Professional Documents
Culture Documents
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
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
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
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.
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
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
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.)
2.5 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
8 GEORGES REBUSCHI AND LAURICE TULLER
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) Ji [[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]
THE GRAMMAR OF FOCUS 9
2.7 The Government and Binding research program was in full swing in the
mid-eighties, with 1986 a landmark. As far as focus is concerned, we must at
least mention the publication (i) of Horvath’s revised version of her 1981
dissertation, in which the term “focus” is now highlighted, (ii) of Abraham and
de Meij’s collection, where the “middle field” of both German and Hungarian is
investigated in great detail, (iii) of Rochemont’s book, an attempt at giving a
semantically unified, but syntactically working, definition of focus as new
information, and (iv) of Selkirk’s study of intonation. The latter takes up
Jackendoff’s hypothesis of an information or focus structure distinct from s–s
and LF, strongly argues in favor of a radical distinction between grammatically
determined stress and the assignment of a to focused items, and
proposes a concomitant a “Phrasal Focus Rule” which allows the percolation of
the focus property of a focused head to either its phrasal projection or to an
internal of that head; Selkirk also proposes a principle establishing a
clear divide between linguistic and metalinguistic use of pitch prominence when
the pertinent unit is the syllable rather than the word (a problem noted, but not
solved, in C 1955 — recall Section 2.1):
“Perhaps the generalization is that pitch accents can be assigned to anything of
level word or below, but that a pitch-accent-bearing element is only interpreted
along the lines of a normal focused constituent when it has an identifiable
separate meaning. When the pitch-accent-bearing element cannot be interpreted
in this way, the presence of pitch accent is interpreted instead in metalinguistic
term.” (p. 271)
10 GEORGES REBUSCHI AND LAURICE TULLER
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
THE GRAMMAR OF FOCUS 11
is still very much alive (see recent publications such as Rizzi 1997, and
Szabolcsi 1997).
2.8 Moving now directly into the most recent developments of generative
syntax, at the beginning of Chapter 4, Chomsky (1995: 220) writes:
Notice that I am sweeping under the rug questions of considerable 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:
12 GEORGES REBUSCHI AND LAURICE TULLER
The same is true of rules that are not rules of sentence grammar at all, e.g.
VP-deletion, which, as observed by Sag and Hankamer […], can apply across
speakers and, correspondingly, is not subject to principles of sentence 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-
THE GRAMMAR OF FOCUS 13
ing of PS rules, the Projection Principle, and d-structure more generally, provides
a more simple means of introducing F (almost) anywhere in a derivation than
was possible in the Standard Theory or its immediate successors (see 2.2).
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
16 GEORGES REBUSCHI AND LAURICE TULLER
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
THE GRAMMAR OF FOCUS 17
thus appears to be consensus on what the focus issues are, which surely amounts
to progress in our understanding of this grammatical phenomenon. The 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.
18 GEORGES REBUSCHI AND LAURICE TULLER
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)
THE GRAMMAR OF FOCUS 19
14. De Rijk (1979) had reached the same conclusions concerning Basque. Note however that, as
indicated by Abraham (1997), focusing need not always be associated with V2 in languages that
have visible wh-movement, since both of the following sentences are grammatical in German:
(i) a. 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
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.
20 GEORGES REBUSCHI AND LAURICE TULLER
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
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.
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.
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
ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE 29
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
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.
ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE 35
5. Other Predictions
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
ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE 37
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
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
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
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 —
ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE 45
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
ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE 47
wh-movement. We will assume, however, that CP, very plausibly, coexists with the projection
where focus is licensed (among other projections) and that the opacity effects manifested in
focus wh- contexts have to be derived in another way.
15. The definition of topic-like element will become clearer below. As we will see, a tarte in (7),
the lexical counterpart of OP in (19), is not a true topic, but just the subject of focus. Very
plausibly languages will differ with respect to the nature of TopicFocusP. Hungarian, for
instance, seems not to allow a topic-like element in Spec,TopicFocusP (only a focus constituent
can occur there), this correlating with the impossibility of sentences like (7a)–(7b) in this
language, which, however is a null subject language (cf. Ambar 1997 and fn.2).
16. For a more detailed treatment of these aspects see the extended version of this work (Ambar
1997).
17. The parametrisation responsible for the differences concerning word order in focus 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).
ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE 49
27. If the copula appears in the position occupied by the verbal trace, it would not precede any
phonetically realized material (a condition for having focus).
28. See Ambar (1997) for detailed discussion.
29. In this case the moved elements form a constituent. Therefore the subject/object asymmetry
manifested in these structures is derived from the fact that the entire VP can be moved (as in
focused subject structures), but the subject + verb cannot, since they do not form a constituent
(as in focused object structures).
30. In fact we do not have an algorithm to evaluate the cost of a derivation.
31. Note that sentence represented in (35), where the subject is in TopicP is more natural than the
ones where TopicP is filled by the object as in (33–34), even with the clitic. Besides the
question of cost of the derivation, pro in subject position is also more natural than in object
position in Portuguese. The contrast then follows.
32. Zubizarreta also proposes that presentational focus constituents have to be under AspectP. In
our analysis, however, we do not claim that the focused element is under AspectP (or
TobjectP); it can be in different positions inside IP (presentational) or in Spec,TopicFocusP
(contrastive). The effects of (36) are derived in some environments by Tense raising at LF.
Note that condition (36) concerns full focused constituents — the ones where the 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
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.
ASPECTS OF THE SYNTAX OF FOCUS IN PORTUGUESE 53
Josef Bayer
Friedrich-Schiller-Universität Jena
Abstract
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
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.
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).
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
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
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.
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
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.
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.
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
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
(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
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
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
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.
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!’
86 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
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
ARE THERE CLEFT SENTENCES IN FRENCH? 87
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
ARE THERE CLEFT SENTENCES IN FRENCH? 89
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
ARE THERE CLEFT SENTENCES IN FRENCH? 91
might argue in favour of (25a) that a “Small Clause” analysis allows a uniform
treatment of the coda in clefts and of the relative clauses of the “third type”
which appear in the complement of verbs of perception, as illustrated in (26):
(26) a. 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
92 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
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.’
ARE THERE CLEFT SENTENCES IN FRENCH? 93
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.
ARE THERE CLEFT SENTENCES IN FRENCH? 95
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%
ARE THERE CLEFT SENTENCES IN FRENCH? 97
declarative marker of the focus. The motivation for this analysis will become
clearer when we consider interrogative sentences later on.
If we did not look at further data, the intonation contour of (40) would be
open to many other analyses. For example, the falling contour of the first group
could be considered an undecomposable marker of focus and the flat low contour
of the coda an equally undecomposable marker of postfocus. In order to motivate
our analysis in terms of tones we must compare the intonation contour of (40)
with that of other focused utterances.
Let us then consider a second very common intonation contour of focused
utterances:
(41) [C’est Jean-Pierre] [qui est sorti].
\/ | |
HL% L L% (see Figure 2)
Here, the primary stressed syllable of the focused group bears a falling tone,
which we suggest is the same as the H and L% tones of (40), both now 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:
(42) [C’est Jean-Pierre] [qui est sorti].
| /\ | |
H HL% L L% (see Figure 3)
Here, the H tones marking the focus are linked to syllables of Jean-Pierre,
forming a high pitch plateau over both syllables of the word. The final L% tone
of the first group is reduplicated as before. This intonation pattern is interpreted
as more emphatic than the preceding ones.23
The duplication of the L% declarative tone may also involve other features
98 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
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.
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
ARE THERE CLEFT SENTENCES IN FRENCH? 99
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).
ARE THERE CLEFT SENTENCES IN FRENCH? 103
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).
ARE THERE CLEFT SENTENCES IN FRENCH? 105
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
108 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
6. Conclusions
Notes
* Thanks to Ileana Comorovski, Denis Creissels, François Lonchamp and Laurie Tuller for their
remarks on an earlier draft. Needless to say, the usual disclaimers apply.
1. (8) would be acceptable if Ça were given a purely exclamative value (cp. Ça alors!), rather than
a deictic one; but such an interpretation is, of course, irrelevant.
2. See however Robert (1993) for a tentative reduction of such sentences to clefts on a purely
pragmatic basis.
3. This fact may be due to independent causes: (3b) is causal with respect to (3a), whereas (4b)
110 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
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).
112 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
23. It is more common in emphatic style, such as used in political speeches for instance.
24. See figures (5) and (6) for a clearer idea of the differences between (44) and (45).
25. In fact, the analysis of French intonation requires still other boundary contour tones, such as the
LHL pattern for the well-known implication countour; such boundary tones have been proposed
by other authors (Mertens 1987; Di Cristo 1996), using other terms.
26. For example, consider the assertion (i.a) and the question (i.b):
(i) a. [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.
ARE THERE CLEFT SENTENCES IN FRENCH? 113
29. The same is true of utterances like (4b), whose syntax and semantics are quite distinct from
those of clefts proper.
30. As the scope of seulement would be ambiguous here, we represent the prosodic prominence
which corresponds to the intended meaning; to render the data more readily comparable, we
also use small capitals to indicate prosodic prominence in the other cases, where it is 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).
114 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
42. The idea of letting the coda bind a predicate variable which is itself part of the semantic content
of a pronoun has been inspired by Cooper’s (1979) work on correlative constructions (which
also involve the adjunction of a CP to an otherwise well-formed sentence). Srivastav (1991)
suggests that correlative pronouns are variables which can be bound by the (translation of the)
correlative clause into a property containg an iota operator — or “remain free and refer
deictically” when there is no correlative CP. That would, however, entail that ce be given two
totally unrelated translations, an unnecessary move at least in the case we are dealing with here
— or even in a case like (i), under one possible analysis, whereby it is now the protasis as such
that would provide the binder of P.
(i) Si quelqu’un fait cela, est un criminel
‘If someone does that, he [lit. ] is a criminal.’
43. We also follow Jacobson in considering atomic individuals to be a subset of plural individuals.
44. See also Chomsky (1995: 170): “The notion of grammatical construction is eliminated, and, with
it, construction-particular rules. Constructions such as verb phrase, relative clause, and passive
remain only as taxonomic artifacts, collections of phenomena explained through the interaction
of the principles of UG, with the values of parameters fixed.”
References
Abraham, W.; Epstein, S. D.; Thráinsson, H. & Zwart, J.-W. (eds.). 1996.
Minimal Ideas. Amsterdam: John Benjamins.
Akmajian, A. 1970. “On Deriving Cleft Sentences from Pseudo-cleft Sentences”.
Linguistic Inquiry 1.2:149–168.
Ashby, W. 1994. “An Acoustic Profile of Right-dislocations in French”. Journal
of French Language Studies 4.2: 127–145.
Bergvall, V. 1987. Focus in Kikuyu and Universal Grammar. Doctoral 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.
ARE THERE CLEFT SENTENCES IN FRENCH? 115
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.
118 ANNE CLECH-DARBON, GEORGES REBUSCHI & ANNIE RIALLAND
Nomi Erteschik-Shir
Ben-Gurion University of the Negev
Abstract
Introduction
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
(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
122 NOMI ERTESCHIK-SHIR
assigned only to constituents for which file cards are available on top of the file,
i.e., cards which have been positioned there by the application of the f-structure
rules to previous utterances.
The constraints on f-structure (f-structure rules I and II) are pragmatic, i.e.,
context determines whether a particular f-structure can be assigned to a sentence.
Yet, f-structure determines interpretation (f-structure rule III) and therefore the
view of grammar proposed here requires a semantic theory which takes 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
FOCUS STRUCTURE AND SCOPE 123
The card index is assigned by a function which maps the set of cards onto the
set of discourse referents. The heading, however, is an attribute of the card (such
as dog). When a card is first introduced the constituent which introduces it
determines the form of this heading. In the new card for dog2, the heading allows
future definite references to this dog such as the dog. Once entries are added to
the card, any other attribute can take the place of this heading, deriving new
headings such as the dog you have (your dog) or the brown dog. Entries can in
this way be viewed as restrictions on the heading.
The entry for the second sentence in (2) can now be made by B. The
pronoun is interpretable only if it is matched with an available card from the top
of the file, i.e., the referent must have been introduced and a card for this
referent must have been positioned on top of the file. Card 2 represents such a
card since the features of its heading in match those of the pronoun. The entry
can therefore be licensed on this card. The following steps are now taken by B:
1. Select card 2 from the top of the file. (TOPIC rule)
2. Evaluate “e is brown” with respect to dog2. (PREDICATION)
3. If 2. yields TRUE, enter “e is brown” on card 2. (UPDATE)
Note that a sentence may have a variety of f-structures depending on
context. (2), for example, could be a response to questions such as
(3) a. 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
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.
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-
FOCUS STRUCTURE AND SCOPE 125
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.
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
FOCUS STRUCTURE AND SCOPE 127
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.
The focus constituent may define a of cards, the students in my class, for
example, forms such a set. Cards which contain a set of cards are called
cards, the set they define is a set and the focus which
introduces them is a focus. A subordinate f-structure can be
assigned to such a restrictive set in a partitive:
(12) [[Two]FOC-sub of [the students]TOP-sub]TOP [are intelligent]FOC
The subordinate f-structure is formed around a card on top of the file which
represents a discoursally available set. A constituent which defines a subset of
128 NOMI ERTESCHIK-SHIR
this topic set is focused triggering the partitioning of this set. The new subset
card is now available as the main topic.
In sum, Topic and Focus are defined within a theory of discourse. Both
individuals and stages provide topics. A rule of Predication is defined which
assigns truth values. This rule is viewed as a relation between the topic of a
sentence and its predicate. The truth value of a statement is determined by
assessing it as putative information about its topic. The f-structure Topic and
Focus rules apply to subordinate f-structures as well, but the rule of Predication
applies only to the main f-structure. Since the topic is the pivot for the assign-
ment of truth value, it follows that every sentence must have a topic.8
2. R-dependencies
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.
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
FOCUS STRUCTURE AND SCOPE 131
contextual restriction of quantifiers fall out automatically. So, for example, the
topic everyone must be interpreted as ‘everyone we are talking about’, it is a
context specified set represented by a card with the heading everyonei.11
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
134 NOMI ERTESCHIK-SHIR
constituent containing Q2 is assessed, i.e., Q1 has scope over Q2. If we follow the
same reasoning for (25b), we get the opposite scope relation. Two scoped readings
thus follow without further ado from applying predication to these f-structures.
In (25c) the topic is a stage topic, hence neither of the quantified NPs is a
topic and a nonscoped reading results. In the next section I show, following
Landman (in prog.), that there are actually 4 nonscoped readings, all of which
have the same focus structure. Under this view, for each case in which a
quantified NP is analyzed as the topic, there are two readings to be derived.
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 →
FOCUS STRUCTURE AND SCOPE 135
The f-structure of all the cases listed in (27) is (25c) which has a stage topic. For
each plural NP, there are two possible interpretations: collective and distributive.
(27a) gets collective readings for both NPs, i.e., the group, ‘2 girls’ arrested the
group ‘3 boys’. (27b) says that the group ‘2 girls’ arrested 3 individual boys.
(27c) involves 2 individual girls who arrest a group of 3 boys. (27d) is the
double distributive reading in which 2 individual girls arrest 3 individual boys,
i.e., there is some pairing between the 2 girls and the 3 boys such that for each
girl there is one or more boys (of the three) that she arrests, and no boy gets left
unarrested. (Since the reading leaves the number of pairings unspecified with a
minimum of three and a maximum of six, the number of ‘arrows’ which
indicates the number of ‘pairings’ has been left open. This reading includes the
‘all-all’ reading in which each of the two girls arrests each of the three boys.
This reading, according to Landman, is the borderline case of (27d).
All of these are assessed with respect to the discourse here-and-now, i.e.,
only event is involved for each reading (and only two girls and three boys
are involved in each case). A distinction is thus made between the number of
pairings, in this case the number of arrests (which ranges between 1 and 3), and
the number of events. Imagine, for example, the following spatio-temporal
parameters: Today, between 6–7 p.m., in Beer Sheva. These parameters define
the stage upon which all the arrests take place. No individual stages for each
separate pairing is made available. This is a requirement of the stage topic
reading evidenced by the presence of an overt stage topic:
(29) a. 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
136 NOMI ERTESCHIK-SHIR
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
138 NOMI ERTESCHIK-SHIR
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.
FOCUS STRUCTURE AND SCOPE 139
The story is the same for the f-structures in (37a–b) except that the ‘arresting’ is
reversed. Here again, properties of the strong quantifier force a subordinate 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.
140 NOMI ERTESCHIK-SHIR
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.
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:
142 NOMI ERTESCHIK-SHIR
(49) In every city that he visited John met someone he went to high
school with.
The ambiguity here results from two subordinate f-structures that can be assigned
to the main object:
(50) a. 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.
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-
FOCUS STRUCTURE AND SCOPE 143
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]]]]]
FOCUS STRUCTURE AND SCOPE 145
The verb has moved out of the surface focus constituent leaving a trace. If focus
is marked on s-structure, the rules that interpret f-structure will involve 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
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
FOCUS STRUCTURE AND SCOPE 147
topic properties are shared by subjects in a number of languages.” Reinhart (1981: 87) also
claims that the syntactic subject is the preferred topic for the following reason: “…it is easier
to interpret the sentence as being about its subject, than, say, about its object, since in the
logical form, something is predicated directly of the subject’s interpretation.” More recently
Lambrecht (1994) argues that the unmarked information-structure sequence is topic followed by
focus and that the subject is the unmarked topic. I make no assumptions here with respect to 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
148 NOMI ERTESCHIK-SHIR
the source of the example. For further discussion of this approach see Eckardt (1994).
21. The indicated f-structure could be the main or the subordinate one.
22. Krifka notes that (53a) from Frey (1993) uses accent on the finite verb to exclude the effect of
focus. This renders the so-called verum focus, which focuses on the truth polarity of the
sentence. In the current framework, such a (metalinguistic) accent masks the intonational
marking of the nonmetalinguistic f-structure but is not predicted to interfere with the scopal
interpretation (cf. Erteschik-Shir 1997).
References
Larry M. Hyman
University of California, Berkeley
Abstract
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.
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
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.
THE INTERACTION BETWEEN FOCUS AND TONE IN BANTU 157
Except for the application of H tone spreading in (7b), the effects of being final
in a focus phrase on a penultimate H are the same as in lu-Haya. As seen in (9a),
(9) 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:
THE INTERACTION BETWEEN FOCUS AND TONE IN BANTU 159
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),
THE INTERACTION BETWEEN FOCUS AND TONE IN BANTU 161
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).
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
THE INTERACTION BETWEEN FOCUS AND TONE IN BANTU 165
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
X Z
[........................]TG
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:
THE INTERACTION BETWEEN FOCUS AND TONE IN BANTU 169
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
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 —
THE INTERACTION BETWEEN FOCUS AND TONE IN BANTU 175
and, as we see in lu-Haya, even within the same language (cf. Mould 1979; Botne, in press;
Nurse 1996).
9. In languages that have the conjoint/disjoint distinction, relative clauses typically show the
properties of the conjoint, i.e. [−focus] form of the verb. This makes sense since relative clauses
are by definition “backgrounded” or “out of focus” with respect to the asserted proposition of
the main clause. In lu-Haya, where the conjoint/disjoint distinction has been replaced by the
morphologized [+F]/[−F] one, the negative and “marked TAM’s” are [+F] independent of the
clause type in which they are found. The one exception is the special behavior of -ka- in the
Past3.
10. The H on the augment prefix /é-/ and /ó-/ is not affected by TR in lu-Haya. Instead, its H tone
is always reduced to L after pause, as in all of the examples in (22); see note 4.
11. Although the exact realization involves other complexities, TR does apply to a verb that is
followed by a word lacking H tone (see Hyman & Byarushengo 1984).
12. It also loses its prefix tí- by a general rule of prefix deletion whenever a noun is followed by
an agreeing modifier other than a numeral (see Hyman 1979).
13. In Hyman (1985) this is accounted for by requiring a proper-government chain from the
affected determiner “slot” up to the top node of the sentence.
14. This section represents a summary of parts of several detailed studies on the interaction of tone
with syntax and with the phonological and grammatical properties of the augment in lu-Ganda.
See especially Hyman, Katamba & Walusimbi (1987), Hyman (1988), Hyman and Katamba
(1990) and Hyman and Katamba (1993b).
15. In fact, the augment reconstructs as *H in Proto-Bantu. Cf. notes 5 and 11.
16. Odden (1991) suggests that a case of direct semantic conditioning of tone exists in Kikongo,
though a syntactic interpretation appears possible here too.
References
Sarah D. Kennelly
Utrecht Institute of Linguistics OTS
Abstract
Introduction
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.
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)]
182 SARAH D. KENNELLY
In Turkish there is evidence that the LF for quantification is the overt structure.
Since it is a head final language, the verb moves to the right under head
movement to check the strong inflectional morphology. If one issue for ‘F’ is
contiguity with the verb, then in order to move to an adjoined position, the
Focused element would also have to move to the right. So it is crucial to
ascertain if verb adjacency is at stake. Cross-linguistically the Wh/Focus position
is consistently verb adjacent: Hungarian (Szabolcsi 1981: 143; Horvath 1986;
Brody 1990), Aghem (Rochemont 1986: 19), Georgian (Nash 1995) and Basque
(Rebuschi 1983; Ortiz de Urbina this vol).
According to Chomsky (1971: 199) “…the Focus is the predicate of the
dominant proposition of the deep structure”. In the following discussion it is
assumed that the verb (predicate adjective/noun/PP)11 is the natural predicate of
an assertion, and hence, in Chomsky’s terms, that it is the natural or ‘neutral’
Focus. When an element other than the verb is Focused, Chomsky doesn’t
mention how it becomes the predicate; what mechanism is at work. Nor does he
elaborate on what the status of the verb is when an argument is Focused, i.e. is
the verb then a ‘secondary’ predicate, is it relieved of its status as the predicate,
or is there a verb-Focus predicate complex? I propose that the motivation for (i)
the verb adjacency constraint and (ii) the adjoined position ‘F’ are distinct: (i)
the Focused element needs to derive the role of predicate from the verb to be
interpreted as ‘Focus’; call it ‘predicate derivation’. That is, given that the natural
predicate of a sentence is the verb, if another element is to become the predicate
it must derive that role from the verb; in Turkish the crucial factor is contiguity.
(ii) P-Focus elements are existential quantifiers which require movement to an
adjoined position for interpretation.
Consider the adjacency constraint. A sentence can be taken as a description
of an event such that the verb is not a direct predicate of the individuals denoted
by the subject but rather it is a predicate of an event while the subject is a
participant in that event (Davidson 1967; Parsons 1990). Then the arguments
have a relation to the event through their theta-roles, as exemplified in (10).
(10) ∃(e)[Lee(e,agent) & hugged(e) & Kim(e,theme)]
‘There was a hugging by Lee of Kim.’
Herburger (1995) notes that this representation of the structure of a sentence may
include the restriction of the existential quantification over events, and conse-
186 SARAH D. KENNELLY
quently the Restrictor of the Event Operator is distinguished from the Matrix. I
am assuming that in a ‘neutral’ sentence the Matrix of the Event Operator is the
verb, the natural predicate.
(11) ∃(e)[Restrictor[Lee(e,agent) & Kim(e,theme)]Matrix[hugged(e)]]
Herburger then proposes that Focus structures the quantification over events such
that the Focused element is the Matrix of the Event Operator. If we link her
proposal with Chomsky’s then we have (12).
(12) P:
The predicate of a sentence is the Matrix of the Event Operator.
If the theme is a NODP, which is P-Focused unless marked as backgrounded, we
get a representation of the structure of the sentence where the Matrix of the
Event Operator is now the theme and the verb constitutes part of the restrictor:
(13) ∃(e)[R[Lee(e,agent) & hugged(e)]M[a child(e,theme)]]
[first attempt]
Perhaps a more precise way of analyzing the step from (11) to (13) (D. Lebeaux
p.c.) is to identify it as a form of raising of the Focus element, leaving the
original Matrix as the matrix of the Restrictor, shown in (14) and (14′). This is
what occurs in the syntax under adjunction to obtain interpretation by the 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′)
(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’.
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;
THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH 189
Vikner 1991: 134) as the driving element for verb movement to I0. Turkish strong
inflection then provides the motivation for the head movement of the verb to I0.
There is no evidence of a CP category in Turkish since there is no lexical
complementizer in the language, subordination is in terms of DP (Kennelly 1990)
and there is no wh-fronting in the usual sense, as seen in (4)/(6).12 Consequently
it is assumed that at Spell-Out the verb resides in I0 rather than higher up. This
analysis is in the spirit of Chomsky (1995) whereby a derivation is structure
building such that the only structure that exists for a given sentence is that
necessary for its derivation, and also in the spirit of Grimshaw’s (1997) claim
that there are no useless projections.
The preverbal position for the NODP is an A-bar position, which cannot act
as an A-binder in a binding relation but is interpreted in situ for A-binding
purposes, as seen in (20).
(20) *Kendi bir hasta-@ 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.
(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.
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-
THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH 191
(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
192 SARAH D. KENNELLY
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.
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.
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.
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’.
194 SARAH D. KENNELLY
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 SYNTAX OF THE P-FOCUS POSITION IN TURKISH 197
the Head Parameter does. If there were a righthand Spec position to host wh-Qs
and NODPs, say Spec, ‘F’P, then multiple occupants due to movement would not
govern their respective traces, resulting in an illegitimate structure. The Focus
Particles, discussed in Sec. VI might be proposed as ‘F’0, but under a righthand
Spec analysis they should occur to the left of the elements in Spec, ‘F’. Instead
they occur to the right of an element in ‘F’. Worse still they are not associated
with one position; they may occur as a clitic on any element in the sentence (e.g.
the Q particle), hence it will be proposed that they are incorporated. The only
possible ‘F’0 would then be an abstract head. If there were an ‘F’P, one would
expect occupants of the ‘F’P projection to take scope over VP adverbs.
(30) a. ?Bu çocuk iki kere üç kitap-@ 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]
198 SARAH D. KENNELLY
(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.
THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH 199
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.
THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH 201
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.
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.
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.
6.2.3 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).
204 SARAH D. KENNELLY
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
THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH 205
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.
THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH 207
1. The head of a RC is to the right and there are postpositions rather than prepositions.
2. It has been proposed there are no nonspecific subjects in Turkish (Kennelly 1997). Turkish
speakers vary in their judgments wrt obliques so that discussion will have to wait for further
research.
3. Drawing on Baker (1970), Cheng (1991: 19) notes a correlation between the presence of a Q
particle in a language and the absence of wh-movement. However Turkish has a yes-no Q
particle, mI, and wh-movement is to the Focus position, first noticed by Güliz 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.
208 SARAH D. KENNELLY
10. According to de Hoop’s (1992) analysis, an argument with Weak Case, the NODP in Turkish,
may move into any A-bar position. I will modify her analysis by claiming that it may move
into an A-bar position as long as occurrence in that position is consonant with its interpretation.
As brought out in Chierchia & McConnell-Ginet (1991: 282) a nonpresupposed element, that is
new information, may also be backgrounded, seen in English in a nonrestrictive relative.
Nonspecific arguments are taken to be new information and hence nonpresupposed. So they
may occur either in the ‘new information’ Focus slot, left verb adjacent in Turkish, or 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.
THE SYNTAX OF THE P-FOCUS POSITION IN TURKISH 209
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.
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)
WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 217
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
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
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,
WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 227
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
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
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
WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 231
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
‘RAM will bring the book, not Sita.’
b. kitaab laayegaa, (siitaa nahii)
Ram book bring- Sita not
‘RAM will bring the book, not Sita.’
Two questions remain regarding Hindi-Urdu scrambling and focus. The first
queries why rightward scrambling, which is analyzed as right-adjunction to VP,
should not yield the same results, since it also results in adjacency between the
232 AYESHA KIDWAI
verb and an element that would not normally appear in that environment. Indeed
for some speakers, preverbal focusing is indeed possible with rightward 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
WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 233
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
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
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
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
WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 239
complement was part of the presuppositional set of the interlocutors prior to the present.” The
focus-relevant notion of presupposition is thus contextually defined.
2. Context-induced contrastive and presentational interpretations are not typical to foci alone, as
the interpretations a topic may receive are identically constrained by context — if a contrast set
of topics is available, the topic will be interpreted contrastively:
A: Tell me about Ram and Sita. Do you like them?
B: , I like (but Ram I cannot tolerate).
These observations hold for morphological topics as well. Miyagawa (1989) and Kidwai (1995)
trace the contrastive interpretation accorded to -wa topics in Japanese and -to topics in 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
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.
WORD ORDER AND FOCUS POSITIONS IN UNIVERSAL GRAMMAR 243
References
Kenesei, István & Irene Vogel. 1989. “Prosodic Phonology in Hungarian”. Acta
Linguistica Hungarica 39.149–193.
Kenstowicz, Michael. 1985. “The Phonology and Syntax of wh-expressions in
Tangale”. Studies in the Linguistic Sciences 15.2.
Kidwai, Ayesha. 1995. Binding and Free Word Order Phenomena in Hindi and
Urdu. Doctoral dissertation, Jawaharlal Nehru University, New Delhi.
Müller, Geroen. & Wolfgang Sternefeld. 1993. “Improper Movement and
Unambiguous Binding”. Linguistic Inquiry 24.461–507.
Rochemont, Michael. 1986. Focus in Generative Grammar. Amsterdam: John
Benjamins.
Tuller, Laurice. 1992. “The Syntax of Postverbal Focus Constructions in Cha-
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
1. Introduction
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
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
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.
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
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
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
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
(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.
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.
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.
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
natural between the focused term and the complementizer or the subject of the embedded clause
(/it was the horse #(that) you bought/).
48. There are more morphemes in (41) than indicated in the gloss. I only spell out what is relevant
to the problem at hand.
49. Ambiclausal items may be rare, but I would be surprised if they were an isolated phenomenon.
For instance, English speakers who accept Who do you wanna win? would seem to make use of
an item wanna which is allowed to straddle the two clauses [who do you want] and [t to win].
The question of course is whether wanna may be considered a Vocabulary item. I think it may,
given the idiosyncrasy of the derivation want to > wanna, and the fact that there are dialects,
especially Afro-American English, where wanna never alternates with want to (see Labov
1972).
50. See The merchant bought only the horse, which does not normally lead us to think that the
merchant refused to buy, say, teapots in addition to the horse, but rather that he didn’t buy
other animals that were also on sale. For only as a focus marker, see Bayer (this volume).
References
Abraham, W. et al. (eds.) 1996. Minimal ideas: Syntactic studies in the minimalist
framework. Amsterdam: John Benjamins
Baker, M. C. 1985. “The mirror principle and morphosyntactic explanation”.
Linguistic Inquiry 16.373–415.
Bennis, H. & L. Haegeman. 1984. “On the status of agreement and relative
clauses in West-Flemish”. In W. de Geest & Y. Putseys (eds.), Sentential
complementation: Proceedings of the International Conference held at
UFSAL, Brussels, June 1983. Dordrecht: Foris, 33–53.
Benveniste, E. 1966. Problèmes de linguistique générale. Paris: Gallimard.
Boilat, D. 1858. Grammaire de la langue woloffe. Paris: Imprimerie impériale.
Brody, M. 1995. Lexico-logical form: A radically minimalist theory. Cambridge,
Mass.: MIT Press.
Chomsky, N. 1993. “A minimalist program for linguistic theory”. In M. Halle &
S.J. Keyser (eds.), 1–52.
Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Church, E. 1981. Le système verbal du wolof. Documents linguistiques de
l’Université de Dakar 27.
Collins, C. 1997. Local Economy. Cambridge, Mass.: MIT Press.
Creissels, D. 1991. Description des langues négro-africaines et théorie syntaxique.
Grenoble: Ellug.
Dialo, A. 1981. Structures verbales du wolof contemporain. Dakar: CLAD.
272 ALAIN KIHM
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
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
278 JACQUELINE LECARME
to have identical truth conditions, and only differ in the location of focus. In
such cases, focus typically has an influence on the felicity condition of the sentence:
(4) a. 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’
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
280 JACQUELINE LECARME
(tops) CP (tops) CP
− nom]
[+ − nom]
[+
C IP (Spec) C′
waa [+acc]
C (IP)
baa
(tops) IP
− nom]
[+
opposition. Supposing that this is so, (8) is about ‘this problem’, and provides the
new information that Ali told me about it. Thus, the entire baa-clause (including
the focus position) is predicated of the topic, introducing new information. This
calls into question both the function of focus markers as an information encoding
device, and the informational status of topics. Topics can occur both inside and
outside the maximal focus-phrase, as is shown in (8), where √ marks other
potential structural positions in which NPs such as arríntan ‘this problem’ may
be adjoined. If these adjuncts were truly informational topics, they would in
principle create a new binary opposition (new vs. given information), and
interpretative conflicts would necessarily arise.
(8) (arríntan) Cáli baa √ iigá warramay √
problem-this Ali me+to-about reported
‘Ali told me about this problem’
As we have already mentioned, and will make more precise below, full NPs in
Somali must appear as adjuncts, which are not necessarily associated with the
same marked interpretation as left dislocation constructions in Romance 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.
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
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
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:
294 JACQUELINE LECARME
Let us, now, turn to the analysis of expletive constructions, which I think provide
the strongest motivation for a Case-based account.
(30) (CP)
C′ (Spec)
NP[+acc]
C IP
waxi aa
...ei ...
(pro/clit.)
(34) CP
ardáydai C′
[+acc]
C IP
baa
ei I′
I VP
V+[T+AgrS]
garán
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
Notes
* I am grateful to participants of the International Workshop on the Grammar of focus for
discussion on the issues dealt with here, to Nomi Erteschik-Shir and Laurie Tuller for detailed
written comments on an earlier draft, and to Bashiir Nuux Keenadiid for linguistic expertise.
Errors are my own responsibility.
1. This article is exclusively concerned with the syntactic structures occurring in standard Somali
and (more generally) the so-called maxaad tiri (lit. ‘what did you say’) dialects, which provide
a natural typology for the phenomena discussed here. Other dialects of Somali may substantially
differ in this respect (see, for example, Lamberti 1983; Tosco 1993). Striking similarities in the
focus system are found in other Cushitic languages, such as Rendille (Oomen 1978), Boni
(Sasse 1981).
2. I follow the national orthography in representing voiced and voiceless pharyngeal fricatives as
c and x, and long vowels by gemination. Somali has a mixed prosodic system that Hyman
(1981) characterized as a ‘tonal accent’ system. Since prosodic facts play a role in the subject
under study, surface high tones that result from the computation of prosodic domains at the
morphological or postsyntactic level are marked on examples in this paper. I will use the
following abbreviations: = focus marker, = restrictive, = extensive, = invariable,
= independent, = subjunctive; = definite determiner, = demonstrative, =
possessive; = expletive; = reflexive pronoun, = impersonal pronoun. Other
pronominal clitics are identified by their features (person, gender, number).
3. Baa and ayaa are strictly equivalent in their distribution and syntactic properties. Only baa
allows phonological coalescence with a left-adjacent nominal ending in a vowel (Cáli + baa →
Caláa. On the origin of these morphemes, see note 19.
4. There is a considerable variation in Cushitic languages. Somali is strictly ‘head initial’ as far as
functional categories are concerned, and is a right-branching language. In the noun phrase,
adjectives, relative clauses and genitive complements follow the head noun. Oromo, Rendille,
Boni also exhibit ‘mixed headedness’, but have a stricter OV order than Somali, and are ‘null
subject’ languages. Other Cushitic languages (e.g. Afar, Sidamo), and other Somali dialects (cf.
Tosco 1993) are consistently head-final.
5. I have suggested in earlier work that there is a correlation between the ‘pronominal argument’
property and the fact that, in Somali, NPs cannot enter in full agreement relation with a Verb,
because number in nominals has a derivational nature: Verbs only agree in number via the
subject clitic system. See the discussion in Lecarme (1995). The correlation between the
‘pronominal argument’ character and the fact that NPs are formally unmarked for number seems
to hold in other languages, as Baker (1996: 121–124) explicitly suggests.
6. For the full justification and motivations of this analysis, I refer the reader to Lecarme (1991,
1994, 1995).
7. The relationship between [+Tense] in finite clauses, [−Tense] in non-finite clauses is quite
different from that of the English or French. The ‘impoverished’ nature of the so-called ‘
Restrictive paradigm’ or ‘negative conjugations’ is the result of reduced patterns of agreement
that lack the Number feature, or of intervening heads in the process of V-movement (See
Lecarme 1995, and Section 2.4.2 below).
FOCUS IN SOMALI 303
8. This slot includes the spatial deictic elements soo/sii by which the action described by the Verb
is related to the speaker’s location.
9. Somali preverbs are highly reminiscent of the preverbal prepositional heads in Classical Greek
and other Indo-European languages. They combine in the strict linear order ú (dative), kú
(locative-instr.), ká (ablative), lá (comitative), and form a morphological X0 complex, the last
(rightmost) prepositional head keeping its lexical accent (kulá, ugú…). More complex surface
forms may result from the PF association of the adjacent object clitics (ií (i+ú), iigú (i+ú-kú),
nóo (na+ú), noogú (na+ú-kú), etc.), with which they form a ‘phonological word’. Unlike
affixes, clitics are not selective of the category of their phonological host, hence may
prosodically attach to categories that do not ‘govern’ them. In the literal glosses, we represent
this phonological contrast with hyphens for affixes and a plus sign for clitics.
10. Such elements include the ‘weak’ negation -an, coordinating particles -na ‘and’, -se ‘but’, etc.
11. As the following paradigms clearly show, strong pronouns are DPs (observe that the definite
determiner retains the -k-/-t- gender distinction in the 3S), while ‘weak’ subject pronouns
represent a clitic version of the nominal part of the DP projection:
Strong weak (subject)
aní-ga aan
adí-ga aad
isá-ga uu(s)
iyá-da ay
(excl) anná-ga aan(u)
(incl) inná-ga (aynu)
idín-ka aad/aydin
iyá-ga ay
12. Following Heine & Reh (1984), ‘term focus’ involves the reanalysis of a copula in a cleft
construction (NP + copula + relative clause), while ‘predicate focus’ is historically derived from
a (copula + predicate) structure. This historical hypothesis is implicitely assumed in most
accounts, and motivates Saeed’s (1984) analysis of focus constructions as synchronic reduced
clefts and pseudo-clefts. See Lecarme (1991) for a critical evaluation of this proposal.
13. When there is a succession of two adverbial expressions, a second baa may occur. See Hetzon
(1965) for more examples.
14. Nomi Erteschik-Shir (personal communication) points out that the focus position described here
might fit her notion of topic. She analyzes contrast as focus within a topic set: contrastive topic
and contrastive focus would then be the same (Erteschik-Shir, in press).
15. As pointed out by Laurie Tuller (personal communication), the facts regarding null third person
pronouns having a definite interpretation, while indefinite meanings requires an overt argument
are parallel in Chadic, where null third person pronouns are also found.
16. For example, waa is understood as a ‘classifier’ by Saeed (1984), who explicitly denies the
pragmatic role of ‘verb focus’ structures in discourse.
17. Well-known exceptions are the short, accented forms of the Past Independent and Past
Comparative (Andrzejewski 1956, pp. 126–129), the imperative, and the exclamative verbal
304 JACQUELINE LECARME
forms. Common to all these ‘root’ assertive paradigms is the absence of both focus markers and
subject clitics.
18. But see (1.3.2) above.
19. Lamberti (1983), following the traditional approach outlined above (see note 12), has proposed
to reconstruct the morphemes baa/waa and ayaa as a (verbal) copula (*awa). There is, however,
much comparative evidence (Arabic huwwa/iyya, Hebrew hu/hi, etc.) that these morphemes
have a pronominal origin.
20. Argument-type CPs behave like NPs or DPs, given abstract nominal properties of their lexical
complementizer (ín). The proposed analysis takes in account the limited range of interpretation
of these constructions. Existential sentences and non-existential locative sentences, which are
verbal in nature, must use verbal roots and full IP projections (cf. Lecarme 1991, 1994). The
predicative vs. presentational interpretation of examples in (14a,b,c) depends on a NP vs. DP
complement (D assumed to be the locus of specificity). Similarly, the function of ‘specification
of necessity’ (Andrzejewski 1965) in (14d,e) is clearly derived here from the nature of the
complement.
21. In adjunct positions (Andrzejewski’s (1964) ‘open configurations’), Case manifests itself both
morphologically and prosodically, affecting the rightmost constituent of the NP. The clearest
manifestation of the [±nom] opposition usually shows up as morphological Case endings (-u/-a)
suffixed to the definite determiner. In focus position (Andrzejewski’s ‘closed configuration’),
there is an invariant, ‘default’ Case (noted here [+acc]) morphologically identical to the
[−nominative] Case in adjunct positions.
22. See, in particular, Vallduví (1990) and Brody (1995) for related discussion.
23. On the absence of WCO effects in focus constructions, see Kennelly (this volume) and Kidwai
(this volume).
24. This includes nominals such as áad ‘muchness’, ág ‘proximity’, hóos ‘bottom’, dhéx ‘middle’,
etc., which, often in connection with the prepositional preverb ú, form syntactically dissociable
adverbial expressions meaning ‘a lot’, ‘near’, ‘under’ ‘among’, etc.
25. Since the language does not have noun complement structures or sentential subjects in argument
position, examples like (22b, c) are the only ‘strong’ islands that can be tested.
26. Significantly, inherently quantificational expressions such as nothing, nobody do not exist in
Somali. Qófna, wáxba are indefinites meaning ‘person, thing’ combined with the enclitic
particle -na, -ba, a polarity element without inherent negative meaning, which has to be
interpreted in the scope of negation.
27. I am setting aside here particular issues concerning relative (and possessive) constructions,
which syntactically involve asymmetric coordination. When there is more than one conjunct
(e.g., when the head Noun is modified), overt coordinating conjunctions must fill the C
position.
28. It is assumed here that the Spec,CP position is filled by ‘base-generation’, an option compatible
with the LCA (Kayne 1994: 165). Somali has further N-to-D raising in the noun phrase (see
Lecarme 1996).
FOCUS IN SOMALI 305
29. I take non-root CPs to be [nominal], a feature of their head (the complementizer ín). The
licencing of CPs is mediated by clitics in argument positions: the expletive (it-like) subject clitic
-ay (3FS), a silent object clitic, and the genitive clitic -eed (Poss3FS).
30. For a full review of other proposals in the literature, and for arguments that postverbal focus
constructions cannot be assimilated to relative clauses or pseudo-cleft sentences, see Lecarme
(1991). Significantly, question phrases, non-referential nominals, ‘understood’ arguments cannot
be extraposed. This fact is one of the main problems facing Saeed (1984)’s analysis of 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 phono- logical 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.
306 JACQUELINE LECARME
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.
Horvath, Julia. 1985. Focus in the Theory of Grammar and the Syntax of Hungari-
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:
Oxford University Press.
Kiss, Katalin É. 1995. “Multiple Topic, one Focus?”. Manuscript, Linguistic
Institute of the Academy of Science. (Talk presented at the 18th GLOW
Colloquium.)
308 JACQUELINE LECARME
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.
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.
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
314 JON ORTIZ DE URBINA
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?’
316 JON ORTIZ DE URBINA
must occupy the position immediately to the left of the matrix verb esan ‘say’.
When, as in (21), some element intervenes, the result is ungrammatical. (22)
shows that exactly the same pattern can be found with wh-words. Both operator
types, foci and wh-words, have scope over the entire structure. This is 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
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
320 JON ORTIZ DE URBINA
Peruk]]
Peru
‘I think Peru said JON wrote the book’
(Lit. ‘That JON wrote the book do I think did Peru say’)
Here the whole clausal complement of say appears now immediately to the left
not of say itself, but of the root think. Notice that we also find here the inversion
effects which we claimed above to be the result of cyclic Spec to Spec 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
As shown throughout the preceding description, wh-words and foci display many
distributional similarities that must be captured somehow. Traditional grammars
of Basque, since Altube (1929), also state that the two elements behave in the
same way, and occupy the same position, which we have identified as left
peripheral, that is, clause initial. Descriptively, the situation described above is
reminiscent of residual V2 phenomena as discussed by Rizzi (1996), in the sense
that we have a certain type of verb movement and adjacency only when an
operator is present, and I have been pointing out the similarities by referring to
the functional nodes to the left as related to CP.
322 JON ORTIZ DE URBINA
However, the main problem we face when dealing with Basque lies in the
fact that the language is rather thoroughly right headed. That is, V2 effects are
achieved straightforwardly, in terms of structure like (36a), provided the operator
and the verbal head move to the same phrase, usually identified as CP or IP, but
not with (36b):
(36) a. XP b. XP
YP X¢ YP X¢
X ... ... X
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
324 JON ORTIZ DE URBINA
in some languages, that is, a lexical head must move to it. This is
achieved in Basque by head-to-head movement of V to Infl. Once the head Infl
is thus lexicalized, the new complex V/I unit must move to C, where 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.
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
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):
330 JON ORTIZ DE URBINA
NegP I
VP Neg
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.
332 JON ORTIZ DE URBINA
2. While examples are given in the standard literary dialect, they primarily describe the situation
found in Biscaine and Gipuzkoan dialects, unless explicitly noted otherwise.
3. In some speakers, the focalized element may occur at the end as in (i), with a correcting
interpretation similar to that of echo questions.
(i) Galdetu didate ZER erosi du-en JONEK
4. An alternative derivation where the wh-operator has been extracted directly over the occupied
embedded specifier position, would also be ruled out as a subjacency violation
5. In the case of foci pied-piping relative clauses, as in (23), we would have to assume that the
relative clause, turned into an operator-like element by the presence of the focalized element,
is moved to the specifier of the DP:
(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
Jamal Ouhalla
Queen Mary, London University
Abstract
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.
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
FOCUS AND ARABIC CLEFTS 339
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
FOCUS AND ARABIC CLEFTS 341
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
FOCUS AND ARABIC CLEFTS 343
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).
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]:
FOCUS AND ARABIC CLEFTS 347
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).’
FOCUS AND ARABIC CLEFTS 349
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,
FOCUS AND ARABIC CLEFTS 351
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
FOCUS AND ARABIC CLEFTS 353
6. Conclusion
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.
FOCUS AND ARABIC CLEFTS 357