Rebuschi Tuller 1999 The Grammar of Focus

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: Abraham@let.rug.nl
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
TM The paper used in this publication meets the minimum requirements of
8
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 99-22380
isbn 90 272 2745 4 (eur) / 1 55619 908 2 (us) (Hb; alk. paper) 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 1

Georges Rebuschi and Laurice Tuller
Aspects of the Syntax of Focus in Portuguese 23
Manuela Ambar
Bound Focus or How can Association with Focus be Achieved
without Going Semantically Astray? 55
Josef Bayer
Are There Cleft Sentences in French? 83
Anne Clech-Darbon, Georges Rebuschi & Annie Rialland
Focus Structure and Scope 119
Nomi Erteschik-Shir
The Interaction between Focus and Tone in Bantu 151
Larry Hyman
The Syntax of the P-Focus Position in Turkish 179
Sarah D. Kennelly
Word Order and Focus Positions in Universal Grammar 213
Ayesha Kidwai
Focus in Wolof: A Study of What Morphology May Do to Syntax 245
Alain Kihm
Focus in Somali 275
Jacqueline Lecarme
Focus in Basque 311
Jon Ortiz de Urbina
vi TABLE OF CONTENTS
Focus and Arabic Clefts 335

Jamal Ouhalla
Index 361
List of Contributors 365
The Grammar of Focus
An Introduction*
Georges Rebuschi Laurice Tuller

University of Paris III 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 incep-
tion. 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 legiti-
mate 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 require-
ments 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 [+interpret-
able], 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.
THE GRAMMAR OF FOCUS 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 senten-
ces), 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)
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. Peter can do it.
b. Peter won’t do it.
c. (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-
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 consid-
erations), 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. John writes poetry in his .
b. 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
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 intro-

duced which “informally” yields the “partially developed logical forms” (10a,b)
when applied to (9a,b) (Chomsky’s (115–118)):
(9) a. Bill likes J
b. B likes John
(10) a. the x such that Bill likes x — is John
b. 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) 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. Georges loves M
(contrastive or noncontrastive)
b. 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. His wife mistreats J
b. 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 informa-
tion (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
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 wh-
movement” 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. his mother loves J
(“J with focal stress”)
b. 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]
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)
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 frame-
work, 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
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 signifi-
cance, notably, questions about what in the earlier Extended Standard Theory
(EST) framework were called “surface effects” on interpretation. These are
manifold, involving -  - , figure-
ground 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 Chom-
sky’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 discuss-
es rules like the one that is concerned with the interpretation of resumptive
pronouns, which violate about every known syntactic constraint: “CNPC, the wh-
island 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:
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 gram-
mar, cf. [(21)]:
[(21)] a. Speaker 1: John didn’t hit a home run
Speaker 2: I know a woman who did
b. 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-
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 languag-
es. 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 dis-
cussed 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 Portu-
guese, 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),
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 interac-
tion/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 character-
ize 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 argu-
ments. 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 config-
urational, 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
in Turkish follows from their position immediately to the left of V. The unam-
biguous wide scope construal of these elements is analyzed as their being right-
adjoined 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 f-
structures (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 phonolo-
gy, 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 PF-
component — 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 (cliticiza-
tion, 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 presenta-
tional 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
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-struc-
ture. 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
thus appears to be consensus on what the focus issues are, which surely amounts
to progress in our understanding of this grammatical phenomenon. The explora-
tion 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 “discourse-
configurational” 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. Péter  Marit
b. Peter  love Mary (contrary to what you said/seem to assume)
c. 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.
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 coindexa-
tion 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)
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. Wir haben gestern   gesehen
we have yesterday the girl seen
‘We sax (lit. have seen) THE GIRL yesterday’
b.   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. Was hast du GELESEN?
‘What have you read?’
b. #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 Interpreta-
tion”. 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 Interpreta-
tion, New York: North-Holland, 163–210.
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. Cam-
bridge, 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.
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 Univer-
sity, 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. Groe-
nendijk et al. (eds.), Formal Methods in the Study of Language 2, Amster-
dam: 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.
Watters, J.R. 1979. “Focus in Aghem”. In L.M. Hyman (ed.) Aghem Grammati-
cal 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 opera-
tions 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 “presenta-
tion” structures, as represented in (1a) vs. (1b) respectively:
(1) a. (S (NP) (VP))
b. (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
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. Comeu a Joana.
Ate the Joana
b. A tarte comeu a Joana
c. *A Joana comeu
d. *A Joana comeu a tarte
e. *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 incompat-
ible 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
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. ISTO fazem os reis quando viajam

This is what kings do when they travel
b. 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 construc-
tions, subject-verb inversion systematically occurs and, when a clitic is present,
we always get the clitic-verb order”, as in (16) below:
(16) a. ISSO lhe disse eu
This is what I told him
b. 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.
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 R-
expression — 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 hypothe-
size 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 licens-
ing. (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 construc-
tions 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) 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 identifica-
tion 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 assign-
ment. 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 representa-
tions (23a) and (23b):
(23) a. [TopicP OPk/a tartek [Topic′[TopicFocusP A Joanai [TopicFocus′ [IP ti
comeuv [ek]]]]]]
b. [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.
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 interpreta-
tion.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 MANUELA AMBAR
5. Other Predictions
5.1 Three-place predicates
Observe how the predictions of our system are borne out in structures of three-
place 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. [TopicP Opj/i [Topic′ [TopicFocusP OPi/j [TopicFocusP ofereceuv [IP ei tv ej
à Joana]]]]]
b. [TopicP o Pedroi [Topic′ [TopicFocusP as floresj [TopicFocusP ofereceuv
[IP proi tv tj à Joana]]]]]
c. [TopicP as floresj [Topic′ [TopicFocusP o Pedroi [TopicFocus′ ofereceuv
[IP ti tv proj à Joana]]]]]
d. [TopicP o Pedroi [Topic′ ofereceuv [TopicFocusP as floresj [TopicFocusP
t′v [IP proi tv tj à Joana]]]]]
e. [TopicP as floresj [Topic′ ofereceuv [TopicFocusP o Pedroi [TopicFocusP
t′v [IP ti tv à Joana]]]]]
f. *[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
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 ]]]]]
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 redupli-
cated 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 successful-
ly 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 exclu-
sive/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. 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 wh-
constituent 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 con-
texts, and with a different interpretation, are well-formed:
(40) a. Isto lhe fazem os reis!
This him do the kings
b. Isso lhe disse eu!
That him said I
42 MANUELA AMBAR
c. Muitos livros lhe ofereceu o Pedro!

Many books him offered Peter
d. 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. Belo trabalho me fizeste tu!
Nice work to me did you
b. 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. Isso disse-lhe eu
This said him I
b. Isso lhe disse eu!
This him said 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 Dis-
course; 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
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 indepen-
dently 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 constitu-
ent — 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 distribu-
tion 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 combina-
tion.” (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 —
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) non-
exclusive/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 c-
command 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 so-
called 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 possibili-
ties, 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 argumenta-
tion 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 qualifica-
tions.
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
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 construc-
tions 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 wh-
extraction 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. 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 …)
b. 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 non-
exclusive/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 topic-
pro 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).
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 exclu-
sive/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 Portu-

guês. PhD Diss., Univ. de Lisboa. Published by Ed. Colibri, Lisboa, 1992.
Ambar, Manuela. 1994. “Aux-to-Comp and Lexical Restrictions Verb Move-
ment”, 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 Portu-
guesa, 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 Lan-
guage”. 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. (Pub-
lished 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.)
Casteleiro, João. 1981. Sintaxe Transformacional do Adjectivo. Lisbon: INIC.

Chomsky, Noam. 1977. “On wh-Movement”. In P. Culicover, T. Wasow & A.
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 compar-
ativa”. In Inês Duarte and Isabel Leiria (eds.), Actas do Congresso Interna-
cional sobre o Português, Vol. I. Lisbon: Associação Portuguesa de Linguís-
tica 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 Refer-
ence to the Order of Meaningful Elements”. In J. Greenberg (ed.), Univer-
sals 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 Hungari-
an. 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 Não-
Massiva. 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. Dor-
drecht: Foris.
Raposo, Eduardo. 1995. “Clitic Position and Verb Movement in European
Portuguese”. Ms., University of California, Santa Barbara.
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 Sentence-
Types. Washington D.C.: Georgetown University Press.
Tuller, Laurice. 1992. “The Syntax of Postverbal Focus Constructions in Cha-
dic”. 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 Linguis-
tica 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 move-
ment 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. [3%n wgd ognlI InvaIt Ám7rI]
b. John would [PrtP only [VP INVITE MARY]]
c. 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. 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
b. 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 when-
ever 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. [? only [DP MARY]] would invite us to her home
b. [? 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. ANNA could [PrtP even / ?*only [VP escape from the prison]]
b. 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 interpreta-
tion 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 differ-
ence 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 c-
command 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) a. Sally likes only HIM / *it
b. Sabine liebt nur IHN / *es
(9) a. Sally likes even HIM / *it
b. 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. Some students smoke even IN the classroom

b. Only LITTLE boys are permitted to use the ladies’ restroom
(13) a. Einige Studenten rauchen sogar IN dem Hörsaal
b. 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. Einige Studenten rauchen [[IN dem Hörsaal] sogar]
b. [[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. Peter hat das Buch nur DURCHGEBLÄTTERT
Peter has the book only skimmed-through
b. 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 scram-
bling trace of the direct object instead of the lexical object itself. This analysis
is sketched in (19):
(19) a. … das Buchi … [nur [VP ti durchgeblättert]]
b. [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. only Bill nur Willi DP
b. only to London nur nach London PP
c. only poor nur arm AP
d. only go to London nur nach London fahren VP
e. only that he goes to London nur daß er nach London fährt 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 63
(21) 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 indistinguish-
able 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. I have only suggested that DOMINGO should sing “Tristan” in
VIENNA (but not that PAVAROTTI should sing it in
SALZBURG)
b. 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) Only MARY would [VP [only
——————
MARY] invite us to her home]
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 repre-
sentation of (4) is not as in (30) but as in (31a); given that the filled Spec-
position 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. Only MARY would [PrtP [only
——————
MARY] [Prt′ [VP [only
——————
MARY]
invite us to her home]]] ( S-O)
b. 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. ———
ANNA could [PrtP even [VP [even
—— ANNA] escape from the
prison]]
b. ———— I guess ————
SALLY SALLY was [PrtP even [VP arrested 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 deriva-
tions 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 conver-
gence. 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 associa-
tion 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 JOSEF BAYER
6. 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. The students in the GDR were required to learn only RUSSIAN
b. 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. Die Studenten in der DDR wurden gezwungen nur RUSSISCH
zu lernen
b. Das Bildungsministerium der DDR verlangte, daß die Studenten
nur RUSSISCH lernten
These sentences are not ambiguous at all. They permit only the false interpreta-
tion, 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. Die Studenten in der DDR wurden nur gezwungen RUSSISCH
zu lernen
b. 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. Die Studenten in der DDR wurden gezwungen [RUSSISCH
nur] zu lernen
b. 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. The GDR education ministry made the suggestion to learn only
RUSSIAN
b. The GDR education ministry made the suggestion that the
students learn only RUSSIAN
(39) a. The conformist student asked the ministry [where to study
[only RUSSIAN]]
b. 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 typo-
logical difference that sets the SOV-language German aside from the SVO-
language 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 Move-
F 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. Wer hat [an wen] gedacht?
who has [at whom thought
‘Who has thought about whom?’
b. [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 comple-
ments in the subjunctive mood.
(44) a. E’ veramente necessario che io parl -i soltanto con Gianni
is really necessary that I speak - only with Gianni
b. 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 catego-
ries 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 agree-
ment 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 Spell-
Out. 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
—— RUS-
SIAN]]]]]]]
(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, how-
ever, 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 possibili-
ties 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. *AUFi hat er nicht ti gehört

intended: “He did not stop”
c. *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) a. The scientist found the dangerous substance in only three organisms
b. Some only weakly gifted students applied for the job
(ii) a. Der nur mit einer Badehose bekleidete Gangster konnte entfliehen
the only with a swim-suit dressed gangster could escape
b. 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 non-
canonical 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, ge-
nau, 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. Cam-
bridge, Mass.: MIT Press.
Jacobs, Joachim. 1983. Fokus und Skalen: Zur Syntax und Semantik der Gradpar-
tikeln 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 agree-
ment”. 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. Lingui-
stische Berichte, Sonderheft 4. Opladen: Westdeutscher Verlag.
Riemsdijk, Henk van. 1996. “The extension of projections”. Ms. Tilburg Univer-
sity/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 Interna-
tional 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 senten-
ces 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. (— Ta fille est tombée dans l’escalier?)
(‘Did your daughter fall down the stairs?’)
b. — Non, c’est le petit qui est tombé dans l’escalier.
‘No, it’s the young one [+masc.] that fell down the stairs.’
(2) a. (— Qui c’est, ce gamin?)
(‘Who’s that kid?’)
b. — C’est le petit qui est tombé dans l’escalier.
‘It’s the young one [+masc.] that fell down the stairs.’
(3) a. (— Tu sembles inquiète. Qu’est-ce qui se passe?)
(‘You look worried. What happened?’)
b. — C’est le petit qui est tombé dans l’escalier.
‘The young one [+masc.] fell down the stairs.’
(4) a. (— Papa a acheté trois gâteaux.)
(‘Dad’s bought three cakes.’)
b. — 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. (— Papa a acheté trois gâteaux.)
(‘Dad’s bought three cakes.’)
b. — C’est le petit qui va être content!
‘The young one’s going to be happy!’
c. ?— 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!’]
d. *— C’est le petit qui va tomber dans l’escalier!
‘The young one’s going to fall down the stairs!’
e. *— C’est le petit qui a été/est content!
‘The young one was/is happy!’
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 interpre-
tations. 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. C’est le garçon à qui/auquel j’ai parlé. []
‘It’s the boy to whom I spoke.’
b. C’est au garçon que j’ai parlé. []
‘It’s to the boy that I spoke.’
(13) a. C’est la maison où/dans laquelle j’ai dormi.[]
‘It’s the house where I slept.’
b. 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 incorpo-
rates 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 post-
focus 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
followed by a relative clause, unless they are preceded by a definite article,

whereas names are just fine in real cleft sentences:
(14) a. C’est *(le) Michel qui parle russe.8 []
b. C’est Michel qui parle russe. []
(15) a. C’est *(le) Michel à qui j’ai parlé. []
b. 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 (restric-
tive) relatives, it is not the case when c-clauses are considered:
(19) 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.’
2. 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 (ordi-
nary) wh-movement: extracting a wh-phrase from inside the object NP is
perfectly natural, whether it is a interrogative pronoun or a relative one, as

shown by the well-formedness of (21a,b):
(21) a. [De qui]i a-t-il épousé [la fille ei]?
‘Whose daughter did he marry?’
. ‘Of-whom did he marry the daughter e?’
b. 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]]].
b. [IP Le  est tombé dans l’escalier].

c. 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 lui-
même 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) a. 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.
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. C’est [CP* le petiti [CP Opi qui [IP ti est tombé]]]
b. 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
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. J’ai vu le petit pleurer.
‘I saw the young one cry/crying’
b. J’ai vu que le petit pleurait.
‘I saw that the young one was crying.’
c. 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 wh-
word will appear in the higher specifier position; but in the case of cleft senten-
ces, 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
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 c-
command its trace:
(29) a. [IP [NP Ce(lui) [CP qui est tombé]] [I′ est le petit]] ⇒
b. [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. What you see is what you’ll get. [/]
b. It’s what you’ll get that you see. [only ]
c. 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. 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. Celui que Marie aime, c’est celui que j’ai vu.
., ‘The one who Mary loves, it’s the one who I saw.’
b. Celui que j’ai vu, c’est celui que Marie aime.
., ‘The one who I saw, it’s the one who Mary loves.’
c. C’est celui que j’ai vu [Ø/*lui que Marie aime].
d. C’est celui que Marie aime [Ø/*lui que j’ai vu].
(35) 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  —
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. C’est dans la maison que j’ai dormi.
‘It’s in the house that I slept.’
b. C’estj [VP tj [PP dans la maison]
[PP Ø [NP Ø [CP Opi que [IP j’ai dormi ti]]]]]20
c. 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 necessar-
ily altogether impossible, all raise difficult questions. We shall now investigate
the prosodic properties of this sentence type.
3. 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 straight-
forward. 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 discus-
sion. 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
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 preposi-
tions (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 secondary-
stressed 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 intona-
tion 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%
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:
\/ | |
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 associat-
ed 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:
| /\ | |
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
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 Jean-
Pierre 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
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.
| | | |
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.
| | | |
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

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.
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. C’est M qui est sorti.
‘It’s Michel that went out.’
b. M est sorti.
c. Seul M est sorti.
(49) 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 descrip-
tion to be complete, we also need to see the degree to which these three con-
structions 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. C’est M qui est sorti, pas J-P (#par exemple).
b. M est sorti, pas J-P (#par exemple).
c. Seul M est sorti, pas J-P #(par exemple).
(52) a. C’est M que j’ai vue, pas J (#par exemple).
b. J’ai vu M, pas J (#par exemple).

c. 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 senten-
ces 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. Seuls M et J-P sont sortis.
b. Seul M est sorti.
(54) a. Ce sont M et J-P qui sont sortis.
b. 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. M et J-P sont sortis.
b. 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
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 quantification Properties of the
≈ uniqueness contextual set
Clefts + closed/small
Association with seul + open/large
Free focus – closed/small
5. 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 c-
clause. Yet another advantage (29) has over some of the rival hypotheses is the
direct semantic interpretation which it provides for the copula (see below).
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 right-
adjoined CP is extracted have a definite, rather than a universal or generic,
reading — or how the extraposition transformation itself could likewise contrib-
ute 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. [IP [IP [DP Un homme ti] est venu] [CP qui voulait te parler]i]
b. [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. [IP C’estv [VP tv [NP le petit]]]
b. [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-phonolo-
gy 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
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. [DP Ce [CP Op que j’ai acheté]](, c’)est un/le dictionnaire
‘What I’ve bought is a/the dictionary’
b. [[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. [D ce] ⇒ lP[lQ[∃x[∀y[P(y) ↔ x = y]&Q(x)]]]39
b. [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. lP[lQ[∃x[∀y[P(y) ↔ x = y]&Q(x)]]](lw[acheter(je,w)])
b. [lQ[∃x[∀y[[lw[acheter(je,w)](y)] ↔ x = y]&Q(x)]]]
c. [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. [lQ[∃x[∀y[[acheter(je,y)] ↔ x = y]&Q(x)]]]
(lz[être-un/le-dictionnaire(z)])
b. [∃x[∀y[[acheter(je,y)] ↔ x = y]&
[lz[être-un/le-dictionnaire(z)]](x)]]]
c. [∃x[∀y[[acheter(je,y)] ↔ x = y]&
[être-un/le-dictionnaire(x)]]]
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. [lQ[∃x[∀y[P(y)× ↔ x = y]&Q(x)]]]
(lz[être-un/le-dictionnaire(z)])
b. [∃x[∀y[P(y) ↔ x = y]&
[lz[être-un/le-dictionnaire(z)]](x)]]
c. [∃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 construc-
tion. 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
(67) a. Ce sont les enfants qui sont tombés []

. ‘it  the children that fell down’
b. C’est les enfants qui sont tombés []
(68) [DP ce] ⇒ [lQ[∃X[∀Y[P(Y) ↔ X = Y]&Q(X)]]]
6. 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 dimen-
sion 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 (inde-
pendently) 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
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 cross-
linguistic variation: compare the properties of English clefts (which admit explicit
relative pronouns in the coda), Wolof clefts (Kihm, this vol.), and more perspicu-
ously 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)
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 unambig-
uous counterparts to (11) in the text, one that corresponds to the restrictive relative interpreta-
tion, 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:
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 correspond-
ing 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 lui-
même 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).
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. [Il est sorti,] [Jean-Pierre.]
| |
L L
. ‘He’s gone out, J.-P.’
b. [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.
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 predict-
able. 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).
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 disserta-
tion, 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: Long-
man, 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.
Chomsky, N. 1970. “Deep Structure, Surface Structure, and Semantic Interpreta-

tion”. 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.
Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, N. 1982. Some Concepts and Consequences of the Theory of Govern-
ment 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.
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 Construc-
tions”. 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.
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 Kommunika-
tionsteori 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 Seman-
tics 1.1:75–116.
Rooth, M. 1996. “Focus”. In S. Lappin (ed.), The Handbook of Contemporary
Semantic Theory. Oxford: Blackwell, 271–297.
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. Groenen-
dijk, 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 princi-
ples. 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 subordi-
nate 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. F-
structure 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-of-
the-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 re-
spect 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
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 f-
structure, 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
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. Do you have a pet?
b. 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
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 f-
structures 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, con-
strained 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 spatio-
temporal 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-
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. #A friend is intelligent.
b. #A student likes linguistics.
Specific indefinites can, however, be topics:
(9) a. A friend of mine is intelligent.
b. A certain student likes linguistics.
c. 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 recog-
nized 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 f-
structure 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
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’ applica-
tion 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 speaker-
specificity. 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 (subordi-
nate) 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
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
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 assign-
ment 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 nega-
tion, 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. John talked to a boy.
b. He criticized her.
c. Two girls arrested three boys.
d. 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-depen-
dence. 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-
dencies are the ones applying to f-structure. R-dependencies thus encompass

those dependencies that follow from the f-structure determined update. Corefer-
ential 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 R-
dependency which receives a specific interpretation, must therefore represent an
existing card in the file. Importantly, sentences with R-dependencies are con-
strained 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 f-
structure 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 confus-
ing 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
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. TOPi [A girl talked to Johni]FOC
(Answers: Tell me about John)
b. 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
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. TWO students are inTELLigent.
b. Two STUdents are inTELLigent.
(18) a. SOME students are inTELLigent.
b. 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 involv-
ing quantifying over undefined sets.
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. Every student is intelligent.
b. Every student in my class is intelligent.
c. 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 restric-
tive 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 suggest-
ed 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 anony-
mous 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 interpreta-
tion 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.
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. [Q1]TOP [V Q2]FOC
b. TOP2 [Q1 V Q2]FOC
c. 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
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. C C {+(a,b)} → {+(1,2,3)}
b. C D {+(a,b)} → 1
→2
→3
c. DC a → {+(1,2,3)}
b →
d. DD a →1
b →2
→3
(28) a. Ds (Co) a → [1,2,3]
b → [4,5,6]
b. Ds (Do) a → 1 b → 4
→2 →5
→3 →6
c. Do (Cs) [a,b] →1
[c,d] →2
[e,f] →3
d. Do (Ds) a → 1 c → 2 e → 3
b→ d→ f →
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. Today, two girls arrested three boys.
b. At 6 o’clock, two girls arrested three boys.
c. 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 candi-
dates 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 object-
topic 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
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. Two GIRLS arrested three BOYS. (sTOPt C C)
b. Two GIRLS arrested THREE boys. (sTOPt C D),(TOPo C)
c. TWO girls arrested three BOYS. (sTOPt D C),(TOPs C)
d. 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 f-
structure 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.
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 b→4
→2 →5
→3 →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. Someone arrested everyone.
b. 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
are left with 3 readings for each sentence represented by the following f-structures:
(36) a. sTOPt [everyoneD arrested someone]FOC
b. sTOPt [everyoneC arrested someone]FOC
c. EveryoneTOP [arrested someone]FOC
(37) a. sTOPt [someone arrested everyoneD]FOC
b. sTOPt [someone arrested everyoneC]FOC
c. 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 f-
structures 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.
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 f-
structure 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. each of them vs. *every of them
b. ??eachone of them vs. every one of them
(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. All the items in the store cost $500.
b. 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 necessi-
tates 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.
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: Most logicians like linguistics.
B: 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
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 f-
structure (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. In every city, John loves someone.
b. In every city, someone loves you.
c. In some city, John loves everyone.
d. In some city, everyone loves you.
None of these are ambiguous as predicted.
As noted by one an anonymous reviewer, the following example  ambiguous:
(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. someoneFOC [he went to high school with]TOP
b. 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 c-
commanding quantifier in surface structure necessarily has wide scope. Excep-
tions to this generalization involve f-structures with stage topics or cases in

which the subject is not a candidate for topichood: Chierchia and McConnel-
Ginet (1990, 117) mention the following exceptions to the c-command constraint:
(52) a. There was a name tag near every plate.
b. A flag was hanging in front of every window.
c. 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) [CPfast jeder Jungei [CHAT
nearly every. boy has
[ti[mindestens einen Roman [gelesen]]]]]
at.least one. novel read
b. (∃∀, ∀∃) 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 f-
structure. (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
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 extreme-
ly 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: [CPe[C′e[IPPeter[das Verbrechen[sofort[zu[gab]]]]]]]
Focus Assignment: [CPe[C′e[IPPeter[das Verbrechen[sofort[zu[gab]]F]]]]]
C0 movement: [CPe[C′gabi[IPPeter[das Verbrechen[sofort[zu[ti]]F]]]]]
Spec-CP movement: [CPPeterj[C′gabi[IPtj[das Verbrechen[sofort[zu[ti]]F]]]]]
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 recon-
struction 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 essen-

tially 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, foreground-
ing 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 f-
structures as its input producing a well-formed discourse representation. F-
structure 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 identifica-
tion is an innate and essential part of conceptual-intentional structure itself.
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
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 f-
structure 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
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 Philoso-
phy 15.111–183.
Chierchia, G. and S. McConnel-Ginet 1990. Meaning and Grammar: An Intro-
duction 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.
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.
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 Universi-
ty Press.
Williams, E. 1988. “Is LF Distinct from S-Structure? A Reply to May”. Linguis-
tic 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 conse-
quences for the linguist interested in studying either phenomenon. The phonolo-
gist 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 morpholog-
ical spell-out, or a phonological process. In all cases that I know, however, the
construction, morphological exponent or phonological process may also charac-
terize 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, morpholo-
gy 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-púnu ‘pig’
e-m-bûzi ‘goat’ o-mú-ti ‘tree’
c. o-mu-kázi a-ka-gw-a d. e-m-punú e-ka-gw-a
e-m-búzi e-ka-gw-a o-mu-tí gú-ka-gw-a
‘a woman/goat fell’ ‘a pig/tree fell’
e. o-mu-kâzi % Káto f. e-m-púnú % Káto
e-m-bûzi % Káto o-mú-tí % Káto
‘a ⁄a , Kato’ ‘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 Internal Before % Before Pause
(2c,d) (2e,f) (2a,b)
a. …H-Ø …H-L …HL-L …HL-L
b. …Ø-H …L-H …H-H …H-L
As is common in Bantu, we propose that vowels in lu-Haya are either underly-
ingly 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 underly-
ing 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 b. σ σ c. σ
= =
H L % L H % 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. 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’
b. 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’
c. 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 penulti-
mate H becomes HL. Now compare this result with the corresponding right-
dislocations 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. [ LD1 [ LD2 [ LD3 [ Assertion ] ] ] ]
………… % % % %
(LD = left-dislocation)
b. [ [ [ [ Assertion ] RD1 ] RD2 ] RD3 ]
…………%
…………………%
…………………………%
…………………………………%
(RD = right-dislocation)
c. [ 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, ci-
Cewa 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-khǔ:ngú
H
c. tí-náa-mu-mê:ny-a % m-khǔ: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.
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) Cewa-Ntcheu and Cewa-Nkhotakota
a. VVCV%
H
b. VVCV%
|
H
Cewa-Nkhotakota
c. 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 se-
Tswana. They offer the minimal pair in (10).
(10) Tonal effect of % in se-Tswana ≠ tonal finality
a. bátSáàb7́r7́ká % líbf̀n7́
‘they1 will work, they1 too’
158 LARRY M. HYMAN
b. bátSáàb7́r7́kà líbf̀n7́
‘they1 will work with them2’
In (10a) the form líbf̀n7́ is separated from the verb by % and means ‘they too’.
In (10b) líbf̀n7́ 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 repre-
sents 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’
c. bámó bá-la-lyá ínsoka (Present, [+focus])
‘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 preced-
ing 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 -la-
had 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. nga mw-aa-tóbá úmutóndó, bálééisaafúlwá [+focus]

‘if you break the pot they will get angry’
b. 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 differ-
ently, 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 spread-
ing 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]
P0 -Ø- … -a/-ye [+TR] -ra- … -a/-ye [−TR]
P1 -a- … -a/-ye [+TR] -aa- … -a/-ye [−TR]
P2 -á- … -a/-ye [+TR] -ára- … -a/-ye [−TR]
The tenses in question are here identified as P0, P1 and P2, roughly the immedi-
ate, 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 ci-
Bemba; 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 LARRY M. HYMAN
3. Tonal marking of [−focus]
The prefix forms in (13) clearly indicate that it is [+focus] that is morpho-
logically 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 lu-
Haya:
(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. b. ‘they tie Káto’ etc.
Present habitual ba-kóm-a ba-kom-a káto
Past1 bá-á-kôm-a ba-a-kom-a káto
Past2 ba-kom-íl-e ba-kom-il-e káto
Past habitual ba-a-kóm-ag-a ba-a-kom-ag-a káto
Future1 ba-laa-kôm-a ba-laa-kom-a káto
Future2 ba-li-kóm-a 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),
(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
Past2 ti-ba-kom-íl-e ti-ba-kom-il-é káto
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
Experiential8 ba-lá-kom-íl-e ba-lá-kom-il-e káto
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
Past3 ti-bá-á-kom-il-e ti-bá-á-kom-il-e káto
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. non-
intrinsically 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)
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 non-
final 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 -
f́ 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 -f́ in Aghem
a. ò mf̀ tà] tí—-bvú — !n7́
he  count dogs today
‘he counted dogs today’
!
b. ò mf̀ tà] n7́ — !tf́
bvú
he 1 count today dogs 
‘he counted dogs today’
c. ò kà tá] bvú — !tf́ n7́
he P1– count dogs  today
‘he didn’t count dogs today’
d. tá] bvú— !tf́ n7́
count dogs OF today
‘count dogs today!’
e. wizí—n wì—l à ò mf̀ tà] bvú — !tf́ n7́
woman   she P1 count dogs OF today
‘the woman who counted dogs today’
In (23b) contrastive focus is marked on the adverb n7́ ‘today’ by placing it in the
IAV position. In this case the following direct object noun bv} ¢ !tf́ 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
adjacent to its c-commanding lexical head. The semantically empty (or “dum-
my”) OF determiner -f́ 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 -f́. However this generalization appears to be violated in (23c–e),
where the object noun appears as bvú!tf́ 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 deter-
miner. 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] constitu-
ents? 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 (Kihung’an)
chair  -fail- Kipes to-buy
‘the chair that Kipese didn’t buy’
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-of-
focused (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

a. X
X Z
[........................]TG
b. e-bi-kópò e-bi-nénè ‘big cups’

e-bi-kópò bi-sátu ‘three cups’
e-bi-kópò bi-rî ‘those cups’
c. e-bi-kópó by-áá=wálúsìmbì ‘cups of Walusimbi’
HØ HL L
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 demon-
strative. 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. tw-áá-làb-à tw-áá-láb-á wálúsììmbì

H L L H ØØ HLL L
‘we saw Walusimbi’ []
e. tú-náá-láb-à tú-náá-láb-á wálúsììmbì
H Ø H L H Ø HØ H LL L
‘we will see Walusimbi’ []
f. tú-lì-làb-a tú-lí-láb-á wálúsììmbì
HL L H ØØ H LL L
‘we will see Walusimbi’ []
g. ne tú-làb-à ne tú-láb-á wálúsììmbì
HL L HØ Ø H LL L
‘and we saw Walusimbi’ []
h. tu-lab-ê tu-lab-é wálúsììmbì
HL HØ HLL L
‘let’s see Walusimbi’ []
i. 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 Walú-
sììmbì. 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. te-tú-ú-láb-è walúsììmbì
H HL HLL L
‘we will not see Walusimbi’ [ ]
f. te-tú-lì-làb-a walúsììmbì
H LL HLL L
‘we will not see Walusimbi’ [ ]
g. ne tú-tà-làb-à walúsììmbì
H L LL HLL L
‘and we didn’t see Walusimbi’ [ ]
h. 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) LTD does not apply in certain affirmative verb tenses, which are [+F]
a. lab-írìr-à walúsììmbì
H L L H LL L
‘look after Walusimbi!’ []
b. tú-kyáá-láb-à walúsììmbì
H Ø H l HLL L
‘we still see Walusimbi’ []
c. tw-aaka-láb-à walúsììmbì
HL HLL L
‘we have just seen Walusimbi’ []
d. 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) 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 right-
dislocated 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. tú-làb-a a-ma-fútà ‘we see oil’

HL HL
c′. 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 [−refer-
ential]
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. 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 augment-
marked 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. y-à-gúl-à e-bi-kópò (bì-nó)
HL H L
‘he bought these cups’ [“even” focus]
b. 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. y-à-láb-á walúsììmbì
H Ø H LL L
‘he saw Walusimbi/Walusimbi’
b. 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. ku-láb-à byàà=mùlímí
H L
‘to see those of the farmers’
b. 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
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. [CG V
|
H → Ø / ____
b. [ 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. Similar-
ly, 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 phrase-
penultimate 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 —
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 Grammat-

ical 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 Linguis-
tica 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.
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. Disserta-
tion, 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 immedi-
ately 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 distinc-
tion 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, maintain-
ing 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 I° Verb j
VP ‘F’
Spec V°tj
Subject
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. Sec-
tion 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 Incorpo-
ration 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. (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]
b. Deniz’i ve Ufuk’u
Deniz- and Ufuk-Acc
[as an answer to (3a)]
c. Deniz Ufuk’u, Uğur Ümit’i ve Remzi’de Ahmet’i

D. Ufuk- U. Ümit- and Remzi-also Ahmet-
gördü.
saw
‘Deniz saw Ufuk, Uğur saw Ümit and Remzi saw Ahmet.’
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 interpreta-
tion. 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) 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 immedi-
ately 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) a. Üç çocuk/Çocuklardan üçü yeni bir araba-@ almış.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 almış.
c. #Üç çocuk/Çocuklardan üçü taze bir çekirdek-@ yemiş.
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]
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.
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-@ kaldFrmiş.
3 girl 3 basket-@ lifted.
‘3 girls lifted 3 (nonspecific) baskets.’
[Cumulative reading:
<girl x, basket y>: 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 görmüş 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. 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-
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 P-
Focus 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 New Matrix
original restrictor original matrix

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
(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;
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-@ öldürmüş.
self a patient-@ killed
‘Self killed a patient.’
When it occurs preverbally the NODP occupies an A-bar position that is left-
adjacent 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
(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 doubt-
ful] 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) theta-
government, government by a theta-marking element, or (b) antecedent govern-
ment. 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-
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 configu-
ration. 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 L-
marked 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 İyileşeceği] son-derece şüpheli 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
VP Adv VP
hastahaneden
(from hosp)
Spec,VP VP VP Adv V°
yürüyerek
(on foot) (emerged)
Spec V° Adv V° DP
(
Iyilesecegi
¸ son derece supehli
¸ olan hasta
(will recover) (extremely) (doubtful being) (patient)
(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
there should also be relativization out of an adjunct clause. This is not attested,
as seen in (24).
(24) *[[[Ayşe ti kırdığı] için] Ahmet’in bağırdığı] bardaki…
[[[A. broken for Ahmet’s yelled glass
‘the glass that Ahmet yelled because Ayşe 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-@ almış.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 quantifi-
cational 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
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 ambigu-
ous 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 adjunc-
tion 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’.
(26) a. Bu şehirde kim kimi arıyordu?

in.this.city who- who- was.seeking
‘Who was looking for who in this city?’
b. Bu şehirde kimi kim arıyordu?
in.this.city who- who- was.seeking
‘Who was looking for who in this city?’
(27) a. 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)]
b. Deniz’i Can arıyordu; Ümit’i Ufuk arıyordu.
[as an answer to (26a/b)]
c. 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. Ufuk kimlere güzel bir hediye-@ vermiş?
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]]]
b. [Can ve Deniz]’e güzel bir hediye-@ vermiş.
C. and Deniz- beautiful a present-@ gave
‘S/he gave a beautiful present to Can and Deniz (together).’
c. Can’a ve Deniz’e güzel bir hediye-@ vermiş.

Can- and Deniz- beautiful a present-@ gave
‘S/he gave a beautiful present to Can and Deniz (together or
separately).’
d. Can’a ve Deniz’e birer hediye-@ vermiş.
Can- and Deniz- one.each present-@ gave
‘S/he gave a present to Can and s/he gave a present to Deniz.’
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-@ vermiş?
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
individual who gave it to each of them]

b. Denize vermişler.
Deniz- they.gave
‘They gave (it) to Deniz.’
[as an answer to (29a) — Turkish is an object drop lang.]]
c. #Can Deniz’e, Uğur Ayşe’ye, ve Ufuk Ümit’e
C. Deniz- U. Ayşe- and U. Ümit-
birer hediye-@ vermiş. [as an answer to (29a)]
each.one present-@ gave
‘Can gave a present to Deniz, Uğur to Ayşe, and Ufuk to
Ümit.’
d. #Can Deniz’e, Uğur Ayşe’ye, ve Ufuk Ümit’e bir
C. Deniz- U. Ayşe- and U. Ümit- a
hediye-@ vermiş. [as an answer to (29a)]
present-@ gave
‘There was one present such that Can gave it to Deniz, Uğur to
Ayşe, and Ufuk to Ümit.’
‘Can gave one present to Deniz, Uğur gave one to Ayşe, and
Ufuk one to Ümit.’
e. #Deniz’e üç hediye-@ vermişler. [as an answer to (29a)]
Deniz- 3 present-@ they.gave
‘They gave 3 presents to Deniz.’
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
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-@ okumuş.
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-@ okumuş.
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]
(30′) IP
Spec I’
VP I° Verb j
VP ‘F’
NODPk
VP Adv
Spec
Subject 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.
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 P-
Focus? 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 clause-
bound 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. [tj Arabayı tamir eden] kadınıj şimdi gördüm.
[tj car- repairing woman- now I.saw
‘I just saw the woman who repaired the car.’
b. [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 şimdi gördüğüm] 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-@ gördüğüm] 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.
While relativization is felicitous out of a RC with a specific head noun, relativi-

zation 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 kime güzel bir hediye-@ vermiş? [=(29a)]
3 child who- beautiful a present-@ gave
‘Who did the three children give a beautiful present to?’
b. *Üç çocuk güzel bir hediye-@ kime vermiş?
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.
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 construc-
tion. 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-@ almış.
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 almış.
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 straightfor-
ward 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-
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) 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-@ almış.
D. new car-@ bought
‘Deniz bought (a) new car.’
(38) a. *Güzel ve eski araba-@ almış.
beautiful and old car-@ s/he.bought
‘S/he bought (a) beautiful and old car.’
b. Güzel ve eski bir araba-@ almış.
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) BAŞ-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 promi-
nence when modified (40c).
(40) a. Deniz -@ almış

D. piano-@ bought
‘Deniz bought (a) piano.’
b. Deniz  piyano-@ almış.
D. old piano-@ bought
‘Deniz bought (an) old piano.’
c. Deniz eski bir -@ almış.
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
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, left-
verb 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. wh-
movement 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
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 quantifi-
cation. 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 contradic-
tion with the empirical facts. Potentially crucial issues such as Case and Incorpo-
ration 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. Kuruoğlu, 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.
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 Kuruoğlu (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 Boğazici 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) Şu tamirci bir arabayı tamir etti.
that mechanic a car- repaired
‘That mechanic repaired a (specific) car.’
(ii) */?Şu 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.
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 back-
grounded 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, right-
adjoined 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 Philoso-
phy 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 Interpre-
tation.” 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.
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. Berke-
ley: 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. Cam-
bridge (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.
Kennelly, S.D. 1996. “Nonspecificity = Theticity An Instantiation of Presenta-

tional 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 Existen-
tials”. In M. Pearson et I. Paul. UCLA Occasional papers in Linguistics 17.
Pinto, M. 1994. “Subject in Italian: Distribution and Interpretation.” In R. Bok-
Bennema 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.
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 Ger-
manic 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 XP-
and 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 sub-
components 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 PF-
scrambling 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 Hindi-
Urdu has largely been ignored in the literature. Section 2 delineates this relation-
ship 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 organiza-
tion 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 Zubizar-
reta (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 presupposi-
tions 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 presenta-
tional 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-of-
the-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 presup-
position 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 non-
canonical 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 wh-
phrases 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. kitaab  laaye-gaa

book Ram bring-
‘It is Ram who will bring the book.’
c. What did Ram bring?
d. What did Ram do to the book?
e. 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. kal 7̃ aaũũ]-gaa
tomorrow I come-
‘I will come tomorrow.’
b.  aaũũ]-gaa 7̃
tomorrow come- I
‘I will come tomorrow.’6
(6) a. anjum-ko  dii nuur-ne
Anjum() book() gave Noor()
‘It was a book that Noor gave Anjum.’
b. 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 proper-
ties exhibited by leftward scrambling — neither are rightward scrambled DPs
interpreted as specific, nor can they serve as potential antecedents for corefer-
ence. 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 crosslin-
guistic 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 verb-
agreement 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 Hindi-
Urdu, 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) a. -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?’
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. uskiii biiwii kis??i-ko bfhft maartii h7
his wife who- a lot hits .
‘Whoi does his??i wife hit a lot (habitually)?’
b. apniii b7h7n-ne mohan*i-ko maaraa
self’s sister- Mohan- hit
‘Selfi’s sister hit Mohan*i.’
c. uskiii-hii biiwii kis??i-ko bfhft maartii h7
his- wife who- a lot hits .
‘Whoi did his??i wife hit a lot (habitually)?’
d. 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 order-
ings 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 XP-
adjunction 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 Case-
driven 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 well-
articulated 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 construc-
tions like PP-extraposition, Directional/Locative Inversion, etc. Moreover, even
positional focus languages may use a heterogeneity of focus-marking mecha-
nisms. 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 distribu-
tion 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 non-
neutral 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 — position-
al, 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 adjacen-
cy 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 discourse-
related, 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 investiga-

tion 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 adjunc-
tion 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 PF-
component 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 PF-
component 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 adjunc-
tion. 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 LF-
interpretability 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 construc-
tions 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 constitu-
ent 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 hypotheti-
cal 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-specifica-
tions 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. kitaab  laayegaa, (siitaa nahii)
book Ram bring- Sita not
b.  kitaab laayegaa, (siitaa nahii)
Ram book bring- Sita not
c. -hii kitaab laayegaa, (siitaa nahii)
Ram- book bring- Sita not
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 Mor-
phology. 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, Case-
assigning 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 verb-
raising 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 require-
ment, 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. UG Requirement on Morphology

Features must be licensed.
b. 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. …X [YP NP[overt] [Y′ …Y…
b. …X [YP trace [Y′ …Y…
c. …X [YP adverb [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 inter-
venes. 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. -hii kitaab laayegaa, (siitaa nahii)
Ram- book bring- Sita not
b.  kitaab laayegaa, (siitaa nahii)
Ram book bring- Sita not
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 scram-
bling, though not preferred. What appears to be happening is that the presup-
positionality 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 conver-
gence, and hence the preference for leftward scrambling amongst native speakers.
The other problem has to do with the oft-claimed “LF-effects” of scram-
bling, 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 Hindi-
Urdu 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 language-
particular 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. Saku aa b6naa kajluwaan
Saku  cook tuwo
‘Saku will cook tuwo.’
b. 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 Spell-
Out 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. gafa-n KE viiridgwar6n

caught who giant-rat
‘Who caught a giant-rat?’
b. [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 wh-
phrase. 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 (right-
ward) 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. wad Billiri ] dooji?
go Billiri who tomorrow?
‘Who will go to Billiri tomorrow?’
b. tui worom mono 
ate beans my she
‘SHE ate my beans.’
c. [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 verb-
raising 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 wh-
questions 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. [CP [Co will-Q [IP you read what]]]
b. [CP what [Co Q [IP you will read]]]
c. [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 presuppo-
sition-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. PF-
movement 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 left-
adjacent 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. Into the forest nude ran  
b. In front of her sat  
Culicover and Rochemont (1990) analyze these constructions as involving VP-
topicalization, 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 configu-
ration. 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: I thought you realized that Mary had a husband
B: 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
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 Hindi-
Urdu 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) m7̃-ne kitaabi - ti dii (Hindi-Urdu)
I() book() Sita() gave
‘I gave the book to SITA.’
(ii) kitaabi m7̃-ne - ti dii
book() I() Sita-() gave
‘I gave the book to SITA.’
(iii) john oru pustakami -6 ti ko2uttu (Malayalam)
John() a book() Sita() gave
‘John gave the book to SITA.’
(iv) oru pustakami john -6 ti ko2uttu
a book() John() Sita() gave
‘John gave the book to SITA.’
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 long-
scrambling 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 non-
finite 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 long-
scrambled, the matrix subject is interpreted as the focus:
(i) kitaabi m7̃-nẽ [- 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 m7̃
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) John only likes 
= It’s only MARY that John likes
(ii) 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
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 left-
adjacent to the verb (Josef Bayer, personal communication):
(i) a. wer hat bei den Hausafgaben  geholfen
who has with the homework who helped
b. ?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 wh-
phrase.
(ii) a. wer hat  nach Hause gefahren
who has whom to home driven
b. *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 characteriza-
tion 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 PF-
component. 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 verb-
raising 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 PF-
features 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. Schoor-
lemmer (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.
an. Dordrecht: Foris.
Jayaseelan, K.A. 1989, 1995. “Question-Word Movement to Focus and Scram-
bling in Malayalam”. Ms., Central Institute of English & Foreign Languag-
es, 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.
dic”. 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 Work-
shop 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 morpholo-
gy, 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 Clech-
Darbon 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 pseudo-
cleft 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. Góor gii di sunu njiit
man this be our chief
‘This man is our chief
b. Góor gii, sunu njiit la
man this our chief .
‘This man is our chief’5
c. 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 incorpo-
rates 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 Predi-
cate-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 complemen-
tary 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’). More-
over, 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 Morpholo-
gy (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 phi-
features (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 comple-
ment 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 corre-
sponding 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 255
(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’
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 predicat-
ed 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 straightfor-
ward, 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 consid-
ered 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 pro-
nounced. 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 articulato-
ry-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 accord-
ing 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 Webel-
huth 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 constitu-
ents, 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 charac-
terized 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 bound-
aries. Another possible example are the West Flemish ‘inflected’ complemen-
tizers 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 aggluti-
nates — 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 obligato-
ry — 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-Morpholo-
gy 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 operator-
determiner /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 pronomi-
nal 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”.
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. Cam-
bridge: 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: Impri-
merie de la Mission.
Labov, W. 1972. Language in the Inner City: Studies in the Black English Verna-
cular. Philadelphia: University of Pennsylvania Press.
Léard, J.-M. 1986. “ ‘Il y a… qui’ et ‘C’est… qui’: la syntaxe comme compa-
tibilité 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 fran-
çais”. 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’em-
phase”. 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”.
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:
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 computa-
tion 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 fundamen-
tal 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 German-
ic. 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. ínam á-yimi
boy F-came
‘The boy came’
b. 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. nínkii wúu/ wáa yimid
man-. +/  came
‘The man (he) came’
b. 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 waa-
clause. 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
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. arríntan Cáli baa (shálay) iigá warramay
problem-.+ Ali F yesterday me+to-about reported
‘Ali told me about this problem (yesterday)’
b. aníga ayuu (Cáli) iigá warramay
me(strong) F+ (Ali) me+to-about reported
(arríntan)
problem-.+
‘Ali told me about it/this problem’
c. arríntan buu (Cáli) iigá warramay
problem-.+ F+ (Ali) me+to-about reported
‘Ali told me about this problem’
d. (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 so-
called ‘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).
FOCUS IN SOMALI 279
– 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 6
1S aan i kay
2S aad ku ú kaa
3S ay(F), uu(s)(M) la Ø kú Ø …
1P aan(u)(incl),aynu(excl) na, ina ká keen, kayo
2P aad idin lá kiin
3P ay Ø Ø
1. Weak subject pronouns (enclitics)
2. Weak impersonal pronoun (proclitic)
3. Weak object pronouns (proclitics)
4. Prepositional preverbs (lexically accented)
5. Weak oblique pronouns
6. (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
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) (CP)
(tops) CP (tops) CP
− nom]
[+ − nom]
[+
C IP (Spec) C′
waa [+acc]
C (IP)
baa
(tops) IP
− nom]
[+
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-.+
FOCUS IN SOMALI 281
b. [cp (Cáli) [cp arríntan b[ip uu (Cáli) iigá

[ Ali problem-.+ . (A.) me+to-about
warramay]]]
told
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 pronom-
inal agreement language: clitics are not inflectional morphemes or incorporated
pronouns. Subject agreement on the Verb does not render superfluous a pronomi-
nal 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 phonologi-
cal 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 catego-
ries 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 Hajičova and Sgall’s current proposals, then we would
expect the distribution of focus markers to directly reflect the binary given/new
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
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 languag-
es, 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 promi-
nence in the informational sense. Neither topic nor focus positions in Somali
have a fixed discourse function, as is shown by many examples from conversa-
tion or texts:
FOCUS IN SOMALI 283
(9) a. 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 mea-
sured their powers of evil’
b. 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. árdaygan baa wuxuu dóonayaa ínuu
student-.+  ++ wants +
arkó warqáddiisa
see- note-...
‘This student wants to see his notes’
b. 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
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. (i dháaf), wax báan akhrínayaa
( let thing + am-reading
‘(let me alone) I am reading’
b. (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. (sáaka) muxúu sameeyay? — Búug buu akhriyay
(today what++ did — book + read
‘What did he do today? — He read a book’
b. (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 com-
monly 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 (left-
or 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. Focus markers are restricted to root clauses, and are obligatory
there17
B. Focus markers are excluded in complement and relative clauses
C. There is only one focus position per clause.18
These properties are clearly syntactic in nature. Focus, as a pragmatically-
construed 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
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 so-
called 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 non-
verbal 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. (nínku) waa macállin/waa hormúudka kulliyádda
(man-the[+nom]  teacher/ dean-. faculty-.
‘This man is a teacher/is the dean of the faculty’
b. (nínku) ma macállin baa?
(man-the[+nom]  teacher 
‘Is this man a teacher?’
c. (nínku) macállin weeye /*wúu yahay
(man-the[+] teacher  + is()
‘This man is a teacher’
d. waa [cp ínuu keeno] (búuggan)
 + bring (book-.+)
‘He must bring it/this book’
FOCUS IN SOMALI 287
e. ma [cp ínuu keenó] baa?

 + 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:
(16) A. There can be multiple topics in a clause. There is an unique

structural focus position.
B. Topics are optional and typically unordered. The focus position
(left-adjacent to baa) is fixed, and must be filled.
C. Topics are possible in matrix clauses and other clause types
(complement, adjunct, relative clauses). Focus is limited to
main declarative clauses.
D. 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.
E. Topics are independent prosodic units at the level of phrasal
phonology. Baa is phrased together with its specifier.
F. 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 rela-
tions (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 disloca-
tions 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 construc-
tions 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. idínka bay idin aragtay
you(str.) +  saw()
‘She saw you (pl.)’
b. 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. Warsamé baa hooyádiis — jeceshahay
W.  mother-.+. pro loves()
‘Warsame hisi mother loves (himi)’
b. Adíga ayaa hooyádaa ku jecéshahay
You  mother-.+.  loves()
‘Youi youri mother loves youi′
This fact is naturally explained, if we assume that it stems from the non-
argumental 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 291
b. 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. Quanti-
fied 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-referen-
tial 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. dóolarku hóos buu ú dhacay
dollar-.[+nom] bottom + to fall
‘The dollar has fallen’
b. hádalkan áad buu ú badányahay
talk-.+ ‘muchness’ + to is-long
‘This talk is very long’
c. 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.
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 preposi-
tional preverb (22c), or a true adjunct (22d):
(22) a. Yáa fahmay su’áashayda?
who+ understood() question-.+.
‘Who understood my question?’
b. Muxúu Biíbisiída ká maqlay?
what++ BBC-. at heard
‘What did he hear at the BBC?’
c. Sidée baad ú mala-awaashay?
way-.+ + (.) guessed
‘How did you guess (it)?’
d. 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 ques-
tioned NP and the resumptive clitics is not sensitive to island constraints (Saeed
1984; Lecarme 1991, 1994)25:
(23) a. Yaa hooyádiis — jecéshahay?
+ mother-.+. pro loves()
‘Whoi does hisi mother loves?’
b. qoraagée ayaad jecéshahay búugga uu qoray?
writer-.+ + like book-.  wrote
‘Which author do you like the book he wrote?’
c. 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 interroga-
tive phrase moves to C, while the others remain in argument position. Multiple-
fronting 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. wáa buuggee?/buuggée waaye?
 book-.+/book-.- 
‘Which book is it?’
b. buuggée baad akhrisay?
book-.+ + read
‘Which book did you read?’
c. kiinnée ayaa akhriyay?
you(pl)+  read()
‘Which one of you read (it)?’
d. 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:
(25) a. má ogi árdey-ga/*-gée shálay yimid

 know() student-./*-.+ yesterday came()
‘I don’t know which student came yesterday’
b. wáxaan is weydíinayaa sí-da/*-dée aad ú
++  am-asking manner-./.+ 2 ()
mala-awaashay
guessed
‘I wonder how you guessed (it)’
c. 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 move-
ment 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. ardáydii[+acc] — shálay hálkaas joogtay
students-. e yesterday place-.-dem stay(R)
‘The students who were there yesterday’
b. xafíiska[+acc] uu wasíirku — kú shaqáynayo
office-.  minister-. pro in is-working
‘The office in which the minister is working’
c. 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 compati-
ble 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 intermedi-
ate 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. The constituent in [Spec,CP] must be [nominal]: thus only
verbs and adjectives are excluded in focus position.
B. The ‘focused’ constituent must be an entire DP (or CP)29
projection.
C. Both Spec and C must be simultaneously occupied, triggering
‘doubly-filled Comp’ effects (contrary to operator-like ele-
ments, by economy): following Platzack (1984) among others,
only lexically filled heads can assign Case.
D. Spec,CP can host a pure expletive, which associates with a
‘postverbal focus’ position.
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 construc-
tions 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. The Verb can be intransitive (29a) or transitive (29b)
B. The postposed DP can be indefinite (29a) or definite (29b,c,d)
C. The subject inversion process is free, although Somali is not a
null-subject language (29a,b,d)
D. Extraposition is not limited to the subject position (29c,e)
E. Expletive constructions can be negative (29d)
(29) a. Waxaa kúu timid gábar
.+ +for came() girl[+]
‘A girl came for you’
b. Waxaa nóo sheegay wárkaas Warsáme
.+ +to said new-.+ W[+]
‘Warsame told us this new’
c. Af Soomáaligu wuxuu ká mid yahay
language(m) Somali-. ++ among one is
afáfka Kushitikáda
languages-. Cushitic-.[+]
‘Somali is a Cushitic language’
d. Waxaan búug keenín árdaygan
.++ book brought student-.+[+acc]
‘This student didn’t bring any book’
FOCUS IN SOMALI 297
e. 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’ constitu-
ents, 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. Exple-
tive 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 struc-
ture32, 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′ (Spec)
NP[+acc]
C IP
waxi aa
...ei ...
(pro/clit.)
The analysis I am proposing is a simple extension of the theory of exple-

tives 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
from its category [N]. Interestingly, while there is inherently V-related, as

suggests its semilocative character, wax ‘thing’ is intrinsically nominal. Assum-
ing 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. wax búu akhriyay
thing() + read
‘He read (lit.: he read thing)’
b. 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 so-
called ‘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. ardáyda baa — garán
students-.[+acc]  — understand()
‘The students are understanding your question’
su’áashaada
question-.+.
b. ardaydée ayaa — joogta?
students-.+  — is-there()()
‘Which students are there?’
FOCUS IN SOMALI 299
c. 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 C′
[+acc]
C IP
baa
ei I′
I VP
V+[T+AgrS]
garán
The focalization of a local subject in (32) is a potential A-movement configura-

tion, assuming the subject has raised from a caseless position. But this explana-
tion is not available for non-local subjects in focus position. The only possibility,
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 wh-
feature, 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] specifica-
tion, and the CP and IP systems remain morphologically unrelated. Free func-
tional 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 A-
binder 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 q-
criterion 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 expres-
sions (bare Nouns) are legitimate in focus position. Clitics are generally under-
stood 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 character-
ized 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 ‘nonconfiguration-
ality’, 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.
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
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 waxaa-
and 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 construc-
tions 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 f-
features, 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 Configuration-
ality. 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.
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. Cam-
bridge: 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 Lan-
guages”. 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 Scandi-
navian Syntax”. Working Papers in Scandinavian Syntax 42.25–43.
Hoop, Helen de. 1989. “Case Assignment and Generalized Quantifiers”. Pro-
ceedings 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. Dor-
drecht: 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.
an. 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:
Kiss, Katalin É. 1995. “Multiple Topic, one Focus?”. Manuscript, Linguistic
Institute of the Academy of Science. (Talk presented at the 18th GLOW
Colloquium.)
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”. Linguis-
tique 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, Univer-
sity 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 Linguis-
tics 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 Seman-
tics 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”. Afrikanis-
tische 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., Universi-
ty 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 immedi-
ately 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. Jonek eskutitza irakurri du (SOV)
Jon letter read has
‘Jon has read the letter’
b. Jonek irakurri du eskutitza (SVO)
Jon read has letter
c. 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 Hindi-
Urdu, 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 wh-
words 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. Nork irakurri du eskutitza?
who read  letter
‘Who has read the letter?’
b. 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
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. 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’
b. MIKELEK uste dut [t idatzi du-ela eskutitza]
Mikel think  write -that letter
‘It is Mikel that I think has written the letter’
(7) 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. Mu-ipan o YANIS ti agorase
Me-told the- Yanis what bought
‘They told me what YANIS bought’
b. 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?’
(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 constitu-
ents that include them:
FOCUS IN BASQUE 317
(16) [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
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 particular-
ly 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 follow-
ing 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. Whose book did you read?
b. Which author’s book did you read?
*The book by which author did you read?
(28) a. Few students knew any answer
b. Few students’ parents met any professor
*The parents of few students met any professor
(29) a. The mother of everyone saw him
b. Everyone’s mother saw him
c. 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):
(30) CP1
CP2 C¢
JONEK 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
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 move-
ment. 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) a. CP b. CP
[wh]
[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
FOCUS IN BASQUE 321
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) a. CP b. CP
[wh]
[wh] C¢
* C¢
C
[wh] 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.
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) a. XP b. XP
YP X¢ YP X¢
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. Jonek uste du [Mikelek eskutitza idatzi du-ela]
Jon think  Mikel letter write -that
‘Jon thinks that Mikel wrote the letter’
b. 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. Zer irakurri du Jonek?
what read  Jon
‘What did Jon read?’
b. 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 ‘partici-
ple + 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
 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 specifier-
head agreement with the feature-bearing operator will satisfy the Operator
Criterion. All of this is represented in (40):
(40) CP
Q C¢
!
Wh-criterion C IP
V/I
t
VP
lexicalization
t
We would then derive the desired adjacency and leftward movement of foci, wh-
words 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 Aux-
to-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. Zer irakurri du Jonek? (Common)
what read has John
‘What has Jon read?’
b. LIBURUA irakurri du Jonek.
book read has John
‘Jon has read the BOOK’
(42) a. Zer du Jonek irakurri? (Northern)
what has Jon read
b. 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
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 obligato-
ry in any way:
(46) a. Jonek liburua irakurri du
Jon book read 
‘Jon has read the book’
b. 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):
(47) CP
C¢
SP C
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. Ba-daki Jonek egia
Ba-knows Jon truth
‘Jon does know the truth’
b. 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. Jonek liburua irakurri du
Jon book read 
‘Jon has read the book’
b. Irakurri du Jonek liburua
read  Jon book
c. 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):
(50) a. Ba-daramazu dirurik?

ba-carry money
‘Do you carry any money with you?’
b. Ba-daki Jonek egia?
ba-knows Jon truth
‘Does Jon know the truth?’
(51) a. Ba-da ogia (cf. Ogia da ‘It is bread’)
ba-is bread
‘There is bread’
b. 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 synthet-
ic 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 forma-
tion, 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):
(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):
(54) IP
NegP I
VP Neg
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 subse-
quent 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 331
b. 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 project-
ing 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, discourse-
oriented 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: :nominat-
ive, :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.
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:
(i) DP
CP D¢
JONEK 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 Pied-
Piping”. 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.
dic”. 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, Moroc-
can 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 pre-
PRON 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 JAMAL OUHALLA
1. 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 Existen-
tial 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 obliterat-
ed 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. ‘allaf-at Zaynab-u RIWAAYAT-an. (Standard Arabic)
wrote-she Zaynab- novel-
‘Zaynab wrote a NOVEL.’
b. RIWAAYAT-AN ‘allaf-at Zaynab-u.
novel- wrote-she Zaynab-
‘It was a NOVEL that Zaynab wrote.’
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 informa-
tion). This can be seen more clearly in examples such as (2a–b) which include
a negative continuation:
(2) a. 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).’
b. 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 Hungari-
an 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. Nadia shr-at KTAB (mashi majalla). (Moroccan Arabic)
Nadia bought-she book (not magazine)
‘Nadia bought a BOOK (not a magazine).’
b. 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
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 computa-
tional 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 wh-
questions 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. 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)
b. 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 f-
phrases 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, wh-
phrases 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 f-
phrase. 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
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 associat-
ed 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. 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)
b. 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 f-
phrase. 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 f-
phrase 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
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 indefi-
nite 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 comple-
ment 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 f-
phrase 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 f-
phrase to the position of the expletive subject, a process of ‘expletive-replace-
ment’ (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. 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.’
b. 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-in-
situ 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
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 wh-
pronoun 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 wh-
pronouns, 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 wh-
pronoun 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 wh-
pronoun 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 focus-
in-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, howev-
er, 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 f-
phrase 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. 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.’
b. 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 equa-
tions, 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 representa-
tion of the form: y[Fiona bought y] = ‘that dictionary’. The analysis extends
naturally to Standard Arabic clefts, where the free relative clearly has a referen-
tial 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. Nadia hiyya l-ra‘isa.
Nadia .she the-president
‘Nadia is the president.’
b. l-ra‘isa hiyya Nadia.
the-president .she Nadia
‘The president is Nadia.’
(9) 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. lli qult-lk huwwa li kayn.
.the told-I-you .it .the is
‘What I told you is what there is.’
b. 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. 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 languag-
es 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 distribu-
tion 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 (Amharic)
-burned-the cloth
‘the cloth that got burned’
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. l-ktab lli shrit-i (Moroccan Arabic)
the-book .the bought-you
‘the book that you bought’
b. l-bnt lli shft-i
the-girl .the saw-you
‘the girl that you saw’
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 who-
ever. 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,
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 quanti-
ficational 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 wh-
phrases, in terms of choice functions to focus certain crucial differences between
existential noun phrases and f-phrases must be taken into consideration. F-
phrases 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
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’ relation-
ship with the relative clause. The relationship enables the relative clause to c-
command 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-command-
ing 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, ex-
plained 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 collective-
ly 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 proper-
ties 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 languag-
es 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. Nadia HIYYA lli ‘llf-at l-ktab.
Nadia .she .the wrote-she the-book
‘It’s NADIA who wrote the book.’
b. 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.
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. KITAAB-an ‘llaf-at Zaynab (Standard Arabic)
book- wrote-she Zaynab
‘It’s a BOOK that Zaynab wrote.’
b. Zinb ‘llf-at l-KTAB. (Moroccan Arabic)
Zinb wrote the-book
‘Zinb wrote the BOOK.’
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 morpho-
logical 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-of-
focus’ 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 JAMAL OUHALLA
6. Conclusion
The interpretive mechanism commonly assumed for focus-preposing and focus-

in-situ sentences assumes that they involve Existential Closure over the presup-
position 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 sentenc-
es, 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.
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. Cam-
bridge: 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”.
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 Rela-
tives: 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.
an. Dordrecht: Foris.
Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cam-
bridge, MA: MIT Press.
Jacobson, Pauline. 1988. “On the Quantificational Force of English Free Rela-
tives”. 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.
Ouhalla, Jamal. 1991. Functional Categories and Parametric Variation. London:

Routledge.
Ouhalla, Jamal. 1994a. “The Syntactic Representation of Arguments and Identifi-
cation”. 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”.
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 Linguis-
tic 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). Cam-
bridge: CUP.
Subject Index
adjacency 229–30, 234–5 Extraposition approach 105, 257

adjectives: formal semantic account 106–8
incorporation 203–4 interrogative clefts 98–9
triggers for f-structure 127 intonational copy 99–101
adjunction 193–200, 196, 200–2, 223, pragmatics of French clefts 102–4
225–6, 236–7 prosodic data 95–101
adverbs: pseudo-cleft 341–6
sentential vs. VP 188–93 raising 90
verbal status in Portuguese 49 semantics of French clefts 102–4
VP adverbs 196–9 small-clause analysis 90–1
Aghem 164, 224 traditional hypotheses 88–94
agreement: typology 84–8
intonational agreement between clitics:
focussed phrase and coda Italian 47
99–100 Portuguese 39
noun-class markers 249 Complete Functional Complex 56–7
Spec-Head 63 Complex-NP Constraint 64
also 60 copular sentences 247ff., 336, 346–51
augment morpheme 169–71, 174 cumulative interpretations 184
Avoid Pronoun Principle, The 47 Cushitic languages 276ff., 302
Basque 110, 219 determiners:
Lekeitio Basque 111 out-of-focus determiner 164–5
northern vs. southern dialects 325, strong 132–3
328 weak 131
BE-focus structures 37–40 Dinka 146
Bemba 158–9 discourse configurationality 281–5
domain discourse 227
Cewa 156–7
Dutch 301
cleft sentences 182, 246–67, 341–6,
349ff. each 139
declarative clefts 95–8 Empty Category Principle 64
362 SUBJECT INDEX
enclisis 41 Focus-Tense Identification 40

English 223, 347 interaction with other areas of
evaluative construction 42–3 grammar 165–73
even 60, 61, 64–5, 69, 77, 240 lowering 66–8
existentional noun phrases 351ff. minimalist theory of 224–35
expletive constructions 296–8 narrow focus 215, 217–8, 242
negative focusing 255, 262–3
file semantics 123–30, 146
neutral focus 215
focal stress-assignment 34
object foucusing 252–3, 256–60
focus:
partial focus 28
assertive focus 153–6
Pied-piping 316–21
bare phrase-structure implementation
presentational focus 38; Portuguese
68–9
25ff.; Hindi-Urdu 239
bound focus 59–66, 240; base-
prosodic focus 46
generation vs. movement 55ff.,
pure focus 33
66
quantificational focusing particles 55
broad event-related focus 85
quantified focus elements 30, 49
broad presentational focus 85
relation with particle 55–6
construction types 56–9
restrictive focus 127–8
contrastive focus 240, 242, 340–1;
scope: German 143–5; Basque 314–6
Arabic 337–8;
structural focus and Case assignment
exclusive/restrictive; vs. new
positions 295–300
information 340;Portuguese
subject focusing 253–5, 260–2
25ff.; value 26; Wolof 246ff.
term focus vs. predicate focus 277
cross-linguistic distribution 219
three strategies in French 104
derived predicate 180ff.
verum focus 148
exclamatory comment 86
wh-words 181–3, 187–8, 194–5,
focus clitics 202–3
204–5, 220–2, 228, 236,
focus effects 282–5
292–5, 312–24
focus movement vs. association with
word order of focus elements 200–5
focus 64–6
(see also tone; negation)
Focus Phrase 331, 337
free inversion 31
focus positions 215–24; 287–300
fusion 264
focus preposing and focus in situ
337–41, 355, 356 Gallician 30
Focus Structure (f-structure) 120ff., Generative tradition 3–13
146–7; f-structure rules 121–4; German 57–9, 61, 64, 71, 72–3, 77–8,
subordinate f-structure 125–7, 143–4, 207, 241–1
138 (see also scope) Greek 79, 314
focus types 13–7
focus-like elements 33
SUBJECT INDEX 363
Hindi-Urdu 201, 217, 220, 221, 222, percolation 320–1

223, 227, 230–3, 239, 240–1 predicational sentences 247
Hungarian 147, 219, 301 preverbs 303
Principle of Exclusion 190
identificational sentences 247
Principle of Lexical Association 60
interpretability 226–9, 241
Principle of Full Interpretation 63
interrogative 158, 292–5; 312–3, 326,
pronouns 129, 146
339
proto-Bantu 170
Italian 32, 35, 47, 74
proximity-to-the-verb requirement 219
ki-Rundi 159, 160 pure topics 33
Kihung’an 165
quantification:
Kirundi see ki-Rundi
adverbial quantifiers 290–1
la-Haya 153–6, 157, 160–3 focus positions 289–95
left-dislocation 155–6 in overt structure 185–8
lu-Ganda 165–73 over Events 186ff.
subjects 189–93
Malayalam 111, 216–7, 227, 230–3,
wh-quantifier as existential 187–8,
239, 240–1
193–4
Minimalist Program 1–2, 56ff., 68
Quantifier Raising 133, 180, 181–4, 190,
minor functional heads 62–3
221–2
Montague Grammar 62
R-dependencies 128–30
negation 168, 171, 326, 327–30
relative clause 191–2, 294, 298–300,
negative imperative 165
350
tonal reduction 160
free relatives 336–7, 346, 348–51,
non-configurationality 301
354
non-restrictive/non-exclusive focus 27ff.
relativization 191–2, 199–200
null operator 31
Relativized Minimality 196
nur 58, 61
Rendille 277
only 56–9, 69, 71, 77, 240
scope:
Parameter Value Preservation measure ambiguous scope 193–5, 196–9
232 and focus in Basque 314–6
particle: cardinal scope 134–7
additive particles 60 f-structure 141–5
attachment to Xmax 60–2 focus in German 143–5
focus-constituent requirement 59–60 Nuclear Scope and rheme 139–40
functional head 62–4 propositional scope 57, 66
Move-F 75–6 quantifier scope 133–45, 147, 183ff.
wide-scope derivation 70–6 scoped stages 141–2
particle phrase 56ff. some–every 137–41, 147
364 SUBJECT INDEX
scrambling 61, 182–3, 218–9, 222, topic:

231–2, 235, 239 scope properties 130–3
se-Tswana 110, 157 Stage Topic 124–5
sogar 58, 61 topic positions 287–8
Spanish 322, 323 Topic vs. Focus: historical overview
syntax-PF interface 263–5 24–5
topic-like elements vs. true topics
Tangale 216, 220, 222, 233–5, 242
32–3
tense:
TopicFocusP: EvaluativeP 41–4;
conjoint vs. disjoint 159, 174, 175
FocusP 29–35
three-place predicates:
Topic Constraint on R-dependencies 129
Portuguese 36–7
topicalization 250
tone 152–73, 242
Tswana see se-Tswana
tonal finality 153
tonal marking of [+focus] 152–9 update semantics 146
tonal marking of [–focus] 160–4
verb raising 235ff.
tonal reduction 160–1
Verb-Second Constraint 58
tonally active 153
too 60 weak crossover 221, 222, 289, 292
Western Bade 216, 222, 233–5
List of Contributors
Manuela Ambar Larry Hyman

Departamento de linguistica Department of Linguistics
Facul. des letras University of California
Universidade de Lisboa 2337 Dwinelle Hall
Cidade Univ. Berkeley, CA 94720 USA
1699 Lisboa codex PORTUGAL hyman@socrates.berkeley.edu
mmfa@fc.ul.pt
Sarah D. Kennelly
Josef Bayer Via Campo Rosso, 4
Institut für Germanistische Sprach- 52010 Ortignano-Raggiolo (AR)
wissenschaft ITALY
Friedrich-Schiller-Universitaet Jena oddnend@elledi.it
Ernst-Abbe-Platz 4
Ayesha Kidwai
07743 Jena GERMANY
Centre of Linguistics & English
josef.bayer@rz.uni-jena.de
School of Languages
Nomi Erteschik-Shir Jawaharlal Nehru University
Ben-Gurion University of the New Delhi — 110067 INDIA
Negev ayesha@jnuniv.ernet.in
Deptment of Foreign Literatures &
Alain Kihm
Linguistics
CNRS-LACITO
PO Box 653
44 rue de l’Amiral-Mouchez
Beer Sheva 84105 ISRAEL
75014 Paris FRANCE
shir@bgumail.bug.ac.il
kihm@vjf.cnrs.fr
366 LIST OF CONTRIBUTORS
Jacqueline Lecarme Anne Clech-Darbon & Georges

2LC-CNRS Rebuschi
1361 route des Lucioles TYGRe (SYLED), Université de
06560 Sophia Antipolis FRANCE Paris III
jl@cra.cnrs.fr 19 rue des Bernardins
75005 Paris FRANCE
Jon Ortiz de Urbina
clechdarbon@ idf.ext.jussieu.fr
Universidad de Deusto
rebuschi@idf.ext.jussieu.fr
Departamento de Filologia Inglesa
Apartado 1 Annie Rialland
48080 Bilbao, SPAIN CNRS
urbina@orion. deusto.es Unité de recherche en phonétique
et phonologie (UPRESA 7018,
Jamal Ouhalla
CNRS-Sorbonne-Nouvele)
Department of Linguistics
19 rue des Bernadins
School of Modern Languages
75005 Paris
Queen Mary and Westfield College
rialland@idf.ext.jussieu.fr
Mild End Road London E1 4NS GB
j.ouhalla@qmw.ac.uk Laurie Tuller
Département de linguistique
Université de Tours
3, rue des Tanneurs
37000 Tours FRANCE
tuller@univ-tours.fr
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 Barri-
ers. 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 Swed-
ish. 1998.
20. LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Ad-
verbs, 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 Move-
ment 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, Morphol-
ogy 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.

Rebuschi Tuller 1999 The Grammar of Focus

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rebuschi Tuller 1999 The Grammar of Focus

Uploaded by

Copyright:

Available Formats

THE GRAMMAR OF FOCUS

Advisory Editorial Board

JOHN BENJAMINS PUBLISHING COMPANY

American National Standard for Information Sciences — Permanence of Paper

Library of Congress Cataloging-in-Publication Data

The Grammar of Focus: An Introduction 1

Focus and Arabic Clefts 335

Georges Rebuschi Laurice Tuller

“There is no reason to suppose that a satisfactory characterization of focus and

number of levels of syntactic representation is reduced to this minimal number

2. Overview of the Evolution of the Study of Focus in the Generative

2.1 If we momentarily accept to define ‘focus’ as ‘prosodic salience at the

2.2 Fifteen years later, the question of the – articulation

2.3 A slightly different (and technically somewhat more sophisticated) approach

2.4 In ‘Conditions on Rules of Grammar” (C 1976), a “rule ” is intro-

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

Radically distinguishing between contrastive and non-contrastive focus and

3. Focus Types and Issues

As the preceding discussion of the evolution of the study of focus in generative

Hindi and Malayam (Kidwai), or to the  of V, as in lu-Ganda and lu-Haya

See Kennelly’s contribution (this volume) for similar data in Turkish.

Chomsky, N. 1977. “On wh-Movement”. In P. Culicover, T. Wasow & A.

Lasnik, H. and M. Saito. 1992. Move-a:: Conditions on its Application and

This paper is a reduced version of a larger work on focus in Portuguese.

2. Topic and Focus — different views

3. Focus in Portuguese — the data

Taking the way information is introduced in the Universe of Discourse as the

appear in their canonical order:

B. Presentational focus — restrictive

question would be What happened? But it can also be an answer to questions

d. *A tarte comeu a Maria

4. The proposal — FocusP vs. TopicFocusP

(15) a. ISTO fazem os reis quando viajam

For concreteness, suppose we assume the following representation, where F

Let us explore the hypothesis outlined in representation (19). A first

(21) O teu vestido, a Joana vestiu-o

Note that the gap in object position in (23a) cannot be bound by an

5.1 Three-place predicates

object — would occupy their respective positions inside IP.

5.2 BE-Focus structures

(28) a. A tarte … a Maria comeu (o que mais comeu não sei…)

We have suggested that to focalize a constituent means to identify an Event

6. TopicFocusP vs. EvaluativeP

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

c. Muitos livros lhe ofereceu o Pedro!

also why they become emphatic elements.

the case of  . If a constituent is marked [topicf] it will be able

Ambar, Manuela. 1988. Para uma Sintaxe da Inversão Sujeito-Verbo em Portu-

Casteleiro, João. 1981. Sintaxe Transformacional do Adjectivo. Lisbon: INIC.

Raposo, Eduardo. 1997. “Some Definite/Zero Alternations in Portuguese:

By the notion “bound focus”, we mean phonological prominence and semantic

is able to take (clausal) scope; (iii) by reconstruction of a phrase such as only

2. Four Types of Constructions

3. Characteristics of bound focus

In this section, I want to discuss characteristics of bound focus, some of which

3.1 PRT requires a focus constituent

3.2 PRT attaches only to a potential Xmax

(12) a. Some students smoke even IN the classroom

interpretation can be derived which has PRT in proper (propositional) scope

3.3 PRT is a “minor” functional head

In which sense is PRT a regular syntactic (lexical or functional) head? Notice

(21) PRT being a M F H cannot project syntactic

(25) is semantically equivalent to the base-generated structure in (1c) and can be