal nature of the structure is entirely erased in the surface string where the ‘pivot’ term nga straddles the two underlying clauses, as it agglutinates — or even fuses in this case — the copular predicate of the matrix clause [fas wi la] ‘it’s the horse’ and the subject and Tense marker of the embedded clause. A noteworthy correlate of this situation, that was already mentioned, is that such sentences must be uttered without pause, “d’une seule émission de voix”, to use Sauvageot’s (1965: 201) terms.47 This may be considered a prosodic (negative) signal of the biclausality of the string. It is not always possible to do without a pause, however, especially when the focused phrase exceeds a certain length, as in the following example from Church (1981: 130): (41)

Bi ma nekkee xale te ma bey sama toolu baay laa when I be child and I cultivate my field father . gis gaynde see lion48 ‘It is when I was a child and I cultivated my father’s field that I saw a lion’

In the English translation, the almost obligatory pause occurs between the last element of the focused phrase, i.e. field, and the complementizer that. The clause boundary is thus ‘audible’. In Wolof, the pause — which is also almost obligatory — can only be inserted between the last element of the focused phrase, i.e. baay ‘father’, and laa. In this way, the ‘I’ subject agglutinated within laa is not prosodically cut off from its predicate gis gaynde ‘saw a lion’. But as a result the focused phrase is separated from the copula and the agglutinated pronominal /l/ that resumes it in a way they could not be in simple predicational sentences (e.g., Sama baay la ‘It’s my father’). Hence an unavoidable mismatch between the syntax and PF due to the ambiclausal constituency of the Vocabulary item laa.49

6.

Conclusion

The evidence from Wolof confirms the extent to which morphological factors may blur common syntactic structures such as clefting. One source of the problem is the fact that the predicational copula la around which the expression is built is not a simple item, but an agglutinative, potentially ambiclausal word. This is the language-specific part of the issue, not limited to Wolof, but obviously not universal. What may have universal value, on the contrary, is the assumption

266

ALAIN KIHM

that clefts are not what they look like, inasmuch as the relative clause they include is not predicated of the focused term, as it seems to be, but of the logical variable spelled out as it in English, as a silent expletive in Wolof. Combining the language-specific and the universal then gives us an explanatorily adequate account of argument focusing in Wolof. Agglutinative words, if they exist, raise serious questions which lie well beyond the scope of this article. I will just outline an interesting consequence, namely that the so-called ‘emphatic pronouns’ such as yaa, etc. now appear as pseudo-categories from the point of view of syntax. Yet, if my reasoning is correct, they exist as Vocabulary items and “creatures” of the Syntax-Morphology interface. This would imply that neither the “External Form”, to use Rouveret & Vergnaud’s (1980) term, nor the Vocabulary are inert reflections or mere embodiments of the underlying structure. Another issue that is crucially raised by the present study concerns the necessity of a Focus projection in the sense of Rizzi (1998) or of a Focus structure in the sense of Erteschik-Shir (this volume). As I am not sure a unitary solution can be brought to this problem, I will limit myself to my subject matter, viz. clefting as one means (English) or the only means (Wolof) to put an argument into focus. Focusing in this case does not seem to require a special operator, insofar as it may be viewed as a pragmatic effect of the explicit designation and contrast brought about by inserting a given element into a ‘it is … that …’ frame. Put differently, focusing p, a member of set {A}, with respect to some predicate q may consist in the explicit statement that q is true of p and false of the complementary of p in {A}. Focusing appears then as a correlate of this logical operation, insofar as the latter has the effect of bringing out one element against the background of all other elements in a given set (also see Léard 1986). Note incidentally that the identity of the set is determined to a large extent by that of the extracted element. Thus, in our pet example ‘It is the horse that…’, horse as the bought element is contrasted not with any buyable entity, but only with buyable  (at least in ordinary circumstances). Since all focusing devices are open to this remark, it does not touch directly upon the question of whether we need a Focus projection or not.50 It shows at least that there is a purely logical side to focusing, which is forcefully demonstrated by (38) if I am right in analyzing this sentence as basically an identificational copular sentence, identifying p as the relevant member of the reference set for property q. In fact, there may be a deep relation between this logical, or perhaps

FOCUS IN WOLOF

267

cognitive, aspect of focusing and the issue of agglutinated words and the complexity that they add. Indeed, one serious problem with such items, if they exist, is their learnability. How does a child recognize that the seemingly simple word yaa means a whole proposition translatable as, ‘The element of set A such that you are this element is the (only) one that…’? A possible answer is that this whole proposition is what s/he innately knows and expects to find in the data, no matter how much reconstruction s/he has to do to successfully pair what s/he hears and what s/he knows. In English, French, etc. the pairing is easy; it is a little more complex in Wolof, but this is the only real difference between the languages.

Notes 1. Wolof is spoken in Senegal and the Gambia, where it is the native language of more than two million people in addition to being a vehicular language for as many more. Wolof belongs to the Northern branch of the Atlantic subfamily of the Niger-Congo family (see Wilson 1989). For descriptions, see in particular Boilat (1858); Kobès (1869); Sauvageot (1965); Church (1981); Njie (1982); Diouf (1985); Robert (1991); Ka (1994); Dunigan (1994). In this list, only Njie (1982) and Dunigan (1994) are syntactic studies in a generative framework, EST the former, P&P the latter. 2. In this article, I am only dealing with argument focusing in matrix clauses. It is worth mentioning, however, that focus in Wolof may be grammatically assigned not only to arguments, but also to predicates. As predicate focusing resorts to quite a different device than argument focusing, however, I will say nothing about it (but see Robert 1986). There is evidence, on the other hand, that focusing is not limited to matrix clauses, as it seems to be in many languages, and that double focusing, i.e. of two arguments at the same time, is even possible. This is something I haven’t really explored, however. 3. I want to emphasize that the present study is strictly morphosyntactic in scope. All the semantics I shall have need for are summarized in the introductory paragraph, and they represent, I believe, the basic signification of what focusing consists of. Yet, it is quite obvious that focused constructions in real usage exhibit a wide range of meanings and functions, the relation of which to the basic semantics is by no means always transparent. This is true generally, and especially so in a language like Wolof, whose speakers make extensive use of focusing, much more so than, e.g., English speakers. For this aspect of things, the reader is directed to Robert’s book (1991, Chapters 4 & 5), which contains a careful study of focused constructions in their ‘enunciative’ context. 4. Example (1a) is from Fal, Santos & Doneux (1990: 62). All other examples except two are of my own devising. In order not to tax the memory of readers with no knowledge of Wolof, I keep to simple sentences with a very limited vocabulary. Unnecessary morphology (for my purpose) is not indicated. For instance, gii translated as ‘this’ is in fact a complex form

268

ALAIN KIHM consisting of a noun class morpheme /g-/ and the proximate demonstrative /-ii/. Official spelling is used throughout: doubled vowels and consonants are long; accents on vowels indicate closure; ë is a mid central vowel close to schwa; j represents the voiced palatal stop, and c its unvoiced counterpart; ñ is the palatal nasal; x is the unvoiced velar fricative.

5. A more literal translation would be the ungrammatical *‘This man, it’s our chief’ (compare French Cet homme, c’est notre chef as opposed to Cet homme est notre chef translating (1a)). 6. Compare Góor gii rey sunu njiit ‘This man killed our chief’, with rey ‘kill’ in the so-called ‘zero’ aspect (see Sauvageot 1965; Dunigan 1994). 7. Past is the traditional interpretation for /-(w)oon/. Dunigan (1994) analyses it as rather a Perfective (i.e. aspectual) marker. This makes no difference for my purpose. 8. See Dunigan (1994) for arguments to the effect that person and number markers in Wolof are clitics rather than direct spell-outs of AGR. 9. I use la as a cover form for the whole paradigm. 10. Dunigan (1994) does not deal with the issue of copular sentences. 11. See lu /L.REL/ ‘what’, lan /L.wh/ ‘what?’, lii /L.DEM/ ‘this (thing)’, etc. I am grateful to Serge Sauvageot for suggesting this idea to me. He is not to be held responsible for what I have made of it. 12. I use ‘subject’ in a strictly structural sense, a necessary remark that will soon become crucial. 13. Paradigm (5) ought to be compared with (a) the paradigm for clitic subject pronouns, viz. ma ‘I’, nga ‘you (sg.)’, mu ‘s/he/it’, nu ‘we’, ngeen ‘you (pl.)’, ñu ‘they’; (b) the affirmative modality paradigm (see Dunigan 1994), e.g. for dem ‘to leave’: dem naa ‘I left’, dem nga ‘you (sg.) left’, dem na ‘s/he/it left’, dem nanu ‘we left’, dem ngeen ‘you (pl.) left’, dem nañu ‘they left’. As it appears, the first and second persons plural are regularly expressed through suffixation of the clitic pronoun to the verbal form (if the ‘particle’ na is analysed as some kind of light verb). First person /-a/ can be deduced from ma ‘I’ via deletion of intervocalic /m/, a process for which there is independent evidence (*/dem nama/ > dem naa). Full absorption of the verb stem (/la/ or /na/) by the second person pronouns is also regular, as is the use of a zero morpheme for third person singular. The fact that subject forms appear in the complement position of the copula instead of object forms (a distinction apparent in the second person singular that has subject nga vs. object la, accidentally homophonous with the predicative copula) further suggests that the copula does not assign Accusative case to its complement, a general feature for this element. 14. We are dealing here with coindexation in a broad sense, since ce (likewise Wolof /l/) does not agree in phi-features with the semantic subject, nor is it coreferential with it in the strict sense of complete identity. On this issue, see Tancredi (1996) and Pica & Tancredi (1996). 15. Yow is the ‘strong’ or ‘autonomous’ form of the second singular pronoun. Strong pronouns occur only in topic position or as complements of prepositions (e.g., ak yow ‘with you’). Note that translating (7) as ‘You, you are our chief’ would add an undue measure of emphasis. 16. Note that, while laa ‘it’s me’ can be phonologically derived from */lama/, no such explanation is available for la ‘it’s her/him/it’ with respect to */lamu/. 17. Similarly, think of the question What is raining? as a reaction to the assertion It is raining.

FOCUS IN WOLOF

269

18. Mu is nevertheless used as a ‘makeshift’ expletive in expressions such as Mu taw ‘It rained’, probably because the pure expletive Ø has no phonological form, and Wolof does not accept phonologically null subjects, except with negative predicates (e.g., Tawul ‘It didn’t rain’, where /-ul/ is the negation). 19. As is well-known, only the first form is deemed correct in normative French. 20. Note that French is like Wolof in rejecting the equivalent of *Yow sunu njiit la, viz. *Toi, c’est notre chef, and accepting only Toi, tu es notre chef (Yow sunu njiit nga). French and Wolof also behave the same with respect to attribute pronouns, so that we find C’est toi (not *Tu es toi) parallel with Yow la (not *Yow nga) meaning ‘It’s you’. This shows the DP nature of strong pronouns in both languages. 21. For some reason, ?Cet homme, notre chef, ce l’est with clitic le resuming notre chef is extremely awkward although not 100% ungrammatical. 22. As for the past or perfective forms la-woon, there are good reasons to think that /-(w)oon/ is an autonomous element rather than an inflection morpheme. For instance, it may be separated from the verb it modifies by various particles (see Church 1981: 195ff.). Whatever the precise status and site of insertion of /-(w)oon/, then, such sequences as lawoon, ngawoon, etc. are formed in the syntax (on this, see Dunigan 1994). 23. I specify ‘adjunct’ with ‘syntactic’ because many arguments that would translate as adjuncts into English are objects in Wolof as a consequence of derivational incorporation to V (see Kihm 1994). Also note that ‘object’ means both accusative and dative object, since double object constructions similar to the English ones exist in Wolof. 24. Similarly, we find Fas wi laa jënd ‘It’s the horse (that) I bought’, Fas wi lanu jënd ‘It’s the horse (that) we bought’, Fas wi ngeen jënd ‘It’s the horse (that) you (pl.) bought’, Fas wi lañu jënd ‘It’s the horse (that) they bought’. 25. Notice that ‘ends in a vowel’ is not a sufficient condition as shown by the sequence /bi di/ in (13). We are thus not dealing with a purely phonological process. 26. Except in artificially slow speech, [bi a] is thus pronounced [be:] as if it were one word. 27. Take, for instance, the second person singular: the subject pronoun is nga, and the autonomous pronoun is yow. Neither of these forms combined with a will yield yaa by any regular process (although the latter would be a more likely candidate than the former). Compare with the entirely regular and transparent phonological process described in the preceding footnote. For discussions, see Dialo (1981); Church (1982); Diouf (1985); Robert (1991). The process itself was discovered by Kobès (1869). 28. On negative focusing, also see Ouhalla, this volume. 29. See, e.g., Rotul ‘She/He/It didn’t fall’ and Dul rot ‘She/He/It won’t fall’ vs. Mu rot ‘She/He/It fell’ and Dina rot ‘She/He/It will fall’. In the last case, the subject, although not apparent, is not null, but fused with the modality morpheme /-na/. Also note that, although the identificational copula di and the Tense morpheme di are homophonous in the positive, they are distinct in the negative, showing up respectively as du and dul. Other persons than the third singular are overtly incorporated to the negation (e.g. rotuma /rot-u-ma/ ‘I didn’t fall’). Negating a predicate also entails extensive Tense-Aspect neutralizations that exceed the scope of this article.

270

ALAIN KIHM

30. I assume the relative clause structure of Kayne (1994). 31. Wu consists of the noun class marker W- agreeing with fas ‘horse’ and the relative operatordeterminer /u/. 32. I have nothing to say about the optional tense concord of the copula. 33. Compare French ce que where both elements are spelled out. 34. Accordingly, parenthetical expressions seem to be excluded from presentational sentences that look like clefts. That is to say, whereas ‘It is the horse, I think, that the merchant bought’ is good as a cleft, it is bad as an answer to ‘What is this?’, unless especially heavy pauses are inserted on each side of ‘I think’. I thank Laurie Tuller for confirming these judgments for me. 35. Especially if we assume that the copula is not represented at LF, so that [the horse… ] is the head of the predicate. 36. This is demonstrated by the fact that there is no direct syntactic relation between a and the determiner bi that precedes it in (16). 37. An implication of (32) is that, contrary to what was suggested earlier, even in copular sentences where only the attribute is realized, as in Fas la ‘It’s a horse’, the silent pure expletive must be present in the higher topic position of the subject, from which it binds its other occurrence within la. 38. The fact that the syntactic object [VP l [V′ a [DP Pro]]] is related to a paradigm in the Vocabulary is of course essential, even though the way to formalize this relation is not obvious to me. 39. In the same configuration in English, C0 can only be left empty if a lexical antecedent is adjacent to it: see It was the horse (that) the merchant bought vs. The horse it was t *(that) the merchant bought. This might support the claim that the silent expletive of Wolof is a pronominal that can be used as an antecedent (see French ce que, Spanish lo que), whereas the trace of the moved DP in English has no such status. 40. Ku consists in the relative pronominal /u/ and the noun class marker K- referring to humans. 41. A frequent, if not universal, requirement of identificational copulas. See French *C’est le docteur de ce village que Jean est ?‘It’s the doctor of this village that John is’. 42. Sentences with an ‘emphatic’ subject pronoun such as Yaa jënd fas wi ‘It’s you (sg.) who bought the horse’ are derived exactly as shown in (38), replacing the lexical DP in Spec,CP and Spec,VP by an abstract form of the second person singular pronoun. I will return to such forms in Section 5. 43. An affirmative answer is expected. 44. I give PF its standard meaning of everything that is concerned with sound, including Morphology. 45. Perhaps this visibility could be captured by a redundancy rule within the Vocabulary. 46. Compare with English that’s, where that and is do form a constituent, viz. VP, and the process of unstressed vowel deletion is transparent. Another interesting item to compare is French qui, if it can really be analyzed as /que-i(l)/. 47. Sauvageot does not elaborate on his observation, but the fact that he feels the need to make it shows he was aware of a problem here. In English, in contrast, a pause comes as perfectly

FOCUS IN WOLOF

271

natural between the focused term and the complementizer or the subject of the embedded clause (/it was the horse #(that) you bought/). 48. There are more morphemes in (41) than indicated in the gloss. I only spell out what is relevant to the problem at hand. 49. Ambiclausal items may be rare, but I would be surprised if they were an isolated phenomenon. For instance, English speakers who accept Who do you wanna win? would seem to make use of an item wanna which is allowed to straddle the two clauses [who do you want] and [t to win]. The question of course is whether wanna may be considered a Vocabulary item. I think it may, given the idiosyncrasy of the derivation want to > wanna, and the fact that there are dialects, especially Afro-American English, where wanna never alternates with want to (see Labov 1972). 50. See The merchant bought only the horse, which does not normally lead us to think that the merchant refused to buy, say, teapots in addition to the horse, but rather that he didn’t buy other animals that were also on sale. For only as a focus marker, see Bayer (this volume).

References Abraham, W. et al. (eds.) 1996. Minimal ideas: Syntactic studies in the minimalist framework. Amsterdam: John Benjamins Baker, M. C. 1985. “The mirror principle and morphosyntactic explanation”. Linguistic Inquiry 16.373–415. Bennis, H. & L. Haegeman. 1984. “On the status of agreement and relative clauses in West-Flemish”. In W. de Geest & Y. Putseys (eds.), Sentential complementation: Proceedings of the International Conference held at UFSAL, Brussels, June 1983. Dordrecht: Foris, 33–53. Benveniste, E. 1966. Problèmes de linguistique générale. Paris: Gallimard. Boilat, D. 1858. Grammaire de la langue woloffe. Paris: Imprimerie impériale. Brody, M. 1995. Lexico-logical form: A radically minimalist theory. Cambridge, Mass.: MIT Press. Chomsky, N. 1993. “A minimalist program for linguistic theory”. In M. Halle & S.J. Keyser (eds.), 1–52. Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Church, E. 1981. Le système verbal du wolof. Documents linguistiques de l’Université de Dakar 27. Collins, C. 1997. Local Economy. Cambridge, Mass.: MIT Press. Creissels, D. 1991. Description des langues négro-africaines et théorie syntaxique. Grenoble: Ellug. Dialo, A. 1981. Structures verbales du wolof contemporain. Dakar: CLAD.

272

ALAIN KIHM

Dialo, A. 1983. Eléments systématiques du wolof contemporain. Dakar: CLAD. Diouf, J.-L. 1985. Introduction à une étude du système verbal du wolof: relation modes, pronoms sujets et autres modalités du prédicat. Dakar: CLAD. Dunigan, M. 1994. “The clausal structure of Wolof: A study of focus and cliticization”. Ms., The University of North Carolina at Chapel Hill. Fal, A., R. Santos & J.-L. Doneux. 1990. Dictionnaire wolof-français. Paris: Karthala. Foley, W. & R.D. Van Valin. 1985. “Information packaging in the clause”. In T. Shopen (ed.), Language typology and syntactic description, Vol. 1. Cambridge: Cambridge University Press, 282–364. Greenberg, J.H. 1978. “How does a language acquire gender markers?”. In J.H. Greenberg (ed.), Universals of human language, Vol. 3. Stanford: Stanford University Press, 47–82. Groat, E. & J. O’Neil. 1996. “Spell-out at the LF interface”. In W. Abraham et al (eds), 113–39. Halle, M. & A. Marantz. 1993. “Distributed morphology and the pieces of inflection”. In M. Halle & S.J. Keyser (eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger. Cambridge, Mass.: MIT Press,111–76. Horvath, J. 1985. FOCUS in the theory of grammar and the syntax of Hungarian. Dordrecht: Foris. Ka, O. 1994. Wolof phonology and morphology. Lanham: University Press of America. Kayne, R.S. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press. Kihm, A. 1994. “Aspects de l’incorporation en wolof: essai comparé des approches minimaliste et autolexicale”. Recherches linguistiques de Vincennes 23.7–24. Kobès, A. 1869. Grammaire de la langue volofe. St Joseph-de-Ngasobil: Imprimerie de la Mission. Labov, W. 1972. Language in the Inner City: Studies in the Black English Vernacular. Philadelphia: University of Pennsylvania Press. Léard, J.-M. 1986. “ ‘Il y a… qui’ et ‘C’est… qui’: la syntaxe comme compatibilité d’éléments sémantiques”. Linguisticae Investigationes 10.85–130. McCawley, J.D. 1988. The syntactic phenomena of English. Chicago: University of Chicago Press. Njie, C.M. 1982. Description syntaxique du wolof de Gambie. Dakar: Nouvelles éditions africaines.

FOCUS IN WOLOF

273

Pica, P. & C. Tancredi. 1996. “(Anti-)epistemicity”. Ms., University of Paris VIII. Pollock, J.-Y. 1983. “Sur quelques propriétés des phrases copulatives en français”. Langue française 58:89–125. Rizzi, Luigi. 1998. “The Fine Structure of the Left Periphery”. In P. Benicà & G. Salvi (eds.), Romance syntax, Budapest studie in Romance linguistics, 112–158. Robert, S. 1986. “Le wolof: un exemple d’expression morphologique de l’emphase”. Bulletin de la Société de linguistique de Paris 81.319–41. Robert, S. 1991. Approche énonciative d’un système verbal: le cas du wolof. Paris: Editions du CNRS. Rouveret, A. & J.-R. Vergnaud. 1980. “Specifying reference to the subject”. Linguistic Inquiry 11.97–202. Sauvageot, S. 1965. Description synchronique d’un dialecte wolof: le parler du Dyolof. Dakar: IFAN. Solà, J. 1996. “Morphology and word order in Germanic languages”. In W. Abraham et al. (eds.), 217–51. Tancredi, C. 1996. “Intricacies of identity”. English Linguistics 13. Webelhuth, G. 1992. Principles and parameters of syntactic saturation. Oxford: Oxford University Press. Wilson, W.A.A. 1989. “Atlantic”. In J. Bendor-Samuel (ed.), The Niger-Congo languages. Lanham: University Press of America, 81–104. Zwart, J.-W. 1997. “The Germanic SOV languages and the Universal Base Hypothesis”. In L. Haegeman (ed.), The new comparative syntax. London: Longman, 246–67.

Focus in Somali* Jacqueline Lecarme CNRS, France

Abstract In a model that raises a set of minimalist questions about the language faculty and the way it interacts with the external systems of language use, the question of what happens in a language that uses morphological ‘focus markers’ is of particular interest. Functional accounts cannot provide a unitary explanation of the distribution of the waa/baa particles in Somali, and of the full inventory of the grammatical options that the language makes in the overt syntax. The solution proposed in this paper is based on the feature specifications of the matrix C node. I argue that both ‘focus markers’ in Somali are overt ‘root’ complementizers, and that focus phenomena are triggered by the [+nominal] strength of C coupled with a case (EPP) feature, which enters in the computation of the unique structural ‘focus’ position and permits the licensing of expletives and semantically unfocusable types of arguments. This analysis precludes reference to a [+f] syntactic feature, assuming that [+f] operator values are set at the LF-discourse interface, an optional process. Within this restricted formal apparatus, the theory can make sense of the most intricate data of Somali, and succeeds in relating its seemingly exotic pattern to categories long since known of other languages. Further, Somali provides concrete evidence for the pivotal role of the ‘assertive’ C node at the interface levels, as point of contact between clause and discourse.

1.

Introduction

The languages of the East Cushitic family provide important data for the investigation of focus and focus phenomena in universal grammar. In most

276

JACQUELINE LECARME

Cushitic languages, we find an explicit (obligatory or optional) system of morphological focus marking. The phenomenon is exclusively a main clause property. Somali1 is a case in point: either baa or waa (so-called ‘term focus’ and ‘predicate focus’ respectively) must obligatorily occur in declarative main clauses. Beginning with Hetzron (1965), linguists working on the languages of this area have adopted a functional, discourse-oriented perspective. According to Heine and Reh (1984), focus markers directly reflect the topic-focus articulation (in the sense of the contemporary Prague school), which would make Somali a clear case of a language which encodes information structure in the grammar. This position has also been accepted by more syntactically-oriented studies, such as Zholkovski (1971), Antinucci (1980), Saeed (1984), Livnat (1984), Svolacchia et al. (1995). In contrast, Andrzejewski’s works (1964, 1975) incorporate the fundamental insight that focus markers play a central role in the Case system of Somali. The purpose of this paper is twofold. First, I will briefly review and evaluate the major insights behind the traditional, discourse-centered approaches, showing that they do not, in fact, provide a plausible account of the role and distribution of focus markers in Somali. In the second part of the paper, I present a syntactic analysis that provides a new perspective on the problem. As a point of departure, we will be assuming the analysis proposed in Lecarme (1991, 1994, 1995), according to which a declarative C, one of the Force indicators (Chomsky 1996), provides the source of focus markers in Somali. The present account is based on the feature specification of the higher CP level. Following much recent work, I will assume that cross-linguistically, the phenomenon of main clause syntax (such as Verb-second in Germanic) depends on the typology of the C position: declarative C is strong, strength being expressed by a categorial feature. I will propose that in Somali C is [nominal], while it is Verb-related in Germanic. I will further argue that the free functional morpheme baa has an intrinsic ‘default’ Case (EPP) feature, which provides a L-related [Spec,CP] position for overt nominals. The focus position has A-position properties, and thus can serve as an A-binder for the pronominal elements internal to the clause, inducing no WCO effects. Such a state of affairs leads, by independent principles, to the formation of expletive constructions (‘postverbal focus’ constructions) at the CP level. The conclusion reached is that given their specific properties, focus constructions cannot be the result of movement triggered by a quantificational [+focus] feature. Rather, the presence of a structural Case position in [Spec,CP] in Somali correlates with the fact that in this language, overt NPs are adjoined to the sentence and linked to pronominal arguments: the focus position then provides a Case

FOCUS IN SOMALI

277

slot for arguments or nominal categories that cannot be interpreted as referential. 1.1 The phenomenon: Overview In most Cushitic languages, the so-called ‘term focus’ vs. ‘predicate focus’ opposition is marked by an enclisis vs. proclisis process. This is the situation, for example, in Rendille (Oomen 1978): (1)

a.

b.

ínam á-yimi boy F-came ‘The boy came’ ínam-é yimi boy-F came ‘The boy came’

Somali is unusual among the Cushitic languages in showing a consonantal (bilabial) b/w opposition. Of particular relevance here is the fact that Somali is exceptional among Cushitic languages in not being a null-subject language: subject clitics (weak pronouns) may combine with the left-adjacent focus marker, as in the following examples, in which the full vocalism baa/waa means a silent clitic (pro) (2a) or a null category (2b):2 (2)

a.

b.

nínkii wúu/ wáa yimid man-. +/  came ‘The man (he) came’ nínkii baa yimid man-.  came ‘The man came’

As is quite well-known, focus is truly the syntactic property of Somali that gives it its distinctive ‘somaliness’: either baa/ayaa3 or waa must obligatorily occur in declarative main clauses. Topics can be adjoined on either side of the waaclause. Their presence is unnecessary for grammaticality: (3)

(Cáli) wuu iigá warramay (arríntan) Ali F+ me+to-about reported problem-.+ ‘Ali told me about (it)/this problem’

Any NP constituent, subject (4a), object (4b), ‘argument’ PP (4c) or ‘adjunct’ PP (4d) may occupy the focus position left-adjacent to baa. These sentences seem

278

JACQUELINE LECARME

to have identical truth conditions, and only differ in the location of focus. In such cases, focus typically has an influence on the felicity condition of the sentence: (4)

a.

b.

c.

d.

arríntan Cáli baa (shálay) iigá warramay problem-.+ Ali F yesterday me+to-about reported ‘Ali told me about this problem (yesterday)’ aníga ayuu (Cáli) iigá warramay me(strong) F+ (Ali) me+to-about reported (arríntan) problem-.+ ‘Ali told me about it/this problem’ arríntan buu (Cáli) iigá warramay problem-.+ F+ (Ali) me+to-about reported ‘Ali told me about this problem’ (Cáli) shálay buu (arríntan) iigá (Ali) yesterday F+ problem-.+ me+to-about warramay reported ‘(Ali) Yesterday he told me about it/this problem’

1.2 The ‘pronominal argument’ typology Cushitic languages are basically OV languages (in Greenberg’s terms), but are far from being consistently ‘head final’.4 Word order in clauses seems to be remarkably free: so far, I have assumed that Somali shares a central feature of nonconfigurational languages discussed by Hale (1983), and reanalyzed by Jelinek (1984, 1995) and Baker (1991, 1995, 1996) in terms of a ‘pronominal argument’ parameter. In a ‘pronominal argument’ language, the argument structure is pronominal. Full NPs, when they appear, are adjuncts external to the major syntactic structure, associated with an internal position that determines the semantic interpretation.5 The following set of assumptions will account for the basic structural properties of the finite clause in Somali:6 – The CP system is overtly represented in root declarative sentences. The socalled ‘focus markers’ baa and waa are pre-sentential particles realized in the C position. – Clitics occur in all argument positions, are assigned q-roles, and perform all the functions of morphological checking of the Verb inside the IP domain. Full argument-type phrases are adjoined to the sentence (either to CP or IP).

279

FOCUS IN SOMALI

–

Both the fact that overt NPs are optional and the fact that they do not have fixed structural positions follow from the claim that they have the status of adjuncts. As a consequence, NP movement and wh-movement are not available options. There are no syntactic passives, no ‘control’ structures, hence no infinitival constructions.7 Since pronominal clitics play a crucial role in the following discussion, some of their main properties must be mentioned. In Somali, all clitic elements must precede the verb in a fixed linear order. Table (5) summarizes this linear structure, which may be thought of as a series of positions, or slots, filled with grammatically specified morpheme sets (including zero morphemes): (5) 1

2

3

4

5

1S

aan

i

2S

aad

ku

ú

kaa

3S

ay(F), uu(s)(M)

Ø

kú

Ø

1P

aan(u)(incl),aynu(excl)

na, ina

ká

keen, kayo

2P

aad

idin

lá

kiin

3P

ay

Ø

1. 2. 3. 4. 5. 6.

la

6

kay …

Ø

Weak subject pronouns (enclitics) Weak impersonal pronoun (proclitic) Weak object pronouns (proclitics) Prepositional preverbs (lexically accented) Weak oblique pronouns (Other clitic elements).8

Pronominal clitics are either enclitics or proclitics, depending on their lexical properties. Yet, they are only loosely attached to a host, if one appears at all. Subject clitics usually attach to focus markers in main declarative clauses, as in (4b,c,d), although the sequence … baa Cáli (uu)… is also allowed in (4b) and (4c). The phonological complex b+uu, b+aan, etc. then cannot be understood as the ‘conjugated form’ of the focus markers, as in Hetzron (1965). Object clitics are proclitics, but need not be adjacent to the verb. They may combine with the right-adjacent prepositional slot into complex phonological words, regardless of thematic head-complement relations.9 Taking VP as the domain in which semantic roles are assigned, (5) at the same

280

JACQUELINE LECARME

time provides a clear picture of the argument structure. Given my assumption that both focus markers occupy the C position, we can think of Somali as a ‘free constituent order’ language with special slots for adjunct arguments before the main sentence, i. e. the IP domain, which only contains pronominal arguments:

(6)

(CP) (tops) [+ − nom]

(CP) CP

C waa

IP

CP (tops) [+ − nom] (Spec) [+acc] C baa

C′ (IP) (tops) [+ − nom]

IP

Within a minimalist approach, the baa/waa complementarity is a matter of optional selection of an element for the numeration. The lexical entry for root C specifies two forms, which are clearly related morphologically, but still have different formal and phonological properties that are coded in their lexical entries. Whether the structure has an obligatory filled focus slot (i.e., a [Spec,CP] position) or not, whether it allows IP-adjunction or not, depends on which option is selected in the numeration. In the syntax-prosody mapping, waa heads a ‘maximal phonological phrase’ (in terms of the prosodic hierarchy of Selkirk 1986) inside of which argument clitics occupy distinct sub-domains. Baa is phrased together with its specifier, subject clitics, and other enclitic categories of clausal scope.10 Under this analysis, no syntactic process moving the subject enclitic with its host is involved. Rather, subject ‘clitic climbing’ effects in the baa-clause must be understood as an optional PF process, moving the subject clitic across the IP-adjoined accented phrase, thus splitting the clitic cluster into two discontinuous subdomains: (7)

a.

[cp (Cáli) [cp w-[ip uu iigá warramay] (arríntan)]] [ Ali + me+to-about told problem-.+ ‘Ali told me about this problem’

FOCUS IN SOMALI

b.

281

[cp (Cáli) [cp arríntan b[ip uu (Cáli) iigá [ Ali problem-.+ . (A.) me+to-about warramay]]] told ‘Ali told me about this problem’

At this point, we are in a position to determine the extent to which the ‘pronominal argument’ analysis applies to Somali. First, Somali is not a pronominal agreement language: clitics are not inflectional morphemes or incorporated pronouns. Subject agreement on the Verb does not render superfluous a pronominal clitic in subject position. As for object clitics, they cannot be taken to be ‘agreement’ either: they do not undergo syntactic cliticization, or pre-syntactic affixation, hence never appear attached to the Verb. Rather, they are phonological clitics attached to their right-adjacent host in a PF process. Secondly, under the DP hypothesis, Somali clitics are not Ds (as in Romance), but purely nominal XP categories: as clear morphological evidence suggests, subject clitics are ‘weak pronouns’ in the sense of Cardinaletti & Starke (1995), that is, nominal categories lacking a D-projection.11 Being NPs or NumPs (not Ds) in the syntax, they behave not like Romance or Semitic clitics, but more like Germanic or Celtic pronouns: they cannot function as syntactic variables, and form representational A-chains that enter into coreference or binding relations with the adjoined NPs. Being XPs and arguments, they satisfy the q-criterion and perform all the functions of morphological checking of the Verb in the IP-internal domain. 1.3 Focus Markers and ‘Discourse Configurationality’ 1.3.1 The Topic-Focus Articulation An idea that has been attractive to many, and is common to both traditional and more recent accounts of focus and focus phenomena in Somali, is that word order in this language is pragmatically determined. More explicitly, according to Heine & Reh (1984), the role of baa and waa is to mark that constituent in the clause which contains ‘new information’ (i.e, either factually new, as an answer to a question bearing on a term or the predicate, or in contrast with what has been said before).12 Supposing languages are allowed the option of encoding the topic-focus articulation (instead of the subject-predicate relation) as an integral part of the deep structure, as in Hajicˇova and Sgall’s current proposals, then we would expect the distribution of focus markers to directly reflect the binary given/new

282

JACQUELINE LECARME

opposition. Supposing that this is so, (8) is about ‘this problem’, and provides the new information that Ali told me about it. Thus, the entire baa-clause (including the focus position) is predicated of the topic, introducing new information. This calls into question both the function of focus markers as an information encoding device, and the informational status of topics. Topics can occur both inside and outside the maximal focus-phrase, as is shown in (8), where √ marks other potential structural positions in which NPs such as arríntan ‘this problem’ may be adjoined. If these adjuncts were truly informational topics, they would in principle create a new binary opposition (new vs. given information), and interpretative conflicts would necessarily arise. (8)

(arríntan) Cáli baa √ iigá warramay √ problem-this Ali  me+to-about reported ‘Ali told me about this problem’

As we have already mentioned, and will make more precise below, full NPs in Somali must appear as adjuncts, which are not necessarily associated with the same marked interpretation as left dislocation constructions in Romance languages, which have a fixed discourse function. Rather, full NPs are in a looser relation of modification (or appositional relation) with the pronominal arguments, and have in most cases the same neutral interpretation as subjects or objects in Romance languages. It is not accidental that topic marking in Somali, as in other languages, tends to make use of adjunction structures that put the topic outside the sentence, but this is a mere tendency, and does not exclude the possibility that this problem in (8) or any other NP in this position introduces a new referent that has not been talked about before. In sum, precisely because Somali must express full nominals as adjuncts, there is no evidence that syntactic structure directly encodes topic-focus articulation. 1.3.2 Focus markers and Focus effects It is commonly assumed in the literature, implicitly or explicitly, that focus markers, like contrastive intonation in English, obligatorily induce focus effects at the discourse level. There is, however, evidence that it is not so. While focus markers clearly delimit a phonological domain, this is not a domain of prominence in the informational sense. Neither topic nor focus positions in Somali have a fixed discourse function, as is shown by many examples from conversation or texts:

FOCUS IN SOMALI

(9)

a.

b.

283

Wáa baa waxaa beló isugú faanáy libáax, góod time  + calamity +of-at were-proud lion snake iyo habár. and old-woman ‘(Once upon) a time, a lion, a snake and an old woman measured their powers of evil’ Libáaxii baa hór hadlay, óo yiri … lion-. F first spoke and said ‘The lion first spoke and said …’

The focus position (left-adjacent to baa) is the typical position for merger of adverbial NPs like wáa ‘time’ that situate the background of the sentence (9a).13 There are also clear cases in which an argument-type NP in focus position functions as an informational topic: libáaxii ‘the lion’ in (9b) clearly does not contain ‘new information’, since it has just been introduced in the previous discourse. Rather, the time-sensitive morpheme -ii affixed to the definite determiner suffices to indicate that ‘the lion’ has been mentioned before, and thus cannot be interpreted as new information. In fact, the characterization of topic and focus positions in terms of discourse functions in Somali would be especially inappropriate, given the intrinsic deictic properties of the determiner system (see Lecarme 1996). Furthermore, in many cases, baa constructions and waxaa-postverbal focus constructions may be superimposed. This is illustrated in (10): (10)

a.

b.

árdaygan baa wuxuu dóonayaa ínuu student-.+  ++ wants + arkó warqáddiisa see- note-... ‘This student wants to see his notes’ wéertan ayaan waxaan ká soo guuriyáy sentence-.+ + + from (.) quoted búuggaas book-.+ ‘This example I quoted (it) from this book’

Here, the [XP baa] position clearly lacks a fixed discourse function: it can be used either to express constrastive focus, contrastive topic (paraphrasable as ‘as for’), or simply the background assumption that the italicized NPs — ‘this student’, ‘this example’ — are under discussion (perhaps provided by the context, perhaps

284

JACQUELINE LECARME

not).14 Clearly, then, the informational status of the NP in focus position cannot be derived from the presence of focus markers, or from constructional meaning. A more serious problem for discourse-centered approaches results from the fact that typically, the [XP baa] position can be occupied by constituents that cannot be interpreted as the focus of the sentence, such as the implicit argument wax ‘thing’ (11a). It is noteworthy here that in Somali, third person object pronoun forms are phonologically null, but semantically have the interpretation of definite pronouns (11b). Thus, an indefinite interpretation can only be obtained by the use of a lexical ‘understood’ argument:15 (11)

a.

b.

(i dháaf), wax báan akhrínayaa ( let thing + am-reading ‘(let me alone) I am reading’ (búuggan) wáan — akhriyay (book-.+ + pro read ‘(this book) I read it’

Given these facts, which will be discussed in more detail below (Section 2.3), it comes as no surprise that many irregularities are found in the question/answer parallelism, which has routinely been taken as a central diagnostic of focus. Although it is true that felicity conditions are met in the general case, it is also true that questions and answers tend to respect a purely syntactic, rather than a pragmatic symmetry: (12)

a.

b.

(sáaka) muxúu sameeyay? — Búug buu akhriyay (today what++ did — book + read ‘What did he do today? — He read a book’ (sáaka) wax má akhriyay? — Máya, wax búu qoray (today thing  read — No thing + wrote ‘Did he read today? — No, he wrote (lit.: he wrote thing)’

Finally, focus markers in most cases do not trigger any special pragmatic effect: both Cáli baa yimid and Cáli wúu yimid ‘Ali came’ are normal, unmarked clauses, simply because it is the only structure that is available. Though commonly used, glosses such as ‘it is Ali who came’ are then inappropriate, since cleft constructions in English or Romance languages are grammaticized structures that are obligatorily interpreted as marking focus. Moreover, it has long been observed that as far as ‘predicate focus’ is concerned, the given/new distinction is irrelevant16.

FOCUS IN SOMALI

285

Summarizing, focus markers in Somali are not discourse markers, and topic and focus positions in this language are not grammaticalized discourse functions. The traditional notion of ‘discourse configurationality’ then only provides useful descriptive correlations. Assuming that the system of morphological ‘focus markers’ instantiates a root C, it is determined both by invariant properties of language and independently motivated parametric choices, not by language use. Yet from a minimalist perspective, there is an answer to the question of why we do expect to find a constituent bearing ‘new information’ in focus position (leftor right-adjacent to baa and waa, respectively), even preferably in certain discourse contexts. Such correlations can be derived from the meaningful elements present in the ‘assertive’ C node, and the information they provide at the interface to the systems of language use.

2.

Syntactic analysis

2.1 Focus Markers as Overt Declarative C As a review of the main distributional properties of focus marking in Somali will immediately reveal, they are systematically incompatible with the pragmatic notion of focus: (13)

A. B. C.

Focus markers are restricted to root clauses, and are obligatory there17 Focus markers are excluded in complement and relative clauses There is only one focus position per clause.18

These properties are clearly syntactic in nature. Focus, as a pragmaticallyconstrued entity, must be considered an optional process (Rooth 1992). It can affect elements at any level of syntactic structure: in particular, embedding does not affect the contrast possibilities expressed by focus. It can have multiple instantiations in a single clause. Within the minimalist framework outlined by Chomsky (1996), pragmatic focus involves operations that apply at the interfaces, i.e., the optimal mapping of PF and LF. Since focus markers are not optional, but either obligatory or excluded in well-defined syntactic environments, they cannot be imposed ‘from the outside’ by the performance systems that make use of the information provided at the interfaces, and must originate in the syntax proper. The C analysis outlined above captures the cluster of properties in (13) in

286

JACQUELINE LECARME

a very simple way. The absence of focus markers in dependent clauses is naturally explained, due to the complementarity between focus markers and other complementizers. The distribution of focus markers in Somali can now be seen as an instance of a more general cross-linguistic pattern. From a minimalist point of view, formal features that determine clause type (declarative, interrogative, etc.) are [Interpretable] features, hence need not appear overtly, unless their presence is required by some structure-building principle, depending on formal, language-specific properties of the functional C. Let us assume that Somali declarative C is strong, as in Germanic, strength being specified by a categorial, [−Interpretable] feature: strength then must trigger an overt operation before spell-out. Strength of C in V-second type languages is partially V-related (Holmberg and Platzack’s (1988) ‘finiteness’ feature), hence the inflected Verb typically raises to C, thereby relating the CP and IP systems. Strength of C being [nominal] in Somali, I-to-C raising is excluded, on grounds of economy. A free functional morpheme (either of the socalled focus markers) is then adjoined to C by Merge, eliminating the strong [nominal] feature of C. The Merge option available in Somali gives support to the view that the CP system is fundamentally distinct from the IP system, in that it is not directly related to the Verb (Rizzi 1997). Since Somali focus markers are essentially nonverbal categories, the higher CP need not (in fact, may not) be associated with an IP projection in purely non-verbal contexts (i.e. [−neg], [−past], [person3]) such as nominal predicates, equative sentences, and ‘small clause’ questions (14a,b,c), in which they function as a pure, nonverbal copula:19 (14)

a.

b.

c.

d.

(nínku) waa macállin/waa hormúudka kulliyádda (man-the[+nom]  teacher/ dean-. faculty-. ‘This man is a teacher/is the dean of the faculty’ (nínku) ma macállin baa? (man-the[+nom]  teacher  ‘Is this man a teacher?’ (nínku) macállin weeye /*wúu yahay (man-the[+] teacher  + is() ‘This man is a teacher’ waa [cp ínuu keeno] (búuggan)  + bring (book-.+) ‘He must bring it/this book’

FOCUS IN SOMALI

e.

287

keenó] baa? ma [cp ínuu  + bring  ‘Must he bring it?’

The Verb ‘to be’ is excluded in such examples, as the ungrammaticality of (14c) shows. This suggests that the structure of non verbal copula constructions involves a root C directly selecting a NP, a DP or a CP complement20. For any specification of tense, person, or negation, the Verb -ah- ‘be’ (and a full IP projection) is required: (15)

a. *adígu waa macállin you(strong)[+nom]  teacher b. adígu macállin baad tahay You(strong)[+nom] teacher + are ‘You are a teacher’ c. Aaminá macallimád má aha Amina teacher()  is ‘Amina is not a teacher’ d. [cp ínuu keenó] bay ahayd [ + bring() + was ‘He had to bring it’

Summarizing, the option illustrated by Somali clearly shows that there are two distinct processes, which often appear superimposed in V-second phenomena: one induced by the strength of the functional C determining clause type, the other triggered by intrinsic properties of the lexical element that checks this feature, either the inflected Verb, or a free morpheme. Supposing that the lexical entry of C in Somali specifies two basic forms baa and waa, both values of C satisfy the [nominal] strength of C. In addition, baa is lexically endowed with a formal feature that allows an extra specifier. Much of the following will have to do with this formal feature, and what appears in the specifier position it makes available. 2.2 Topic and Focus positions In Somali as in many languages in which topic and focus are structurally expressed, there is an evident asymmetry between topic and focus positions:

288

JACQUELINE LECARME

(16)

A. B. C.

D.

E. F.

There can be multiple topics in a clause. There is an unique structural focus position. Topics are optional and typically unordered. The focus position (left-adjacent to baa) is fixed, and must be filled. Topics are possible in matrix clauses and other clause types (complement, adjunct, relative clauses). Focus is limited to main declarative clauses. A binary system of morphological Case [± nominative]21 is available in topic position; the morphological Case of the topic is q-related in that it always corresponds to the Case of the clause-internal pronominal argument it is connected to. The NP in focus position bears an invariant, structural [+acc] Case (the ‘default’ Case of the Noun in citation form), whatever is the Case of the pronominal argument in the clause. Topics are independent prosodic units at the level of phrasal phonology. Baa is phrased together with its specifier. The relation between topics and pronominal elements in the clause is coreference. The association between focus and pronominal elements can be expressed in terms of local relations (binding).

The fundamental asymmetry reflected in (16) has long been observed across languages: at a descriptive level, topics have a tendency to make use of dislocations or adjunctions that put the topic element outside the sentence, whereas focus is often marked by sentence-internal devices. I will assume then, as in earlier works, that topics are adjoined to CP or IP, while focus occupies a fixed, unique specifier position. Without additional mechanisms, the view that topics and focus in Somali occupy specifier positions in a recursive CP, as has been recently proposed by Svolacchia et al. (1995), is empirically inadequate to account for the properties listed in (16). An approach assuming full TopP and FocP projections at the ‘left periphery’, in the spirit of Rizzi (1997), would not be appropriate either, since movement in the Spec of these projections has to be triggered by the satisfaction of focus and topic criteria, which are by definition universal interpretive requirements. In fact, it is particularly dubious to make the semantic notion of focus and topic relevant in the syntax, if focus and topic positions are the normal mode of licensing NPs in Somali.

FOCUS IN SOMALI

289

2.3 Focus position is not quantificational 2.3.1 On some topic/focus non-asymmetries Most syntactic analyses assume a parallelism between focus constructions and other operator-variable constructions in human language. In the approach put forth by Chomsky (1976) and common to most recent analyses, focus translates as a syntactic feature accessible to the computational system, which turns a constituent into an operator moving overtly or covertly in the syntax. But as observed by É. Kiss (1995) and others,22 the semantic notion of operator cannot explain the uniqueness of the focus position: there can be both multiple topics and multiple foci in a clause. ‘More than one topic, one focus’ is then only a structural requirement: questions and focalizations typically allow no more than one constituent to be questioned or focalized in the syntax, because these constructions crucially make use of the C position, which is unique for a given clause. In fact, there is strong evidence that in Somali, the focus position has no quantificational properties. Focus constructions do not display the set of specific properties of focus chains when compared with topic chains, which, following Cinque (1990) and Rizzi (1997), are revealing of the presence vs. absence of an operator. In both topic and focus constructions, pronominal clitics (or their null counterpart, which will be assumed here to be pro) are required in all argument positions: (17)

a.

b.

idínka bay idin aragtay you(str.) +  saw() ‘She saw you (pl.)’ idínka wáy idin aragtay you(str.) +  saw() ‘She saw you (pl.)’

Furthermore, Somali focus constructions show no WCO effects (cf. Lecarme 1991)23: (18)

a.

b.

Warsamé baa hooyádiis — jeceshahay W.  mother-.+. pro loves() ‘Warsame hisi mother loves (himi)’ Adíga ayaa hooyádaa ku jecéshahay You  mother-.+.  loves() ‘Youi youri mother loves youi′

290

JACQUELINE LECARME

This fact is naturally explained, if we assume that it stems from the nonargumental status of the NPs in Somali. Descriptively, WCO effects arise from LF configurations where a pronoun and a trace are both bound by a quantifier, and the pronoun is contained within an argument NP that c-commands the trace. Since pronouns occupy argument positions, the LF representation of (18) contains no variable: supposing the NP has raised in (18), the potential bindees are the clitic and the trace of the focalized NP, neither of which qualifies as a syntactic variable. Assuming that a bare quantifier must bind a variable, and that pronouns cannot be directly A′-bound, strongly quantified NPs such as everything, nothing, nobody, etc. are predicted not to exist in Somali, since the pronoun cannot be interpreted as a variable bound by the dislocated element at LF. But as Baker observes, strongly quantified NPs (including wh-phrases) are not entirely absent in ‘pronominal argument’ languages, although each quantified expression requires a separate discussion. Somali bare quantifiers kúlli, dhammáan ‘all’, qóf-na ‘no one’, etc. are excluded both in topic and focus position, but are fully acceptable in either position when the quantified expression includes a lexical restriction, a genitive pronoun (19c,d), or a full genitive DP (19b): (19)

a. *dhammáan baa timid entirety()  came() ‘All came’ b. dhammáan ardádii baa timid entirety() students-.  came()() ‘All the students came’ c. dhammáantood baan aruuriyay (buugáagta) entirety-.+. + gathered-up books-. ‘I all gathered them up (these books)’ d. dhammáanteen wax báan akhrinnnay entirety-.+. thing + read ‘We all read’

Similarly, adverbial quantification (always, never) is also expressed by a nominal such as wéli (m) ‘time’ associated with a genitive pronoun coindexed to a pronominal argument in the clause: (20)

a.

wéligay dúhurkii ayaan wax cunaa time-.+. noon-. + thing eat ‘I always eat at noon’

FOCUS IN SOMALI

b.

291

hílib doofaar wéligay má cunin meat pig/warthog time-.+.  eat() ‘I have never eaten warthog meat’ (cf. ‘on my life I didn’t…’)

Adverbial quantifiers, like existential quantifiers, are categorially Nouns, project a DP, and syntactically function as the head of a genitive construction. Quantified expressions are then syntactically complex, consisting of a head Noun and a ‘possessor’, a pronominal or a full NP they quantify over, which typically bears a distinct index from the DP as a whole, and in turn is connected to an argument position within the clause. Assuming quantifier raising at LF, QR covertly raises the quantificational feature, adjoining it to D. The quantifier then binds a variable in topic position, which in turn is connected to a pronominal argument. Finally, non referential indefinites provide a very interesting set of facts. As we have noted, the Somali lexicon is typically ‘nominal’ in character, and includes a number of nominal elements that cannot refer, or have a non-referential reading, such as implicit arguments, dissociable compounds, or idiom chunks.24 Such nominals are therefore excluded in adjunct positions. They only occur either in focus position (21a, 21b), or in the immediately preverbal position (21c): (21)

a.

b.

c.

dóolarku hóos buu ú dhacay dollar-.[+nom] bottom + to fall ‘The dollar has fallen’ hádalkan áad buu ú badányahay talk-.+ ‘muchness’ + to is-long ‘This talk is very long’ hádalkii baa áad ú badnaa talk-.  ‘muchness’ to was-long ‘This talk was very long’

An NP which is part of an idiomatic construal, such as áad ‘muchness’ in áad…ú ‘a lot’ expressing adverbial quantification, has no meaning of its own, and never occurs as an independent NP. That such ‘meaningless’ elements would be quantified presents a problem, since they do not have NP-type meanings. In fact, under the assumption that focus constructions involve syntactic movement driven by a quantificational [+focus] feature, these examples become a complete mystery. Supposing our analysis to be correct, such idiomatic construals as (21) follow directly from the availability of a structural Case position in [Spec,CP] in Somali, into which idioms chunks and other non-referring nominal expressions can be licenced in a local c-command configuration.

292

JACQUELINE LECARME

2.3.2 On wh-movement and the [wh] feature In Somali, so-called wh-expressions have the same distribution as focused phrases, and form ordinary-looking constituent questions, in which the wh-phrase might be the subject (22a), the object of the Verb (22b), the object of a prepositional preverb (22c), or a true adjunct (22d): (22)

a.

b.

c.

d.

Yáa fahmay su’áashayda? who+ understood() question-.+. ‘Who understood my question?’ Muxúu Biíbisiída ká maqlay? what++ BBC-. at heard ‘What did he hear at the BBC?’ Sidée baad ú mala-awaashay? way-.+ + (.) guessed ‘How did you guess (it)?’ Xaggée bay ahayd? place-.+Q + was() ‘Where was it?’

However, Somali interrogatives differ in a fundamental way from those in English. As a subcase of focus constructions, they do not display the properties of overt wh-movement. Constituent questions, like focalizations, lack WCO effects (cf. Lecarme 1991), and the the relation between the focused or questioned NP and the resumptive clitics is not sensitive to island constraints (Saeed 1984; Lecarme 1991, 1994)25: (23)

a.

b.

c.

Yaa hooyádiis — jecéshahay? + mother-.+. pro loves() ‘Whoi does hisi mother loves?’ qoraagée ayaad jecéshahay búugga uu qoray? writer-.+ + like book-.  wrote ‘Which author do you like the book he wrote?’ ardaygée ayaad dóonaysaa ínaad — lá hadashid student-.+ + want + pro to talk íntuusan tegin before-.++ go() ‘Which student do you want to talk to before he goes?’

FOCUS IN SOMALI

293

These facts can be partially unified with those discussed in the preceding section. If Somali is a pronominal argument language, then wh-expressions cannot occur in argument positions. In situ questions are therefore excluded. It also follows that there are no multiple interrogatives, as in English, where only one interrogative phrase moves to C, while the others remain in argument position. Multiplefronting of wh-phrases, as in Hungarian, is not allowed either: as there is only one focus position per clause, only one wh-phrase can be licensed in the [Spec,CP] position. Somali question words and D-linked wh-expressions are also excluded in adjoined positions, given their indefinite, non-referential nature. Some variant of the in situ strategy must then be used: interrogative elements are obligatorily associated with the root C node, and licenced in Specifier or complement position: (24)

a.

b.

c.

d.

wáa buuggee?/buuggée waaye?  book-.+/book-.-  ‘Which book is it?’ buuggée baad akhrisay? book-.+ + read ‘Which book did you read?’ kiinnée ayaa akhriyay? you(pl)+  read() ‘Which one of you read (it)?’ maxaa igú dhacay? +thing+ me+to happened ‘What happened to me?’

I interpret these facts in the following way. As in many languages, Somali question words combine a wh-element (ma, -ee) and an indefinite: for instance, maxaa (ma+wax+baa) is simply an indefinite NP wax ‘thing’ in the scope of an interrogative particle. Languages differ in which element must be checked in overt syntax. Since there is no evidence for a strong [wh] feature in Somali, what is checked in focus position is the categorial feature of the indefinite nominal.26 In fact, there are strong reasons to think that wh-movement operates only covertly in the language. Embedded interogatives do not involve a [+wh] C, but display the same structure as relative clauses. Interestingly, the head of this kind of relative clauses must take a definite article, which in turn may not support an overt wh-morpheme:

294

JACQUELINE LECARME

(25)

a.

b.

c.

má ogi árdey-ga/*-gée shálay yimid  know() student-./*-.+ yesterday came() ‘I don’t know which student came yesterday’ wáxaan is weydíinayaa sí-da/*-dée aad ú ++  am-asking manner-./.+ 2 () mala-awaashay guessed ‘I wonder how you guessed (it)’ ma ógtahay wáqtiga iyo méesha ay ká dhacayso  you-know time-. and place-.  at is-falling xafládda reception-. ‘Do you know when and where the reception will take place?’

Relative clauses themselves bring no independent evidence supporting a movement analysis. Rather, the Somali relativization strategy involves in situ pronouns in the positions of the variable. There is no overt C,27 no relative pronoun, and no relative operator: (26)

a.

b.

c.

ardáydii[+acc] — shálay hálkaas joogtay students-. e yesterday place-.-dem stay(R) ‘The students who were there yesterday’ xafíiska[+acc] uu wasíirku — kú shaqáynayo office-.  minister-. pro in is-working ‘The office in which the minister is working’ ardáyda[+acc] aan magacyádooda qórayo (wáy students-.  names-.+. am-writing (they gudbeen) succeeded ‘The students whose names I am writing (succeeded)’

Note that the head of the relative clause must be marked by the same invariant, ‘default’ [+acc] Case as the NP in focus position, whatever the grammatical function of the associated pronoun within the relative. Assuming Kayne’s (1994) reanalysis of relative clauses as [D CP] structures, the ‘head’ of the resumptive pronoun relatives in (26) is in [Spec,CP].28 Suppose further, slightly anticipating the conclusions in the following section, that the [Spec,CP] position is L-related in Somali, the most natural conclusion is that relativization only implies A-positions and relations. Summarizing, the Somali facts give no support to the view that focus

FOCUS IN SOMALI

295

configurations arise from a movement process driven by an operator-like [+focus] feature. There are also strong reasons to think that overt wh-movement is not an available option in the language. We will therefore assume that quantificational features are optionally adjoined to C at a covert level. In the light of these facts, it seems necessary to seek an alternative explanation for the structural properties of the focus position in Somali. 2.4 Focus position has A properties Another fundamental assumption in current syntactic theory is that there is a close parallelism between the nature and origin of structural focus positions and of structural Case assignment positions. This second approach, which is compatible with the facts discussed here, also retains a classical idea in generative grammar: focus movement and other ‘surface effects’ on interpretations are relegated to the PF component (Chomsky 1971), or expressed at some intermediate level that is concerned with the ‘phonological part’ of the Case Filter, the overt licensing of arguments, and configurational word order (Kayne 1994; Chomsky 1996). Horvath (1984, 1995) and Tuller (1992) explicitly express the idea that focus assignment and structural Case assignment are analogous processes. However, as Kenesei (1995) observes, focus and (structural) Case crucially differ in their grammatical effects: a) Case is obligatory and unique (in the sense of the Case Filter); focus is optional. b) Case is always assigned to entire DPs (or clauses); focus can be assigned to any phrasal nodes. c) Case has no semantic consequence; focus does have semantic effects. Assuming these contrasts are indeed revealing, it follows, for reasons already discussed, that the focus position in Somali is indeed a Case position. This evidence is confirmed by the following salient properties, as listed in (27): (27)

A. B. C.

D.

The constituent in [Spec,CP] must be [nominal]: thus only verbs and adjectives are excluded in focus position. The ‘focused’ constituent must be an entire DP (or CP)29 projection. Both Spec and C must be simultaneously occupied, triggering ‘doubly-filled Comp’ effects (contrary to operator-like elements, by economy): following Platzack (1984) among others, only lexically filled heads can assign Case. Spec,CP can host a pure expletive, which associates with a ‘postverbal focus’ position.

296

JACQUELINE LECARME

Let us, now, turn to the analysis of expletive constructions, which I think provide the strongest motivation for a Case-based account. 2.4.1 Postverbal focus: Expletive constructions An important subset of the baa-constructions allows a waxaa-construction, in which the focused constituent appears extraposed to the right, following (i.e. right-adjacent to) either the Verb or a postverbal adjunct NP. As in earlier work, I will assume that waxaa phonologically combines baa/ayaa and the expletive element wax ‘thing’ (Lecarme 1991, 1994).30 The main property of expletive constructions holds, namely, the associate of the expletive, if interpreted as the subject, determines agreement (gender/person only: see 2.4.2 below) with the Verb. But crucially, Somali expletive constructions do not show the set of restrictive properties of their English (and Romance) counterparts, such as Definiteness Restriction (DR)-effects, or ergativity/ intransivity constraints: (28)

A. B. C. D. E.

(29)

a.

b.

c.

d.

The Verb can be intransitive (29a) or transitive (29b) The postposed DP can be indefinite (29a) or definite (29b,c,d) The subject inversion process is free, although Somali is not a null-subject language (29a,b,d) Extraposition is not limited to the subject position (29c,e) Expletive constructions can be negative (29d) Waxaa kúu timid gábar .+ +for came() girl[+] ‘A girl came for you’ Waxaa nóo sheegay wárkaas Warsáme .+ +to said new-.+ W[+] ‘Warsame told us this new’ Af Soomáaligu wuxuu ká mid yahay language(m) Somali-. ++ among one is afáfka Kushitikáda languages-. Cushitic-.[+] ‘Somali is a Cushitic language’ Waxaan búug keenín árdaygan .++ book brought student-.+[+acc] ‘This student didn’t bring any book’

FOCUS IN SOMALI

e.

297

Wáxaa layskú wada raacsán yahay ín + +-on together agreeing are  falálkaasi (ay) ká soo jeedaan (ásal) verbs-.+[+nom]  from are coming origin af Semétig ah language Semitic is(R) ‘Everyone agrees that these verbs have a Semitic origin’

It is noteworthy here that focus-related effects, even with ‘light’ constituents, are clearly less marked than in English presentational there-constructions.31 I interpret these facts in the following way: since Somali expletive constructions are construed at the CP level, and since there is no I-to-C movement in the language, it follows that the associate is not c-commanded by the Verb. Expletive constructions then do not provide either the adjacency requirement for inherent (partitive) Case, in line with Belletti’s (1988) analysis, or the syntactic basis of so-called presentational sentences, namely, following Guéron (1989), a state operator contained in a verbal chain, and binding the the indefinite NP’s empty position. Somali expletive constructions can be represented by the following structure32, in which the focus position is a right-hand specifier, in the spirit of Belletti & Shlonsky (1995)’s analysis of postpositional subject constructions in Italian, thereby accounting for the fact that the postverbal focus is unique and structurally higher than postverbal DPs, which I assume to be adjoined to IP in (30). In expletive-associate chains, relations are typically local: in Somali, the link between the expletive and the associate is mediated by a pronominal element33:

(30)

(CP) C′ C waxi aa

(Spec) NP[+acc] IP ...ei ... (pro/clit.)

The analysis I am proposing is a simple extension of the theory of expletives included in the minimalist program, in effect, a generalization of EPP to C. As a ‘pure’ expletive, wax (like there) has no semantic or formal features34 apart

298

JACQUELINE LECARME

from its category [N]. Interestingly, while there is inherently V-related, as suggests its semilocative character, wax ‘thing’ is intrinsically nominal. Assuming that wax is merged in the [Spec,CP] position, the categorial [N] feature of the expletive satisfies the strong N-feature of C (EPP), but not its Case feature. An associate is thus required in the structure, and provides the Case at LF. Further evidence for this analysis is supported by the following contrasting examples, in which wax is either a pure expletive (31b), or an implicit (indefinite ‘understood’) argument(31a): (31)

a.

b.

wax búu akhriyay thing() + read ‘He read (lit.: he read thing)’ wúxuu akhriyay búug ++ read book ‘He read a book’

As an implicit argument, (ie, a Case-related category), wax is assigned Case in [Spec,CP] and no associate can be licensed in the structure. As an expletive, wax has no Case and an associate is required, in order to check the Case feature of C. The distinction is made phonologically explicit: in (31b), the expletive combines with both the focus marker and the adjacent subject clitic; in (31a), no contraction nor assimilation processes occur. 2.4.2 Binding relations As in many languages, the focalization (or relativization) of a subject NP in Somali triggers special syntactic effects. Descriptively, a) a subject in focus position bears an invariant, ‘default’ [+acc] Case; b) the Verb displays the socalled ‘restrictive paradigm’,35 whose crucial feature is the lack of number agreement (Lecarme 1995); c) no overt clitic appears in subject position. Subject relative clauses (32c) show a similar pattern of Case and agreement: (32)

a.

b.

ardáyda baa — garán students-.[+acc]  — understand() ‘The students are understanding your question’ su’áashaada question-.+. ardaydée ayaa — joogta? students-.+  — is-there()() ‘Which students are there?’

FOCUS IN SOMALI

c.

299

kooxáha — Xámar kú dagaallámaya groups-.[+] — Mogadiscio in is-fighting()() ‘The factions who are fighting in Mogadiscio’

If the NP in focus position is a non-local subject, as in (33), a coreferential subject clitic, fully agreeing with the Verb, must occur in the lower clause: (33)

ardáyda ayaan ú maléynayaa ínay students-.[+] + to think -[+nom] búuggan sóo akhriyeen book-.+ () read ‘I think that the students have read this book

Distributionally, the paradigm illustrated by (32) corresponds to typical cases of ‘extraction’ from the subject position, but differs in fundamental aspects. In standard operator movement, typically, the trace must be in a Case marked position, and the operator must inherit Case from it. If the empty category in subject position in (32) were a trace, then we would expect the NP in focus position to be marked [+nominative]. The empirical evidence also suggests that the ‘long movement’ configuration (33) is not a reflex of A′ movement, given Case considerations. The only possible conclusion, then, is that the NP in [Spec,CP] position is assigned [+acc] Case in situ in both (32) and (33). In line with proposals made in Lecarme (1994, 1995), I assume the following representation:

(34)

CP ardáydai [+acc]

C′

C baa

IP ei

I′

I V+[T+AgrS] garán

VP

The focalization of a local subject in (32) is a potential A-movement configuration, assuming the subject has raised from a caseless position. But this explanation is not available for non-local subjects in focus position. The only possibility,

300

JACQUELINE LECARME

then, is that the default [+acc] ‘focused’ NP is merged in a [Spec,CP] position. Assuming that the [Spec,CP] position is L-related in the sense of Chomsky (1993), the structural relation of e to its antecedent is made compatible with Condition A of the Binding Theory. In our terms, the empty category in [Spec,IP] is not a trace, but a [+anaphoric, +pronominal] category that shares the non-referential properties of PRO and its restriction to subject position, and checks the ‘restrictive’ agreement features of the Verb in a Spec-head relation. A similar account is available for relative clauses, given our former assumptions (Section 2.3.2). There is, then, no process of movement involved in the derivation of these constructions. This is expected, given the non-argumental status of NPs in the language. Viewing a chain as the abstract, discontinuous representation of an argument, Somali chains are not derived by movement, but rather by relating two base-generated positions. The relation between the NP in focus position and the thematic positions of the clitics is typical of A-relations. Covert raising of the whfeature, or optional adjunction of the [+focus] feature in LF, do not modify the status of e: although e is interpreted as a bound variable, it is not directly A′-bound. To sum up, in Somali, the strength of C has a purely [nominal] specification, and the CP and IP systems remain morphologically unrelated. Free functional morphemes waa/baa then adjoin to C by Merge to satisfy the non interpretable [strength]. Assuming baa/ayaa is lexically endowed with a Case feature, the focus position has A-position properties, and thus can serve as an Abinder for the pronominal elements internal to the clause, inducing no WCO effects. I further assume that the ‘focused’ element is inserted (substitution) by Merge directly in [Spec,CP], by economy: there is no FI violation, since the qcriterion is satisfied by the pronominal arguments.

3.

Summary and Conclusions

It has been proposed here that in Somali, the focus system is a Case-related device making structural slots available for the licensing of certain types of arguments. Assuming adjacency effects in focus constructions do imply a Case feature, the clear asymmetry between topicalization and focusing in Somali can be expressed in terms of a difference in Case realization. In minimalist terms, topics are adjuncts that do not participate in the ‘core’ computational system, and bear independent, uncheckable Cases. Focus is a structural Case position that

FOCUS IN SOMALI

301

participates in checking relations. I have also suggested that the grammar of Somali specifies certain positions for nominal categories according to their referential capacities. Assuming D to be the locus of specificity, DPs in adjunct positions are directly connected to the discourse, given their intrinsic deictic properties. Nonspecifics, wh-phrases, as well as non-referring nominal expressions (bare Nouns) are legitimate in focus position. Clitics are generally understood as referential, but cannot be connected to the discourse, being non-maximal functional projections (NumPs): clitics then are Case-licensed IP-internally, in a feature-checking relation. There are several important issues about the correlation between Case and the ‘definiteness effect’. Belletti (1988) claims that the NPs that are characterized as semantically definite are assigned structural Case, whereas non-definite NPs are assigned inherent Case. A similar point is made in Enç (1989) for the accusative marker in Turkish, which is claimed to correlate with specificity (i.e. discourse linking). Following Kennelly (this volume), the parallel interpretation between nonspecifics, wh-phrases, and focused arguments in Turkish is reflected in their structural position, i.e. the immediately preverbal, so-called focus position, to the right of argument positions. Horvath (1985, 1995), É. Kiss (1995), Szabolcsi (1995) have independently shown that Hungarian distinguishes different positions in its surface syntax for different types of arguments and quantifiers. De Hoop (1989, 1990), argues that there is a correlation between the type of Case assignment and the strength/weakness of determiners, which is based upon intrinsically semantic properties: in Dutch, scrambled objects, which are no longer adjacent to the Verb, are assigned structural Case and necessarily have a strong reading. Objects that are assigned Case inherently have to remain adjacent to the Verb. As in German (Fanselow 1996), wh-phrases and focused NPs cannot be scrambled. This in turn recalls the well-known correlation between ‘nonconfigurationality’, the V-second property, ‘focus prominence’, and OV order (Koopman 1984; Abraham & de Meij 1986). Although a full typological study of this correlation is clearly outside the scope of this paper, it is clear that the Somali focus system is a Case-licensing device providing configurational positions in a ‘non-configurational’ language. At a more abstract level, our analysis also offers an important source of evidence for the root ‘assertive’ C node and its role in interpretive strategies of language use, and provides significant clues for further investigation of the syntax-pragmatics interface.

302

JACQUELINE LECARME

Notes * I am grateful to participants of the International Workshop on the Grammar of focus for discussion on the issues dealt with here, to Nomi Erteschik-Shir and Laurie Tuller for detailed written comments on an earlier draft, and to Bashiir Nuux Keenadiid for linguistic expertise. Errors are my own responsibility. 1. This article is exclusively concerned with the syntactic structures occurring in standard Somali and (more generally) the so-called maxaad tiri (lit. ‘what did you say’) dialects, which provide a natural typology for the phenomena discussed here. Other dialects of Somali may substantially differ in this respect (see, for example, Lamberti 1983; Tosco 1993). Striking similarities in the focus system are found in other Cushitic languages, such as Rendille (Oomen 1978), Boni (Sasse 1981). 2. I follow the national orthography in representing voiced and voiceless pharyngeal fricatives as c and x, and long vowels by gemination. Somali has a mixed prosodic system that Hyman (1981) characterized as a ‘tonal accent’ system. Since prosodic facts play a role in the subject under study, surface high tones that result from the computation of prosodic domains at the morphological or postsyntactic level are marked on examples in this paper. I will use the following abbreviations:  = focus marker,  = restrictive,  = extensive,  = invariable,  = independent,  = subjunctive;  = definite determiner,  = demonstrative,  = possessive;  = expletive;  = reflexive pronoun,  = impersonal pronoun. Other pronominal clitics are identified by their features (person, gender, number). 3. Baa and ayaa are strictly equivalent in their distribution and syntactic properties. Only baa allows phonological coalescence with a left-adjacent nominal ending in a vowel (Cáli + baa → Caláa. On the origin of these morphemes, see note 19. 4. There is a considerable variation in Cushitic languages. Somali is strictly ‘head initial’ as far as functional categories are concerned, and is a right-branching language. In the noun phrase, adjectives, relative clauses and genitive complements follow the head noun. Oromo, Rendille, Boni also exhibit ‘mixed headedness’, but have a stricter OV order than Somali, and are ‘null subject’ languages. Other Cushitic languages (e.g. Afar, Sidamo), and other Somali dialects (cf. Tosco 1993) are consistently head-final. 5. I have suggested in earlier work that there is a correlation between the ‘pronominal argument’ property and the fact that, in Somali, NPs cannot enter in full agreement relation with a Verb, because number in nominals has a derivational nature: Verbs only agree in number via the subject clitic system. See the discussion in Lecarme (1995). The correlation between the ‘pronominal argument’ character and the fact that NPs are formally unmarked for number seems to hold in other languages, as Baker (1996: 121–124) explicitly suggests. 6. For the full justification and motivations of this analysis, I refer the reader to Lecarme (1991, 1994, 1995). 7. The relationship between [+Tense] in finite clauses, [−Tense] in non-finite clauses is quite different from that of the English or French. The ‘impoverished’ nature of the so-called ‘ Restrictive paradigm’ or ‘negative conjugations’ is the result of reduced patterns of agreement that lack the Number feature, or of intervening heads in the process of V-movement (See Lecarme 1995, and Section 2.4.2 below).

FOCUS IN SOMALI

303

8. This slot includes the spatial deictic elements soo/sii by which the action described by the Verb is related to the speaker’s location. 9. Somali preverbs are highly reminiscent of the preverbal prepositional heads in Classical Greek and other Indo-European languages. They combine in the strict linear order ú (dative), kú (locative-instr.), ká (ablative), lá (comitative), and form a morphological X0 complex, the last (rightmost) prepositional head keeping its lexical accent (kulá, ugú…). More complex surface forms may result from the PF association of the adjacent object clitics (ií (i+ú), iigú (i+ú-kú), nóo (na+ú), noogú (na+ú-kú), etc.), with which they form a ‘phonological word’. Unlike affixes, clitics are not selective of the category of their phonological host, hence may prosodically attach to categories that do not ‘govern’ them. In the literal glosses, we represent this phonological contrast with hyphens for affixes and a plus sign for clitics. 10. Such elements include the ‘weak’ negation -an, coordinating particles -na ‘and’, -se ‘but’, etc. 11. As the following paradigms clearly show, strong pronouns are DPs (observe that the definite determiner retains the -k-/-t- gender distinction in the 3S), while ‘weak’ subject pronouns represent a clitic version of the nominal part of the DP projection: Strong weak (subject)  aní-ga aan  adí-ga aad  isá-ga uu(s)  iyá-da ay (excl) anná-ga aan(u) (incl) inná-ga (aynu)  idín-ka aad/aydin  iyá-ga ay 12. Following Heine & Reh (1984), ‘term focus’ involves the reanalysis of a copula in a cleft construction (NP + copula + relative clause), while ‘predicate focus’ is historically derived from a (copula + predicate) structure. This historical hypothesis is implicitely assumed in most accounts, and motivates Saeed’s (1984) analysis of focus constructions as synchronic reduced clefts and pseudo-clefts. See Lecarme (1991) for a critical evaluation of this proposal. 13. When there is a succession of two adverbial expressions, a second baa may occur. See Hetzon (1965) for more examples. 14. Nomi Erteschik-Shir (personal communication) points out that the focus position described here might fit her notion of topic. She analyzes contrast as focus within a topic set: contrastive topic and contrastive focus would then be the same (Erteschik-Shir, in press). 15. As pointed out by Laurie Tuller (personal communication), the facts regarding null third person pronouns having a definite interpretation, while indefinite meanings requires an overt argument are parallel in Chadic, where null third person pronouns are also found. 16. For example, waa is understood as a ‘classifier’ by Saeed (1984), who explicitly denies the pragmatic role of ‘verb focus’ structures in discourse. 17. Well-known exceptions are the short, accented forms of the Past Independent and Past Comparative (Andrzejewski 1956, pp. 126–129), the imperative, and the exclamative verbal

304

JACQUELINE LECARME forms. Common to all these ‘root’ assertive paradigms is the absence of both focus markers and subject clitics.

18. But see (1.3.2) above. 19. Lamberti (1983), following the traditional approach outlined above (see note 12), has proposed to reconstruct the morphemes baa/waa and ayaa as a (verbal) copula (*awa). There is, however, much comparative evidence (Arabic huwwa/iyya, Hebrew hu/hi, etc.) that these morphemes have a pronominal origin. 20. Argument-type CPs behave like NPs or DPs, given abstract nominal properties of their lexical complementizer (ín). The proposed analysis takes in account the limited range of interpretation of these constructions. Existential sentences and non-existential locative sentences, which are verbal in nature, must use verbal roots and full IP projections (cf. Lecarme 1991, 1994). The predicative vs. presentational interpretation of examples in (14a,b,c) depends on a NP vs. DP complement (D assumed to be the locus of specificity). Similarly, the function of ‘specification of necessity’ (Andrzejewski 1965) in (14d,e) is clearly derived here from the nature of the complement. 21. In adjunct positions (Andrzejewski’s (1964) ‘open configurations’), Case manifests itself both morphologically and prosodically, affecting the rightmost constituent of the NP. The clearest manifestation of the [±nom] opposition usually shows up as morphological Case endings (-u/-a) suffixed to the definite determiner. In focus position (Andrzejewski’s ‘closed configuration’), there is an invariant, ‘default’ Case (noted here [+acc]) morphologically identical to the [−nominative] Case in adjunct positions. 22. See, in particular, Vallduví (1990) and Brody (1995) for related discussion. 23. On the absence of WCO effects in focus constructions, see Kennelly (this volume) and Kidwai (this volume). 24. This includes nominals such as áad ‘muchness’, ág ‘proximity’, hóos ‘bottom’, dhéx ‘middle’, etc., which, often in connection with the prepositional preverb ú, form syntactically dissociable adverbial expressions meaning ‘a lot’, ‘near’, ‘under’ ‘among’, etc. 25. Since the language does not have noun complement structures or sentential subjects in argument position, examples like (22b, c) are the only ‘strong’ islands that can be tested. 26. Significantly, inherently quantificational expressions such as nothing, nobody do not exist in Somali. Qófna, wáxba are indefinites meaning ‘person, thing’ combined with the enclitic particle -na, -ba, a polarity element without inherent negative meaning, which has to be interpreted in the scope of negation. 27. I am setting aside here particular issues concerning relative (and possessive) constructions, which syntactically involve asymmetric coordination. When there is more than one conjunct (e.g., when the head Noun is modified), overt coordinating conjunctions must fill the C position. 28. It is assumed here that the Spec,CP position is filled by ‘base-generation’, an option compatible with the LCA (Kayne 1994: 165). Somali has further N-to-D raising in the noun phrase (see Lecarme 1996).

FOCUS IN SOMALI

305

29. I take non-root CPs to be [nominal], a feature of their head (the complementizer ín). The licencing of CPs is mediated by clitics in argument positions: the expletive (it-like) subject clitic -ay (3FS), a silent object clitic, and the genitive clitic -eed (Poss3FS). 30. For a full review of other proposals in the literature, and for arguments that postverbal focus constructions cannot be assimilated to relative clauses or pseudo-cleft sentences, see Lecarme (1991). Significantly, question phrases, non-referential nominals, ‘understood’ arguments cannot be extraposed. This fact is one of the main problems facing Saeed (1984)’s analysis of waxaaand baa-constructions respectively as pseudo-clefts and (derived) clefts. 31. See Hetzron’s (1975) comparative study of the ‘presentative function’ of these constructions, in relation to there- constructions and other cases of constructional focus. However, it must be noted that as in many languages, complement and relative clauses typically occur in postverbal focus position because of their ‘heaviness’, and that in most cases, object-postponed constructions superficially look like the unmarked VO constructions of Romance languages and lack substantive discourse function. 32. I slightly modify an earlier version of this analysis (Lecarme 1991) in which the expletive occupies [Spec,CP] and the associate is right-adjoined to IP. 33. Alternatively, we may suppose that the expletive can be null during the computational process, and added only at the phonological level (Halle et Marantz 1993). Assuming Kayne’s reanalysis of rightwards adjunction, what looks like a right-hand postverbal focus position would actually involve a stranded left-hand specifier. 34. The two other syntactic functions of the same morpheme wax are derived from different semantic, formal and phonological specifications listed in separate lexical entries. As a full nominal category (DP) wáx ‘thing’ has semantic and formal features, including Case and ffeatures, can combine with a (masculine) definite determiner (wáx-a, wíx-ii), can bear a plural suffix (waxyaaló, waxyaalá-ha), can be modified, and is able to function as the antecedent of a relative clause. As an implicit argument, wax projects a NP (not a DP) and has a low tone (but is not a clitic). 35. The terminology is due to Andrzejewski (1956). Note, however, that ‘paradigm’ is inappropriate and misleading, since the phenomenon involves a construction-specific pattern of agreement, which does not affect the morphological finiteness (tense and aspect) of the Verb.

References Abraham, Werner & Sjaak de Meij (eds.) 1986. Topic, Focus, and Configurationality. Amsterdam: John Benjamins. Andrzejewski, Bogumil W. 1956. “Accentual Patterns in verbal forms in the Isaaq dialect of Somali”. BSOAS 13(1).103–129. Andrzejewski, Bogumil W. 1964. The Declensions of Somali Nouns. London: School of Oriental and African Studies.

306

JACQUELINE LECARME

Andrzejewski, Bogumil W. 1975. “The Role of Indicator Particles in Somali”. Afroasiatic Linguistics 1. Antinucci, Francesco & Annarita Puglielli. 1980. “The Syntax of Indicator Particles in Somali: Relative Clauses Constructions”. Afroasiatic Linguistics 7(3).85–102 Baker, Mark. 1995. “On the absence of certain quantifiers in Mohawk”. In E. Bach et al. (eds.), Quantification in Natural Languages. Kluwer: Dordrecht. Baker, Mark. 1996. The Polysynthesis Parameter. Oxford: Oxford University Press. Belletti, Adriana. 1988. “The Case of Unaccusatives”. Linguistic Inquiry 19.1–34. Belletti, Adriana & Ur Shlonsky. 1995. “The order of verbal complements: a comparative study”. Natural Language and Linguistic Theory 13(3).489–526. Brody, Michael. 1996. Lexico-logical Form: a Radically Minimalist Theory. Cambridge, Mass.: MIT Press. Cardinaletti, Anna & Michal Starke. 1995. “Deficient Pronouns: a View from Germanic”. Geneva Generative Papers 3(1). Chomsky, Noam. 1971. “Deep Structure, Surface Structure, and Semantic Interpretation”. In D. Steinberg & L. Jakobovits (eds.), Semantics. Cambridge: Cambridge University Press. Chomsky, Noam. 1976. “Conditions on rules of grammar”. Linguistic Analysis 2.303–351. Chomsky, N. 1993. “A minimalist program for linguistic theory”. In M. Halle & S.J. Keyser (eds.), 1–52. Chomsky, Noam. 1996. The Minimalist Program. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1990. Types of A′ Dependencies. Cambridge, Mass.: MIT Press. Enç, M. 1989. “The Semantics of Specificity”. Linguistic Inquiry 22(1).1–25. Erteschik-Shir, Nomi. 1997. The Dynamics of Focus Structure Cambridge: Cambridge University Press. Fanselow, Gisbert. 1996. “Features and Free Constituent Order in a Minimalist framework”. GLOW Colloquium, Athens. Gueron, Jacqueline. 1989. “Subject, Tense, and Indefinite NPs”. In Juli Carter and Rosemarie Dechaine, eds, Proceedings of NELS 19, University of Mass., Amherst: GLSA. Hale, Kenneth. 1983. “Warlpiri and the Grammar of Nonconfigurational Languages”. Natural Language & Linguistic Theory 1.5–49.

FOCUS IN SOMALI

307

Halle, M. & S.J. Keyser (eds.) 1993. The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger. Cambridge, Mass.: MIT Press, Halle, M. & A. Marantz. 1993. “Distributed morphology and the pieces of inflection”. In M. Halle & S.J. Keyser (eds.), 111–76. Holmberg, Anders & Christer Platzack. 1988. “The Role of Inflection in Scandinavian Syntax”. Working Papers in Scandinavian Syntax 42.25–43. Hoop, Helen de. 1989. “Case Assignment and Generalized Quantifiers”. Proceedings of NELS 19, University of Mass., Amherst: GLSA, 176–90. Hoop, Helen de. 1995. “On the characterization of the weak/strong distinction”. In Emmon Bach et al. (eds.), Quantification in Natural Languages. Dordrecht: Kluwer. Heine, Bernt and Mechthild Reh. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Hetzron, Robert. 1965. “The Particle baa in Northern Somali”. Journal of African Languages 4(2). 118–130. Hetzron, Robert. 1971. “Presentative Function and Presentative movement”. Studies in African Linguistics, Suppl. 2.79–105. Horvath, Julia. 1985. Focus in the Theory of Grammar and the Syntax of Hungarian. Foris, Dordrecht. Horvath, Julia. 1995. “Structural Focus, structural Case and the Notion of Feature-assignment”. In Katalin É. Kiss (ed.). Hyman, Larry. 1981. “Tonal Accent in Somali”. Studies in African Linguistics 12(2). Jelinek, Heloïse. 1984. “Empty categories, case, and configurationality”. Natural Languages and Linguistic Theory 2.39–76. Jelinek, Heloïse. 1995. “Quantification in Straits Salish” In Emmon Bach et al. (eds.), Quantification in Natural Languages. Kluwer: Dordrecht. Kenesei, István. 1996. “On the Syntax of Focus. Manuscript, University of Szeged (Hungary). Kennelly, Brenda. 1995. “Focus Position in Turkish”. Manuscript, Rutgers University. Talk presented at Langues et Grammaire 2, Université de Paris 8, Juin 1995. Kiss, Katalin É. (ed.). 1995. Discourse Configurational Languages. Oxford: Oxford University Press. Kiss, Katalin É. 1995. “Multiple Topic, one Focus?”. Manuscript, Linguistic Institute of the Academy of Science. (Talk presented at the 18th GLOW Colloquium.)

308

JACQUELINE LECARME

Koopman, Hilda. 1984. The Syntax of Verbs. Dordrecht: Foris. Lamberti, Marcello. 1983. “The Origin of the Focus Particles in Somali”, in Rainer Vossen and Ulrike Claudi, eds, Sprache, Geschichte und Kultur in Afrika, Vorträge, gehalten auf dem III Afrikanistentag, Köln, 14–15 Oktober 1982. Hamburg: Helmut Buske, 57–112. Lecarme, Jacqueline. 1991. “Focus en somali: syntaxe et interpretation”. Linguistique Africaine, 7.33–64. Lecarme, Jacqueline. 1994. “Focus et effets ‘verbe second’ en somali” In L. Picabia (ed.), Syntaxe des langues africaines, Recherches Linguistiques de Vincennes 24. Lecarme, Jacqueline. 1995. “L’accord restrictif en somali”, Langues orientales Anciennes Philologie et Linguistique 5–6. (Paper presented at the Journées d’études chamito-sémitiques, Paris 1992). Lecarme, Jacqueline. 1996. “Tense in the Nominal System: the Somali DP” In J. Lecarme, J. Lowenstamm & U. Shlonsky (eds.), Studies in Afroasiatic Grammar. The Hague: Holland Academic Graphics. Livnat, Michal. 1984. Focus Constructions in Somali. Ph.D dissertation, University of Illinois at Urbana Champaign. Mous, Maarten. 1993. A Grammar of Iraqw. Cushitic Languages Studies 9, Hamburg: Helmut Buske. Oomen, A. 1978. “Focus in the Rendille Clause”. Studies in Afroasiatic Linguistics 9(1). Pillinger, Owen. 1989. Accent, Tone, and Prosodic Structure in Rendille (with particular reference to the nominal system). Ph.D.dissertation, SOAS, University of London. Platzack, Christer. 1984. “The Position of the Finite Verb in Icelandic” In Wim de Geest & Yvan Putsey (eds.), Sentential Complementation. Dordrecht: Foris. Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery”. In Liliane Haegman (ed.), Elements of Grammar: Handbook of Generative Grammar. Dordrecht: Kluwer. Rooth, M. 1992. “A Theory of Focus Interpretation”. Natural Language Semantics 1(1). Saeed, John. 1984. The Syntax of Topic and Focus in Somali. Hamburg: Helmut Buske. Sasse, Hans Jurgen. 1981. “Basic Word Order and Functional Perspective in Boni”. Folia Linguistica 15.253–290.

FOCUS IN SOMALI

309

Selkirk, Elizabeth. 1986. “On Derived Domains in Sentence Phonology”. Phonology Yearbook 3.371–405. Svolacchia, Marco et al. 1995. “Aspects of Discourse Configurationality in Somali”. In Katalin É. Kiss (ed.) Speas, Margaret. 1990. Phrase Structure in Natural Language. Dordrecht: Kluwer. Szabolcsi, Anna. 1995. “Strategies for Scope Taking”. Working Papers in the Theory of Grammar 2(1). Research Institute for Linguistics, Hungarian Academy of Sciences, Budapest. Tosco, Mauro. 1993. “Copula, Focus, Cleft, etc. in Southern Somali”. Afrikanistische Arbeitspapiere 36. Tuller, Laurice. 1992. “The Syntax of Postverbal Constructions in Chadic”. Natural Language and Linguistic Theory 10.303–334. Vallduví, Enric. 1990. The Information Component. Ph.D. Dissertation., University of Pennsylvania. Zholkovsky, A. K. 1971. Syntaxis Somali: (Glubinïe i Poverxnostnïe Strukturï). Moscow: Izdatel’stvo ‘Nauka’.

Focus in Basque* Jon Ortiz de Urbina Universidad de Deusto

Abstract This article presents the basic facts of focalization in Basque and reviews some of the approaches pursued in recent years. Like wh-words, foci occur immediately to the left of the inflected verb, apparently in clause initial position. Also like wh-words, they can be extracted to higher clauses, trigger pied-piping and seem to target the same position. The V2-like effects prompted by foci (and wh-words) are difficult to analyze in a symmetric syntactic analysis because Basque is largely head-final. If the focus-verb adjacency is to be analyzed along familiar V2 lines, we need a left-headed phrase to which foci and inflected verb may move. The article reviews in detail two proposals for such phrase: CP and Laka’s Sigma Phrase, pointing out the problems they raise.

1.

Word Order and Focalization

Basque is a designated position focus language where focalized phrases must occur immediately to the left of the verb, although I will have to qualify this statement as we proceed. In fact, this adjacency between foci and verbal elements is one of the clearest observational features of word order in Basque, and some traditional grammarians have claimed it is the only word order restriction to be found in this language. This is so because, leaving operators of this type aside, word order is extremely flexible in Basque. However, abstracting away from numerous descriptive and analytical problems, I will follow standard assumptions and claim that the neutral basic word order is of the SOV type, with right-headed phrases.1 Thus, the sentence in (a) can be interpreted as a neutral

312

JON ORTIZ DE URBINA

statement where everything is new information, while that in (b) or (c) cannot. Rather, (1b,c) must be interpreted as containing a focalized subject: (1)

a.

b. c.

Jonek eskutitza irakurri du (SOV) Jon letter read has ‘Jon has read the letter’ Jonek irakurri du eskutitza (SVO) Jon read has letter Eskutitza, Jonek irakurri du (OSV) letter Jon read has

The position occupied by the focalized element plus verb unit is clause initial. Where, as in (c), any element precedes the focus, it is intonationally separated from the latter by a pause and interpreted as a topic, with properties rather different from foci, as we will see (for a discussion of similar effects in HindiUrdu, Malayalam, Western Bade and Tangale, see Kidwai (this volume), where a minimalist PF-adjunction approach is offered). In this presentation I will review first the basic data on focalization in Basque2 and then some of the typical approaches that have been pursued in the last years. The discussion will concentrate on the exact structural position of both the focus and the verb. After describing the basic facts in Section 2, I will show that a) there is a single functional projection in root and embedded clauses to which both foci and whwords move and b) that this position is the specifier position of the highest functional layer of the clause (Section 3). Section 4 will review some of the problems in identifying this functional projection.

2.

The Basic Facts

Foci in Basque share the same distributional properties as wh-words: both occur in a clause initial position, optionally preceded by topics, and immediately followed by the verb and inflection. Thus, parallel to the examples in (1) we find the interrogative structures in (2): (2)

a.

b.

Nork irakurri du eskutitza? who read  letter ‘Who has read the letter?’ Eskutitza, nork irakurri du?

FOCUS IN BASQUE

313

In both questions and focalized clauses, this movement to the left with verb second-like effects is obligatory, so that a sentence like (3) where the focalized constituent remains in situ is ungrammatical: (3)

*JONEK eskutitza irakurri du Jon letter read  ‘It is John that read the letter’

The operator/focus adjacency is then, at least descriptively, quite similar to the residual V2 phenomena in better known languages, and I will be using this label in what follows. Foci, like wh-words, can also undergo cyclic movement with bridge verbs, as in the following example: (4)

JONEK uste dut [t esan du-ela Mikelek [t idatzi du-ela Jon think  say -that Mikel write -that eskutitza letter ‘It is Jon that I think Mikel has said has written the letter’

This movement seems to be cyclic, perhaps through intervening C complexes, since on top of the left adjacency with the matrix verb uste ‘think’, we also find a preferred verb initial pattern in both the most deeply embedded source clause and the intervening one. This may indicate that in each clause we find focus movement to the left periphery and left adjacency with the verb, producing an apparent verb initial pattern once the operator has moved on. Thus, the traces in the previous example stand for this probable movement of the focalized element through a functional projection we can provisionally label CP at this stage. The same pattern has been described for wh-words, which also produce apparent V1 effects in intervening clauses when extracted: (5)

NORK uste duzu [t esan du-ela Mikelek [t idatzi du-ela Jon think  say -that Mikel write -that eskutitza letter ‘Who do you think Mikel has said has written the letter?’ (Lit. ‘Who do you think has Mikel said has written the letter?’)

One difference in this area between wh-words and foci is that wh-extraction is of course obligatory in this context: the bridge verbs in (5) are all [−wh] and do

314

JON ORTIZ DE URBINA

not subcategorize for an indirect question, so that the question operator is acceptable only if it takes scope over the whole sentence as a direct question. On the other hand, there is no subcategorization for focalization, and the focalized element may remain in either the embedded clause or move to the matrix. This is shown in (6) and (7): (6)

a.

b.

(7)

Nik uste dut [MIKELEK idatzi du-ela eskutitza] I think  Mikel write -that letter ‘I think that it is Mikel that has written the letter’ MIKELEK uste dut [t idatzi du-ela eskutitza] Mikel think  write -that letter ‘It is Mikel that I think has written the letter’

a. *Nik uste dut [NORK idatzi du-ela eskutitza] I think  who write -that letter *‘I think who has written the letter’ b. NORK uste duzu [t idatzi du-ela eskutitza] Mikel think  write -that letter ‘Who do you think has written the letter?’

(6) shows that, at least in overt syntax, the focalized element may occupy an operator position in its own clause or at a higher clause. Tsimpli (1995) shows that in Greek foci must have matrix scope at LF, regardless of their overt location. This accounts for (8), where a wh-word in the matrix clause cannot cooccur with a focalized element in the embedded clause: (8)

*Pjos ipe oti tin MARIA sinantise? who- said that the- Maria met ‘Who said that met MARIA?’

If there is one single functional projection to host both wh-words and foci at the root level, and the embedded focalized element must take matrix scope, the two operators in (8) would be competing for the same position, accounting for the deviant status of the clause. In Basque, however, the sentence corresponding to (8) is perfectly grammatical: (9)

Nork esan du [MIREN topatu du-ela antzoki-an]? who say  [Miren meet -that theater-at ‘Who said that it was Mary (s)he met at the theater?’

This might indicate, within Tsimpli’s assumptions, that either a) focus in Basque

FOCUS IN BASQUE

315

need not take matrix scope or b) that there are two different landing sites for these operators at the root level. The first option seems to be preferable, since, as (10) shows, wh-word and focus are incompatible in the same clause: (10) *Nork ikusi du MIREN antzoki-an? Who see  Mary theater-at ‘Who saw MARY at the theater’ Notice by the way that (10) also shows that absorption is not possible between the two operator types, just as it is impossible between wh-words and yes/no operators. Otherwise, one would expect (10) to be as acceptable as (11): (11)

Nork ikusi du nor antzoki-an? who see  who theater-at ‘Who saw whom at the theater?’

Tsimpli also shows that two different positions are available in embedded clauses in Greek (12a), so that focus and wh-word are compatible in such contexts. In Basque (12b), though, this is as bad as (10): (12)

a.

b.

Mu-ipan o YANIS ti agorase Me-told the- Yanis what bought ‘They told me what YANIS bought’ Galdetu didate (*JONEK) zer (*JONEK) erosi du-en3 ask  what buy - ‘They have asked me what JOHN bought’

The focalized element may not be moved to the left periphery of these clauses, either preceding or following the wh-word, suggesting again there is one single position available. The same assumptions account for the extraction facts in (13) and (14). These show that the embedded clause may include neither a focalized constituent nor a wh-word, since a wh-word has originated there, occupying the only operator position available before moving to the matrix clause:4 (13)

??Zer

uste du Mikelek [ETXEAN aurkitu du-ela Jonek]? What think  Mikel [home-at find -that Jon ‘What does Mikel think that JON has found at home?’

316

JON ORTIZ DE URBINA

(14)??/*Zer galdetu du Mikelek [NON aurkitu du-en Jonek]? What ask  Mikel [where find -that Jon ‘What has Mikel asked where Jon has found’ Similarly, in (15) an adjunct may not be overtly extracted from an indirect clause: (15)

??ETXEAN

galdetu du Mikelek [Jonek liburua aurkitu du-en] home-at ask  Mikel [Jon book find - ‘It is at home that Mikel has asked whether Jon has found the book’

The only possible interpretation is for the operator to originate in the matrix clause. The violation is less severe here than in other cases, and this might be related to the fact that the empty yes/no operator in the embedded interrogative clause in (15) does not even trigger verb movement, as opposed to wh-operators. The previous facts seem to indicate that both foci and wh-words actually move to the same position in both matrix and embedded clauses in Basque, and that a single position is available for them in both contexts.

3.

Pied-piping and Focalization

Let us assume, as in many analyses of focalization (Brody 1995; Rizzi 1997), that the relevant morphological feature that triggers movements is a strong [+F] feature which can be checked by some functional head. The parallelism between focalization and question formation leads us to expect that, in the same way as interrogative operator features can percolate up in pied-piping configurations, something similar may happen with focal operators, and in fact this is so. This property is interesting in that such percolation typically involves some feature transfer process from the highest specifier position of a projection or from a position inside that specifier. Its availability with foci (and wh-words) might be taken to indicate that these elements occupy a position inside the highest functional specifier position. To begin with, let us examine simple structures where the focalized constituent occurs inside a larger phrase. Just as wh-words can pied-pipe constituents within which they appear, foci also can (indeed must) drag constituents that include them:

FOCUS IN BASQUE

(16)

317

[JONEN lagunek] idatzi zuten eskutitza [Jon’s friends write  letter ‘JON’s friends wrote the letter’

(17) *JONEN lagunek eskutitza idatzi zuten Jon’s friends letter write  ‘JON’s friends wrote the letter’ In (16) the focalized constituent is the genitive modifier of the head noun, and the whole DP must occur in the initial position, left adjacent to the verbal element, as shown by the ungrammaticality of (17). The same data are found in the following two examples, where the focalized element originates deep inside the subject DP. Again, the whole constituent must occur initially and pre-verbally: (18)

[JONEN etxe-ko teilatuak] izan ditu itoginak [Jon’s house-of roof have  leaks ‘The roof of JON’s house (has) leaks’

(19) *[JONEN etxeko teilatuak] itoginak izan ditu leaks have  Notice that an adjunct is being focalized here. There is evidence, though, that the focus occupies a derived specifier position in Basque, in keeping with the previous examples. Since extraction from DPs is not possible in Basque, the sentences above are in fact the obligatory patterns. The parallelism with question formation extends to less usual pied-piping patterns, like the ones in the following examples: (20)

[JONEK idatzi du-ela liburua] esan du Peiok [Jon write -that book say  Peio ‘Peio said that JON wrote the book’ (‘That JON wrote the book has Peio said’)

(21) *[JONEK idatzi duela liburua] Peiok esan du Peio say  (22)

[NORK idatzi du-ela liburua] esan du Peiok? [who write -that book say  Peio ‘Who wrote the book has Peio said?’

In (20), the embedded clause contains the subject JONEK as a focalized element. As such this occupies the position immediately to the left of the embedded verb. However, there is yet another verb-second phenomenon: the whole clause itself

318

JON ORTIZ DE URBINA

must occupy the position immediately to the left of the matrix verb esan ‘say’. When, as in (21), some element intervenes, the result is ungrammatical. (22) shows that exactly the same pattern can be found with wh-words. Both operator types, foci and wh-words, have scope over the entire structure. This is particularly clear in the case of wh-operators. Observe that (22) is a direct question, even though the wh-word still occupies a position inside the embedded clause. In fact, it could not be an embedded question, since the matrix verb here does not select interrogative complements, and such clausal pied-piping is not possible with interrogative complements. Thus, the wh-word must either move alone to the matrix pre-verbal position or pied-pipe the whole embedded clause to that same position. Another example of clausal pied-piping can be observed in the following examples, where the focalized constituent occurs within a relative clause. Again, the focus is preverbal with respect to the verb of the relative clause, and now the whole DP containing the noun and its modifying clause must occur immediately to the left of the matrix verb: (23)

[[JONEK idatzi du-en] liburuak] izan ditu salmenta onak [[Jon write - book have  sale good ‘The book that JON has written sold well’

(24) *[[JONEK idatzi duen] liburuak] salmenta onak izan ditu sale good have  (25)

[[NORK idatzi du-en] liburuak] izan ditu salmenta onak? [[who write - book have  sale gook ‘The book that who wrote had good sales?’

Notice that the possibility of focalizing or questioning inside these islands is possible because there is no extraction in the overt syntax. Thus, (26) shows the same pattern with focalization inside an adjunct: (26)

a.

[MINTEGIA egin ondoren] joan ziren afaltze-ra [workshop do after go  dinner-to ‘They went for dinner after having the WORKSHOP’ (Lit. ‘After having the WORKSHOP did they go for dinner’) b. *[MINTEGIA egin ondoren] afaltze-ra joan ziren dinner-to go 

One possible analysis of these structures might go as follows. It is well known that operators in the specifier position can have scope outside of that position over the phrase they specify. Some well-known instances of this phenomenon

FOCUS IN BASQUE

319

can be observed in simpler pied-piping structures in English (27), in NPI licensing (28) and in bound pronoun variable interpretations ((29); Reinhart 1983; May 1985): (27)

a. b.

Whose book did you read? Which author’s book did you read? *The book by which author did you read?

(28)

a. b.

Few students knew any answer Few students’ parents met any professor *The parents of few students met any professor

(29)

a. b. c.

The mother of everyone saw him Everyone’s mother saw him Every student’s mother saw him (?)

Whatever the explanation, the fact seems to be that operator features in specifier positions can percolate up to the maximal projection they specify, and have therefore scope over the c-command domain of that phrase. If foci and wh-words occupy a specifier position of the highest relevant functional category in their clause, we would have a parallel situation. Assuming for the time being that the position to which these operators move is CP, foci and wh-words would also have scope over anything c-commanded by CP itself. Thus, in (20) above, the focus JONEK would move to Spec of the embedded clause CP, and the whole CP2 itself would be pied-piped to the Specifier of the matrix CP1, as in (30):

CP1

(30)

CP2 JONEK

C¢ C¢

This position is in effect parallel to that of the negative element that licenses the NPI in the English example (28) above. Here the scope of focus is the whole root CP (the upper one). A confirmation that the embedded clause as a whole occupies the Spec,CP position can be found in the fact that the clause containing the focus can also undergo cyclic movement as in (31): (31)

[JONEK idatzi du-ela liburua] uste dut nik [t esan du-ela [Jon write -that book think  I say -that

320

JON ORTIZ DE URBINA

Peruk]] Peru ‘I think Peru said JON wrote the book’ (Lit. ‘That JON wrote the book do I think did Peru say’) Here the whole clausal complement of say appears now immediately to the left not of say itself, but of the root think. Notice that we also find here the inversion effects which we claimed above to be the result of cyclic Spec to Spec movement. These facts provide some evidence that indeed the clause containing the focus ends up in the Spec of the relevant functional category, CP.5 In any event, and pending more refined analyses of this complex phenomenon, it seems desirable to assume that the landing site of both operators is similar, so that the mechanisms that account for clausal pied-piping in one can be extended to the other at no cost. In Ortiz de Urbina (1993), I assumed that these structures actually involve percolation of an operator feature from Spec up to CP, in such a way that the feature actually moved, and Spec lost it: (32)

CP

a. [wh]

b.

CP [wh]

C¢ C

C¢ C

In this way, I accounted for the fact that the wh-word in (22) occurs inside a declarative complement of a verb that does not subcategorize for questions. The apparent mismatch between the selected [−wh] complementizer head of the complement clause and the wh-word in its Specifier is resolved after percolation, where the [wh] feature physically disappears from Spec and moves up to CP itself. This also explains why clausal pied-piping patterns are not possible with subcategorized questions, as in the relevant interpretation of (33): (33) *[Nor etorri d-en] galdetu du Jonek? [who arrive - ask  Jon ‘Who has arrived has Jon asked?’ If the complement C is [+wh], the operator criterion requires that it be matched by a [+wh] element in its Specifier position. This will be possible if the wh-word

321

FOCUS IN BASQUE

retains its operator feature. If the latter percolates up to CP as in the pied-piping structure in (33), a mismatch occurs with the resulting ungrammaticality: (34)

CP

a.

b. C¢

[wh] C [wh]

CP [wh] *

C¢ C [wh]

Since focalization is not involved in selection, clausal pied-piping should be possible with foci inside embedded questions. This is so, and an interesting case is actually that of emphatic wh-words, which we may assume contain both a [wh] feature and a [F] feature. The former may not be percolated, as indicated, but the latter may, as shown in (35): (35)

[NOR etorri d-en] galdetu du Jonek who arrive - ask  Jon ‘Jon asked WHO has arrived’ (Lit. ‘WHO has arrived has Jon asked’)

Again, nothing may intervene here between focus and verb. Notice that the operator feature that is being percolated here is the focal one, and the crucial difference with (33) is that (35) is not a direct question. The operator feature that has matrix scope is focalization, not an interrogative one.

4.

The Analytic Problem: Head Hunting

As shown throughout the preceding description, wh-words and foci display many distributional similarities that must be captured somehow. Traditional grammars of Basque, since Altube (1929), also state that the two elements behave in the same way, and occupy the same position, which we have identified as left peripheral, that is, clause initial. Descriptively, the situation described above is reminiscent of residual V2 phenomena as discussed by Rizzi (1996), in the sense that we have a certain type of verb movement and adjacency only when an operator is present, and I have been pointing out the similarities by referring to the functional nodes to the left as related to CP.

322

JON ORTIZ DE URBINA

However, the main problem we face when dealing with Basque lies in the fact that the language is rather thoroughly right headed. That is, V2 effects are achieved straightforwardly, in terms of structure like (36a), provided the operator and the verbal head move to the same phrase, usually identified as CP or IP, but not with (36b): (36)

XP

a. YP

X¢ X

XP

b. YP ...

X¢ ...

X

YP-X adjacency is not immediately evident in a right headed language like Basque, where absolutely all lexical material would intervene in (b) between YP and its corresponding head position X. More precisely, while focus movement to the left is easy, since targets abound in that area, the main problem is to derive the apparent verb movement to the left, because in principle one does not expect to find heads in that area. I will then concentrate on this standard research line, one which tries to identify some left-headed functional category which may serve as landing site for both operators and heads in the appropriate left periphery. Other alternatives will only be briefly mentioned later on in the text. 4.1 Left-headed CP Before the proliferation of functional categories, options for left heads were severely limited, and in fact the only likely candidate I found when first looking into this problem was CP itself, linking the Basque facts with similar ones in neighboring languages like Spanish (see Ortiz de Urbina (1995) for a review). This line is possible because of the fact that Basque complementizers are not free morphemes, but bound clitics always attached to inflection. When inflection is final, they will also be final, as in (37a), but when it is not, as in many of the preceding examples and (37b), they will not either: (37)

a.

b.

Jonek uste du [Mikelek eskutitza idatzi du-ela] Jon think  Mikel letter write -that ‘Jon thinks that Mikel wrote the letter’ Jonek uste du [MIKELEK idatzi du-ela eskutitza] ‘Jon thinks that MIKEL wrote the letter’

FOCUS IN BASQUE

323

In examples like (37b), the complementizer will occur attached to the verb in second position, followed by any remaining clausal material. I will return to the basic assumption later, but let us see first how this hypothesis can account for the data. In fact, the analysis sounds rather familiar, from work by Rizzi (1996), Brody (1990), (1995), Tuller (1992), Horvath (1995), etc. The operator feature of wh-words and foci in Spec,CP must be matched by a corresponding feature in the C head. This independent feature must be supplied by some head provided with that feature and moving to C. Rizzi locates the [wh] feature in Infl, and we can follow Horvath (1995) and Tuller (1992), and assume that the syntactic feature [Focus] can be hosted by some functional head, among them Infl itself. Thus, if Spec of CP is occupied by an element bearing an operator feature, C will have to possess that feature to agree with its specifier, and movement of the functional head hosting the feature will supply it. These hypotheses are not enough by themselves, since the head moving to C in Basque V2 sentences like (38) is not Infl alone, but Infl and V, whether amalgamated in synthetic forms or as the complex head of periphrastic verbs: (38)

a.

b.

Zer irakurri du Jonek? what read  Jon ‘What did Jon read?’ LIBURUA irakurri du Jonek book read  Jon ‘Jon has read THE BOOK’

In effect, something similar happens in the Romance languages, as in the Spanish (39), with a periphrastic verb: (39)

a.

¿A quién ha visto María? ¿who has seen Mary ‘Who has Mary seen?’ b. *¿A quién ha María visto?

Unlike similar examples in Italian, in Basque, as well as in Spanish, the ‘participle + auxiliary’ unit may not be broken, and this could be captured by an extended head-to-head movement of V to C through Infl. We must therefore find a reason for V to move to Infl first; once there, further movement to C will be prompted by the Operator Criterion. For this purpose, I will adopt here the analysis of Focus phenomena developed in Tuller (1992) and Horvath (1995). The functional head hosting the syntactic features [wh] and [Focus] must be

324

JON ORTIZ DE URBINA

 in some languages, that is, a lexical head must move to it. This is achieved in Basque by head-to-head movement of V to Infl. Once the head Infl is thus lexicalized, the new complex V/I unit must move to C, where specifierhead agreement with the feature-bearing operator will satisfy the Operator Criterion. All of this is represented in (40):

CP

(40)

Q ! Wh-criterion

C¢ C V/I

IP t VP

lexicalization t

We would then derive the desired adjacency and leftward movement of foci, whwords and verbs. Anything to the left of Spec,CP would then be a topicalized constituent, intonationally separate from the main clausal structure. Lexicalization of INFL, which accounts for the need to move the participle along with the auxiliary, cannot be as such considered to be a parameter, since this runs into some facts from French Basque dialects. These dialects admit Auxto-C raising in questions and focalizations, leaving behind the lexical verb and producing V2 contexts where the second element is the auxiliary head. The two patterns are exemplified in: (41)

a.

b.

(42)

a. b.

Zer irakurri du Jonek? (Common) what read has John ‘What has Jon read?’ LIBURUA irakurri du Jonek. book read has John ‘Jon has read the BOOK’ Zer du Jonek irakurri? (Northern) what has Jon read LIBURUA du Jonek irakurri. book has Jon read

FOCUS IN BASQUE

325

Notice that the dialectal distribution is not V+INFL movement in some dialects versus INFL movement in the others.6 Rather, Southern dialects must move V+INFL in these cases, whereas northern dialects have both options. This cannot be a parameter, in the sense that the options cannot be simultaneously positive and negative. I will leave this issue open here, but I will return to what I consider to prompt the V to I movement later on. From a descriptive perspective it seems then that northern dialects show both a Romance and an English type of residual V2 pattern. This analysis rests on the assumption of a left headed CP. There are several clause initial non-clitic subordinating elements, as in the following sentences: (43)

Galdetu du [ea Mikel heldu d-en] ask  Mikel arrive - ‘He has asked whether Mikel has arrived’

(44)

Entzunik [ezen hil z-ela diruak behar zituen zapatagina… hearing that die - money need  shoemaker ‘upon hearing that the shoemaker who needed the money had died…’

In fact, causal ezen does behave like a clitic in some French Basque dialects, often attaching to the right of the first clausal constituent: (45)

xerriak ezen lakhet du zikhinpean pig since like  in the dirt ‘since pigs like to be in the dirt’

However, at least ezen can be analyzed (diachronically) as phrasal, and could occupy the Spec position, rather than initial C, so these elements do not provide strong evidence in favor of independent left complementizers. The only fact that can be used in this line of research, other than the evidence in question and other cases of leftward head movement we will turn to, is that the clitic nature of real complementizing particles does not offer any strong evidence to the contrary. This is rather scant for language acquisition, as Uriagereka (1992) points out. 4.2 Laka’s Σ-phrase With the proliferation of functional categories, more possibilities emerged. The hard-core inflectional categories themselves, AGR and TNS cannot help, since they are plainly right headed. However, there is another candidate for a left head in Basque, namely, Laka’s (1990) S-phrase. The main motivation for this left

326

JON ORTIZ DE URBINA

functional head comes from another conspicuous movement to the left, that found in negative clauses. As the example in (46) shows the negative morpheme in root clauses appears to the left of the clause, fused with the tense-bearing element. The subject usually precedes negation, as in (b), but this is not obligatory in any way: (46)

a.

b.

Jonek liburua irakurri du Jon book read  ‘Jon has read the book’ Jonek ez du liburua irakurri J   book read ‘Jon has not read the book’

In order to analyze negation facts, Laka assumes a structure like that in (47):

CP

(47)

C¢ C

SP S

IP

In this account, Infl would move up to the S head, in order to c-command it, separating the auxiliary from the participle, which would remain in situ. We thus have another possible target for heads and phrases appropriately located in the left periphery of the clause. Although Laka does not deal with focalization per se or with question formation, she explicitly links this structure to one type of focalization, emphatic affirmation, so that the S head would be a polarity head for positive and negative poles. The positive counterpart of negative ez would be the prefix ba, diachronically related to the affirmative bai ‘yes’, exemplified in (48): (48)

a.

b.

Ba-daki Jonek egia Ba-knows Jon truth ‘Jon does know the truth’ Ez daki Jonek egia N know Jon truth ‘Jon does not know the truth’

FOCUS IN BASQUE

327

These patterns are described in traditional grammars as verb focalization patterns, where this means not contrastive verb focalization, but positive emphasis similar to the one indicated in the English glosses. Observe the parallelism with negative formation, in that the infected form, here an amalgamated synthetic form also including the root for the main verb, has been displaced to the left periphery of the clause. There are some problems with this extension of the negative analysis to incorporate positive focalization. First, there is a certain asymmetry which may or may not be important. This is the fact that while the negative value of S is neutral with respect to emphasis, the positive value must be emphatic. So two parameters seem to be involved in this value of S: polarity plus emphasis or focalization. It may be an empirical question whether the positive pole of this head is so combined with focalization while the negative pole is not, but a priori there is a surprising asymmetry here. In fact, I think that there are empirical problems with the idea of a positive counterpart to the negative value of this head. The similarity in (48) is in part deceptive. Notice that the particle ba is only found in focalization of synthetic verbs, those which like the ones in (48) combine the lexical root with all of the inflectional morphemes usually carried on by the auxiliary. Focalization of a periphrastic verbal form like the one in (49a) is as in (b) rather than as in (c): (49)

a.

b. c.

Jonek liburua irakurri du Jon book read  ‘Jon has read the book’ Irakurri du Jonek liburua read  Jon book Ba-du Jonek liburua irakurri

The pattern in (c) does exist, but it is dialectal, restricted in distribution, and far more marked than the general one in (b). This just involves movement of the participle plus auxiliary unit to the left periphery again, as opposed to movement of the auxiliary by itself. Finally, there is a more important empirical problem. This is that ba can be shown not to be by itself an emphatic marker. In fact, it shows up in a variety of contexts which do not necessarily have anything to do with emphasis. Thus, it not only shows up in emphatic affirmation, but also in yes/no questions like (50) and in existential sentences like (51) (from Oyharçabal 1984):

328

JON ORTIZ DE URBINA

(50)

a.

b.

(51)

a.

b.

Ba-daramazu dirurik? ba-carry money ‘Do you carry any money with you?’ Ba-daki Jonek egia? ba-knows Jon truth ‘Does Jon know the truth?’ Ba-da ogia (cf. Ogia da ‘It is bread’) ba-is bread ‘There is bread’ Lagunak ba-ditut Friends ba-have (cf. Lagunak ditut ‘They are friends of mine’/‘Ce sont des amis’) ‘I have friends’ (‘J’ai des amis’)

In fact, in northern dialects, this particle is also found, according to Rebuschi (1983), with synthetic verbal forms in apparently unmarked contexts. Although the previous contexts do not have much to do with emphasis, they do involve, at least arguably so, verb movement to the left of the clause, and, just like the emphatic affirmation sentences discussed by Laka, they are restricted to synthetic verbal forms. This indicates that the ba in emphatic affirmation, which also involves verb movement of synthetic forms, is to be analyzed along with the previous cases, rather than with negation. What seems to be going on here is actually a problem located at the syntax/phonology interface, related to the fact that tensed forms, whether auxiliaries or synthetic verbs, are clitics in Basque. As a result, in root contexts they always cliticize to the element to the left, forming a prosodic word. So even though word order is, as indicated above, very flexible, unlike periphrastic forms, synthetic ones may not occur clause initially or form the only overt element of the utterance: (52)

a. *Daki Jonek egia. *Daki knows Jon truth. (He) knows (it) b. Jakin du Jonek egia. Jakin du know  Jon truth. know  ‘Jon has found out the truth. He has found out about it’

Although I think there is also a syntactic factor of licensing of tense features involved here, I will not pursue this issue. It is enough to point out that tense-

FOCUS IN BASQUE

329

bearing elements are clitics which require a host to their left in root contexts. Thus, the structures with this particle considered by Laka to show the positive value of the S head can be analyzed, along traditional lines, as verb focalization structures, which, like the constituent focalization patterns we have examined, involve verbal movement to the left periphery. The appearance of the particle is then a by-product of this movement, a last resort, language specific mechanism to provide a basis to the clitic element in initial position (see Ortiz de Urbina 1994). Summarizing, the distribution of ba is both narrower and wider than that of the purported negative counterpart. It is narrower in that it is exclusively found with synthetic verbal forms, and it is wider in that it appears in a variety of structures other than positive emphasis. Let us pursue an analysis of these verb focalization structures along the preceding lines. As indicated above, the type of verb focalization involved here is positive emphasis, as opposed to contrastive emphasis. This means that we should perhaps not assume that the feature [+F] originates in V, if we are to maintain the same ‘contrastive’ interpretation for this feature throughout. Rather, we may have a positive empty operator, similar to the empty yes/no question operator, moving to the Spec position to the left and triggering the usual verb movement pattern to the head position. The particle ba, as usual, would supply the phonetic base for the tensed clitic. We return now to the same problem we had before, namely, the identity of the left headed functional phrase that serves as target for these movements. We began by presenting the possibility of another such phrase in the appropriate position, namely SP. However, the motivation for the positive polarity value of S is not very compelling, as I have just shown, and with this the motivation for S itself is weakened. When considering negative sentences only, the main motivation for Laka’s S head is the displacement of tensed elements to the left. But we have already seen many other such movements. We might think that we have two different landing sites to the left for tensed elements, one in negative sentences identified by Laka and a different one in other contexts which we are trying to locate, but, in a pre-Rizzi (1995) framework, this would be redundant: if it is difficult enough to motivate one, proposing two is a much more unwarranted move. It seems to me more economical in the traditional sense to posit that we have movement to the same CP complex in all these cases — question formation, focalization and negation, and that negation does not therefore originate in that position. An alternative position for negation is that displayed in some structures where verb movement seems not to take place, as in relatives like (53):

330

JON ORTIZ DE URBINA

(53)

Jonek Miren-i idatzi ez dio-n eskutitza Jon Miren-to write  - letter ‘the letter that John did not write to Miren’

The order of heads to the right is actually the expected one in a right headed language: V-NEG-INFL-COMP, where the latter is a clitic which may have originated somewhere else. This suggests an IP structure like (54):

IP

(54)

NegP

I Neg

VP V

Movement of N might be prompted by scopal reasons, picking up inflection on its way to the left-headed phrase above IP. Focalization in negative clauses follows the expected pattern: the focalized constituent is followed by the negative particle plus the auxiliary: (55)

MIKELEK ez du … irakurri Mikel   read ‘It is Mikel that has not read …’

This means that the negative particle can lexicalize [+F] in Infl, and the subsequent movement to the functional head serves the double purpose of producing the required feature matching in a specifier-head configuration and of assigning scope to negation. This analysis of negation in Basque makes this language look quite similar in this respect to neighboring languages, a result which in principle looks appropriate. And the similarity might extend even further if we reconsider focalization in the light of the discussion of positive emphasis above. Notice that I showed that the particle ba is also found in yes/no questions with synthetic verbal forms, as illustrated in the following examples: (56)

a.

Ba-daki Jonek egia? ba-knows Jon truth ‘Does Jon know the truth?’

FOCUS IN BASQUE

b.

331

Irakurri du Jonek eskutitza? read  Jon letter ‘Has Jon read the letter?’

These apparent verb-initial structures pattern with the verb-second ones in focalization and question formation described above if we assume the presence of an empty yes/no operator in the relevant position, as is fairly standard. The parallelism between questions and focalized clauses is quite close, as may have been observed throughout. Both of them raise the same question of what the target is, and thus whatever we say for one should seemingly also be extended to the other. 4.3 Focus Phrases In recent years, proposals for another functional category, a Focus head projecting its own phrase, whether included in the CP complex as in Rizzi (1995) or more or less independent from it (Uriagereka 1995; Brody 1995), have provided a natural alternative worth exploring. The basic issue discussed here, the apparently exceptional phrase required by this line of research, receives a new perspective in these analyses. It may be the case that, quite generally, discourseoriented non-lexical heads such as Topic or Focus, do not display the same directionality found in the V-related phrases, even in a ‘symmetric syntax’ framework like the one assumed here. A proliferated CP analysis along the lines of Rizzi (1995), then enables us to separate the clitic complementizer, which may be a finiteness marker properly belonging to a right-headed CP, from the Focus, Topic (and perhaps Force) left-headed phrases hosting operator features and may prove fruitful to account for the data presented here.

Notes * I would like to thank the participants at the Paris Table Ronde Internationale sur la Grammaire du Focus, and, especially, Laurie Tuller and Georges Rebuschi, for their attention and comments. Usual disclaimers apply. The following abbreviations have been used: :nominative, :accusative, :auxiliary, :complementizer, :negation. 1. This general view implies that it is possible to have neutral sentences without focalized elements, a claim that has been recently challenged (Hualde, Elordieta & Elordieta 1994): in dialects where foci receive a distinct pitch accentual pattern, all sentences can be shown to contain one element so marked as focus. I will skip over this issue here.

332

JON ORTIZ DE URBINA

2. While examples are given in the standard literary dialect, they primarily describe the situation found in Biscaine and Gipuzkoan dialects, unless explicitly noted otherwise. 3. In some speakers, the focalized element may occur at the end as in (i), with a correcting interpretation similar to that of echo questions. (i)

Galdetu didate ZER erosi du-en JONEK

4. An alternative derivation where the wh-operator has been extracted directly over the occupied embedded specifier position, would also be ruled out as a subjacency violation 5. In the case of foci pied-piping relative clauses, as in (23), we would have to assume that the relative clause, turned into an operator-like element by the presence of the focalized element, is moved to the specifier of the DP:

DP

(i)

CP JONEK

D¢ C¢

6. At least in modern times. In older forms of the language, auxiliary-only fronting was also found in Southern dialects more extensively than today, when it is largely fossilized.

References Altube, Seber. 1929. Erderismos. Bermeo. (2nd. edition, Bilbao 1975). Brody, Michael. 1990. “Remarks on the Order of Elements in the Hungarian Focus Field”. In István Kenesei (ed.), Approaches to Hungarian 3. Szeged: JATE. Brody, Michael. 1995. “Focus and Checking Theory”. In István Kenesei (ed.), Approaches to Hungarian 5. 30–43. Szeged: JATE. Horvath, Julia. 1995. “Structural Focus, Structural Case, and the Notion of Feature Assignment”. In Katalin É. Kiss. (ed.), 28–64. Hualde, José Ignacio, Gorka Elordieta & Arantzazu Elordieta. 1994. The Basque Dialect of Lekeitio. ASJU Supplements. Bilbao: University of the Basque Country. Kidwai, Ayesha. This volume. “Word-order and Focus Positions in Universal Grammar”. Kiss, Katalin É. 1995. Discourse Configurational Languages. Oxford: Oxford University Press.

FOCUS IN BASQUE

333

Laka, Itziar. 1990. Negation in Syntax. Doctoral dissertation, MIT. May, Robert. 1985. Logical Form: Its Structure and Derivation. Cambridge, Mass.: MIT Press. Ortiz de Urbina, Jon. 1993. “Operator Feature Percolation and Clausal PiedPiping”. In José Iignacio Hualde & Jon Ortiz de Urbina (eds.), Generative Studies in Basque Linguistics. Amsterdam: John Benjamins. Ortiz de Urbina, Jon. 1994. “Verb Initial Patterns in Basque and Breton”. Lingua 94.125–153. Ortiz de Urbina, Jon. 1995. “Residual Verb Second and Verb First in Basque”. In Katalin É. Kiss ed., 99–121. Oyharçabal, Beñat. 1984. “Le préfixe BA d’assertion positive en basque”. Bulletin du Musée Basque 105. 161–190. Rebuschi, Georges. 1983. “A Note on Focalization in Basque”. Journal of Basque Studies 4(2).29–42. Reinhardt, Tanya. 1983. Anaphora and Semantic Interpretation. London: Croom Helm. Rizzi, Luigi. 1996. “Residual Verb Movement and the wh-Criterion”. In A. Belleti & Luigi Rizzi (eds.), Parameters and functional heads. New York: Oxford University Press, 63–90. Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery”. In Liliane Haegman (ed.), Elements of Grammar: Handbook of Generative Grammar. Dordrecht: Kluwer. Tuller, Laurice. 1992. “The Syntax of Postverbal Focus Constructions in Chadic”. Natural Language and Linguistic Theory 10.303–334. Tsimpli, Ianthi Maria. 1995. “Focusing in Modern Greek”. In Katalin É. Kiss 1995, 176–206. Uriagereka, Juan. 1992. “The Syntax of Movement in Basque”. In Joseba Lakarra & Jon Ortiz de Urbina (eds.), Syntactic Theory and Basque Syntax. ASJU supplements. San Sebastián: Diputación Foral de Gipuzkoa. Uriagereka, Juan. 1995. “An F Position in Western Romance”. In Katalin É. Kiss (ed.), 153–175.

Focus and Arabic Clefts Jamal Ouhalla Queen Mary, London University

Abstract Arabic makes use of a formula for expressing argument focus which is identical to that of simple equative sentences. The formula consists of two DP’s linked together by a pronominal copula (PRON). In argument focus constructions, the pre-PRON DP (the focused argument) is a normal DP denoting an individual and the post-PRON DP has the properties of a free relative, i.e. a relative noun phrase with a null head. This similarity between argument focus constructions and simple equatives suggests that Heycok and Kroch’s (1996) analysis of English pseudo-clefts as equative can perhaps be extended to Arabic argument focus constructions. This would involve treating the free relative as denoting an individual rather than a set. However, Moroccan Arabic free relatives do not lend themselves to this interpretation; they consistently appear to have a set reading. On the basis of this fact, among others, it is argued that the focus constructions in question are best treated as existential statements (or assertions) over choice functions, whereby the prePRON DP (the focus-phrase) is the individual obtained by applying a choice function to the post-PRON (free relative) DP interpreted as denoting a set. The function variable is associated with the whole sentence, more specifically its head PRON, and is bound by an existential operator. It is argued further that this interpretive formula possibly underlies cleft constructions of the type found in English as well. These acquire the Arabic-like formula subsequent to LF movement of the focus phrase (the clefted category) and replacement of (or adjunction to) the expletive IT.

336 1.

JAMAL OUHALLA

Introduction

This paper evaluates the properties of clefts in Arabic with the aim of identifying the interpretive mechanism underlying them compared to the one commonly assumed for focus-preposing and focus-in-situ sentences (Chomsky 1971; Jackendoff 1972 and Williams 1980a). The latter takes focus to involve Existential Closure over the presupposition of the focus-sentence where the focus phrase (f-phrase) is replaced with a variable (Section 2). Arabic clefts have a structure which consists of the f-phrase, a pronominal copular element (PRON) and a free relative: [f-XP PRON FR]. It is argued that they differ minimally from English clefts, which have the format [It COP f-XP RC], in that they lack the expletive subject ‘it’. This difference can be obliterated at LF if English clefts are assumed to undergo raising of the f-phrase to the position of the expletive (preceding the copula), and if the ‘relative clause’ of English clefts is analysed as a free relative, that is, a relative noun phrase headed with a null pronoun. According to this analysis, the ‘out-of-focus’ constituent of clefts is not an open sentence, but, rather, a noun phrase with the semantics of a noun phrase. This makes it difficult to see how the interpretive mechanism assumed for focus-preposing and focus-in-situ sentences can extend to them. These issues are discussed in Section 3. A possible analysis for clefts is suggested by the fact that they bear a strong resemblance to simple equative copular sentences in Arabic. The analysis suggested by this similarity would be to treat clefts as equations involving two categories of the same (denotational) type, the f-phrase and the free relative. An analysis along these lines has been suggested for English pseudo-clefts, arguably a species of focus-constructions (Schachter 1973), by Heycock and Kroch (1996). However, this analysis turns out not to work for at least Moroccan Arabic clefts, on the grounds that Moroccan Arabic free relatives cannot enter into a true equative relationship with a noun phrase. They seem invariably to have a quantificational reading irrespective of the context in which they occur. These issues are discussed in Section 4. The properties of Moroccan Arabic free relatives, and clefts in general, are then taken as the basis for outlining an alternative analysis which exploits the notion of choice functions discussed in Reinhart (1995). The analysis is based on the idea that the free relative constituent of clefts denotes a set and the f-phrase denotes an individual member of that set. The relationship between them is a functional one, whereby the f-phrase is selected as the value of the variable

FOCUS AND ARABIC CLEFTS

337

associated with the free relative. Structurally, the functional relationship between the two constituents is mediated by the category PRON/INFL, assumed to translate as a function variable by virtue of carrying the feature [+f]. This function variable is bound by an existential operator introduced via Existential Closure. The analysis is argued to extend to focus-preposing and focus-in-situ sentences as well. They too include the feature [+f] under their INFL which undergoes raising to a pre-IP head-position, overtly in focus preposing sentences (translating into the well-known adjacency between the f-phrase and the verbal complex) and covertly in focus-in-situ sentences. The INFL of these sentences translates as a function which selects the f-phrase as the value of the variable associated with the open sentence. These issues are discussed in Section 5. The overall conclusion is that focus involves a functional relation between the f-phrase and a constituent with a variable such that the f-phrase represents the value of that variable. Focus sentences are basically existential statements over choice functions.

2.

Focus-preposing and Focus in situ

First we start with focus-preposing and focus-in-situ sentences. In Standard Arabic, as in many other languages (see Kiss 1995 for an overview), f-phrases can be found either in-situ or preposed to the initial position of the sentence. This is shown in examples (1a–b), where f-phrases are written in capital letters, following common practice. In pronunciation, f-phrases are marked with a special pitch accent (or focal stress) called l-nabr in the Arabic Linguistic Tradition: (1)

a.

b.

‘allaf-at Zaynab-u RIWAAYAT-an. wrote-she Zaynab- novel- ‘Zaynab wrote a NOVEL.’ RIWAAYAT-AN ‘allaf-at Zaynab-u. novel- wrote-she Zaynab- ‘It was a NOVEL that Zaynab wrote.’

(Standard Arabic)

Moutaouakil (1989) argues that there is a difference in ‘pragmatic function’ between f-phrases in-situ and the preposed ones. The former are instances of ‘new information focus’, used in contexts where the speaker is giving new information. The latter, on the other hand, are instances of ‘contrastive focus’,

338

JAMAL OUHALLA

used in contexts where the speaker gives information which is in conflict with existing information (i.e. contexts where the speaker gives corrective information). This can be seen more clearly in examples such as (2a–b) which include a negative continuation: (2)

a.

b.

RIWAAYAT-AN ‘allaf-at Zaynab-u (laa (S. Arabic) novel- wrote-she Zaynab- not QASIIDAT-AN). poem- ‘It was a NOVEL that Zaynab wrote (not a POEM).’ LAYLAA ’ashiqa Qays-un (laa ZAYNAB-A). Laylaa loved-he Qays- not Zaynab- ‘It was LAYLAA that Qays loved (not Zaynab).’

Moutaouakil’s analysis builds on the idea originating in the Arabic Tradition that the function of l-takhsiis ‘specification’ associated with preposed f-phrases and which he understands to mean ‘contrastive focus’, correlates syntactically with the process of l-taqdiim ‘preposing’. The stated correlation essentially amounts to the claim that Standard Arabic is a focus-movement language of the Hungarian type, meaning that phrases with a ‘contrastive focus’ reading are obligatorily preposed to the ‘focus field’. Moroccan Arabic differs significantly from Standard Arabic in this respect. The equivalents of (2a–b) have a marginal status at best. The more natural strategy is for the f-phrase to be left in-situ and designated with a pitch accent, or, alternatively, included in a cleft-structure which is close, though not identical, to the structure of English cleft-sentences (see Section 3): (3)

a.

b.

Nadia shr-at KTAB (mashi majalla). (Moroccan Arabic) Nadia bought-she book (not magazine) ‘Nadia bought a BOOK (not a magazine).’ shaf l-BNT (mashi l-WLD). saw-he the-girl not the-boy ‘It was the GIRL he saw (not the BOY).

Moroccan Arabic, therefore, seems to be a focus in-situ language of the English-type. Since Chomsky (1971, 1977), a common way of dealing with focus in-situ is to assume that the f-phrase undergoes covert movement at LF, presumably to the position occupied by overtly preposed f-phrases in languages such as Standard Arabic and Hungarian. This analysis assumes the interpretive scenario


339

outlined in Chomsky (1971), Jackendoff (1972) and Williams (1980a), whereby focus involves Existential Closure over the presupposition of the sentence where the f-phrase is replaced with a variable. An example such as (2a) would have the representation: ∃x, x a book, [Nadia bought x]. Covert movement of the f-phrase, an A′-movement process which leaves behind a variable trace, is the computational equivalent of the process of replacing the f-phrase with a variable. The LF representations of focus sentences being identical, whether the f-phrase is overtly preposed or remains in-situ, the difference between focus-movement languages and focus in-situ languages reduces to overt syntax. The variation is basically the same as for wh-phrases in wh-questions between wh-movement languages and wh-in-situ languages, although it is not clear how Chomsky’s (1995) recent view that all languages undergo movement of a q-morpheme of the wh-phrase in whquestions and that variation reduces to pied-piping can be extended to what seems to be the same type of variation in focus. The parallelism with wh-questions suggests that the position targeted by moved f-phrases is Spec,C. However, the Standard Arabic examples (4a–b), where the preposed f-phrase in the embedded clause appears following the complementizer ‘anna ‘that’ suggest that the position in question may be a different one: (4)

a.

b.

dhanan-tu ‘anna-(hu) KITAAB-an qara‘-at Zaynab-u believe-I that-it book- read-she Zaynab- ‘I believe that it was a BOOK that Zaynab read.’ (S. Arabic) ya-bduu ‘anna(-hu) QASIIDAT-an ‘alqaa Zayd-un it-seems that-it poem- read-he Zayd- ‘It seems that it was a POEM that Zayd read.’ (S. Arabic)

On the basis of parallel examples in other languages, a number of linguists (e.g. Choe 1987; Brody 1990 and Tsimpli 1990, 1995) have argued that preposed fphrases target the Spec position of a functional head which exists over and above C, often called F(ocus) (its projection being therefore Focus Phrase or FP). This is the view that will be adopted in this paper, with the qualification that the projection in question is not specific to f-phrases, but hosts all categories (generally operators) which express information over and above the propositional content of the sentence. This includes, possibly among others, f-phrases, whphrases and modality particles (see Culicover 1991 and Ouhalla 1994b). FP is therefore taken to mean an unspecified Functional Projection.

340

JAMAL OUHALLA

How about the distinction between ‘new information focus’ and ‘contrastive focus’? The standard analysis briefly outlined above does not make a distinction between different types of focus. However, such a distinction needs to be made in view of the fact that Standard Arabic seems to distinguish between at least the two types of f-phrases mentioned in terms of their syntactic distribution. A possible way of incorporating this distinction into the analysis is by manipulating the relevant features encoded in the representation of focus-sentences and which serve as the basis for their interpretation at LF. Following Jackendoff (1972) and much subsequent work, I will assume that f-phrases are designated with the feature [+f] which is spelled out as focal stress or, as is the case in some Standard Arabic constructions, a constituent focus-marker (see Ouhalla 1994b). This instance of the feature [+f] is paired with another instance encoded in the functional head F in relation to which f-phrases are interpreted. This is to say, that f-phrases are interpreted in the position associated with F, which amounts to saying that F marks the scope of f-phrases (outside the presuppositional matrix). With this in mind, let us go back to the distinction between ‘contrastive focus’ and ‘new information focus’. Suppose that while ‘contrastive focus’ entails the presence of the feature [+f] under F, as we have assumed, ‘new information focus’ does not. This distinction amounts to saying that phrases with a ‘contrastive focus’ reading have the broad scope normally associated with focus, whereas phrases with a ‘new information focus’ reading do not. It follows that a phrase with a ‘contrastive reading’ must move to Spec,F if not in the syntax then at LF for interpretive reasons. No such requirement is placed on a phrase with a ‘new information’ reading. It is fair to assume that since phrases with a ‘new information’ reading do not move overtly in Standard Arabic, they do not move covertly either, at least not for the same reason. The distinction just made also amounts to the claim that ‘new information’ is arguably a pragmatic notion without a reflex in the computational system. The term ‘contrastive focus’ is itself confusing and in some instances perhaps outright inaccurate. Not all f-phrases necessarily have a ‘contrastive reading’, as a closer look at examples such as (3a–b) reveals. Their ‘contrastive reading’ is actually a function of the negative continuation rather than of the fphrase. Without the continuation, (3a), for example, has the reading that among the things that the participants assume Nadia could have bought, the speaker asserts that thing to be a book. In other words, (3a) presupposes a set of things that Nadia could have bought, and the speaker specifies ‘book’ as the appropriate individual member of that set. To the extent that this reasoning is correct, focus


341

is probably best understood in terms of the notion ‘specification’ as explained, with ‘contrastive focus’ being perhaps a subtype dependent on the context. This claim is further developed over the next sections. For the moment, note that ‘specification’ is probably the closest possible rendition of the Arabic word l-takhsiis used in the Arabic Linguistic Tradition to describe the reading associated with preposed f-phrases.

3.

Clefts and Pseudo-clefts

In addition to the strategies discussed in the previous section, both Standard and Moroccan Arabic make use of an alternative strategy which involves a structure close, but not identical, to the structure of clefts in English. The structure consists of the f-phrase followed by a pronominal copula (PRON) of the type found in equative copular sentences (more on this later on) followed by a relative clause marked with the relative marker (RM): [f-DP PRON RC]: (5)

a.

b.

ZAYNAB-u hiyya llatii ?allaf-at l-riwaayat-a. Zaynab- .she RM wrote-she the-novel- ‘It was ZAYNAB who wrote the novel.’ (Standard Arabic) L-WLAD huma lli sarrd-at (-hum) Nadia. the-children .they  sent-she (-them) Nadia ‘It was the CHILDREN that Nadia sent.’ (Moroccan Arabic)

Arabic clefts are restricted to definite argument phrases only. Indefinite noun phrases, prepositional phrases as well as categories with an adverbial function, when focused, can only make use of the in-situ strategy in both Standard Arabic and Moroccan Arabic and of the preposing strategy in Standard Arabic. The restrictive scope of the focus position in Arabic clefts may well have to do with the involvement of the pronominal copula. I will not have much to say about this property of Arabic clefts, except to warn that it will limit our discussion of clefts in English to examples which involve a focused definite noun phrase. Despite the surface differences, Arabic clefts can be analysed along the lines of their English counterparts. The difference between them is rather minimal and only affects the (semantically superfluous) expletive subject it which has no equivalent in the Arabic clefts. One could assume that the expletive subject of English clefts deletes at LF, possibly along with the copula on the grounds that it is a ‘verbal expletive’ inserted for the purpose of supporting

342

JAMAL OUHALLA

inflection (see Ouhalla 1991; and Cinque 1999). What is left subsequent to this ‘pruning’ ([It was] the CHILDREN that Nadia sent) is a sequence which consists of the f-phrase followed by a clause, which is basically the sequence found in focus-preposing sentences of Standard Arabic (and Hungarian) and focus-in-situ sentences of Moroccan Arabic and English subsequent to LF-movement of the fphrase. Moreover, relative clauses are often assumed to be open sentences (Williams 1980b), and therefore do not differ in any relevant way from the open sentences which follow preposed f-phrases. Existential Closure can apply to both, yielding the focus reading along the lines described in the previous section. Although there are no necessary reasons for the two types of cleft sentence to have the same structure in order to receive the same focus interpretation, it is possible to assign them approximately the same LF structure. One of the crucial steps is to determine the nature of the structural relation between the f-phrase (the clefted phrase) and what seems to be a relative clause. One obvious possibility is that the two categories form a relative noun phrase with the fphrase as the head N. As far as Moroccan Arabic cleft sentences are concerned, this hypothesis immediately faces the challenge of explaining how the two categories come to appear separated from each other by PRON. Two possibilities suggest themselves: i) either the head N is raised to the subject position out of the relative noun phrase situated following PRON, or ii) the relative clause is extraposed to the final position out of the relative noun phrase situated in the subject position preceding PRON. Neither option seems plausible. The first option presupposes legitimate DP-movement out of a relative noun phrase which is otherwise an island to movement, including the comparatively less restricted wh-movement. The second option presupposes the existence in the language of the process of extraposition of a clause for which there is no evidence. There are additional reasons internal to English clefts which cast doubt on the view that the f-phrase and the relative clause form a relative noun phrase with the f-phrase as the head, some of which are discussed in Schachter (1973). For example, names can easily be clefted/focused (e.g. It’s Marsha that John loves), but they do not make good heads of restrictive relatives (except, of course, in restricted contexts involving more than one individual with the same name, e.g. The Marsha that John loves…). Secondly, clefted/focused categories invariably receive nuclear stress, whereas heads of relatives do not, a difference which has structural implications at least in the framework outlined in Cinque (1993). Thirdly, clefts invariably have an existential presuppositional reading even when they are indefinites, whereas indefinite relatives may or may not have


343

an existential presuppositional reading (e.g. I am looking for a man who runs faster than the speed of light from Schachter 1973). The latter may or may not have an existential reading presumably for the same reasons that simple indefinite noun phrases in general may or may not have such a reading. On the other hand, the necessary existential reading of clefts may be due to the involvement of Existential Closure in the interpretation of focus sentences described earlier. Let us assume that these differences (among others discussed in Green and Ouhalla 1996) are sufficient reasons for assigning the cleft sequence [f-phrase RC] a structure different from that of relatives. The more appropriate structure is one where the f-phrase and RC form a small clause constituent in the complement position of I: [IP e I [SC f-XP RC]] along the lines sometimes suggested for copular sentences (see Moro 1990). The same structure can actually be assumed to underlay Arabic clefts too, with the difference in their respective surface form attributed to the application of a movement process or lack thereof in the mapping onto the surface structure (Spell-out). In English clefts, the f-phrase remains in-situ, thereby prompting insertion of the expletive in the subject position. In Arabic clefts, however, the f-phrase moves to the subject position (Spec,I), and hence the lack of an expletive subject as well as the fact that the fphrase agrees with the pronominal copula and precedes it in linear order. The two representations can be said subsequently to go back to having the same representation at LF, though one different from the underlying representation. It is plausible to assume that English clefts undergo covert movement of the fphrase to the position of the expletive subject, a process of ‘expletive-replacement’ (Chomsky 1986) which results in the derivation of an LF representation identical to the surface representation of Arabic clefts. Although this analysis arguably succeeds in assimilating Arabic and English clefts to each other, the presence of the unattached relative clause remains mysterious. One obvious possibility is that the relative clause actually implies the presence of a whole relative noun phrase, though one which is structurally independent of the f-phrase for the reasons discussed earlier. The relative noun phrase in question would presumably be a ‘free/headless relative’. There is evidence from Arabic which points towards this conclusion. In both Standard Arabic and Moroccan Arabic, the relative clause alone (i.e. without a head-N) can fulfil the function of a DP, as shown in (6a–b). This suggests that relative clauses are probably DP’s rather than CP’s, a conclusion made all the more plausible by the fact that in both languages the relative marker includes a definite article:

344

JAMAL OUHALLA

(6)

a.

b.

wasal-a lladhi haddath-ta-nni ’an-hu. (Standard Arabic) arrived-he .the-he talked-you-to-me about-him ‘The one you talked to me about has arrived.’ ma kansahb ghar lli ‘nd-u flus. (Moroccan Arabic) not I-befriend but .the has money ‘I only befriend those who have money.’

Despite the similarity illustrated in (6a–b), there are crucial differences in the distribution and interpretation of the RC/DP in the two languages. In Standard Arabic, RC/DP has a fairly productive distribution and a referential reading akin to those of simple definite noun phrases. Their counterparts in Moroccan Arabic, however, have a comparatively limited distribution and a quantificational reading. Intuitively, this difference is a function of the difference in the properties of the relative marker in the two languages. The Standard Arabic relative marker includes, in addition to the definite article, pronominal inflection. Its Moroccan Arabic counterpart lacks pronominal inflection, and consists of the definite article only. This situation is reminiscent of the situation with respect to null subjects and objects discussed in Rizzi (1986). Null subjects and objects which are identified by rich inflection have a referential reading, whereas null subjects and objectives which are not so identified receive an arbitrary/quantificational reading. To the extent that the parallelism is genuine, the implication is that Arabic RC/DP’s are arguably relative noun phrases with a null pronoun as the head: [dp pro [RC]]. This is the conclusion reached independently for free relatives in English and other languages (see Groos and van Riemsdijk 1981). A pronominal head can arguably be taken as the defining property of free relatives, which apparently they share with relatives having an overt pronominal head found in the archaic English formula: He who lives in a glass house…. If this conclusion is correct, Arabic clefts have the structure [DP PRON DP], which is identical to the structure of equative copular sentences as we will see in the next section. However, the conclusion has major implications for the attempt to assimilate Arabic and English clefts to each other, as well as the attempt to extend the interpretive mechanism for focus-movement and focus-insitu sentences to clefts. For the parallelism with English clefts to hold, the ‘relative clause’ in English clefts must also be assumed to be a relative noun phrase with a null pronoun as the head (i.e. a free relative). There are some obvious difficulties with such an assumption, some of which have to do with the fact the relative clause of clefts has properties which align it with its counterpart in relative noun phrases but not with free relatives. For example, the relative


345

clause of clefts admits the complementizer that alone (e.g. It’s Marsha that John loves) whereas free relatives do not; they must have a wh-pronoun (e.g. I ordered what/*that he ordered). Also, the relative clause of clefts does not admit the whpronoun what (e.g. It’s that book *what/that John likes) on a par with the relative clause in relative noun phrases (e.g. The book *what/that John likes…) Free relatives, however, admit the wh-pronoun what. There is a sense in which the first difference follows from the idea that the relative clause in clefts modifies a null pronoun head, and as is clear from the discussion of Arabic, also identifies the content of the null pronoun. English whpronouns, though not rich in pronominal inflection, overtly encode the feature [+/- Human], and therefore identify at least this feature of the null pronoun. The complementizer that does not encode any features, and hence perhaps the reason it is excluded from free relatives in non-cleft contexts. The reason that is allowed in clefts may have to do with the proximity of the f-phrase, which identifies the null pro via linear adjacency (see Ouhalla 1994a on this type of identification). The second difference, i.e. the fact that the ‘relative clause’ in clefts does not admit what, unlike free relatives, is less easy to explain. Of course, there are dialects of English, widely spoken in England, where clefts admit the whpronoun what (e.g. It’s the Sun what/wot won it, to cite a notorious headline), although it is not clear how this bears on the issue. As far as the better known dialects are concerned, one could perhaps explain the unacceptable status of examples such as It’s The Sun what won it by appealing to the ban on over-identification suggested in Ouhalla (1994a), a principle of economy. Because the null pronominal head is identified by the adjacent f-phrase (the ‘antecedent’), the wh-pronoun what is not needed, and therefore excluded. Obviously, this explanation does not extend to examples with who (e.g. It’s the Prime Minister who won it), unless one assumes that who is the default whpronoun unmarked for the feature [+/- Human]. There is fairly clear evidence that in the dialects where examples such as It’s the PM’s soap-box wot won it are acceptable, the wh-pronoun wot does not encode the feature [+/- Human], as shown by the fact it is routinely used with human ‘antecedents’, e.g. It’s the Prime Minister wot won it, It’s John wot did it. The extension of the interpretive mechanism for focus-movement and focusin-situ sentences to clefts was made on the assumption that the ‘relative clause’ is an open sentence with an open position accessible to binding from outside (by the existential quantifier introduced via Existential Closure). As it turned out, however, such clauses are, at least in Arabic, noun phrases rather than open sentences.

346

JAMAL OUHALLA

This is also true for English clefts on the assumption that their ‘relative clause’ is a free relative noun phrase. English free relatives are widely assumed to have the semantics of noun phrases (see Larson 1987; and Jacobson 1988, among others). It seems that clefts require a different and peculiar interpretive mechanism after all. Two such mechanisms are discussed in the subsequent sections, and one of which is argued to extend to focus-movement and focus-in-situ sentences as well. The conclusion reached for clefts arguably brings English pseudo-clefts as well into the picture. English pseudo-clefts, sometimes classified as focus constructions (see Schachter 1973), resemble clefts as analysed here in that they too consist of a free relative, a copula and an f-phrase (e.g. What this country needs is a five-cent cigar from Schachter 1973) where a five cent cigar is the f-phrase by virtue of specifying an individual of the set defined by the free relative (more on this later on). As Higgins (1979) explains, at least specificational pseudo-clefts (e.g. What he promised was to reform himself) have properties very similar to those of simple specificational copular sentences (e.g. His promise was to reform himself). This suggests a derivation for pseudo-clefts which is similar to the derivation we have adopted for copular sentences in general and Arabic clefts. The free relative and the f-phrase form a small clause complement of I, and either the free relative or the f-phrase can move to the subject position, essentially the analysis for copular sentences suggested in Moro(1990) with the difference that the small clause is assumed here to be the complement of I rather than V. Movement of the free relative yields examples of the type What this county needs is a five-cent cigar, whereas movement of the fphrase yields (‘non-inverse’?) examples of the type A five-cent cigar is what this country needs. The latter are to all intent and purposes similar to Arabic clefts in their surface form. English clefts differ from pseudo-clefts in that none of the two members of the small clause moves to the subject position in overt syntax. In cases where the free relative includes the complementizer that (e.g. It’s Marsha that John loves) movement of either constituent is prevented by the need for the null pronoun head of the free relative to be locally identified, a PF requirement which does not apply to LF (see Ouhalla 1994a for discussion).

4.

Equatives and Equations

It was pointed out above that Arabic clefts are identical in structure to simple equative copular sentences such as (7b). They both have the format [DP PRON DP]:


(7)

a.

b.

347

Nadia hiyya lli ‘llf-at l-ktab. (Moroccan Arabic) Nadia .she wrote-she the-book ‘It’s Nadia who wrote the book.’ ‘Nadia is the one who wrote the book.’ Nadia hiyya l-mu‘llifa. Nadia .she the-author ‘Nadia is the author (of an understood book).’ ‘Nadia is the one who is the author (of an understood book).’

This obvious link with equative copular sentences suggests that the mechanism involved in the interpretation of Arabic clefts may be the same as the one for equative copular sentences. In other words, Arabic clefts may simply be equations, and therefore different in their interpretive properties from focus-preposing and focus-in-situ sentences. An analysis based on equations is suggested by Heycok and Kroch (1996) for English pseudo-clefts such as What Fiona bought was that dictionary. Heycock and Kroch take the constituent what Fiona bought to denote a plural individual of which the sentence Fiona bought y holds: y[Fiona bought y]. Analysed as such, the constituent in question is identical in its denotational properties to the other constituent of the pseudo-cleft that dictionary. The consequence is that the pseudo-cleft can be said to have an equative representation of the form: y[Fiona bought y] = ‘that dictionary’. The analysis extends naturally to Standard Arabic clefts, where the free relative clearly has a referential reading similar to that of definite noun phrases, as we have seen. The analysis arguably also extends to English clefts on the view suggested above that the ‘relative clause’ is actually a free relative noun phrase, together with the widely assumed view that English free relatives, at least the ones where the wh-pronoun is not modified by -ever, have the semantic properties of definite noun phrases. It is not clear, though, that the analysis extends to Moroccan Arabic clefts. Unlike their English counterparts, Moroccan Arabic free relatives cannot enter into a true equative relationship with a definite noun phrase, examples such as (7a) notwithstanding. A defining property of equative sentences is that their two constituents can be reversed in order with no apparent effect on meaning, as is well-known. The English pseudo-cleft What Fiona bought is that dictionary has That dictionary is what Fiona bought as an alternative, and the copular sentence Nadia is the author has The author is Nadia as an alternative. The situation is different in Moroccan Arabic. While the two constituents of a simple equative copular sentence can easily be interchanged, the constituents of a cleft sentence

348

JAMAL OUHALLA

cannot be so interchanged. Only the order where the free relative follows PRON is possible: (8)

a.

b.

(9)

Nadia hiyya l-ra‘isa. Nadia .she the-president ‘Nadia is the president.’ l-ra‘isa hiyya Nadia. the-president .she Nadia ‘The president is Nadia.’

a.

hadak l-ktab huwwa lli shr-at (u) Nadia. that the-book .it .the bought-she (it) Nadia ‘That book is what Nadia bought.’ b. *lli shr-at (u) Nadia huwwa hadak l-ktab. .the bought-she (it) Nadia .it that the-book ‘What Nadia bought was that book.’

A free relative can only appear to the left of PRON if the category to the right of PRON is also a free relative (a true equation): (10)

a.

b.

lli qult-lk huwwa li kayn. .the told-I-you .it .the is ‘What I told you is what there is.’ lli kat-shuf huwwa lli kayn. :the you-see .it .the is ‘What you see is what there is.’

Examples (9a–b) and (10a–b), taken together, seem to suggest that Moroccan Arabic free relatives do not have the same denotational properties as definite noun phrases, and hence the fact that they cannot enter into a true equative relationship with a definite noun phrase given that equations require sameness of denotational type. Rather, Moroccan Arabic free relatives seem to denote a set, a conclusion which is consistent with the observation made in the previous section that they invariably have a quantificational reading. This is true for all contexts in which they occur, including the two illustrated in (11a–b): (11)

a.

ma kansahb ghar lli mzlut. not I-befriend but .the destitute ‘I only befriend the destitute (with a singular reading, person).’


b.

349

lli ’ndu flus kaydxxlu-h .he has money cause-enter-they-him ‘He who has money is allowed in.’ ‘Those who have money are allowed in.’

If the reasoning so far is on the right line, Moroccan Arabic clefts cannot be treated as equative, and therefore require a different interpretive mechanism. Presumably, the mechanism in question is unlikely to be peculiar to them, to the exclusion of clefts in Standard Arabic and English. The attempt to assign clefts an equative analysis was based on the properties of Standard Arabic and English clefts, but turned out not to extend to Moroccan Arabic clefts. An obvious alternative strategy is to set up an interpretive mechanism on the basis of the properties of Moroccan Arabic clefts and see whether it extends to clefts in Standard Arabic and English. First, we need to solve a puzzle presented by Moroccan Arabic free relatives. Their non-definite reading appears to be hopelessly inconsistent with the fact that they include the definite article, as we saw in the previous section. Associating the definite article with the whole free relative noun phrase would amount to a contradiction, given its non-definite reading. However, there are good reasons to believe that the definite article is not associated with the whole relative noun phrase but with the relative clause only. In a study of relatives and possessives in the Semitic languages, Ouhalla (1996) observe that in the languages where the relative clause carries the definite article, e.g. Arabic and Amharic, the definite article of the relative clause is independent of the head-N. In Amharic, the definite article of the relative clause is in complementary distribution with the definite article of the head-N identical to the complementary distribution between the definite article of the possessor and of the head-N known to exist in Arabic and Hebrew Construct State possessives and English genitive possessives (e.g. the girl’s (*the) toy). However, the definite article is clearly associated with the relative clause rather than with the head-N as it precedes the head-N and follows the relative clause, the normal sequence being [RC Det N]. Amharic definite articles follow the category they are linked to, the normal sequence for a noun phrase being [N Det]: (12)

a.

yä-täk’at’älä-w lebs -burned-the cloth ‘the cloth that got burned’

(Amharic)

350

JAMAL OUHALLA

b.

bet-u house-the (Halefom 1994: 83/79)

The situation is even clearer in Arabic, where the definite article of the relative clause co-occurs with the definite article of the head-N, as shown in (13a–b). The situation in Arabic relatives is similar to the situation in Arabic and Hebrew Free State possessives and English of-possessives to the extent that they exist (e.g. the roof of the house): (13)

a.

b.

l-ktab lli shrit-i the-book .the bought-you ‘the book that you bought’ l-bnt lli shft-i the-girl .the saw-you ‘the girl that you saw’

(Moroccan Arabic)

Arabic relatives show clearly that the definite article of the relative clause is independent of the definite article of the head-N and the relative noun phrase as a whole. Arabic relatives with a head-N have the structure: [NP Det N [RC Det IP]]. Apart from the fact their head is a null pro, free relatives have exactly the same structure: [NP pro [RC Det IP]], where the definite article is bracketed together with the relative clause. Thus, the fact that the relative clause in free relatives carries a definite article does not necessarily mean that free relatives are definite noun phrases. They are basically pronouns whose reference is restricted by the relative clause, and the fact that they receive a non-definite reading in Moroccan Arabic is due partly to this property as we will see shortly. Moroccan Arabic free relatives seem to have different non-definite readings depending on the context. In (11a), it has the reading associated with exceptive phrases licensed by negation more familiar from languages like French (Je n’ai vu que Jean ‘I not saw but Jean’). Whatever the exact nature of this reading, it is clearly due to external binding by negation. In (11b), it seems to have a universal reading of the type associated with English free relatives with whoever. However, there is a sense in which (11b) has the flavour of a general statement, and in fact can easily be modified with the adverb ‘generally’. In view of this, it is perhaps more accurate to describe the reading of the free relative in (11b) as generic, due to binding by a specially inserted operator Gen usually assumed to be a default operator (see Diesing 1992 and Ouhalla 1996). In clefts, free relatives have an existential reading in a sense to be explained in detail later,


351

which is due to binding by an existential operator introduced by Existential Closure. The picture that emerges is that free relatives do not have inherent quantificational properties, but derive their various quantificational readings by binding by external factors. This view is consistent with the view about English free relatives expressed in Srivastav (1991) and Grosu (1994, 1996), among others, that their universal reading is due to external factors. English free relatives are also often claimed to have a definite reading (e.g. I ordered what he ordered). Moroccan Arabic free relatives lack this property, as we have seen, possibly for the reason explained in Section 3 above. On the other hand, English free relatives are not usually claimed to have an existential reading. However, this is due to the fact that the discussion of free relatives tends to exclude the ‘relative clause’ of clefts which is not considered a free relative. If the proposal made here that the ‘relative clause’ of clefts is a free relative is plausible, then English free relatives too can have an existential reading in the appropriate context.

5.

Focus and Choice Functions

Let us now see how the interpretation of clefts works out. As described above, Moroccan Arabic clefts consist of a noun phrase which denotes a set (the free relative) and a definite noun phrase which denotes an individual (the f-phrase). A closer look reveals that the relationship between the two is not one of equation, but, rather, one of inclusion. That is to say, that the individual that the f-phrase denotes is a member of the set that the free relative denotes. In (9a), for example, the free relative denotes a restricted set, ‘the set of things that Nadia could have bought’, while the f-phrase denotes an individual member of that restricted set. Thus, the interpretive mechanism required is one whereby a function applies to a set and yields an individual member of that set, essentially a ‘choice function’ of the type applied by Reinhart (1995) to existential noun phrases with an indefinite/weak determiner, including wh-phrases. The traditional analysis for existential noun phrases is to translate the weak determiner as an existential quantifier which binds the variable associated with N, so that a noun phrase such as some woman has the representation: [(∃x) (woman(x))]. Reinhart suggests to allow the variable associated with N to be bound by forming a set as the translation of N, so that some woman has the representation: [f {x|woman(x)}]. In this representation f is a ‘function variable’ which is bound via Existential Closure, thereby doing away with the need for QR

352

JAMAL OUHALLA

at LF. A sentence such as John met some woman has the representation: ∃f [John met f(woman)], with the existential noun phrase bound in-situ. Reinhart argues that the analysis extends to wh-phrases such as which woman, which is assigned a representation basically identical to that of some woman. wh-phrases in-situ are then ‘function variables’ which are bound in-situ via Existential Closure. The question reading of wh-questions is due the presence of an additional question operator (Q), which defines the set of sentences that represent true answers to the question (Karttunen 1977). In other words, the question reading is not inherent in the wh-phrases, as shown by the fact that in many languages they can have a non-wh-question word reading (see Kuroda 1965 and Watanabe 1991 for Japanese and Aoun and Li 1993 for Chinese). In extending Reinhart’s analysis of existential noun phrases, including whphrases, in terms of choice functions to focus certain crucial differences between existential noun phrases and f-phrases must be taken into consideration. Fphrases can be definite (with a strong determiner) or even names as shown in numerous examples in this paper. These noun phrases, obviously, cannot be said merely to introduce variables, at least in Reinhart’s system (but see Heim 1982) and therefore cannot be said to translate as choice functions. Of the two DP constituents of a cleft, it is the free relative that is more amenable to an analysis in terms of choice functions, although this is not so clear as we will see shortly. The common analysis for restrictive relatives with a head-N is based on the idea that the relative clause and the head-N denote sets which intersect, with the output of the intersection quantified over by Det. Free relatives lack N, and according to the analysis adopted here, they consist of the relative clause and a pro with the status of a Det in the sense that it is located under D: [DP pro [RC]]. This means that pro quantifies directly and exclusively over the relative clause to bind its open (relativized) position. However, it is not easy to see how pro could bind the open position of the relative clause across the definite article in the head position of Arabic relative clauses. The problem is actually broader and involves relatives with a head-N in Arabic, which also include a definite article in the head position of the relative clause. There is no apparent reason why the definite article on the relative clause should not have the same closing effect as it does on simple definite noun phrases. Amharic definite relatives, which lack the definite article on the head-N, pose a different but related problem. It is not clear what quantifies over the intersection of the relative clause and the head-N in the absence of a definite article. For these reasons, among others, Ouhalla (1996) suggested an alternative


353

analysis to relatives whereby the relative clause, first, is a closed expression where the open relativized position is bound by an operator in its own head position (the definite article in Arabic and Amharic and the wh-/relative pronoun in other languages), and secondly, functions as an operator which binds the variable associated with the head-N. In languages where the head-N carries a definite article separately from the relative clause, as is the case in Arabic and English (e.g. the book that Mary wrote), the definite article of the head-N is considered an ‘expletive determiner’ which is in an ‘expletive-associate’ relationship with the relative clause. The relationship enables the relative clause to ccommand and bind the head-N, possibly subsequent to raising of the relative clause and replacement of the expletive. The pattern found in Amharic, where the head-N lacks the definite article and the relative clause c-commands the head-N, characteristic of N-final relatives, is taken to be the more basic pattern. According to this analysis, free relatives involve binding of the head pro by the relative clause, an instance of A′-binding which results in a variable status for the head pro. The relative clause can be assumed already to be in a c-commanding position with respect to pro, for the purposes of identification (via spec-head agreement) discussed earlier or moves there at LF, resulting in the representation: [DP [RC] pro]. Thus, the set/open reading of free relatives in Moroccan Arabic clefts is due to their being A′-bound pronouns. What remains to be added is that while binding by a relative clause of a head-N yields an individual member of the set defined by N, binding by a relative clause of a pro yields a sub-set or a restricted set rather than an individual member. This means that although free relatives involve internal binding of their head, they still have a free variable associated with them open to binding from outside the relative noun phrase. The analysis extends to the use of free relatives in non-cleft contexts on the grounds, explained earlier, that their various readings tend to be due to external factors. Assuming this reasoning to be on the right line, the next step is to explain the functional relationship between the free relative and the f-phrase in clefts, that is, the relationship whereby the f-phrase is selected as the value of the restricted variable associated with the free relative. The solution lies with PRON, the constituent of clefts ignored so far. This element has attracted a lot of attention in the literature on Arabic and Hebrew, and the main challenge it presents is as follows: although the elements collectively referred to as PRON are independent (subject) pronouns, in clefts they consistently display properties which indicate that they are under I(NFL). The

354

JAMAL OUHALLA

body of empirical evidence in favour of the latter conclusion is quite substantial (see Doron 1983, 1986 for Hebrew and Eid 1983, 1991, 1992 and Benmamoun 1992 for Arabic, among many others). These two apparently conflicting properties of PRON are somewhat reconciled by assuming that it is the spell-out of the AGR element of I, that is, the pronominal part of I. However, this analysis fails to explain why the usual subject agreement inflection is not used instead of independent pronouns. The analysis also fails to bring into the picture the fact the pronouns which are collectively referred to as PRON actually have an emphatic/contrastive reading, like strong pronouns in general (at least in languages with a weak series as well such as Arabic and Hebrew). Both problems would be resolved, if the AGR/I of clefts were assumed to have the distinctive property of including the focus feature [+f], in addition to the agreement features. Strong pronouns differ from weak/clitic pronouns and agreement inflection precisely in that they have the extra focus feature. This is the reason they are used in clefts, instead of agreement inflection. The conclusion that PRON carries the feature [+f] is supported by the fact that in Moroccan Arabic clefts, it can carry focal stress instead of the f-phrase. (14a–b) are both possible pronunciations, where capital letters designate the location of focal stress. There is no apparent difference in interpretation between the two examples: (14)

a.

b.

Nadia HIYYA lli ‘llf-at l-ktab. Nadia .she .the wrote-she the-book ‘It’s NADIA who wrote the book.’ NADIA hiyya lli ‘llf-at l-ktab. Nadia .she .the wrote-she the-book ‘It’s NADIA who wrote the book.’

The idea that the feature [+f] is associated with INFL is somewhat puzzling in view of the assumption that categories which are assigned this feature are the ones interpreted as in focus. It is arguably possible to make sense of it if focus is understood essentially to involve a function, more precisely, an existential choice function, which applies to a set and yields an individual member of that set. Let us assume that PRON/INFL translates as a functional variable, and that the f-phrase related to it via spec-head agreement is the member it selects. Clefts can then be said to have the following representation, using examples (14a–b): ∃f: f(Nadia) [DP [RC] prox], which says that a function exists whereby ‘Nadia’ is selected as the value of the variable associated with the free relative.


355

There is a fairly straightforward sense in which the analysis in terms of choice functions extends to focus-preposing and focus-in-situ constructions, illustrative examples of which are reproduced in (15a–b): (15)

a.

b.

KITAAB-an ‘llaf-at Zaynab book- wrote-she Zaynab ‘It’s a BOOK that Zaynab wrote.’ Zinb ‘llf-at l-KTAB. Zinb wrote the-book ‘Zinb wrote the BOOK.’

(Standard Arabic)

(Moroccan Arabic)

The analysis does not necessarily require the ‘out-of-focus’ constituent to be a free relative. The attempt to analyse the ‘relative clause’ of English clefts as a free relative rather than a relative clause which forms a relative noun phrase with the f-phrase is motivated by independent grounds and was intended to bring English clefts into line with their Arabic counterparts. What the analysis requires is for the ‘out-of-focus’ constituent to have a variable (open position) associated with it. The ‘out-of-focus’ clause in (15a) clearly includes an open position (the position vacated by the f-phrase) and (15b) includes a parallel open position subsequent to LF-raising of the f-phrase. Like clefts, (15a–b) include the feature [+f] under their INFL, with morphological reflexes in a number of languages (see Green and Ouhalla 1996). To obtain the appropriate representation, where INFL is outside the ‘out-of-focus’ clause, INFL could be said to raise to F, overtly in (15a) and covertly in (15b). Movement of INFL (or the complex I+V) also has a reflex in the syntax usually in the form of an adjacency requirement between the preposed f-phrase and the verbal complex (see Brody 1990; Tsimpli 1990; and É. Kiss 1995). This leads to a representation which is basically similar to that of clefts, where INFL is in a spec-head relation with the f-phrase and the two are followed by the ‘out-offocus’ constituent. Since the spec-head relation between INFL and the f-phrase involves FP (rather than IP), it is plausible to assume that it only involves the features relevant to this projection, namely [+f] (the two categories could have different pronominal agreement features). The existential operator is located under C immediately above F: [CP ∃f [FP f(the book) [IP Nadia wrote x]]]. The latter says that a function exists whereby ‘the book’ is selected as the value of the variable associated with the ‘out-of-focus’ constituent.

356 6.

JAMAL OUHALLA

Conclusion

The interpretive mechanism commonly assumed for focus-preposing and focusin-situ sentences assumes that they involve Existential Closure over the presupposition of the sentence where the f-phrase is replaced with a variable. It is not clear how this interpretive mechanism can be extended to Arabic cleft-sentences, which consist of the focus-phrase (f-phrase) a pronominal copular (PRON) and a free relative (FR): [f-XP PRON FR]. This is because in the place of an open sentence clefts include a noun phrase with the semantics of a noun phrase. Arabic clefts bear a strong resemblance to simple equative copular sentences, suggesting that they perhaps are interpreted as equations. However, this analysis is shown not to work for Moroccan Arabic clefts where free relatives cannot enter into a true equative relationship with a definite noun phrase. It has been argued that Arabic clefts, and by extension their English counterparts, are best analysed as involving a choice function whereby the f-phrase is selected as the value of the variable associated with the free relative noun phrase. Their representation includes the feature [+f] associated with INFL/PRON which results in INFL/PRON being interpreted as a function which links the f-phrase to the variable associated with the free relative. The analysis extends to focus-preposing and focus-in-situ sentences since they too include, in addition to the f-phrase, a constituent with an open position. Their structural description involves the feature [+f] under INFL which subsequent to its raising to F/C results in linking the f-phrase in Spec,F to the variable associated with the clause. The main conclusion is that focus in general, irrespective of the syntactic properties of the focus construction, involves Existential Closure over a choice function.

References Aoun, Joseph and Yen-hui Audrey Li. 1993. “On Some Differences between Chinese and Japanese wh-elements”. Linguistic Inquiry 24: 365–371. Baker, Christopher Lee. 1970. “Notes on the Description of English Questions: The Role of an Abstract Question Morpheme”. Foundations of Language 6.197–219.


357

Benmamoun, Elabbas. 1992. Functional and Inflectional Morphology: Problems of Projection, Representation and Derivation. PhD dissertation, University of Southern California. Brody, Michael. 1990. “Some Remarks on the Focus Field in Hungarian”. University College London Working Papers in Linguistics 2.201–225. Choe, Hyon-Sook. 1987. “Restructuring Parameters and Scrambling in Korean and Hungarian”. Ms., MIT. Chomsky, Noam. 1971. “Deep Structure, Surface Structure and Semantic Interpretation”. In D. Steinberg and L. Jacobovits (eds.), Semantics. Cambridge: CUP. Chomsky, Noam. 1977. “On wh-Movement”. In Peter Culicover, Tom Wasow and Adrian Akmajian (eds.), Formal Syntax. New York: Academic Press. Chomsky, Noam. 1986. Knowledge of Language. New York: Praeger. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1993. “A Null Theory of Phrase and Compound Stress”. Linguistic Inquiry 24.239–297. Cinque, Guglielmo. 1999. Adverbs and Functional Heads. Oxford University Press. Culicover, Peter. 1991. “Topicalization, Inversion, and Complementizers in English”. Ms., Ohio State University. Diesing, Molly. 1992. “Bare Plural Subjects and the Derivation of Logical Representations”. Linguistic Inquiry 23.353–380. Doron, Edit. 1983. Verbless Predicates in Hebrew. PhD Dissertation, University of Texas. Doron, Edit. 1986. “The Pronominal Copula as an Agreement Clitic”. In Hagit Borer (ed.), The Syntax of Pronominal Clitics: Syntax and Semantics. Vol. 19.313–332. Eid, Mushira. 1983. “The Copula Function of Pronouns”. Lingua 59.197–207. Eid, Mushira. 1991. “The Copula Pronoun in Arabic and Hebrew”. In Bernard Comrie and Mushira Eid (eds.), Perspectives on Arabic Linguistics III. Amsterdam: John Benjamins, 31–61. Eid, Mushira. 1992. “Pronouns, questions and agreement”. In Mushira Eid, John McCarthy and E Borselow (eds.), Perspectives on Arabic Linguistics IV. Amsterdam: John Benjamins, 107–142. Groos, A. and Henk van Riemsdijk. 1981. “Matching Effects with Free Relatives: A Parameter of Core Grammar”. In A. Belletti, L. Brandi, and L. Rizzi (eds.), A Parameter of Core Grammar”. Theory of Markedness in

358

JAMAL OUHALLA

Generative Grammar: Proceedings of the 1979 GLOW Conference. Pisa: Scuola Normale Superiore, 171–216. Grosu, Alexander. 1994. Three Studies in Locality and Case. London: Routledge. Grosu, Alexander. 1996. “The Proper Analysis of Missing-P Free Relative Constructions”. Linguistic Inquiry 27.257–293. Halefom, Girma. 1994. The Syntax of Functional Categories: A Study of Amharic. PhD dissertation, University of Quebec at Montreal. Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD Dissertation, University of Massachusetts, Amherst. Heycock, Caroline and Anthony Kroch. 1996. “Pseudo-cleft Connectivity: Implications for the LF Interface Level”. Edinburgh Occasional Papers in Linguistics 96–1. Higginbotham, James. 1985. “On Semantics”. Linguistic Inquiry 16.547–593. Higgins, Francis Roger. 1979. The Pseudo-Cleft Construction in English. New York: Garland. Horvath, Julia. 1986. Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jacobson, Pauline. 1988. “On the Quantificational Force of English Free Relatives”. In E. Back, Eloise Jelinek, Angelica Kratzer and Barbara Partee (eds.), Quantification in Natural Languages. Dordrecht: Kluwer, 451–486. Karttunen, Lauri. 1977. “The Syntax and Semantics of Questions”. Linguistics and Philosophy 1.3–44. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kiss, Katalin É. 1995. “Introduction”. In Kiss (ed.), 3–27. Kiss, Katalin É. (ed.) 1995. Discourse Configurational Languages, Oxford Studies in Comparative Syntax, Oxford University Press Larson, Richard. 1987. “Missing Prepositions and the Analysis of Free Relative Clauses”. Linguistic Inquiry 18.239–266. Moro, Andrea. 1990. “There-raising: Principles across Levels”. Paper presented at the 1990 GLOW Colloquium, Cambridge. Moutaouakil, Ahmed. 1982. Réflections sur la théorie de la signification dans la pensée linguistique arabe. Rabat: Publications of the Faculty of Letters and Human Sciences. Moutaouakil, Ahmed. 1989. Pragmatic Functions in a Functional Grammar of Arabic. Dordrecht: Foris.


359

Ouhalla, Jamal. 1991. Functional Categories and Parametric Variation. London: Routledge. Ouhalla, Jamal. 1994a. “The Syntactic Representation of Arguments and Identification”. Seminar presented at the ASG-group, Max Plank Gesellschaft, Berlin. Ouhalla, Jamal. 1994b. “Focus in Standard Arabic”. Linguistics in Potsdam 1.65–92. Ouhalla, Jamal. 1996. “Remarks on the Binding Properties of wh-pronouns”. Linguistic Inquiry 27.676–707. Reinhart, Tanya. 1995. “Interface Strategies”. Manuscript, Research Institute for Language and Speech, University of Utrecht. Rizzi, Luigi. 1986. “Null Objects in Italian and the Theory of pro”. Linguistic Inquiry 17.501–558. Schachter, Paul. 1973. “Focus and Relativization”. Language 49.19–46 Srivastav, Vaneeta. 1991. “The Syntax and Semantics of Correlatives”. Natural Language and Linguistic Theory 9.637–686. Tsimpli, Ianthi-Maria. 1990. “The Clause Structure and Word Order of Modern Greek”. University College London Working Papers in Linguistics 2.226–258. Tsimpli, Ianthi-Maria. 1995. “Focusing in Modern Greek.”. In Kiss (ed.), 176–206. Watanabe, Akira. 1991. “Wh-in-situ, Subjacency and Chain-formation”. Ms., MIT. Williams, Edwin. 1980a. “Remarks on Stress and Anaphora”. Journal of Linguistic Research 1.71–82. Williams, Edwin. 1980b. “Predication”. Linguistic Inquiry 11.203–238. Williamson, Janis. 1987. “An Indefiniteness Restriction for Relative Clauses in Lakhota”. In Eric Reuland and Alice ter Meulen (eds.), The Representation of (In)definiteness. Cambridge: MIT Press, 168–190. Wright, W. 1896. A Grammar of the Arabic Language (Third Edition). Cambridge: CUP.

Subject Index

adjacency 229–30, 234–5 adjectives: incorporation 203–4 triggers for f-structure 127 adjunction 193–200, 196, 200–2, 223, 225–6, 236–7 adverbs: sentential vs. VP 188–93 verbal status in Portuguese 49 VP adverbs 196–9 Aghem 164, 224 agreement: intonational agreement between focussed phrase and coda 99–100 noun-class markers 249 Spec-Head 63 also 60 augment morpheme 169–71, 174 Avoid Pronoun Principle, The 47

Extraposition approach 105, 257 formal semantic account 106–8 interrogative clefts 98–9 intonational copy 99–101 pragmatics of French clefts 102–4 prosodic data 95–101 pseudo-cleft 341–6 raising 90 semantics of French clefts 102–4 small-clause analysis 90–1 traditional hypotheses 88–94 typology 84–8 clitics: Italian 47 Portuguese 39 Complete Functional Complex 56–7 Complex-NP Constraint 64 copular sentences 247ff., 336, 346–51 cumulative interpretations 184 Cushitic languages 276ff., 302

Basque 110, 219 Lekeitio Basque 111 northern vs. southern dialects 325, 328 BE-focus structures 37–40 Bemba 158–9

determiners: out-of-focus determiner 164–5 strong 132–3 weak 131 Dinka 146 discourse configurationality 281–5 domain discourse 227 Dutch 301

Cewa 156–7 cleft sentences 182, 246–67, 341–6, 349ff. declarative clefts 95–8

each 139 Empty Category Principle 64

362

SUBJECT INDEX

enclisis 41 English 223, 347 evaluative construction 42–3 even 60, 61, 64–5, 69, 77, 240 existentional noun phrases 351ff. expletive constructions 296–8 file semantics 123–30, 146 focal stress-assignment 34 focus: assertive focus 153–6 bare phrase-structure implementation 68–9 bound focus 59–66, 240; basegeneration vs. movement 55ff., 66 broad event-related focus 85 broad presentational focus 85 construction types 56–9 contrastive focus 240, 242, 340–1; Arabic 337–8; exclusive/restrictive; vs. new information 340;Portuguese 25ff.; value 26; Wolof 246ff. cross-linguistic distribution 219 derived predicate 180ff. exclamatory comment 86 focus clitics 202–3 focus effects 282–5 focus movement vs. association with focus 64–6 Focus Phrase 331, 337 focus positions 215–24; 287–300 focus preposing and focus in situ 337–41, 355, 356 Focus Structure (f-structure) 120ff., 146–7; f-structure rules 121–4; subordinate f-structure 125–7, 138 (see also scope) focus types 13–7 focus-like elements 33

Focus-Tense Identification 40 interaction with other areas of grammar 165–73 lowering 66–8 minimalist theory of 224–35 narrow focus 215, 217–8, 242 negative focusing 255, 262–3 neutral focus 215 object foucusing 252–3, 256–60 partial focus 28 Pied-piping 316–21 presentational focus 38; Portuguese 25ff.; Hindi-Urdu 239 prosodic focus 46 pure focus 33 quantificational focusing particles 55 quantified focus elements 30, 49 relation with particle 55–6 restrictive focus 127–8 scope: German 143–5; Basque 314–6 structural focus and Case assignment positions 295–300 subject focusing 253–5, 260–2 term focus vs. predicate focus 277 three strategies in French 104 verum focus 148 wh-words 181–3, 187–8, 194–5, 204–5, 220–2, 228, 236, 292–5, 312–24 word order of focus elements 200–5 (see also tone; negation) free inversion 31 fusion 264 Gallician 30 Generative tradition 3–13 German 57–9, 61, 64, 71, 72–3, 77–8, 143–4, 207, 241–1 Greek 79, 314

SUBJECT INDEX Hindi-Urdu 201, 217, 220, 221, 222, 223, 227, 230–3, 239, 240–1 Hungarian 147, 219, 301 identificational sentences 247 interpretability 226–9, 241 interrogative 158, 292–5; 312–3, 326, 339 Italian 32, 35, 47, 74 ki-Rundi 159, 160 Kihung’an 165 Kirundi see ki-Rundi la-Haya 153–6, 157, 160–3 left-dislocation 155–6 lu-Ganda 165–73 Malayalam 111, 216–7, 227, 230–3, 239, 240–1 Minimalist Program 1–2, 56ff., 68 minor functional heads 62–3 Montague Grammar 62 negation 168, 171, 326, 327–30 negative imperative 165 tonal reduction 160 non-configurationality 301 non-restrictive/non-exclusive focus 27ff. null operator 31 nur 58, 61 only 56–9, 69, 71, 77, 240 Parameter Value Preservation measure 232 particle: additive particles 60 attachment to Xmax 60–2 focus-constituent requirement 59–60 functional head 62–4 Move-F 75–6 wide-scope derivation 70–6 particle phrase 56ff.

363

percolation 320–1 predicational sentences 247 preverbs 303 Principle of Exclusion 190 Principle of Lexical Association 60 Principle of Full Interpretation 63 pronouns 129, 146 proto-Bantu 170 proximity-to-the-verb requirement 219 pure topics 33 quantification: adverbial quantifiers 290–1 focus positions 289–95 in overt structure 185–8 over Events 186ff. subjects 189–93 wh-quantifier as existential 187–8, 193–4 Quantifier Raising 133, 180, 181–4, 190, 221–2 R-dependencies 128–30 relative clause 191–2, 294, 298–300, 350 free relatives 336–7, 346, 348–51, 354 relativization 191–2, 199–200 Relativized Minimality 196 Rendille 277 scope: ambiguous scope 193–5, 196–9 and focus in Basque 314–6 cardinal scope 134–7 f-structure 141–5 focus in German 143–5 Nuclear Scope and rheme 139–40 propositional scope 57, 66 quantifier scope 133–45, 147, 183ff. scoped stages 141–2 some–every 137–41, 147

364

SUBJECT INDEX

scrambling 61, 182–3, 218–9, 222, 231–2, 235, 239 se-Tswana 110, 157 sogar 58, 61 Spanish 322, 323 syntax-PF interface 263–5 Tangale 216, 220, 222, 233–5, 242 tense: conjoint vs. disjoint 159, 174, 175 three-place predicates: Portuguese 36–7 tone 152–73, 242 tonal finality 153 tonal marking of [+focus] 152–9 tonal marking of [–focus] 160–4 tonal reduction 160–1 tonally active 153 too 60

topic: scope properties 130–3 Stage Topic 124–5 topic positions 287–8 Topic vs. Focus: historical overview 24–5 topic-like elements vs. true topics 32–3 TopicFocusP: EvaluativeP 41–4; FocusP 29–35 Topic Constraint on R-dependencies 129 topicalization 250 Tswana see se-Tswana update semantics 146 verb raising 235ff. Verb-Second Constraint 58 weak crossover 221, 222, 289, 292 Western Bade 216, 222, 233–5

List of Contributors

Manuela Ambar Departamento de linguistica Facul. des letras Universidade de Lisboa Cidade Univ. 1699 Lisboa codex PORTUGAL [email protected] Josef Bayer Institut für Germanistische Sprachwissenschaft Friedrich-Schiller-Universitaet Jena Ernst-Abbe-Platz 4 07743 Jena GERMANY [email protected] Nomi Erteschik-Shir Ben-Gurion University of the Negev Deptment of Foreign Literatures & Linguistics PO Box 653 Beer Sheva 84105 ISRAEL [email protected]

Larry Hyman Department of Linguistics University of California 2337 Dwinelle Hall Berkeley, CA 94720 USA [email protected] Sarah D. Kennelly Via Campo Rosso, 4 52010 Ortignano-Raggiolo (AR) ITALY [email protected] Ayesha Kidwai Centre of Linguistics & English School of Languages Jawaharlal Nehru University New Delhi — 110067 INDIA [email protected] Alain Kihm CNRS-LACITO 44 rue de l’Amiral-Mouchez 75014 Paris FRANCE [email protected]

366

LIST OF CONTRIBUTORS

Jacqueline Lecarme 2LC-CNRS 1361 route des Lucioles 06560 Sophia Antipolis FRANCE [email protected] Jon Ortiz de Urbina Universidad de Deusto Departamento de Filologia Inglesa Apartado 1 48080 Bilbao, SPAIN urbina@orion. deusto.es Jamal Ouhalla Department of Linguistics School of Modern Languages Queen Mary and Westfield College Mild End Road London E1 4NS GB [email protected]

Anne Clech-Darbon & Georges Rebuschi TYGRe (SYLED), Université de Paris III 19 rue des Bernardins 75005 Paris FRANCE clechdarbon@ idf.ext.jussieu.fr [email protected] Annie Rialland CNRS Unité de recherche en phonétique et phonologie (UPRESA 7018, CNRS-Sorbonne-Nouvele) 19 rue des Bernadins 75005 Paris [email protected] Laurie Tuller Département de linguistique Université de Tours 3, rue des Tanneurs 37000 Tours FRANCE [email protected]

In the series LINGUISTIK AKTUELL/LINGUISTICS TODAY (LA) the following titles have been published thus far, or are scheduled for publication: 1. KLAPPENBACH, Ruth (1911-1977): Studien zur Modernen Deutschen Lexikographie. Auswahl aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene Malige-Klappenbach. 1980. 2. EHLICH, Konrad & Jochen REHBEIN: Augenkommunikation. Methodenreflexion und Beispielanalyse. 1982. 3. ABRAHAM, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983. 4. ABRAHAM, Werner & Sjaak De MEIJ (eds): Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks, Groningen, 1984. 1986. 5. GREWENDORF, Günther and Wolfgang STERNEFELD (eds): Scrambling and Barriers. 1990. 6. BHATT, Christa, Elisabeth LÖBEL and Claudia SCHMIDT (eds): Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences. 1989. 7. ÅFARLI, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. 8. FANSELOW, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. 9. GELDEREN, Elly van: The Rise of Functional Categories. 1993. 10. CINQUE, Guglielmo and Guiliana GIUSTI (eds): Advances in Roumanian Linguistics. 1995. 11. LUTZ, Uli and Jürgen PAFEL (eds): On Extraction and Extraposition in German. 1995. 12. ABRAHAM, W., S. EPSTEIN, H. THRÁINSSON and C.J.W. ZWART (eds): Minimal Ideas. Linguistic studies in the minimalist framework. 1996. 13. ALEXIADOU Artemis and T. Alan HALL (eds): Studies on Universal Grammar and Typological Variation. 1997. 14. ANAGNOSTOPOULOU, Elena, Henk VAN RIEMSDIJK and Frans ZWARTS (eds): Materials on Left Dislocation. 1997. 15. ROHRBACHER, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and pro-drop. 1999. 16. LIU, FENG-HSI: Scope and Specificity. 1997. 17. BEERMAN, Dorothee, David LEBLANC and Henk van RIEMSDIJK (eds): Rightward Movement. 1997. 18. ALEXIADOU, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. 19. JOSEFSSON, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. 20. LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. 21. KLEIN, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. 22. ALEXIADOU, Artemis and Chris WILDER (eds): Possessors, Predicates and Movement in the Determiner Phrase. 1998. 23. GIANNAKIDOU, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. 24. REBUSCHI, Georges and Laurice TULLER (eds): The Grammar of Focus. 1999. 25. FELSER, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999. 26. ACKEMA, Peter: Issues in Morphosyntax. 1999.

° 27. RUZICKA, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. 28. HERMANS, Ben and Marc van OOSTENDORP (eds.): The Derivational Residue in Phonological Optimality Theory. 1999. 29. MIYAMOTO, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 1999. 30. BEUKEMA, Frits and Marcel den DIKKEN (eds.): Clitic Phenomena in European Languages. 2000. 31. SVENONIUS, Peter (ed.): The Derivation of VO and OV. 2000. 32. ALEXIADOU, Artemis, Paul LAW, André MEINUNGER and Chris WILDER (eds.): The Syntax of Relative Clauses. 2000. 33. PUSKÁS, Genoveva: Word Order in Hungarian. The syntax of È-positions. 2000. 34. REULAND, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. 35. HRÓARSDÓTTIR, Thorbjörg. Word Order Change in Icelandic. From OV to VO. 2000. 36. GERLACH, Birgit and Janet GRIJZENHOUT (eds.): Clitics in Phonology, Morphology and Syntax. 2000. 37. LUTZ, Uli, Gereon MÜLLER and Arnim von STECHOW (eds.): Wh-Scope Marking. 2000. 38. MEINUNGER, André: Syntactic Aspects of Topic and Comment. 2000. 39. GELDEREN, Elly van: A History of English Reflexive Pronouns. Person, ‘‘Self’’, and Interpretability. 2000. 40. HOEKSEMA, Jack, Hotze RULLMANN, Victor SANCHEZ-VALENCIA and Ton van der WOUDEN (eds.): Perspectives on Negation and Polarity Items. 2001. 41. ZELLER, Jochen : Particle Verbs and Local Domains. n.y.p. 42. ALEXIADOU, Artemis : Functional Structure in Nominals. Nominalization and ergativity. n.y.p. 43. FEATHERSTON, Sam: Empty Categories in Sentence Processing. 2001. 44. TAYLAN, Eser E. (ed.): The Verb in Turkish. n.y.p. 45. ABRAHAM, Werner and C. Jan-Wouter ZWART (eds.): Issues in Formal German(ic) Typology. n.y.p 46. PANAGIOTIDIS, Phoevos: Pronouns, clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. n.y.p.

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Experimental Pragmatics Semantics (Linguistik Aktuell Linguistics Today)

Multiple Wh-Fronting (Linguistik Aktuell Linguistics Today)

Parentheticals (Linguistik Aktuell Linguistics Today, 106)

Argument Structure (Linguistik Aktuell Linguistics Today)

Phase Theory (Linguistik Aktuell Linguistics Today)

Causatives in Minimalism (Linguistik Aktuell Linguistics Today)

Cyclical Change (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Continuity and Change in Grammar (Linguistik Aktuell Linguistics Today)

The Derivation of Anaphoric Relations (Linguistik Aktuell Linguistics Today)

The Emergence of Order in Syntax (Linguistik Aktuell Linguistics Today)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

Current Issues in Generative Hebrew Linguistics (Linguistik Aktuell Linguistics Today)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

The Grammar of Repetition: Nupe grammar at the syntax-phonology interface (Linguistik Aktuell Linguistics Today)

Movement Theory of Control (Linguistik Aktuell Linguistics Today)

Syntatic Aspects of Topic and Comment (Linguistik Aktuell Linguistics Today)

Adverbials and the Phase Model (Linguistik Aktuell Linguistics Today, 177)

Scrambling and the Survive Principle (Linguistik Aktuell Linguistics Today)

Quantifier Scope in German (Linguistik Aktuell Linguistics Today)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

The Linguistics Enterprise: From knowledge of language to knowledge in linguistics (Linguistik Aktuell Linguistics Today)

Balkan Syntax and Semantics (Linguistik Aktuell Linguistics Today)

Theoretical Approaches to Universals (Linguistik Aktuell Linguistics Today)

Grammaticalization As Economy (Linguistik Aktuell Linguistics Today, LA 71)

Perspectives on Negation and Polarity Items (Linguistik Aktuell Linguistics Today)

Copular Clauses: Specification, Predication And Equation (Linguistik Aktuell Linguistics Today)

Agreement Systems (Linguistik Aktuell Linguistics Today, Volume 92)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)