Sei sulla pagina 1di 63

Clitic Placement after Syntax: Evidence from Wolof Person and Locative Markers

Author(s): Anne Zribi-Hertz and Lamine Diagne

Source: Natural Language & Linguistic Theory, Vol. 20, No. 4 (Nov., 2002), pp. 823-884
Published by: Springer
Stable URL:
Accessed: 01/03/2009 08:10

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact

Springer is collaborating with JSTOR to digitize, preserve and extend access to Natural Language & Linguistic



ABSTRACT. This paper is an empiricalcontributionto the typology of functional nom-

inals (pronouns),and to the theory of clitics. Its primarygoal is to present an adequate
descriptive analysis of the attached person and locative markersof Wolof, a language
whose 'special clitics' partlypatternlike those of Berber,describedby Dell andElmedlaoui
(1989), Ouhalla(1989), andBoukhris(1998). Ourstudyleads us to discardan all-syntactic
account of the special position of clitics, of the sort developed by Ouhalla and Boukhris
for Berber,and by Njie (1982) and Dunigan (1994) for Wolof, and to adopt an approach
crucially separatingthe syntactic and morphophonologicalpropertiesof clitics, along the
lines set by variousscholarsworkingon the syntax-phonologyinterface.


The description of clitics a priori seems to involve several components

of grammaticaltheory: Phonology, since clitics are typically spelt out as
'leaners', in the sense of Zwicky (1982); Morphology, since they have
affixal properties;and Syntax, since special clitics, as initially defined by
Zwicky (1977), generally exhibit Subjacencyeffects. However,the relat-
ive prominenceof each module in availabledescriptionsof clitic systems
varies from one study to the next: depending on the author'stheoretical
backgroundand on the empiricalevidence supportingthe study,the linear

* Parts of the researchthat led to this article were presentedat the Langues & Gram-
maire seminar of UPRESA 7023 (Paris-8/CNRS)and at the Clitic Works'hoporganized
by Universite Paris-7 in February 1999, whose audiences came up with a lot of inter-
esting questions and remarks.We thank several individual colleagues and students for
their feedback while this work was in progress:NadiraAljovic, Stephen Anderson, Car-
men Dobrovie-Sorin,Celia Jakubowicz,Makoto Kaneko,Philip Miller, Lea Nash, Alain
Rouveret, Liliane Tasmowski,and Arnold Zwicky. We thankStephaneRobert and Serge
Sauvageot for kindly supplying us with copies of their own works on Wolof. We owe
special debts of gratitudeto Alain Kihm and Patrick Sauzet, who shared with us some
of their expertise on Wolof and linearity (respectively),and to Joan Maling and our four
NLLTreviewers,who providedus with manypages of demandingbut supportivefeedback
on previousdraftsof our text.

i Natural Language & LinguisticTheory 20: 823-884, 2002.

? 2002 KluwerAcademicPublishers. Printedin the Netherlands.

position of special clitics is regardedas pertainingto syntax (1), phonology

(2), lexical morphology(3), or morphosyntax(4):

(1) All-SyntacticAccounts of Special Clitics (A Few Illustrative

a. Kayne(1991)
* (Romance) clitic pronouns are generated in an argument
position and move leftwardto adjoin to a functionalhead.
Procliticsadjointo a head to which the verb has previously
raised;enclitics adjointo a head priorto verb raising:
proclisis: cl-[V+X?]

enclisis: V-[cl+Y?]

b. Shlonsky (1994)
* Each (Semitic) clitic pronounspells out agreementfeatures
generatedin the Agr-headof its maximalprojection.A lex-
ical head moves to adjoin to the closest Agr-head,which
containsthe clitic: an enclisis configurationfollows:
V/N/P/A-[AGR Cl]

c. Cardinalettiand Starke(1999)
* Clitics are phrases which are deficient for some syntactic
projections.Consequently,they must move in syntaxto ad-
join to a functionalhead, this strategymaking up for their
inherentdeficiency.The prosodicdeficiency of clitics is but
a reflex of theirdeepersyntacticdeficiency.

d. Nash and Rouveret(1999), Rouveret(1999)

* Clitics spell out argument features that could not be
checked in their lexical domain and must therefore be
checked further up, in the functional domain. A feature
which is not checked in situ (lexical domain) moves up to
the closest accessible functionalhead. If thereis no access-
ible functional head, a 'proxy' category is created during
the derivation. Distinctive person inflection on the verb
triggersproclisis (as on Italianfinite verbs);nondistinctive
person inflection triggers enclisis (as on Semitic verbs or

(2) Phonological Accountsof Special Clitics

a. Wackernagel's(1892) Law
In Ancient Greek (and, more generally, in Proto-Indo-
European), unstressed elements (including unstressed pro-
nouns) attach to the right of the first prosodic word in their
clausal domain. [All clitics consideredare thereforeenclitics.]

b. Sauzet's (2000) Nontaxic Analysis of clitics

Special clitics are nontaxic expressions, i.e., expressions that
are not positioned in the output of syntax and are linearized
postsyntacticallyby syllabificationrules that either adjoin or
incorporatethem to a taxic expression. Simple clitics are taxic
expressions that lack stress and thus behave phonologically as

(3) (Special) Clitics as LexicalAffixes

Miller (1992), Miller and Sag (1997)
* (French) clitic pronouns spell out argument features of
lexical heads.
* Correlatively,each clitic surfacesas an affix thatattachesto
the lexical head (or extendedhead) whose argumentfeature
it spells out.
[The contrast between enclisis and proclisis boils down
to that between suffixation and prefixation- an inherent
propertyof each affixalelement.]

(4) (Special) Clitics as Phrasal Affixes

Klavans(1985), Anderson(1992, 1993)1
* Clitics are affixes which spell out phrasefeatures.
* Each clitic is specifiedin the lexicon for
- its domain (the syntacticconstituentwhose featureit spells
- its anchor (FIRST, HEAD, LAST)
- its orientation(PRECEDESor FOLLOWSits anchor)
* The same set of principles accounts for both clitic place-
ment and affix placement.

1 Our
phrasing in (4) is based on Anderson (1992, p. 203). Klavans (1985), whose
approachis very similar,derives the proclitic/encliticcontrastfrom yet anotherparameter

These are but a few representativesamples2of availableaccountsof (spe-

cial) clitics, but they will be sufficient for the present discussion. The
assumption that (special) clitics either undergo syntactic movement or
occupy special syntactic positions seems to be dominant, although not
unanimous(cf. Miller 1992; Miller and Sag 1997), among syntacticians.It
is cruciallybasedon the fact thatspecial clitics exhibitSubjacencyeffects,3
as prototypicallyillustratedby the following Frenchexamples:

(5)a. Je sais que Marie voit cette mouche.

I knowthat Marie sees this fly

b. Je sais que Marie la- voit.

I knowthat Marie it + sees
I know that Mariesees it.
(right-vs. left-attachment).Andersonarguesthat this distinctionfollows from a particular
language'srule(s) of StrayAdjunction,andis thus independentfromthe issue of clitichood.
2 Othersyntacticaccounts may be found in Emonds (1975), Jaeggli (1982), Bouchard
(1982), Borer (1986), Burzio (1986), Roberge (1990), Sportiche(1996), Dobrovie-Sorin
(1999), among others.For a phonological accountof Portugueseclitics, see Frota(1992).
3 Cf. Kayne (1989, 1991), Ouhalla(1989). Furtherempiricalevidence takenas support-
ing the syntacticmovement of clitics is drawnfrom the past-participle-agreement facts of
French:in this language,a past-participleagrees with the objectclitic which surfacesto its
left, not with the object DP which surfacesto its right:

(i)a. Pierre a mis cette chemise.

Pierre has put-pp-DEFAULTthis shirt-Fsg
Pierreput on this shirt.

b. *Pierre a mise cette chemise.

Pierre has put-pp-Fsg this shirt-Fsg

c. (cette chemise) Pierre 1' a mise.

this shirt-Fsg Pierre 3Fsg-ACChas put-pp-Fsg
lit. (this shirt)Pierreit (Fem)-puton
= This shirt,Pierreput it on.

Assuming that agreementmust involve a spec-head relation(a well-spreadbut in no way

necessary assumption, cf. Miller and Sag 1997), it follows that the object pronoun of
(i-c) must not occupy the same syntactic position as the object DP of (i-a). It does not
necessarily follow, however, that the clitic ever moved in syntax to its surface position
in (i-c): the pronounand the DP could be linearizeddifferentlyfor morphophonological
reasons, with agreementoccurringafterlinearization(in morphology).

c. *Jela- sais que Marie voit.

I it + know that Marie sees
The special position of the accusative pronounin (5b), contrastingwith
the accusativeDP in (5a), is what led Kayne (1975) to propose the syn-
tactic rule of Clitic Movement- the ancestorof the more recent analyses
mentionedin (1) and fn. 3. The ungrammaticalityof (5c) providesfurther
supportfor this analysis by suggesting thatClitic Movementcannot cross
the CP barrier,a typically syntacticconstraint.However,both the special
clitic position exemplifiedby (5b) and the Subjacencyrestrictionexempli-
fied by (5c) are straightforwardlyaccountedfor under the phonological,
lexical, and morphologicalanalyses sketchedin (2), (3), and (4). The syl-
labificationrules which are assumed to position special clitics under(2b)
are crucially local, as are all phonological rules. The lexical approachin
(3) correctlypredicts that the clitic of (5b) must be phoneticallyrealized
as an affix on the (extended) lexical head whose argumentit represents;
the Klavans/Andersontheory in (4) similarly predicts that the clitic of
(5b) should be anchoredto a high-visibility (head, initial, or final) locus
within its scopal constituent (its domain). The crucial contrastbetween
the syntactic, phonological, and morphologicalaccounts lies in the way
they explain the 'special' linear position of clitics: under all-syntactic
analyses, special clitics occupy or move up to special structuralpositions
becausethey have special syntacticproperties,e.g., they instantiatespecial
(agreement)features (Shlonsky), they have a deficient internalstructure
(Cardinalettiand Starke), or a peculiar (Kayne, (Nash and) Rouveret)
featurecontent. Under phonological and morphologicalanalyses, special
clitics do not occupy special structuralpositions; they are either thought
of as regularly-positioneditems which move in phonology, due to their
prosodic deficiency (Wackernagel),or as 'nontaxic' features which are
not structurallypositioned in the output of syntax and are consequently
linearizedaftersyntax (Sauzet, Miller, Klavans,Anderson).
The apparentlycomplex distributionof Wolof' person and locative
markers,presentedbelow, is obviously an interestingchallenge for a gen-
4 Wolof is a language spoken in Gambia and Senegal; Greenberg(1966) classifies it
in the NorthernWest-Atlanticbranch of the Niger-Congofamily. This paper is based on
the Senegalese varietyspoken by one of the co-authors,Lamine Diagne, born in Dakarin
1968 and raised in Dakar and Saint-Louisby Saint-Louisparentsand relatives. Our data
hence exhibit some typical features of Saint-LouisWolof, which, in gross sociolinguistic
terms,we may characterizeas 'upper-classconservative':Saint-Louisspeakersare known
to cultivatean awarenessof Good Wolof and, for instance, hold on to the full paradigm
of noun classifiers([b/g/k/w/m/lIs/j],in the singular),which Dakarspeakerstend to reduce
to one ([b]). This researchwas conducted in French (the two authors' common tongue),
which accountsfor occasional referencesto Frenchin the phrasingof our description.

eral theory of pronouns and pronominalclitics. We shall focus here on

clitics, startingout with no modularbias: the central issue will precisely
be to determinewhich module(s) of grammaticaltheory should account
for the variouspropertiesobserved,a task which can only be achievedon
We shall first survey the structureand typology of Wolof clauses (sec-
tion 2), a necessary preliminarystep for the study of clitic placement.
Section 3 will arguethat an all-syntacticanalysis of Wolof clitics, such as
thatdevelopedby Dunigan(1994), shouldbe discarded.Section 4 will pro-
pose an alternativeanalysis which deals separatelywith the phonological,
syntactic,and morphologicalpropertiesof personandlocativemarkersand
accountsfor the linearposition of 'special clitics' at the syntax-phonology
interface.Section 5 concludes thatthe Wolof evidence globally supportsa
modularapproachto clitics and cliticization.


Before we take up the issue of the best analysis for clitics, we need to
provide the reader with a general understandingof Wolof clause struc-
ture. For convenience's sake, we shall use zeros in the transcriptionof
our examples to indicatethe unmarkedvalues of functionalfeatures.Our
presentationof the Wolof datadepartsin variousrespectsfrom otheravail-
able syntactic descriptions of Wolof clause structure(particularlyNjie
1982; Robert 1991; Dunigan 1994). Some discrepancies between other
authorsand ourselves will be pointedout along the way.
Let us begin with independent,simplex, affirmative,transitiveclauses,
exemplifiedin (6) and (7):5

(6)a. xale -yi lekk -oon -na -fnuceeb -bi

child DFpl eat +pst +F 3pl rice DFsg
The childrenhad eaten the rice.
5 The following abbreviationswill be used in the glosses: C = complementizerposition;
COP = copula;DF = definite;DM = demonstrative;EXPL = explicative; F = finite;IMP =
imperative;ipf = imperfective;LOC = locative;neg = negation;0 = object;obv = obviative;
PRES = presentative;pst = past tense;pI = plural;sg = singular;STR = strong;WK = weak;
1, 2, 3 = first, second, thirdperson. Hyphens in examples note phonological attachment
(which is always left-orientedin Wolof):thus, the morpheme[na] is transcribedna when it
is unattached,-na when it is enclitic. Bracketsin the examplesindicatesegmentswhich are
deleted in phonology. Hyphens in glosses note the negative value of a functionalfeature:
thus, [-pst] representsthe negativevalue of the tense feature,[?pst].

b. xale -yi lekk 0 -na -nluceeb -bi.

child DFpl eat -pst +F 3pl rice DFsg
The childrenhave eaten the rice.

(7)a. xale -yi d(i)6 -oon -na -flu lekk ceeb -bi
child DFpl +ipf +pst +F 3pl eat rice DFsg
The childrenwere eating rice.

b. xale -yi di 0 -na -nu lekk ceeb -bi.

child DFpl +ipf -pst +F 3pl eat rice DFsg
The childrenwill eat rice.

The sentences in (6) and (7) contrast as to their aspect feature

(?imperfective); the feature [+imperfective] is spelt out by the auxiliary
di. Depending on aspectualvalue, the linear orderof constituentsis SVO
(cf. (6)) or S AuxiliaryVO (cf. (7)). The morphemena, which we gloss by
[+finite], hosts a personmarkerwhich we shall identifybelow (section 4)
as an inflectionalaffix;7the [+finite] affix typically occurs in independent
6 The final vowel of the imperfectiveauxiliary(di) is truncatedwhen this auxiliaryhosts
a vowel-initialenclitic or suffix:di+oon > doon, di+u(l > du(l), di+e6e> dee. Furthermore,
the imperfectiveauxiliarydi loses its initial consonantand is spelt out as an enclitic when
it is precededby a vowel and does not itself host any affix, thus:... Moodu di lekk ... >
Moodu-(d)i lekk [Moodu-ylekk].
7 Our analysis here is at odds with Dunigan's (1994), who treatsthe person markerof
such examples as (6) and (7) as a subject clitic. This issue will be discussed in sections 3
and 4.
8 Dunigan glosses oon by 'perfect', di by 'present', 'future',or 'inversive',depending
on context, and na by 'affirmative'.We gloss oon by 'past' (as do Church1981; Ka 1994;
Fal et al. 1990; Fal 1991; and Robert 1991), di by 'imperfective'(as do Ka 1994; Fal et al.
1990; and Robert 1991) and na by '+finite' (which Church1981 glosses by 'enonciatif',
Ka 1994 by 'no-focus', Fal et al. by 'terminative',and Robert 1991 by 'perfect'). We
believe that our own descriptionresults in a minimally 'exotic' general picture of the
Wolof inflectional system. Regardingna, Robert's 'perfect' gloss is due to the fact that
(6b) reportsan accomplished,therefore'past' event. With stativeverbs, however,na does
not triggera past interpretation,cf:

(i) xale -yi begg 0 -na -flu ceeb.

child DFpl want -pst +F 3pl rice
The childrenwant rice. (#&The childrenwantedrice)

There is no a prioriincompatibilitybetween atelic verbs and the perfect tense, as shown

by English: The children wanted rice. But there are naturaland widespreadrestrictions

In the examples in (8), negation is spelt out as an inflectionalaffix on

the verband it surfacesbetween Tense andFiniteness:

(8)a. xale -yi lekk 0 -u(l)9 0 -nu ceeb -bi.

child DFpl eat -pst +neg +F 3pl rice DFsg
The childrenhave not eaten the rice.

b. xale -yi d(i) -oon -u(l) 0 -flu lekk ceeb -bi.

child DFpl +ipf +pst +neg +F 3pl eat rice DFsg
The childrenwere not eating the rice.

This first set of data leads us to distinguishfive functionalfeatureswithin

the inflectional domain, labeled Person, Finiteness, Polarity,Tense, and
Aspect. Adopting Baker's (1988) IncorporationTheory for description's
on the [atelic]x[imperfective]combination.Underour assumptions,the semanticconstrast
between lekk-na-nuceeb 'they have eatenrice' andbegg-na-nuceeb 'they wantrice' stems
from the fact thatthe telic verb lekkis inflectedfor the [-imperfective] feature(triggering
an accomplishedreading),whereasthe atelic verbbegg is left uninflectedfor imperfectivity
and simply read as 'nonpast',in this context. In any case, na has nothing to do with tense
and aspect.
Dunigan's 'affirmative'gloss of na misleadingly suggests that it is a polarity marker,
which it clearly is not, as hinted by Dunigan herself (cf. Dunigan 1994, pp. 33-34). It is
indeed truethatna does not combine with negation,an interestingrestrictionfor which we
have yet no explanationto offer. However,many Wolof clauses which may host negation
may not host na, e.g., relativeclauses:

(ii)a. ceeb -bi xale -yi lekk- ul

rice DFsg childrenDFpl eat +neg
the rice which the childrenhave not eaten

b. *ceeb -bi xale -yi lekk- na

rice DFsg childrenDFpl eat ?

If na should spell out an affirmativepolarityfeature,it should not be missing from affirm-

ative subject-focus,object-focus, and TS clauses (see below), which may all be negated.
Furthermore,na and negationhave very differentfeaturecontents.As will appearthrough
furtherexamples, the functionalhead which hosts na is semanticallycorrelatedwith the
UtteranceSource (person, viewpoint, UtteranceTime). We assume that na and NEG are
generatedin two distinctfunctionalheads, respectivelyFiniteness and Polarity.The finite-
ness assumptionregardingna, which is explicated in Zribi-Hertzand Diagne (1999), is
backed up by linguisticintuitionsexpressedby severalSenegalese linguists, among whom
are Diallo (1981b), Ka (1982), and Fal (1991).
9 The negationmarker[ul] is truncatedof its finalconsonantwhen it hosts a consonant-
initial suffix.

sake, we assume that each one of these functionalfeaturesheads its own

syntacticprojectionand is incorporatedinto the verb or auxiliaryby verb
(or auxiliary)movement.Following Collins (1997) and Chomsky (1999),
we, however, assume that the incorporationprocess pertainsto Morpho-
logy (word structure),ratherthan to syntax. We might have alternatively
assumed that Person, Finiteness, Polarity,Tense, and Aspect were 'clit-
ics' in the sense of Klavans (1980, 1985) and Anderson (1992, 1993),
i.e., phrasalfeatureswhose linearizationwould occur postsyntactically,in
Morphology.We shall arguein section 4 that the above featuresare hier-
archicallyorganizedin syntax, contrastingin this respect with accusative
and locativepronouns,the 'special clitics' of Wolof.
The [+past] specificationis not always spelt out as an affix as in (6),
(7), (8). Thereare interestingconstraintson the distributionof the [+past]
affix thatdeserve furtherprobing.One such restriction,for which we have
no explanationto offer, is thatit cannotcooccur with the negativeaffix on
the same verb root; thus, the [+past] counterpartof (8a) is not (9a), but
(9b), where the [+past] markersurfacesoutside the inflectedverb:

(9)a. *xale -yi lekk -oon -u(l) 0 -nu ceeb -bi.

child DFpl eat +pst +neg +F 3pl rice DFsg

b. xale -yi lekk -u(l) 0 -flu woon ceeb -bi

child DFpl eat +neg +F 3pl +pst rice DFsg
The childrenhad not eaten the rice.

To capturethis fact, we assume that the tense feature can be generated

in either the head of TP (surfacingas an inflectional affix) or in spec,TP
(surfacing as an adverbial,uninvolved in verb movement). Independent
evidence in supportof this analysis is the fact that tense can be spelt out
both inflectionally and adverbiallywithin the same clause, as in (10a),
inspired by Sauvageot (1965, p. 126). Similarly, the [+ipf] specifica-
tion may be spelt out twice within a clause, as in (lOb), triggering a

(lO)a. xale -yi d(i) -oon -u(l) 0 -flu woon lekk ceeb -bi.
child DFpl +ipf +pst +neg +F 3pl +pst eat rice DFsg
The childrenwould not have eaten the rice in the past.

b. xale -yi di 0 0 -na -flu -(d)i lekk ceeb -bi.

child DFpl +ipf -pst -neg +F 3pl +ipf eat rice DFsg
The childrenare usually eating the rice.

This firstset of descriptiveresultsis summarizedin diagram(11):

(1 1) Wolofclause structure:the lexical and inflectionaldomains

Pers? FinP
spec /
Fin' PolP
Pol0 TP
To AspP

(woon) (di)
[3pl] [+F] (+neg] [+pst] [+ipf] ~ 'lekk

xale -flu -nalo -u(l -oon (d)(i)


The positive value of the aspectfeature,[+imperfective],is spelt out by

the auxiliarydi, which supportsthe tense, polarity,finiteness,and person
affixes.10Anotherinterestingauxiliaryis explicativeda (a termwe borrow
from Njie 1982),1"exemplifiedby (12):

10 A similar descriptionmay be proposed for the auxiliatedfinite verb forms in, e.g.,
French- except that Frenchauxiliariesspell out the [+accomplished] feature,ratherthan
11 Dais glossed verbfocus in Ka (1994), predicatefocus in Dunigan(1994), emphatique
du verbe in Church (1981) and Robert (1991), processif in Fal et al. (1990). The term
explicative more accuratelysuggests the semantic effect correlatedwith this element: da
is typically inserted to indicate that the clause provides an explanation for a previous

(i)a. xale -yi liggey -u(l) -fnutey, da -nu oop 0 0 0

child DFpl work +neg 3pl today EXPL3pl be ill -ipf -pst -neg
The children did not work today, (it is because) they are ill. (adaptedfrom
Njie 1982, p. 143)

(12)a. xale -yi da -nlulekk 0 0 0 ceeb -bi.

child DFpl EXPL3pl eat -ipf -pst -neg rice DFsg
(It is because) the childrenhave eaten the rice.

b. xale -yi da -flu -(d)i 0 0 lekk ceeb -bi.

child DFpl EXPL3pl +ipf -pst -neg eat rice DFsg
(It is because) the childrenare eating the rice.

Da clauses are inflected for aspect, tense, and polarity,but these features
are spelt out below da, on the verb or imperfectiveauxiliary.Da itself is
only inflected for person, and does not combine with na: we assume that
da is generatedin the same position as na, i.e., the Finiteness head, and
incorporatesthe Personfeature.12
The Wolof clauses so far considered all host a person markeron the
verb or auxiliaryand are, correlatively,positively specified for finiteness.
A [+finite] clause may be embedded;it may then be introducedby the

(13) jigeen -ni xam 0 0 0 -na -fu -ni xale -yi

woman DFpl know -ipf -pst -neg +F 3pl that child DFpl
lekk 0 0 0 -na -flu ceeb -bi.
eat -ipf -pst -neg +F 3pl rice DFsg
The women know thatthe childrenhave eaten the rice.13

Let us now turn to clause types which, unlike the ones above, do not
include a person markeron the verb or auxiliary.We begin with clauses

12 In other words, na and da are both 'finiteness' markers,whose semanticimportper-

tains to the UtteranceSource. That da is related to na is independentlysuggested by the
occurrenceof (i) and (ii) as dialectal variants:

(i) na dem Dakar. [Dakararea]

(ii) da dem Dakar. [Saint-Louisarea]

(It is my wish ) that (you) go to Dakar.
13 Dependingon discoursecontext, the featurecombination[-imperfective] x [-past]
may be translatedby the English preterite(ate) or by the presentperfect (e.g., have eaten).
In treatingthe 'perfect' reading as unmarked,Wolof is similar to, e.g., Classical Arabic
(thanksto an anonymousreviewerfor this remark).

that stand as independentutterancesand do not overtly activate the left

(14) optativeclauses
na xale -yi lekk (*-oon) ceeb -bi!
+F child DFpl eat +pst rice DFsg
(It is my wish that we) let the childreneat the rice!
Optativeclauses (which are called injunctivein Ka 1994, desiderativein
Fal et al. 1990, and obligative in Church 1981 and Dunigan 1994) are
unspecifiedfor polarity,tense, aspect,andperson.They includethe morph-
eme na, which we gloss as [+finite], a descriptiveassumptionconfirmed
by interpretation:contrastingwith imperatives,exemplified by (17) be-
low, optativeclauses cruciallyinvolve the speaker'sperspective(cf. Diouf
1982; Zribi-Hertzand Diagne 1999).
Narrativeclauses14form a second type of independentclauses uninflec-
ted for person, whose syntax does not activatethe left periphery:
(15)a. narrativeclauses
gaynde -bi lekk Aram.
lion DFsg eat Aram
So the lion eats Aram.15

b. mu lekk Aram.
3sg eat Aram
So he eats Aram.

c. *gaynde-bi -(di) lekk Aram

lion DFsg +ipf eat Aram

d. taynde-bi lekk -oon Aram.

lion DFsg eat +pst Aram

e. *gaynde-bi lekk -ul Aram.

lion DFsg eat +neg Aram

14 Clauses such as (15) are called minimal by Church

(1981). The term narrative is
borrowed from Robert (1991), who similarly labels what we shall describe below as
nonpast/nonimperfectivetense-settingclauses, exemplifiedby (16). As revealedby (15c-
f), however, such examples as (15a, b) include no inflectional specification, whereas the
embeddedclause in (16) is negatively specifiedfor tense and aspect.
15 We translateWolof narrativeclauses by So + PRESENT TENSE, which is intendedto
convey the 'vivid narrative'semanticeffect.

f. *gaynde-yi lekk -nluAram

lion DFpl eat 3pl Aram

As revealedby (15c-f), the verb of a narrativeclause goes uninflectedfor

tense, aspect, polarity,finiteness,and person. This leads us to assume that
this type of clause includesno inflectionalprojection.Narrativeclauses are
productively(cf. Robert 1991) used in a certain type of 'vivid' narrative
style and are, in this respect, formally similar to the so-called infinitifde
narration (narrativeinfinitive) of French (e.g., Et le loup de manger le
chasseur 'So the wolf eats the hunter'.)
In Wolof, however, narrativeclauses crucially contrast with what
we shall call dependent-tenseclauses, which may translateas English
infinitivals,in some contexts:

(16)a. DT-clauses(DT = DependentTense)

xale -yi begg 0 0 -na -ilu [Aramlekk 0 0 ceeb
child DFpl want -pst -neg +F 3pl Aram eat -ipf -pst rice
The childrenwant [Aramto eat the rice].

b. kxale -yi begg 0 0 -na -flu [Aram lekk 0 0

child DFpl want -pst -neg +F 3pl Aram eat -ipf -pst
-ul ceeb -bi].
+neg rice DFsg

c. xale -yi gis 0 0 0 -na -flu [Aramdi 0 lekk

child DFpl see -ipf -pst -neg +F 3pl Aram +ipf -pst eat
ceeb -bi].
rice DFsg
The childrensaw [Arameating the rice].

d. xale -yi gis 0 -oon 0 -na -flu [Aramd(i) -oon lekk

child DFpl see -ipf +pst -neg +F 3pl Aram +ipf +pst eat
ceeb -bi].
rice DFsg
The childrenhad seen [Arameating the rice].

e. xale -yi gis 0 -oon 0 -na -nu [Aramdi 0 lekk

child DFpl see -ipf +pst -neg +F 3pl Aram +ipf -pst eat
ceeb -bi].
rice DFsg

f. *xale -yi gis 0 0 0 -na -fnu[Aramd(i) -oon lekk

child DFpl see -ipf +pst -neg +F 3pl Aram +ipf +pst eat
ceeb -bi].
rice DFsg

The bracketedclauses in (16) may not host a finitenessmarker,or person

inflection,or polarity(as witnessedby the ill-formednessof (16b)): we as-
sume that, like narrativeclauses, they are unspecifiedfor these features.'6
They are, however, specified for tense and aspect, but their tense feature
must agree with that of the matrix clause. Interestingly,the subject of
both narrativeand DT-clausessurfacesin the nominativeCase, as revealed
below by Table (33) (see section 3), which suggests that the nominative
should be identifiedin this languageas a defaultCase specification,rather
thanas a propertydependenton tense and/orfiniteness.
Relativeclauses also lack person inflection,but overtly activatethe left
periphery.Theirtopmostfunctionalhead hosts a definitedeterminer(bi, in
(17)). Like DT-clauses,relativeclauses are specifiedfor tense, aspect, and
polarity,but not for finitenessand person:

(17)a. relativeclauses
ceeb -bi xale -yi lekk 10 /-oon 0
rice DFsg child DFpl eat -pst /+pst -neg
the rice which the children (have/hadI eaten

b. ceeb -bi xale -yi di I0 /-oon 0 lekk

rice DFsg child DFpl +ipf -pst/+pst -neg eat
the rice which the children ( are/wereI eating
16 The fact that DT-clauses may not be inflected for negative polarity still needs to be
explained. To translatean English negative infinitival,Wolof must use a noninflectional

(i) xale -yi begg-na-niu [Arambani lekk ceeb -bi]

child DFpl want +F 3pl Aram refuseeat rice DFsg
lit. The childrenwant Aram (to) refuse (to) eat the rice.
= The childrenwant Aramnot to eat the rice.

c. ceeb -bi xale -yi lekk 0 0 -ul

rice DFsg child DFpl eat -ipf -pst +neg
the rice which the childrenhave not eaten

Whatwe shall call tense-seitingclauses form a thirdtype of embedded

clauses lackingperson inflection:

(18)a. Tense-Setting(TS) clauses

bi xale -yi lekk 0 0 -ee ceeb -bi
Co,bvchild DFpl eat -ipf -neg -F rice DFsg
When the childrenate the rice...

b. bi xale -yi d(i) 0 -ee lekk ceeb -bi...

child DFpl +ipf -neg -F eat rice DFsg...
While the childrenare eating the rice...

c. bi xale -yi lekk 0 -ul -ee ceeb -bi...

C0obvchild DFpl eat-ipf +neg -F rice DFsg
When the childrendid not eat the rice...

TS-clauses are adverbialclauses which, in initial position, set the tense-

frame of the whole sentence. Their temporalvalue is not dependenton
the matrix tense: in the above examples, it is the other way around.TS-
clauses characteristicallyinclude the affix -e'e,whose linearposition to the
right of the negation affix suggests that it fills the same structuralslot as
na, i.e., the Finiteness head. This accounts for our tentativeglossing of
-ee as [-finite], a feature value we take as indicating that the temporal
anchoringof TS-clauses is set as independentfrom the UtteranceTime.
The global interpretationof TS-clauses depends on their aspectual and
We now turnto clause types which are unspecifiedfor person, activate
the left periphery,but occur as root sentences. This is first instantiated
by imperatives.As their homologues in other more familiar languages,
Wolof imperativeclauses do not exhibitthe same syntax in the affirmative
and negativepolarities.Affirmativeimperativeshave a specific imperative
morphology(-al) and a deficientperson+numberinflection(pragmatically
limitedto the second person),which does not license an externalargument.
17 Thus, (18a) and (18c) aretranslatedas 'When.. ',due to theirnonimperfectiveaspect;
whereas(18b) is translatedas 'While. . . ', due to imperfectivedi.

Negative imperativeshave no specific imperativemorphology,are unin-

flectedfor person,and may host a subjectargument.The structuralanalysis
we (very tentatively)suggest in (19) is stronglyinspiredby Harris's(1998)
analysis of Spanishimperatives:

(19)a. imperativeclauses
0 lekk0 -al 0 ceeb-bi!
C eat -neg IMP -pl rice DFsg
Eat the rice! (singularaddressee)

b. 0 lekk 0 -al -(l)een ceeb -bi!18

C eat -neg IMP +pl rice DFsg
Eat the rice! (pluraladdressee)

c. b -u(l) 0 lekk ceeb -bi!

C +neg -pl eat rice DFsg
Do not eat the rice!

d. b -u(l) -leen lekk ceeb -bi!

C +neg 2pl eat rice DFsg
Do not eat the rice!

e. b -u(l) xale -yi lekk ceeb -bi!

C +neg child DFpl eat rice DFsg
Let the childrennot eat the rice!

f. b. -u(l) -fnulekk ceeb -bi!

C +neg 3pl eat rice DFsg
lit. Let they not eat the rice!

18 As rightlyobservedby a shrewdanonymousreviewer,a genuineDakarspeakerwould

be likely to realize (19b) as in (i), where the imperativemarkeris deleted to the left of the

(i) 0 lekk 0 -leen ceeb -bi!

C eat IMP 2pl rice DFsg

Our example (19b), with its overt imperative marker, stands as a typical 'Saint-
Louisianism',cf. fn. 1.

Subject-focus and presentative clauses involve an enriched peripheral

structureincluding a copular morpheme,glossed by COP. Subject-focus
clauses are exemplifiedby (20):

(20)a. xale -y(i>a) -a lekk 0 0 0 ceeb -bi

child DFpl COP eat -ipf -pst -neg rice DFsg
It is the childrenwho ate the rice.

b. xale -y(i>a) -a lekk 0 -oon 0 ceeb -bi.

child DFpl COP eat -ipf +pst -neg rice DFsg
It was the childrenwho had eaten the rice.

c. xale -y(i>a) -a (d)i 0 0 lekk ceeb -bi.

child DFpl COP +ipf -pst -neg eat rice DFsg
It is the childrenwho are eating the rice.

d. xale -y(i>a) -a lekk 0 0 -ul ceeb -bi.

child DFpl COP eat -ipf -pst +neg rice DFsg
It is the childrenwho did not eat the rice.

Subject-focus clauses are inflected for tense, aspect, and polarity: we

thereforeassume theirinflectionaldomainto be saturated.
Presentativeclauses19 characteristicallyinclude the copular element
ng-, which combines with the insertion of a to its left (plausibly, the
subject-focusmarker,as hinted by Church 1981), and supportsan obvi-
ation vowel suffix (-i = [-obv], -a = [+obv]) (cf. Sauvageot 1965; Church

(21) presentativeclauses
a. xale -y(i>a)+a -ngi lekk 0 0 ceeb -bi.
child DFpl COP/-obveat -ipf -pst rice DFsg
The childreneat the rice.

b. xale -y(i>a)+a -ngi -(d)i 0 lekk ceeb -bi.

child DFpl COP/-obv+ipf -pst eat rice DFsg
The childrenare eating the rice.
19 Our terminologyhere follows Sauvageot (1965), Church(1981), and Robert (1991).
Ka (1994) calls such clauses sentence-focus, and Fal et al. (1990), situative.

c. xale -y(i>a)+a -ngi d(i) -oon lekk ceeb -bi.

child DFpl COP/-obv+ipf +pst eat rice DFsg
The childrenwere eating the rice.

d. xale -y(i>a)+a -nga d(i) -oon lekk ceeb -bi.

child DFpl COP/+obv +ipf +pst eat rice DFsg
The childrenwere eating the rice (long ago).

Presentativeclauses cannotbe negated:we assumethey areunspecifiedfor

polarity.Depending on their aspectualfeature,presentativeclauses may
be used to translateEnglish generic/stative(-ipf), or progressive(+ipf),
Object-focusis markedby means of yet anothercopular morpheme,
la, a nominal functional element made up of [1], the [-count] classifier
(Sauvageot 1965), and [a], the [+obv] spatializer:

(22) focalized-object clauses

a. ceeb -bi -la xale -yi lekk 0 0 0.
rice DFsg COP child DFpl eat -ipf -pst -neg
It is the rice, which the childrenate.

b. ceeb -bi -la xale -yi lekk 0 -oon 0.

rice DFsg COP child DFpl eat -ipf +pst -neg
It was the rice, which the childrenhad eaten.

c. ceeb -bi -la xale -yi -(d)i 0 0 lekk.

rice DFsg COP child DFpl +ipf -pst -neg eat
It is the rice, which the childrenare eating.

d. ceeb -bi -la xale -yi lekk 0 0 -ul.

rice DFsg COP child DFpl eat -ipf -pst +neg
It is the rice, which the childrendid not eat.

Like subject-focusclauses, object-focus clauses have a saturatedinflec-

tionaldomain.We assumehere (for lack of evidence suggestingotherwise)
thatthe subject-focusmarkera and the object-focusmarkerla occupy the
same syntactic(Focus) head, whose specifierhosts the presentativesubject
or focalized (subjector object)phrase,and we shall leave open the precise

syntactic status of the presentativemarkerngi/nga, which incorporatesto

the focus copula a in Morphology.
Topics typically surface as dislocated phrases (more on this below,
section 3):

(23) xale -yi, nu -a lekk 0 0 0 ceeb -bi.

child DFpl 3pl COP eat -ipf -pst -neg rice DFsg
The children,it is they who ate the rice.

To accommodate this second set of descriptive results, we assume

with Rizzi (1997) that the left peripheryof a Wolof clause includes sev-
eral functional projections, among which are Force, Topic, Focus, and

(24) Wolofclause structure

spec Force'
Forc0 TopicP
spec Top'
Topo FocusP
spec Foc'

ni Foc? PresP
bi 0 Pres'
ba Pres?



fnu n ul n di DOMAIN(vP)



Both the left peripheryand the inflectionaldomain are variously sat-

urated,dependingon the type of clause: for instance,the complementizer
ni selects a Person Phrase for its complement;the determiner(in relat-
ive clauses) selects a Polarity Phrase;the complementizerof TS-clauses
selects a FinitenessPhrase.Table(25) summarizesWolof clause types:

(25) Wolofclause types

type e
inflectedfor V to Infl xale-yilekk(-oon)-na-fAu
person Aux(+ipf)to Infl xale-yidi(-oon)-na-iulekkceeb-bi
____________________ Aux (+F) to Infl xale-yida-fiulekk(-on) ceeb-bi
imperative lekk-al(-een)
optative naxale-yilekkceeb-bi!
narrative xale-yilekkceeb-bi
no dependent-tense ...[xale-yilekk(-oon)ceeb-bil
person relative ceeb-bixale-yilekk(-oon)
inflection tense-setting bi xale-yilekk-eeceeb-bi
negativeimperative buxale-yilekkceeb-bi!
subjectfocus xale-ya-alekk(-oon)ceeb-bi
presentative lekk(-oon)ceeb-bi
object focus ceeb-bi-la xale-yi lekk(-oon)


3.1. Personand LocativeMarkers:A GeneralSurvey

The functional nominals of Wolof relevant for our study of clitics in-
clude person and locative markers,which may first be subclassifiedinto
two phonologicalparadigms:strong-stressforms, and weak-stressforms,
a point we shall returnto below in section 4. Weak forms may be fur-
ther subdividedinto (a) subjectmarkers(a termwe borrowfrom Dunigan
1994) and (b) accusativeand locative markers.The three types are briefly

(26) strongperson and locative markers

a. Moodu liggey 0 0 0 -na 0 ak i{foom/xale-yi}.
Moodu work -ipf-pst -neg +F 3sg with 3pssTR/childDFpl
Moodu has workedwith {them/thechildren).

b. Moodu lekk 0 0 0 -na 0 tfoofu/ ci seez -bil.20

Moodu eat -ipf-pst-neg +F 3sg LOCSTR/on chair DFsg
Moodu ate Ithere(STR)/on the chair}.

(27) subjectmarkers
a. xale -yi lekk 0 0 0 -na -iiu ceeb -bi.
child DFpl eat -ipf -pst -neg +F 3pl rice DFsg
The childrenhave eaten the rice.

b. ceeb -bi -la -inulekk 0 0 0.

rice DFsg COP 3pl eat -ipf -pst -neg
It is the rice, which they ate.

c. nu -a (>fioo) lekk 0 0 0 ceeb -bi.

3pl COP eat -ipf -pst -neg rice DFsg
It is they, who ate the rice.

(28) objectpronounsand weak locatives

a. Moodu lekk 0 0 0 -na 0 -leen.
Moodu eat -ipf -pst -neg +F 3sg 3pl/O
Moodu has eaten them.

b. Moodu lekk 0 0 0 -na 0 -ci.

Moodu eat -ipf -pst -neg +F 3sg LOCWK
Moodu has eaten there (WK).

As shown by Table (29) below, object pronouns are morphologically dis-

tinguished from subject markers in the second and third persons: those that
are distinguished will be specified as 'objects' (0) in our glosses. The term
'subject marker' which we use for convenience's sake does not, in our
opinion, cover a unitary syntactic class of nominative pronouns, a view
we shall make more explicit in section 4: for this reason we only gloss
subject markers by their person/number features, with no further 'subject'
or 'nominative' specification.

(29) Wolof person and locative markers


EA~TURES PERSONANDLOCAIIVE sujt M:.- s objet:pronoun.

MARKERS and weak. locatives

I sg man (nUa ma
3sg ya__ __ a_ _

LOG, -dx -obv cii

LOG, -dx. ca

LOC, +dx, -obv iii =
LO|, +dx, +obv faa. toofu

Functional locatives are specified for obviation (obv) and deixis (dx): loc-
atives glossed as [-dx] may be read as bound variables, locatives glossed
as [+dx] behave as r(eferring)-expressions (the exact nature of the deictic
feature is not relevant to the present study).

20 Interestingly,the morphemewe gloss and translatehere by 'on' is none otherthanthe

weak locative ci of Table (29).
21 Some Wolof dialects exhibit a nonpalatalinitial nasal in Ipl pronounsnoom and nu,
which are hence phonologically distinguishedfrom 3pl fioom and fnu. This distinctionis
absentfrom the Wolof varietyunderconsideration.

3.2. StrongForms
Strong functional nominals occur in positions otherwise open to lexical
DPs (cf. (26a)) and locative phrases (cf. (26b)). Strong person mark-
ers occur in P-governed (30a), dislocated (30b) and object-focus (30c)

(30)a. Moodu liggey 0 0 0 -na 0 ak noom.

Moodu work -ipf -pst -neg +F 3sg with3plSTR
Moodu has workedwith them.

b. nioom,fnu -a (> fioo) lekk 0 0 0 ceeb -bi

3plSTR3plWKCOP eat -ipf -pst -neg rice DFsg
lit., They/them,it is they who ate the rice.

c. noom -la Moodu gis 0 0 0.

3plSTR COP Moodu see -ipf -pst -neg
It is them thatMoodu saw.

Strong personal pronounscannot occur as direct objects (cf. (31a)), and

when they seem to occur in subject position, as in (31b), they are felt
as optional emphatic adjuncts,similar to dislocated topics ratherthan to
proper subject arguments.22Their nonargumentalstatus is confirmedby
22 The same remarkapplies to French strong pronouns which also seem to occur in
subjectposition in, e.g., (i):

(i) Eux ont mange le riz.

3Mpl (STR)have eaten the rice
THEY have eaten the rice.

In French (i) as in Wolof (31b), the strongpronounmust bear contrastivestress, which is

neithertrue of the P-governedstrongpronounof (ii) (the Wolof analogueof (30a)), nor of
the lexical subjectof (iii):

(ii) Jean a travailleavec eux.

Jean has worked with them.

(iii) Les enfants ont mange le riz.

The childrenhave eaten the rice.

This suggests that the markedreading of the strong pronounin (i) does not derive from
some inherentfeatureof the pronounandthatthe strongpronounof (i) does not occupy the

the fact that theiromission does not cause ungrammaticality,as shown by

(31b)-(31c), unlike thatof weak subjectmarkers,as shown by (32):

(31)a. *Moodulekk 0 0 0 -na 0 noom.

Moodu eat -ipf -pst -neg +F 3sg 3plSTR

b. inoom lekk 0 0 0 -na -nu ceeb -bi.

3plSTReat -ipf -pst -neg +F 3pl rice DFsg
THEY have eaten the rice.

c. lekk 0 0 0 -na -ilu ceeb -bi.

eat -ipf -pst -neg +F 3pl rice DFsg
They have eaten the rice.

(32)a. nu -a(>fnoo)lekk 0 0 0 ceeb -bi.

3pl COP eat -ipf -pst -neg rice DFsg
It is they who ate the rice.

b. *-a lekk 0 0 0 ceeb -bi.

COP eat -ipf -pst -neg rice DFsg

These data suggest that the descriptive generalizationput forward by

Kayne (1999) regardingthe strongpronounsof Frenchmay be exported
to Wolof: for reasonsstill unclearto us, strongpronounsin both languages
are banned from structural-casepositions, and are thereforerestrictedto
nonargumentand P-governedpositions. Underthis assumption,the strong
pronoun of (3 ib) must not occur in subject position, but higher up in a
caseless, nonargumentposition. This assumptionis confirmedby the fact
that the strongpronounof (3 lb) is associatedwith a marked(contrastive)
subjectposition, but a higherposition in the clause periphery.Furtherevidence in support
of this assumption,regardingFrench, is the ungrammaticalityof liaison in (i) = (iv-a),
contrastingwith (iv-b):

(iv)a. Eux [*z] ont mang6le riz,

THEY have eaten the rice.

b. ils [z] ont mange le riz.

They have eaten the rice.

reading,whereas the lexical subjectin the same construction(e.g., (27a))

is construedas neutral.23

3.3. WeakPronounsas SyntacticClitics

3.3.1. TheDistributionof WeakPronouns:A First Survey SubjectMarkers. Dependingon the construction,subjectmarkers
undergo morphological attachmentto their left or (so it seems) to their
right, or remainunattached.The complete paradigmis given in Table(33)
for the 3pl subjectmarker,fiu:

(33) The distributionof Wolof's 'subject markers',exemplifiedby

fiu (3pl)24
construction example translation

V-to-lnfl xale-yi lekk-na-nfu) ceeb-bi. The children have eaten the rice.
lekk-na-ii(u) ceeb-bi. They have eaten the rice.
ipf-to-Infl xale-yi di-na-fi(u) lekk ceeb-bi. The children will eat the rice.
di-na-ii(u) lekk ceeb-bi. They will eat the rice.
EXPL-to-Infl xale-yi da-ni(u) lekk ceeb-bi. (It is because) the children have eaten the
L-attachment rice.
da-fi(u) lekk ceebbi. (It is because) they have eaten the rice.

optative na-iiu lekk ceeb-bi! (It is my wish) that they eat the rice!
relative ceeb-bi-iinulekk the rice which they have eaten
TS bi-fiu lekk6e ceeb-bi... when they ate the rice...
neg. imp. bu-fiu lekk ceeb-bi! Let them not eat the rice!
__________ _ -object focus ceeb-bi-la-nu lekk. It is therice, which e ate.
R-attachment subject focus fiu-a(>rioo) lekk ceeb-bi. It is they who ate the rice.
presentative fiu-a-ngi(>fioongi) lekk ceeb-bi They eat the rice.

unattached narrative ilu lekk ceeb-bi. So they eat the rice.

_ DT . lekk ceeb-bi)
[lnu [them to eat the rice]

Right-attachedsubjectmarkersarenot proclitics(Wolof is an initial-stress,

all-enclitic language).In sentence-initialposition, they beardefaultinitial
stress(moreon this in section 4), andundergobonding,25i.e., phonological
coalescence with their rightwardcontext. Bonding does not specifically
23 Our descriptionof strongpronounsis odds
at with Dunigan's, who assumes that the
strongpronounin (31b) occupies the same structuralposition as would a lexical subjectin
the same construction.This issue, however,is a minor one with respect to the analysis of
24 Segments between bracketsundergocontextualdeletion.
25 We borrowthis termfrom Keenan(1996), who uses it to describe Malagasy.

concernweak pronouns;it appliesto lexical DPs as well, as shownby (34)

and (35):26

(34) subjectfocus
a. xale -yi -a (> xaleyaa) lekk ceeb -bi.
child DFpl COP eat rice DFsg
It is the children,who ate the rice.

b. fnu -a (>noo) lekk ceeb -bi.

3pl COP eat rice DFsg
It is they, who ate the rice.

(35) presentative
a. xale -yi -a -ngi (> xaleyaangi)lekk ceeb -bi.
child DFpl COP PRES eat rice DFsg
The childreneat the rice.

b. nu -a -ngi (> ioongi) lekk ceeb -bi.

3pl COP PRES eat rice DFsg
They eat the rice.

Left-attachedsubjectmarkers,on the otherhand, are unstressed,and their

morphologicalattachmentis specific to weak morphemes.In (36), for in-
stance,we see thatthe subjectmarkerundergoesleft-attachment,while the
lexical subjectxale-yi does not:

(36)a. ceeb -bi -la -flu lekk.

rice DFsg COP 3pl eat
It is the rice, which they ate.

b. ceeb -bi -la xale -yi lekk.

rice DFsg child DFpl eat
It is the rice, which the childrenate.
26 Furtherdetails on Wolof bondingmay be found in Ka (1994). To make the examples
typographically shorter, we shall from now on omit from our transcriptionsthe zero
inflectional features which have no relevance for the properties under discussion. Zero
inflectionis relevantfor the analysis of syntacticstructure,but it is irrelevantas regardsthe
positioning and attachmentof clitics.

Furthermore,we shall see below (Table (37)) that left-attached sub-

ject markers serve as morphological supports for object pronouns and
weak locatives (hereafter:OLCs: Object and Locative Clitics), whereas
right-attachedand unattachedsubjectmarkersdo not. Object Pronouns and WeakLocatives. Unlike subject markers,

OLCs are always enclitic. Like subject markers,however,they surface in
differentlinearpositions dependingon the construction.A survey of their
distributionis given below in Table (37), taking as samples the 3sg object
markerko and the [-obviative] weak locative ci:

(37) The distributionof Wolof's OLCs

construction [lexical subject [functional subject

V-to-Infl xale-yi lekk-na-fi(u) -ko. lekk-na-fi(u)-ko.

The children have eaten it. They have eaten it.
Ipf-to-Infl xale-yi di-na-fi(u) -ko lekk. di-na-fi(u) -ko lekk.
The children will eat it. They will eat it.
EXPL-to-Infl xale-yi da-fi(u) -ko lekk. da-fi(u) -ko lekk.
(It is because) the children have eaten it. (It is because) they have eaten it.

optative na-ko xale-yi lekk! na-fiu-ko lekk!

It is my mish that the childreneat it! It is my wish that they eat it!
relative ceeb-bi-ci xale-yi lekk ceeb-bi-fiu-ci lekk
the rice which the children ate there the rice which they ate there
TS bi-ko xale-yi lekk&e... bifisu-ko lekk6e...
when the childrenate it... when they ate it...
neg. imp. bu-ko xale-yi lekk! bu-fiu-ko lekk!
Let the children not eat it! Let them not eat it!
object focus ceeb-bi-la-ci xale-yi lekk. ceeb-bi-la-fiu-ci lekk.
It is the rice, which the children ate there. It is the rice, which they ate there.

subject focus xale-yi-a(>yaa)-ko lekk fiu-a(>fioo)-kolekk.

It is the children who ate it. It is they who ate it.
presentative xale-yi-a-ngi(>yaangi)-ko lekk. fiu-a-ngi(>fioongi)-kolekk.
The childreneat it. They eat it.
narrative xale-yi lekk-ko. hlulekk-ko.
So the childreneat it. So they eat it.
DT ..lxale-yi lekk-ko] ...[iu lekk-ko]
(want)...thechildrento eat it. (want)...them to eat it.

The constructionslisted in Table (37) are subdividedinto three subtypes

(from the top down). In the firstsubtype,which includes the V-to-Infl,Ipf-

to-Infl,and Expi-to-Inflconstructions,OLCs encliticize to the rightof the

subjectmarker,which itself encliticizes to the inflected verb or auxiliary,
and the behaviourof OLCs is insensitive to whetheror not a lexical DP
occursfurtherup in subjectposition.In the second subtype,which includes
the optative,relative,TS, negative-imperative,and object-focusconstruc-
tions, OLCs are linearized differently when the subject is realized as a
lexical DP and when it is realized as a functional 'subject marker'.The
third subtype, which includes subject-focus,presentative,narrative,and
DT constructions,is characterizedby the fact thatOLCs do not encliticize
to subject markers,but are separatedfrom them by a copula or a lexical

3.3.2. SyntacticClitics and Clitic Movement

Dunigan (1994) infers from the data in Table (29) that subject markers
and OLCs form a class, distinctfrom strongpronouns.Since left-attached
subjectmarkersexhibitenclitic properties(cf. (36)), and since they further-
more clusterwith OLCs (cf. (37)), she concludes that subjectmarkersare
themselves clitics, which, from an all-syntacticperspective, means syn-
tactic clitics. To describe formally the distributionof clitics in (37), she
assumes thattheir clitic behaviouris an effect of Clitic Movement,under-
stood as syntactic Head Movement. In two subclasses of cases, however,
we saw that subject markersdo not cliticize and occur in linearpositions
otherwise open to full DPs (cf. (34), (35)). To account for these cases
withoutdiscardingthe unitary'syntactic-clitic'analysis,Duniganborrows
from Sportiche (1996) the idea that clitics are a special type of morph-
emes which head their own syntactic projections, called clitic phrases.
This assumptionmakes it possible to smooth out the paradoxicalfact that
some clitics are not cliticized: cliticized clitics are claimed to behave as
the heads of their Clitic Phrases,and hence to undergoClitic Movement,
while uncliticized clitics are claimed to behave as XPs, i.e., as full Clitic
Phrases,positioned in specifier positions. Applying this theory to Wolof,
Dunigan assumes that all the weak pronounsand locatives of Table (29)
are syntactic clitics, but that only those which behave as enclitics in (33)
and (37) undergoClitic Movement:it follows thatOLCs all undergoClitic
Movement,while those subject markerswhich do not surfaceas enclitics
(cf. (34b), (35b)) areassumedto behavesyntacticallyas full (clitic) phrases
occurring in specifier positions.

3.3.3. The Targetof CliticMovement TheAttraction-by-Inflection Theory. An all-syntacticanalysis of
clitics requiresthat the target of Clitic Movementbe specified in strictly

syntacticterms.In independentfinite clauses such as (38a) and (38b), the

subject markerfollowed by OLCs attachto the tensed verb or auxiliary.
In DT-clauses,OLCs similarlyattachto the tensed verb (39a) or auxiliary

(38)a. xale -yi lekk -na -flu -ko.

child DFpl eat +F 3pl 3sg/O
The childrenhave eaten it.

b. xale -yi di -na -flu -ko lekk.

child DFpl +ipf +F 3pl 3sg/O eat
The childrenwill eat it.

(39)a. Arambegg -oon -na [xale -yi lekk -oon -kol.

Aram want +pst +F child DFpl eat +pst 3sg/O
Aram wantedthe childrento eat it.

b. Aram gis -oon -na [xale -yi d(i) -oon -ko lekk].
Aram see +pst +F child DFpl +ipf +pst 3sg/O eat
Aramhad seen the childreneating it.

c. *Aramgis -oon -na [xale -yi d(i) -oon lekk -ko].

Aram see +pst +F child DFpl +ipf +pst eat 3sg/O

This firstset of datais probablywhatled Njie (1982) to assumethatWolof

OLCs move in syntax underthe attractionof Tense. This idea could also
27 In Wolof as in, e.g., Italian or Classical French, a few modal and movement
behave like auxiliarieswith respect to OLC placement,cf.:

(i)a. xale -yi begg -na -fnu-ko lekk.

child DFpl want +F 3pl 3sg/O eat
lit., The childrenit-wantto eat.
(cf. Classical French:Les enfantsl'ont voulu manger.)

Gambian-Wolofsemi-auxiliaries(a termwe borrowfrom Zagona 1982) are listed by Njie

(1982, p. 123) and include begg 'want', dem 'go', now 'come', safi 'dare', bani'refuse',

accountfor the special behaviourof OLCs in double-objectconstructions,

exemplifiedby (40c, d, e), contrastingwith (40a, b):

(40)a. xale -yi wan -na -flu Aram ceeb -bi.

child DFpl show +F 3pl Aram rice DFsg
The childrenhave shown Aramthe rice.

b. *xale -yi wan -na -flu ceeb -bi Aram.

child DFpl show +F 3pl rice DFsg Aram

c. xale -yi wan -na -flu -ko -ko.

child DFpl show +F 3pl 3sg/O 3sg/O
The childrenhave shown Ihim/herI Ihim/her/it}.

d. xale -yi wan -na -flu -ko Aram.

child DFpl show +F 3pl 3sg/O Aram
(i) The childrenhave shown Aramto Ihim/her}.
(ii) The childrenhave shown Ihim/her/itI to Aram.

e. *xale -yi wan -na -flu Aram (-)ko.28

child DFpl show +F 3pl Aram 3sg/O

28 The distributionof Wolof OLCs in double-objectconstructionsis partlyparallel to

that of English unstressedpronouns in similar contexts - as described and analysed by
Zwicky (1986):

(i) *Johnshowed Maryhim. (cf. (40e))

(ii) John showed him Mary. (cf. (40d), reading(i))

Zwicky's assumptionthatunstressedpronounscannotform a prosodicphraseof theirown

may be extended to Wolof to account for the ill-formednessof (40e). Wolof OLCs, how-
ever, furthercontrastwith English pronounsin that they may not bear contrastivestress;
thus (iii) is well-formedin English, while (iv) is not in Wolof, for ko is a weak-stressitem,
and the only availablestress for a Wolof weak-stressitem is default, domain-initialstress
(see section 4 below):

(iii) John showed Mary HIM.

(iv) *Moodu wan -na 0 AramKO.

Moodu show +F 3sg Aram 3sg/O


In Wolof as in English canonical double-objectconstructions,the Goal

must precede the Theme (cf. (40a) vs. (40b)). Correlatively,the linear
orderin the Wolof translationof (40d-ii) should be: Aram>pronoun.This,
however, is made impossible (cf. (40e)) by the fact that the OLC must
attach to the inflected verb, which may be taken as evidence that Tense
attractsthe clitic.
However, the attraction-by-tenseidea runs into empirical problems.
In explicative clauses, OLCs attach to the subject markerwhich itself
encliticizes to the explicativeauxiliary,locatedabove Tense and Aspect:

(41) xale -yi da -nlu-ko lekk -oon.

child DFpl +F 3pl 3sg/O eat +pst
(It is because) the childrenhad eaten it.

We might consider the alternativeassumptionthat OLCs are attractedby

Finiteness,ratherthanby Tense. This could also accountfor the behaviour
of OLCs in optativeclauses:

(42) na -ko xale -yi lekk!

+F 3sg/O child DFpl eat
(It is my wish that)the childreneat it!

The attraction-by-finiteness idea is, however,provedincorrectby the fact

that OLCs also surface in special positions in nonfinite clauses. In DT-
clauses such as (39b), the imperfective auxiliary targeted by OLCs is
not inflected for finiteness. And in narrativedouble-objectconstructions,
OLCs seem just as attractedby the noninflectedverb as they were by the
finite one in (40):

(43)a. xale -yi wan -ko -ko.

child DFpl show 3sg/O 3sg/O
So the childrenshow {him/her} Ihim/herfit}.

b. xale -yi wan -ko Aram.

child DFpl show 3sg/O Aram
(i) So the childrenshow Aramto {him/her}.
(ii) So the childrenshow {him/her/it}to Aram.

c. *xale -yi wan Aramko.

child DFpl show Aram 3sg/O

In (nonfinite)clauses which activatea Focus projection, OLCs attachto

the focus marker,whose clausal domainwe have shown to be nonfinite:

(44)a. xale -yi -a -ko lekk. [yi-a>yaa]

child DFpl COP 3sg/O eat
It is the children,who ate it.

b. xale -yi -a -ngi -ko lekk. [yi-a-ngi>yaangi]

child DFpl COP PRES3sg/O eat
The childreneat it.

c. ceeb -bi -la -ci xale -yi lekk.

rice DFsg COP LOC child DFpl eat
It is the rice, which the childrenate there.

Finally, in TS-clauses such as (45), OLCs targetthe initial complement-

izer/determinermorpheme,which is beyond the Finitenessprojection:

(45) bi -ko xale -yi lekk -e'e...

C?obv3sg/O child DFpl eat -F
When the children ate it ... The TopmostFunctional Head Theory. The above data led

Dunigan(1994) to propose the following alternativegeneralization:

(46) Dunigan's Theoryof WolofClitics

Wolof clitics move in syntax to the topmost functional head
within the g(rammatical)-projection
of V.

The g-projectionof V is definedas the maximalfunctionalprojectionof V,

leaving out CP.The reasonfor leaving out CP is thatenclitics do not attach
to the complementizerni which introducesfinite complementclauses, as
witnessedby (47b), contrastingwith (47a):

(47)a. Aram wax -na -ni xale -yi lekk -na -flu -ko.
Aram say +F that child DFpl eat +F 3pl 3sg/O
Aram said thatthe childrenate it.

b. *Aram wax -na -ni -ko xale -yi lekk -na -flu.
Aram say +F that 3sg/O child DFpl eat +F 3pl

Generalization(46) is devised to predict the distributionof both OLCs

and enclitic subject markers,assumed to cluster on the same head. In all
constructionsbut two (DT and narrativeclauses), clitics are assumed to
targeta single functionalhead, labeled E, the locus of 'clausal particles'
which include the na morphemeof finiteandoptativeclauses, the negative-
imperativemarkerbu, the initialcomplementizer/determiner morphemeof
TS-clauses, the determiner-headof relativeclauses, and the copula morph-
eme of focus and presentativeclauses. In DT and narrativeclauses, which
do not containa E projection,clitics are assumedto targetthe Tense head,
to which the verb or auxiliaryhas raised. Dunigan'sanalysis is consistent
with the set of data summarizedin Table(48):

(48) Dunigan'ssyntacticanalysis of Wolof clitics

Dunigan's construction locus of enclitics translation
analysis ype
V-to-Infl xale-yi lekk-na-fiu-ko The childrenhave eatenit.
ipf-to-Infl xale-yi di-na-fiu-kolekk. The childrenwill eat it.
F-to- Infl xale-yida-fiu-ko lekk. (It is because)the childrenhaveeatenit.
Clitic Movement optative na-fiu-kolekk! (It is my wish) thatthey eat it!
targets relative ceeb-bi-iu-ci lekk the rice whichthey ate there
s TS bi-fiu-kolekkeie... Whentheyate it...
neg. imp. bu-fiu-kolekk! Let themnot eat it!
objectfocus ceeb-bi-la-fiu-c lekk. It is the rice, whichtheyate there.
subject focus fiu-s(>oo)-ko lekk. It is they who ate it.
______________ presentative fiu-a-ngi(>fioongi)-ko lekk They eat it.

Clitic Movement narrative fiulekk-ko. So they eat it.

targets T DT .*linulekk-kol [them(to) eat it)

Dunigan's syntactic generalization (46) captures the intuition that in

most cases, OLCs cluster with one or several monosyllabic functional
morphemes.It also correctlypredictsthe contrastin (49):

(49)a. Aram gis -oon -na [xale -yi d(i) -oon -ko lekk].
Aram see +pst +F child DFpl +ipf +pst 3sg/O eat
Aram had seen the childreneating it.

b. Aramgis -oon -na [xale -yi -ko d(i) -oon lekk].

Aram see +pst +F child DFpl 3sg/O +ipf +pst eat
Aram had seen the childrenwho were eating it.

In (49a), where the object clitic is attachedto the inflected auxiliary,the

bracketedstring can only be read as a phrase headed by Tense: a (DT)

clause; in (49b), wherethe object clitic is attachedto the determineryi, the

bracketedstring can only be read as a phraseheaded by yi: a relativized

3.3.4. Discussion of Dunigan's Analysis

Dunigan'sanalysis of Wolof clitics startsout with a syntacticbias regard-
ing cliticization:as most syntacticiansdealing with this issue since Kayne
(1975), Dunigan postulates without demonstrationthat special clitics, in
the sense of Zwicky (1977), are moved to their 'special' position via
syntacticHead-Movement.She does not find it necessaryto examine and
discuss otherpossible or availableapproachesto cliticization,and actually
provides no compelling evidence thatWolof clitics undergomovementin
syntax. As pointed out in section 1, the fact that clitics are positioned
within some local domain with respect to the verb (what she calls the
g-projectionof V) could just as well be predictedunder a lexical, mor-
phological, or phonological analysis. Dunigan reasons that since enclitic
pronouns and WK locatives cluster on other monosyllabic morphemes
plausiblygeneratedin syntacticheads, they must attachto them as a result
of Head Movement, hence undergoingClitic Movement in syntax. This
conclusion, however, is in no way necessary,for the clustering of clitics
might be sensitive to morphophonological,ratherthan to syntactic,prop-
erties. Consider,for instance,the contrastbetween the lexical subject and
the subjectmarkerin (36), repeatedbelow:
(36)a. ceeb -bi -la -nlulekk.
rice DFsg COP 3pl eat
It is the rice, which they ate.

b. ceeb -bi -la xale -yi lekk.

rice DFsg child DFpl eat
It is the rice, which the childrenate.
This pair of examples shows that in the object-focus construction,the
subjectmarker(inu)encliticizes to the focus-markingcopula, whereasthe
lexical subject (xale-yi) remains unattached.Under Dunigan's view, this
shows thatthe subjectmarkeris a clitic, thereforea syntacticclitic, which
raises up to E (filled by the copula) via syntacticClitic Movement,while
the lexical subjectis not a clitic, hence remainsin spec,TP.It is, however,
equallypossible to assume, as we shall in section 4, thatthe lexical subject
andthe functionalsubjectoccupy the same structuralposition in the output
of syntax- arguably,spec,TP - and thatthe subjectmarkerundergoesen-
clisis for phonologicalreasons- because it is a weak-stressmonosyllable,

unlike the lexical subject. But Dunigan does not even consider such an
The all-syntactic approachwhich she develops furthermoreraises a
numberof descriptiveproblems.First,recall thatthe stringa+ngi which is
targetedby OLCs in presentativeclauses is likely to be made up of two
morphemes generated in two distinct syntactic heads (cf. Church 1981
and section 2 above); under this assumption,the behaviour of OLCs in
presentativeclauses violates generalization(46). Second, note that the
existence of a syntacticrule of Clitic Movementmay be disputedon the-
oretical grounds, since we may regardas doubtfulthat Head Movement,
as a whole, should pertainto syntax (cf. Collins 1997; Chomsky 1999).29
Furthermore,the Clitic Phraseassumptionon which Dunigan'sanalysis is
based is only motivatedby an a prioriconceptionof the syntax-phonology
interface,accordingto which weak-stresson functionalnominalsis a reflex
of syntax, ratherthan an independentphonological property.Under the
all-syntacticview, weak-stressfunctionalnominals must form a syntactic
class, thereforemustbe 'clitics' even when they do not cliticize. The Clitic
Phrase analysis results in various descriptiveassumptionswhich receive
no independentjustification.Considerin particularthe subject markerof
finite clauses such as (50), which we proposed to analyse above as the
V-to-Infl,Ipf-to-Infland Expl-to-Inflconstructions:

(50)a. V-to-Infl
xale -yi lekk 0 0 0 -na -flu ceeb -bi.
child DFpl eat -ipf -pst -neg +F 3pl rice DFsg
The childrenhave eaten the rice.

29 Collins and Chomskyrightlypoint out thatHead-to-Headmovementin syntaxresults

in illegitimate traces. Thus, if V raises up (i.e., adjoins) to T in syntax, it does not c-
commandits trace,as exemplifiedbelow by the Frenchinflectedform mange-ait,(eat/PST)
'was eating' or 'used to eat':


V0 TO V0
mange, -ait tz

To avoid this problem, Collins (1997) and Chomsky (1999) suggest that Head-to-Head
movementshould occur in Morphologyand leave no trace.

b. Ipf-to-Infl
xale -yi di 0 0 -na -fnulekk ceeb -bi.
child DFpl +ipf -pst -neg +F 3pl eat rice DFsg
The childrenwill eat the rice.

c. Expl-to-Infl
xale -yi da -flu lekk 0 0 0 ceeb -bi.
child DFpl EXPL3pl eat -ipf -pst -neg rice DFsg
(It is because) the childrenhave eaten the rice.

These and only these constructionsinstantiatesaturatedfinite clauses, i.e.,

independentclauses specifiedfor the full range of inflectionalfeatureslis-
ted in section 2: Finiteness,Aspect, Tense, and Polarity.In these and only
these constructions,the subject markeroccupies a linear position which
is closed to a lexical subject. And in these and only these constructions,
the lexical subject (xale-yi) is construed as neutral (i.e., untopicalized,
unfocalized). These propertiesare quite naturallyconsistentwith the ana-
lysis proposed above in section 2, which treats the subject markeriiu in
(50) as an inflectionalelement (a person-inflectionor 'subject-agreement'
marker),crucially correlatedwith (the positive value of) finiteness, and
generatedin the topmostfunctionalhead of the inflectionaldomain.Such
analysis is, however,in starkconflict with the assumptionthat all enclitic
pronouns start out in the heads of Clitic Phrases generatedin argument
positions and move up to some functional head in syntax. In order to
discard an inflectional analysis of the subject markersof (50), Dunigan
arguesthat subjectmarkersobviously cannotbe analysedas the spell-outs
of Person inflectionin such cases as (51la)(where fiu agrees with a distant
antecedent),or (5 lb) (whereinusurfaceson the same side of the copula as
its antecedent):

(5 1)a. objectfocus construction

xale -yi ceeb -bi -la -n-ulekk.
child DFpl rice DFsg COP 3pl eat
The children,it is the rice, which they ate.

b. subjectfocus construction
xale -yi, nu -a lekk ceeb -bi. [fnu-a>fnoo]
child DFpl 3pl COP eat rice DFsg
The children,it is they who ate the rice.

Dunigan rightly assumes that in such constructions,the subject marker

behaves like a resumptivepronouncoindexed with a dislocatedtopic. But
she then goes on to argue that since subject markersbehave as subject
argumentsin (51), they must be subject argumentswhereverthey occur.
Since in (50), the antecedentof the subjectmarkermay be shown to behave
as an argument,Dunigan proposes to identify flu, in this example, as a
'clitic-doubler',coindexed with a DP in argumentposition - an analysis
that, incidentally,fails to accountfor the cruciallyfinite characterof (50),
since clitic-doublingis known to be insensitive to finiteness.30Dunigan's
reasoningis based on the postulatethatall instancesof, e.g., fnu,must spell
out the same syntacticentity.Should we refuse this a prioriassumption,it
is quite possible to arguethat flu spells out Personinflectionin (50) while
it is generatedin (51) in an argumentposition - as a nominativepronoun.
An alleged advantageof Dunigan'sanalysis is thatit providesa unitary
theory of Wolof clitics, phrased in (46). This unity is based on the as-
sumptionthatClitic Movementoccurs in syntax and targets- wheneverit
is present - a single locus, X, a functional Propositional-Operator head
placed above Tense and below CP in the syntactic tree. It strikes us,
however, as unlikely that Wolof clause structureshould involve a single
projection (EP) where other languages usually distinguish at least three
(Topic, Focus, Polarity- cf. Rizzi 1997). Dunigan's E hypothesisresults
in presenting Wolof as a strongly exotic language set apart from more
familiar ones, and furthermorerequiresseveral disputabledescriptiveas-
sumptions. For instance, to correctly predict (under the E analysis) the
distributionof enclitic subjectmarkersin negativeclauses, Dunigan is led
30 Cf. the following Spanishexamples:

(i) Dative clitic-doubler

a. Mariale dio un regalo a Juan. [finite clause]
Maria gave a present to Juan.

b. Serfamejor darleun regalo a Juan. [nonfiniteclause]

It would be better to give a present to Juan.

(ii) Accusativeclitic-doubler(Rio de la Plata Spanish)

a. Marialo visito a Juan. [finite clause]
Maria visited Juan.

b. Serlamejor visitarloa Juan. [nonfiniteclause]

It would be betterto visit Juan.

to assumethatnegationis generatedin E when the subjectmarkerattaches

to it (bu-fiu lekkceeb-bi! 'Let them not eat the rice!'), but in a lower pos-
ition when the subjectmarkersurfaces'to its left (bi-fiu lekk-ul-eeceeb-bi
'when they did not eat the rice'). To achieve a unitaryanalysis of clitics,
Dunigan is led to assume that narrativeclauses contain an empty tense
head, a claim for which empirical evidence is lacking since, as pointed
out in section 2, no overt tense or aspect markercan ever occur in such
clauses. There is no obvious unity in the claims that clitics behave either
as headsor as phrases,thatthey may targeteither E or tense, or thatsubject
markersare either clitic doublers or resumptiveclitics. Our own analysis
will lead us to conclude that Wolof clitics indeed do not form a class at
any level of representation.But while this lack of unity is expected undera
modularapproachwhich treatsphonological,morphological,and syntactic
propertiesas largelyindependentfromone another,it is not expectedunder
the assumptionthatclitics form a syntacticclass.

3.3.5. Possible Updatingsof Dunigan's Theory

Dunigan'sanalysis is basically in keeping with more recent syntacticthe-
ories of special clitics such as those of Cardinalettiand Starke (1999),
Rouveret(1999), or Nash and Rouveret(1999), which propose theoretical
groundsfor such generalizationsas (46): underCardinalettiand Starke's
assumptions,Wolof OLCs would be forcedto raise by theirinherentstruc-
turaldeficiency; under (Nash and) Rouveret's,they would be by the fact
that, for some reason,the argumentfeaturewhich they spell out could not
be checked within the VP and would thus have to be 'externalized'on a
higherfunctionalhead. (Nash and) Rouveret'sfeature-checkingapproach,
however,cruciallypredictsthatclitics should always move up to the func-
tional head which is closest to their argumentposition, which is not what
happensto the objectpronounin (52):

(52) xale -yi -a -ko d(i) -oon lekk [yi-a>yaa]

child DFpl COP 3sg/O +ipf +pst eat
It was the childrenwho were eating it.

Furthermore,(Nash and) Rouveret's assumption that enclisis generally

correlateswith a nondistinctiveverbalinflectiondoes not exportto Wolof,
an all-enclisis language with a transparentlydistinctive verbal morpho-
31 According to (Nash and) Rouveret, enclisis obtains when distinctive verbal
phology is lacking. Verbalmorphology is viewed as distinctive if it involves an explicit
person-numbermarking,segmentally distinct from the verb root, as in Italian canta-0-no

Cardinaletti and Starke's theory is based on the assumption that

morphophonologicaldeficiency reflects a deficient syntactic featurecon-
tent. It crucially predicts that weak and clitic pronouns should be 'less
referential' than strong pronouns, in one sense or another.As regards
Wolof, however, this prediction does not seem supportedby empirical
evidence: for instance, both strong and clitic pronouns may refer to an-
imates or inanimates,as shown by (53); conjoinedpronouns(which must
be strong) may refer to both animatesand inanimates(cf. (54)); and both
strong and clitic pronounsmay be read as bound variables,as shown by

(53)a. {Moodu/ordinater-bi}, Arambegg -na -ko.

MoodulcomputerDFsg Aram like +F 3sgAWK
Moodu/thecomputer,Aramlikes {himlit}.

b. {Moodulordinater-bi }, Aramjende -na buumngir moom.

MoodulcomputerDFsg Aram buy +F wire for 3sgSTR
Moodu/thecomputer,Aramhas bought a wire for it/himI.

(54)a. (Moodu) Aramgis -na Nafi ak moom

Moodu Aram see +F Nafi and 3sgSTR
(Moodu) Aramsees both Nafi and him.

b. (ordinater-bi) Aramgis -na teeri -bi ak moom.

computer DFsg Aram see +F book DFsg and 3sgSTR
(The computer)Aramsees both the book and it.

(55)a. Aram, begg -na Moodu {gis -ko [liggey ngir moom}, ak
Aram want +F Moodu see 3sg/O//work for 3sgsTR and
Maryamaktam (zlk)
Maryama too
Aram, wants Moodu to {see herz/work for herz}, and
Maryamaktoo (z/k).
'sing/PRS/3pl'. Correlatively,clitics procliticizeto the Italianfinite verb, but they encliti-
cize to the Italian infinitive verb, which does not bear distinctive morphology.The same
assumptionis meantto accountfor the fact thatenclisis is generalizedin Semiticlanguages,
which typically exhibit a nondistinctive(nonconcatenative)verbalmorphology.
32 These violations of Cardinalettiand Starke'stheory are not specific to Wolof: strong
pronounsmay also refer to inanimateswhen conjoined, and be read as bound variablesin,
e.g., Malagasy (cf. Zribi-Hertzand Mbolatianavalona1999) and French (cf. Zribi-Hertz

b. Ordinater -bii lacc -na Moodu {gis -ko, /liggey ak

computer DM need +F Moodu see 3sg/O /fwork with
moomz}, ak teere -biik tam (z/k).
3sgSTR andbookDM too
This computerzneeds Moodu to Isee itz/workwith it, }, andthis
bookz too (z/k).

3.3.6. Conclusion
We concludethatthe all-syntacticanalyses of clitics reviewedabove either
make wrong predictions for Wolof or must sacrifice explanatoryto de-
scriptiveadequacy.In the next section, we shall arguethat the Wolof data
should be accountedfor in a modularframeworknot involving syntactic
clitics, nor Clitic Phrases,nor Clitic Movement.


We shall now argue that the phonological, syntactic, and morphological

propertiesof Wolof person and locative clitics are independentfrom one
another, and that clitics do not form a linguistic class at any level of
grammaticalrepresentation.From a phonological perspective,Wolof clit-
ics exhibitone of the two regularbehavioursof all weak-stressitems in this
language;from a syntacticperspective,they spell out either theta-marked
expressionsor inflectionalpersonfeatures;and from a morphologicalper-
spective, they instantiateweak words or word-levelaffixes or phrase-level
affixes. We shall propose a global analysis of Wolof functionalnominals
which fits togetherthese independentpropertiesin the syntax-phonology

4.1. Phonology: Strongvs. WeakStress

The Wolof items which Duniganidentifiesas (syntactic)clitics are subject
markerson the one hand, and object pronouns and functional locatives,
on the other:in other words, the weakforms of Table (29). We shall first
argue that the phonological propertiesof these items are common to all
weak-stressmorphemes,in Wolof.

Wolof words regularly bear initial stress (cf. Ka 1978, 1994; Diallo
1981a), a property independent from vowel length, as pointed out by
Sauvageot(1965). This is illustratedin (57) by a few examples:

(56) Wolofinitial stress

short-short [jabar] 'wife'
long-short ['saaku] 'bag'
short-long ['ginaar]'chicken'
short-short-short ['kakator]'chameleon'
short-long-short ['banaana]'banana'
short-short-long ['asamaan]'sky'

Monosyllablesmay be subdividedinto two groups.Some, among them all

lexical monosyllables (e.g., ceeb 'rice', lekk 'eat'), are like polysyllables
(cf. (56)) in that they never behave as leaners, hence always surface as
completeprosodicwords:we assumethattheybearinherent,lexical stress,
and analysethem as strong-stress(STR) items. Othermonosyllables,all of
which are of a functionalnature,33behave as left-attachedleaners (enclit-
ics), except when they occur in initial position in their prosodic domain.
This latterbehaviouris illustratedin (57) by na (labeled above a finiteness
marker),bi (determiner/complementizer), and the 3pl subjectmarkerniu:

(57) WKitems
a. noninitialposition: ATTACHMENT(enclisis)
['lekk-na-niu]'they have eaten', ['ceeb-bi] 'the rice'

b. domain-initialposition: STRESS
['na xale-yi lekk] 'thatthe childreneat'
['iiu lekk ceeb] '. . . them (to) eat rice'
['bi xale-yi lekk-ee. . . ] 'when the childrenate. . .

Due to their bearingstress, domain-initialWK items are legitimatephon-

ological supportsfor enclitics, e.g. ['bi-nu lekk-ee... ] 'when they ate...'.
We assumethatWK items have no stressof theirown, and, in this straight-
forwardlyinitial-stresslanguage, receive default stress in domain-initial
33 WK items all have a functional content, as assumed by, e.g., Selkirk (1984). Con-
versely, however, all functionalmorphemesare not WK items, as witnessed by the strong
pronounsand locatives of Table (29).

position. In noninitial positions, they fail to be stressed and must un-

dergo attachmentin order to enter a phonologically well-formed word.
This double behaviourmay be accountedfor in purelyphonologicalterms
and is not specific to the items listed in Table (29), which include neither
na nor bi. Another relevant example is ni, which occurs either as a
complementizer-determiner heading a TS construction,as in (58a), or as
the head of a complementclause, as in (58b):

(58)a. ni xale -yi lekk -ee ceeb -bi...

as child DFpl eat -F rice DFsg
As the children ate the rice . . .

b. Aram wax -na -ni xale -yi lekk -na -fnuceeb -bi.
Aram say +F that child DFpl eat +F 3pl rice DFsg
Aram said thatthe childrenate the rice.

As rightlypointedout by an anonymousreviewer,ni spells out two distinct

items in (58a) and (58b): in (58a), it has the same internalmakeup(classi-
fier+spatializer)as otherWolof determinermorphemes(e.g., bi); in (58b),
ni is etymologically a verb meaning 'say',34which is also used to intro-
duce the complementclause of some verbs (e.g., wax, which also means
'say', or xam 'know', but not gis 'see'). Ourpoint here is thatthe nonclitic
or clitic behaviour of ni in (58a) and (58b) follows the general pattern
sketchedin (57) for WK items: ni receives default stress in domain-initial
position, as do na, fiu, and bi in (57b); and it encliticizes to the adjacent
word (the inflected matrixverb) in (58b), as do na, fiu, and bi in (57a). A
relevantquestionis: why is the complementclause of (58b) not treatedas
an independentprosodicdomain,withinwhich ni would bearinitialstress?
We assumethatin linkingthe complementclause to the matrixdomain,the
prosodic rules of Wolof acknowledge,in the syntax/phonologyinterface,
the selectionalrelation- accuratelypointed out by Dunigan- between ni
and the matrixverb.
We conclude the following: (i) the 'clitic' behaviourof the weak items
of Table (29) is in no way specific to the subtype of functionalnominals
that Dunigan analyses as 'clitic phrases';(ii) Wolof clitics do not form a
class even at the phonological level: they exhibit one of the two possible
andpredictablebehavioursof Wolof WK items, andthe relevantgrammat-
ical distinction for this language is between STR and WK items, rather
thanbetween clitics and nonclitics.
34 Underthis readingni has a ne variant,which it does not have in (58a).

4.2. Syntax:SubjectPronounsvs. Person Inflection

4.2.1. Clitic SubjectMarkersin XP Positions
Consider the pair of examples given in (59), which illustrate the TS-
(59)a. bi xale -yi lekk -ee ceeb -bi...
C-obv child DFpl eat -F rice DFsg
When the childrenate the rice ...

b. bi -nlulekk -ee ceeb -bi...

C-obv 3pl eat -F rice DFsg
When they ate the rice...

These two examples contrastas to the featurecontentof the subject:(59a)

contains a lexical subject (xale-yi 'the children'), while (59b) contains a
functionalsubject(the 3pl markerfiu). As transcribedby hyphenation,the
subject markerof (59b) is spelt out as enclitic, while the lexical subject
of (59a) is not. Under the all-syntacticanalysis of clitics put forwardby
Dunigan(1994), this contrastis due to the fact thatthe subjectmarker(but
not the lexical subject) is a syntactic clitic, and hence undergoes Clitic
Movementin syntax. However,there does not seem to exist any evidence
- other than phonological attachment- that the lexical subject and the
subject markerdo not occupy the same position in syntax. Syntacticevid-
ence rathersuggests the opposite, since both the lexical subject and the
subject markerin (60) must be linearly adjacentto bi (to their left) and
to the tensed item (to their right), as witnessed by the distributionof the
adverbialtalaata 'Tuesday':

(60)a. *bi talaata Iflu /xale -yi} lekk -ee ceeb ...
when Tuesday3pl Ichild DFpl eat -F rice

b. *bi I{nu/xale -yi I talaata lekk -ee ceeb ...

when 3pl IchildDFpl Tuesdayeat -F rice

c. *bi (flu /xale -yi} lekk -ee talaata ceeb ...

when 3pl IchildDFpl eat -F Tuesdayrice

d. bi Infu/xale -yi I lekk -ee ceeb talaata...

when 3pl Ichild DFpl eat -F rice Tuesday
When the children ate rice (on) Tuesday ...

e. talaata, bi f flu/xale -yi I lekk -ee ceeb ...

Tuesdaywhen 3pl Ichild DFpl eat -F rice
(On) Tuesday,when the childrenate rice ...
The adverbial may only surface in a left or right peripheralposition
within the clause. The linear position of adverbialsis constrainedin the
same fashion in optative, relative, negative imperative,and object-focus
constructions,as shown by (61):
(61)a. na (*talaata)Iflu/xale -yi} lekk ceeb (talaata)!
+F Tuesday 3pl Ichild DFpl eat rice Tuesday
It is my wish that {they/thechildrenI eat rice on Tuesday!

b. ceeb -bi (*talaata)Iflu /xale -yi I lekk (talaata)

rice DFsg Tuesday 3pl IchildDFpl eat Tuesday
the rice which f they/thechildrenI ate on Tuesday

c. bu (*talaata)Iflu /xale -yi I lekk ceeb (talaata)!

C+neg Tuesday3pl IchildDFpl eat rice Tuesday
Let Ithey/thechildrenI not eat rice on Tuesday!

d. ceeb -la (*talaata)Ifu /xale -yi I lekk (talaata).

rice COP Tuesday3pl IchildDFpl eat Tuesday
It is rice, which {they/thechildrenI ate on Tuesday.
Empiricalevidence thus fails to indicate that subject markersand lexical
subjects occupy different syntactic positions in the five classes of cases
underconsideration.These data are, however,readily consistent with the
assumptionthat lexical and functionalsubjectsoccupy the same syntactic
position35 in these constructionsand that the enclitic behaviourof subject
markersregularlyderivesfrom pattern(57a).
Lexical subjectsand subjectmarkersdo contrast,in the above construc-
tions, in that the former,but not the latter,may be left-dislocated(a point
acknowledgedby Dunigan'sanalysis):
(62)a. Ixale -yi/ *hul, bi -flu lekk -ee ceeb, ...
child DFpl 3pl when 3pl eat -F rice
IThe children/*they}, when they ate rice ...
35 We assume this position to be the canonical subjectposition, which may be identified
as the specifierof the topmostprojectionwithinthe inflectionaldomain.The specific nature
of this projectiondependson the constructionitself (cf. section 2, diagramsin (24)).

b. {xale -yi/ *nu}, na -fnulekk ceeb!

child DFpl 3pl +F 3pl eat rice
tThe children/*they}, it is my wish thatthey eat rice!

c. {xale -yi/ *iu}, bu -ilu lekk ceeb!

child DFpl 3pl C+neg 3pl eat rice
IThe children/*they1, let them not eat rice!

d. {xale -yi/ *nuI, ceeb -bi -flu lekk...

child DFpl 3pl rice DFsg 3pl eat
{the children/*they), the rice which they ate ...

e. {xale -yi/ *nfu},ceeb -bi -la -ilu lekk.

childDFpl3pl rice DFsg COP3pl eat
IThe children/*they}, it is the rice, which they ate.
This regularitymay, however, equally be derived from the phonological
contrastbetweenWK and STR items, if we assume thatdislocatedphrases
must, as such, bear strongstress. In Wolof as, for instance,in French,only
strong-stressexpressionsmay occur as independentdislocatedtopics.
Now consider the contrastbetween the noncliticized subjectmarkerin
(63a) and the enclitic subjectmarkerin (63b):

(63)a. tflu/xale -yi} -a lekk ceeb -bi. [flu-a>floo;yi-a>yaa]

3pl /child DFpl COP eat rice DFsg
It is {they/thechildren}, who ate the rice.

b. bi I-fnu/xale -yi I lekk -ee ceeb -bi...

when 3pl/ child DFpl eat -F rice DFsg
When Ithey/thechildrenI are the rice...

In the subject-focusconstructionexemplifiedby (63a), the subjectmarker

fiu is not cliticized, while it is in the TS-constructionexemplifiedby (63b).
Under Dunigan's all-syntacticanalysis of clitics, both instances of fnuin
(63) are 'syntactic clitics', i.e., the heads of Clitic Phrases, but the clitic
pronounof (63a) raises as an XPto a specifierposition open to any subject
phrase,whereasthe clitic pronounof (63b) raises as a head to right-adjoin
to bi. We just saw, however,that there is no independentevidence in sup-
port of the assumptionthat fiu undergoes syntactic movement in (63b).

Assuming that it does not and that it occupies in (63) the same syntactic
position as the lexical subject xale-yi, the contrastbetween the nonclitic
vs. clitic behaviourof niuin (63a) and (63b) boils down to the phonolo-
gical patterndescribedin (57): iiu bears default-initialstress in (63a), and
undergoesattachmentin (63b).

4.2.2. Clitic SubjectMarkersin SyntacticHeads

Now consider the contrastbetween the optative sentences in (64) and the
independentdeclarativesentences in (65):

(64)a. na Ixale -yi/ -nu } lekk ceeb!

+F child DFpl 3pl eat rice
It is my wish that Ithe children/theyI eat rice!

b. xale -yi, na fiu lekk ceeb!

child DFpl +F 3pl eat rice
The children,it is my wish thatthey eat rice!

(65)a. xale -yi lekk -na -ilu ceeb.

child DFpl eat +F 3pl rice
The childrenhave eaten rice.

b. lekk -na -niuceeb.

eat +F3pl rice
They have eaten rice.

c. *lekk-na xale -yi ceeb.

eat +F child DFpl rice

In both (64) and (65), the subjectmarkerflu surfacesas an enclitic. How-

ever, several contrastsmay be observed between (64) and (65). In (64),
the lexical subject and subject markeralternateto the right of na, while
such is not the case in (65), as witnessedby the ungrammaticalityof (65c).
In (64b), the lexical DP xale-yi which occurs to the left of na exhibits
properties characteristicof dislocated topics, while in (65a) it behaves
like a subject argument(an issue discussed above, section 3). To account
for these data, Dunigan assumes that the subject markeris generatedin
both (64) and (65) as the head of a Clitic Phrase in subjectposition, and
that enclisis results in all cases from syntactic Clitic Movementto the E

head, occupied by na. In (65), the lexical subjectxale-yi is assumedto be

generatedin the specifierof the Clitic Phraseand to move leftwardto an
argumentposition, with the subject markerstandingas a Case-absorbing
'clitic doubler'. In (64), xale-yi is assumed to occupy a non-argument
position, with the subject markerstandingas a Case-markedresumptive
clitic. The differentstatusof xale-yi in (64b) and (65a) is derivedfrom the
differentfeaturecontentof na, assumedto instantiatean epistemic modal
in (65a), but not in (64b) (with the specifierof an epistemic modal defined
as an argumentposition).
As already pointed out in section 3, this very complicated analysis
of subject markersis crucially motivatedby the (in no way necessary)
assumptionsthat all weak subject markersshould form a syntactic class
and thatthe phonologicalattachmentof clitics shouldresultfrom syntactic
movement.Should we give up this a prioriconceptionof clitics, it is quite
reasonableto assume, as we did above in section 2, thatthe subjectmark-
ers of (65), but not those of (64), are inflectionalperson markerswhich
surface at the right periphery of a +finite verb or auxiliary as a result
of an incorporationprocess. This analysis straightforwardlyaccounts for
the facts (i) that the lexical subject of (65a) exhibits argumentproperties,
(ii) that the subject markerof (65a, b) occupies a linear position which
is closed to lexical DPs (cf. (65c)), and (iii) that the subject markerof
(65a, b) cruciallycorrelateswith finiteness.The threeconstructionswhich
containan inflectionalsubjectmarkerarethe ones labeled 'V-to-Infl','Ipf-
to-Infl' and 'EXPL-to-Infl'in Table (33), which we may refer to as finite
clauses.36 Underour own account,the acceptabilityof (65b) indicatesthat
rich person inflection licenses a null subject in Wolof as it does in more
familiar 'pro-drop'languages.

4.2.3. WolofSubjectMarkersin Phonology-FreeSyntax

Summarizing,we assume that the subject markerof (64a) occupies in
syntax the same structuralposition as the lexical subjectwith which it al-
ternates,whereasthe subjectmarkerof (65a, b) is relatedto finitenessand
is generatedin an inflectionalhead closed to DPs. This boils down to say-
ing that the Wolof items which we (as Dunigan)have been calling subject
markershave two distinct syntactic statuses, dependingon the construc-
tion: they are eithersubjectpronounsor the spell-outsof personinflection,
36 By glossing na as +F in all of its occurrences,we suggest thatoptativeclauses
such as
(64a) arefinite,which is not perfectlysatisfactory.Such clauses indeedcontainna, which is
+finite in the sense thatit is Speaker-oriented(as arguedin Zribi-HertzandDiagne 1999).
But they fail to be thoroughlyfinite because theirinflectionaldomainis not saturated(it is
unspecifiedfor Tense, Aspect, and Polarity).Saturatedfinite clauses containboth na AND
a complete inflectionaldomain.

and only the former should be assumed to be specified for nominative

Case. The above syntacticdistinctionis quiteindependentfromphonology,
since both the subjectpronounof (64a) and the person-inflectionmarkers
of (65a, b) regularlyundergoenclisis under(57a).
Now consider the subject markers listed in Table (33) as 'right-
attached'and 'unattached'.These occur in the four constructionsexem-
plified by (66):

(66)a. subject-focus
Ixale -yi/ nflu -a lekk ceeb. [yi-a>yaa;niu-a>nioo]
child DFpl 3pl COP eat rice
It is Ithe children/they), who ate rice.

b. presentative
xale -yi/ flu) -a -ngi lekk ceeb.
child DFpl 3pl COP PRESeat rice
( The children/they
) eat rice.

c. narrative
Ixale -yi/ nu
f lekk ceeb.
child DFpl 3pl eat rice
So Ithe children/they
) eat rice.

d. DT
Arambegg -na Ixale -yi/ nu u lekk ceeb.
Aram want +F child DFpl 3pl eat rice
Aram wants {the children/they)to eat rice.

The 'right-attached'subjectsand subjectmarkersof (66a, b) actuallyserve

as supportsfor the following left-attachedcopula, which regularlybehaves
as a WK item. In all four cases exemplifiedby (66), the subjectmarkerniu
occurs domain-initially;hence it does not undergo enclisis and receives
default-stressunder(57b). In all four cases, the subjectmarker(inu)altern-
ates with a lexical subjectas it does in (59) and (61), and we shall therefore
assumeit to instantiatea nominativepronoun,ratherthanpersoninflection.
In otherwords,the subjectmarkersof (59)-(6 1), on the one hand,and(66),
on the other,have exactly the same syntactic status (they are nominative
pronouns),althoughthey do not exhibit the same phonological behaviour
(they undergoattachmentin (59)-(61), not in (66)).

Table (67) below gives the full paradigm of Wolof subject markers
subclassifiedaccordingto our own syntacticassumptions:

(67) Wolof subjectmarkers37

syntactic construction
status type l sg 2sg 3sg Ipi 2pl 3p}
inflectional V-to-Infl (m)a nga 0 fl(u) ngeen fl(u)
person ipf-to-Infl (m)a nga 0 fl(u) ngeen fl(u)
markers F-to-Infl (m)a nga 0 fl(u) ngeen fl(u)
optative (m)a nga mu flu ngeen flu
relative (m)a nga mu nlu ngeen nu
nominative TS (m)a nga mu nu ngeen nu
neg. imp. ma nga (m)u nu ngeen nu
pronouns obj.focus (m)a nga 0 nu ngeen nu
presentative ma ya mu flu yeen nu
subj.focus ma ya mu flu yeen flu
narrative ma nga mu nu ngeen nu
DT ma nga mu flu ngeen flu

This table does not specify whetheror not each item undergoesattach-
ment, a propertywe have shown to be independentlyderivablefrom the
phonological patternin (57). Three of the forms listed in (67) only spell
out nominativepronouns((m)u, ya, and yeen), one only spells out person
inflection (fi). In most cases, however, nominativepronounsand person-
inflectionmarkershaveidenticalphonologicalspell-outs,a predictablefact
underthe reasonableassumption(Giv6n 1972, 1976) thatthe latterare but
a diachronicdevelopmentof the former.

37 Bracketedsegments in (67) undergocontextualdeletion.

[ma] > [a]/[a] - - (e.g.: na-ma > na-a)

[mu] > [u]l/[u] - - (e.g.: bu-mu> bu-u)

nga (2sg) and ngeen (2pl) become ya, yeen when initially-stressedand supportingan ad-
jacent WK item (e.g., in subject-focusand presentativeconstructions).Besides, the initial
consonantin nga is vocalized after [u]:

[nga] > [ual/[u] - - (e.g.: bu-nga > bu-u-a > boo)


4.3. Morphology:Words,Affixesand Clitics

4.3.1. Simple Clitics vs. InflectionalAffixes
Considerthe two examples in (68), which both containthe enclitic subject
(68)a. bi -nlu lekk -ee ceeb -bi....
when 3pl eat -F rice DFsg
When they ate the rice...

b. lekk -na -n~uceeb -bi.

eat +F 3pl rice DFsg
They have eaten the rice.

Underour own assumptions,fiu in (68a) occupies a nominativeposition in

syntax and attachesto the adjacentword under(57a). In (68b), we assume
that fiu is generatedin the topmost head of the inflectional domain (cf.
diagram(24a)), which is targetedby the inflectedverb.The terminalstring
resulting from both attachmentprocesses is similar (fiu is spelt out as a
suffix within an initially-stressedword), but the morphologicalrelation
between nluand its phonologicalsupportis differentin each case. In (68a),
attachmentis triggeredby purely phonological reasons (WK+adjacency)
and results in a prosodic word which includes constituentsunrelatedin
syntax (bi+fiu):fiu exhibits in this case the prototypicalbehaviourof what
Zwicky (1977) named simple clitics. In (68b), attachmentis triggeredby
incorporation,a morphologicalprocess which forms an inflected word
from a set of featurespositionedin syntaxwithin an x-barstructure:nlube-
haves here as a word-level,inflectionalaffix, as definedby Zwicky (1985)
and Halpem (1995).

4.3.2. Taxicvs. NontaxicPositioning OLCsas Special Clitics. In most cases, OLCs do not surface in
the same linearpositions as lexical objects and stronglocatives. They thus
behave as 'special clitics', as recalledby (69):

(69)a. xale -yi di -na -fnulekk ceeb -bi.

child DFpl +ipf +F 3pl eat rice DFsg
The childrenwill eat the rice.

b. xale -yi di -na -fnu-ko lekk.

child DFpl +ipf +F 3pl 3sg/O eat
The childrenwill eat it.

This contrastis what cruciallymotivatesDunigan'sClitic Movementana-

lysis. OLCs are WK morphemeswhich undergophonological attachment,
as do a subclass of subjectmarkers,but some properties,examinedbelow,
suggest that the phonological attachmentof OLCs must follow that of
subjectmarkers. Attach: (i) Person Inflection,(ii) OLCs. That OLCs must attach
to the verb or auxiliaryafterpersoninflectionis suggested by severaltests
laid out by Zwicky (1985) to distinguishclitics from inflectional affixes.
OLCs may form clusters,and linearizationwithin clustersmust follow the
patternexplicited in (70):

(70) Linear orderingwithin WolofOLCclusters38

a. 1st person > 2nd person> 3rdperson> Locative

b. long vowel > shortvowel

Restriction(70a) is exemplifiedby (71):

(71)a. xale -yi lekk -na -flu [-ko -ci]/ *[-ci -ko].
child DFpl eat +F 3pl 3sg/O LOC LOC3sg/O
The childrenate it there.

b. xale -yi wan -na flu [-ma -la]/ *[-la -ma].

child DFpl show +F 3pl Isg 2sg/O 2sg/O ]sg
lit. The childrenhave shown me you.
= (i) have shown me to you, or
= (ii) have shown you to me.

Although we have seen that the canonical order of constituents in a

double-objectconstructionis Goal>Theme(cf. (40)), this restrictionseems
overrunby (70a) in clitic clusters:in (71b), 1st-personma must be linear-
ized to the left of 2nd-personla, whateverthe thematic roles associated
with each clitic.

38 Church (1981, pp. 94-95) spells out the constraintin (70a), but not that in (70b).
According to him, a furtherrule should linearize singularpronouns to the left of plural
ones - but we have found no evidence of this in the dialect underconsideration.

Restriction(70b) is exemplifiedby (72):

(72)a. *xale-yi wan -na -fnu[-ko -leen].

b. xale-yi wan -na -flu [-leen -ko].

child DFpl show +F 3pl 3pl/O 3sg/O
lit. The childrenhave shown them Ihim/her/it).
= (i) have shown them to [him/her), or
= (ii) have shown Iher/him/itIto them.

In (72), the two OLCs,both 3rd-person,are linearizedby (70b).

The following examples show that the linearizationconditionsin (70)
apply only within the limits of OLC clustersand do not extend to adjacent
inflectionalpersonaffixes.For example,in (73), the shortinflectionalaffix
fiu is not linearizedby (70b) to the right of the long accusativeclitic leen;
similarlyin (74), the 1sg object clitic ma is not linearizedby (70a) to the
left of the 3pl inflectionalaffix fiu:

(73)a. xale -yi gis -na -nu -leen.

child DFpl see +F 3pl 3pl/O

b. *xale -yi gis -na -leen -fnu.

child DFpl see +F 3pl/O 3pl
The childrenhave seen them

(74)a. xale -yi gis -na -flu -ma.

child DFpl see +F 3pl Isg

b. *xale -yi gis -na -(m)a -flu.

child DFpl see +F Isg 3pl
The childrenhave seen me.

Note that if inflectional person markersare positioned as a result of in-

corporation,hence aftersyntax,it follows thatOLCs, which are linearized
afterthem, must also be positionedaftersyntax. Attach:(i) NominativePronouns,(ii) OLCs. The examplesin (75)

show that nominativepronouns,like person inflection,precedeOLCs and

are unaffectedby the linearizationconstraintswhich apply within OLC

(75)a. bi -fnu-ma -ko -ci wan ee...
when 3pl Isg 3sg/O LOC show -F
When they showed me to {him/her} there...
When they showed Iit/him/her}to me there...

b. bi -ilu -leen -ko -ci wan -ee

when 3pl 3pl/O 3sg/O LOC show -F
When they showed them to [him/her) there...
When they showed lit/him/hernto them there...

c. *bi -ma -nu -ko -ci wan -ee...

when Jsg 3pl 3sg/O LOC show -F

Two issues must now be settled regardingOLCs:(a) how should we char-

acterize their 'special' linear position ? (b) how and at what stage in the
derivationdo they reachthis position? The Target of OLCs. Dunigan's answer to our question (a),

which is contained in her generalization(46), is that OLCs target the
topmost functional head within the extended projectionof V. Under our
own descriptiveassumptions,this hypothesisruns into at least two sets of
First, OLCs may occur in narrativeclauses, such as (76), where they
encliticize to the verb:
(76) xale -yi lekk -ko.
child DFpl eat 3sg/O
So the childreneat it.

To predictthis patternunderDunigan's analysis, we must assume that the

verb of a narrativeclause raises up to a functionalhead which provides a
target for the object clitic. However, as alreadypointed out in section 3,
overt evidence thatthis functionalhead is presentfails to exist, and if such
a functionalheadis lacking,we are led to assumethatOLCsdirectlyattach
to the lexical verbin (76), which weakensthe all-syntacticapproachsince
lexical heads are not assumedto 'attractfeatures'in syntax.
The second set of problematicevidence is the behaviourof OLCs in
optative, relative, TS, negative imperative,and object-focus clauses. As

shown by Table (37), their linear position in these constructionsseems

to crucially depend upon the clitic vs. nonclitic natureof the subject, as
exemplifiedby (77)-(78):
(77)a. bi -ko (-ci) xale -yi lekk -ee...
when 3sg/O LOC child DFpl eat -F
b. *bi xale -yi -ko (-ci) lekk -ee...
when child DFpl 3sg/O LOC eat -F
When the childrenate it (there)...
(78)a. *bi -ko (-ci) -nu lekk -ee...
when 3sg/O LOC 3pl eat -F
b. bi -nu -ko (-ci) lekk -ee...
when 3pl 3sg/O LOC eat -F
When they ate it (there)...
Under Dunigan's analysis, all clitic pronouns and locatives in these ex-
amples are positioned in syntax by Clitic Movement, which right-adjoins
them to bi, assumed to fill the E head. This theory correctlypredictsthat
OLCs should occur to the left of the lexical subject in (77), and that they
should cluster with the subject clitic in (78) - their position with respect
to the subject markercould be taken care of by an external linearization
restriction(subjectmarker>OLC),to be addedto (70). This view, however,
conflicts with our own descriptiveassumptions(cf. 4.2.1), according to
which the subject markerof (78b) occupies in syntax the same structural
position as the lexical subjectof (77a) and undergoesphonological attach-
ment under(57a). If the positioning of OLCs in (77)-(78) is sensitive to
the attachedbehaviourof the subject,as suggested by the data, it follows,
under our own assumptions,that they must also be linearized after syn-
tax, and that they must furthermoreundergoattachmentafter nominative
pronouns(this confirmsthe resultsof section In other words,the
OLCs of (78b) targetthe string bi-iiu, which is NOT a syntactichead, but
a prosodic word made up of a syntactichead (bi) followed by a functional
nominal(nu) occupying in syntax an argument(specifier)position.
It would, however, be inaccurateto characterizethe target of Wolof
OLCs in purelyphonologicalterms,as consideredin (79):
(79) The Targetof Wolof OLCs: A Purely Phonological Tentative
Attach OLCs to the leftmostprosodic word within the prosodic

This descriptiveassumptionmight suffice to accountfor a subset of data,

among which are (77)-(78), but it would fail to accountfor many others,
e.g., (69b), (71), (72b), (74a), or (76).
The alternative phrasing we propose in (80) in effect collapses
Dunigan's syntactic assumptionwith the phonological restrictionin (79)
and correctlypredictsall the datareviewedabove:

(80) The Targetof WolofOLCs:TheSyntax-PhonologyInterface

Attach OLCs to the prosodic word which containsthe topmost
head of theirextended-Vdomain.

Our crucial assumptionin (80) is that the targetof OLCs must be defined
neither in syntax alone, nor in phonology alone, but in the syntax-
phonology interface.OLC attachmentoccurs in Phonology and thus first
targetsa prosodic word, but it is neverthelesssensitive to syntactic x-bar
structuresince it must cruciallyidentify the topmosthead within diagram
(24). In some cases, such as (77a), the topmost head of the extended-V
domainforms a word of its own at PF, and OLCs simply attachto it (e.g.,
bi+ko). In othercases, OLC attachmentmust see througha prosodic word
to identify its syntactictarget:thus in (78b) the topmostsyntactichead (bi)
is contained in the prosodic word bi-fiu (>bi-fiu+ko),and in (69b), (71),
(72b), (74a), the topmost syntactic head na is containedin the inflected
verb or auxiliaryresultingfrom incorporation(e.g., di-na-fiu+ko).Unlike
the object clitics of Frenchanalysedby Miller (1992) and Miller and Sag
(1997), Wolof OLCs do not necessarily spell out a syntactic feature of
their syntactic target (thus, the OLCs of (77a) do not spell out features
of bi); their syntactic targetis left unrestrictedas to both category (it is
functionalin, e.g., (69b): di-na-nu+ko,and lexical in (76): lekk+ko) and
syntacticposition (it may fill V, T, Pol, F, Foc, or C). Because it is primarily
identifiedas a phonological entity (a prosodic word), the targetof OLCs
may include material which does not occupy a head position in syntax
(e.g., iiu in bi-fiu, cf. (78b)). The result of OLC attachmentmay thus
indeed be phonologically characterizedas a well-formed prosodic word,
regularlybearing initial stress. However, their targetmay not be defined
in the purely phonological terms consideredin (79), as clearly witnessed
above by (49), which shows that the linearizationof OLCs is sensitive to
That OLCs may not attachto the cliticized complementizerni, as ex-
emplifiedby (47), repeatedunder(81), suggests thatgeneralization(80) is
subjectto a uniquenessrequirement:

(81)a. Aramwax -na -ni xale -yi lekk -na -nu -ko.
Aram say +F that child DFpl eat +F 3pl 3sg/O
Aramsaid thatthe childrenate it.

b. *Aramwax -na -ni -ko xale -yi lekk -na -flu.

Aram say +F that 3sg/O child DFpl eat +F 3pl

The complementizerni satisfiesthe syntacticcomponentof (80), since it is

the topmosthead of the clause (extended-Vdomain)containingthe object
pronoun.However,we see in (81b) that the object pronouncannot attach
to ni. We assumethatthis restrictionarisesfrom the fact thatni encliticizes
(by (57a)) to the adjacentmatrixinflected verb, resultingin the prosodic
word wax-na-ni, which includes both the topmost head of the matrix do-
main (na) and the topmost head of the embedded domain (ni), and hence
fails to provide a unique syntactic targetfor OLC attachment.Under our
analysis, the ungrammaticalityof (81b) thus results from a conflict which
typically involvesthe syntax-phonologyinterface,ratherthansyntaxalone
('subjacency'). OLCs as Nontaxic WeakItems. We have arguedabove that en-

clisis, in Wolof, occurs after syntax and that OLCs undergo attachment
afterboth person inflection and subjectclitics: this suffices to predictthat
OLCs always follow subjectmarkerswithinclitic clusters,with no need to
add a subject>objectlinearizationconstraintto (70).
The fact thatOLCs generallyfollow subjectmarkersmay be accounted
for in terms of the taxic/nontaxicdistinctionproposed by Sauzet (1996,
1998, 1999, 2000), which largely converges with the analyses of special
clitics developed by Klavans (1980, 1985) and Anderson (1992, 1993).
We have arguedabove thatthe person-numberfeaturesspelt out by Wolof
subjectmarkersare either generatedas nominativeargumentsor as inflec-
tional markers.Whichevertheirsyntacticstatus,all Wolof subjectmarkers
share one common property:their linear position within their clause is
straightforwardly determinedin syntax.Nominativepronounsoccur in the
linear positions generally open to subject phrases;and person inflection
is spelt out at the right peripheryof the finite verb or auxiliary,which -
underBaker's(1988) incorporationapproach- reflectsits topmostposition
in the inflectionaldomain. Generalizing,we may thus characterizeWolof
subject markersas taxic items. Contrastively,OLCs may be identified as
nontaxic, i.e., items that are left unlinearized in the output of syntax and
must consequentlybe positioned after syntax, in morphophonology.Fol-
lowing Klavans(1980, 1985) and Anderson(1992, 1993), we assume that

nontaxic items spell out phrasal features- in the case of Wolof OLCs:
object and locative features pertainingto the extended projection of V.
Underthe taxic/nontaxicdistinction,we may assumethattaxic WK items,
as a class, undergo attachmentprior to nontaxic WK items - a natural
assumptionif we think of Phonology as a layeredcomponentwhose first
activatedstratumis the closest to the outputof syntax(cf. Zwicky andPul-
lum 1983; Selkirk 1984; Zwicky 1985; Pullumand Zwicky 1988; Inkelas
and Zec 1990; Kaisse 1983, 1990; Miller et al., 1997, among others).
Our own analysis of Wolof OLCs crucially contrastswith Dunigan's
in the way it explains the 'special' linear position of these items: under
Dunigan'sgeneralization(46), it resultsfromthe topmostg-headattracting
the clitic, a magnetic phenomenon which might be due to the inherent
strengthof the g-head, to the inherentweakness of the clitic, or to a com-
binationof both. Under our own generalization(80), it stems out from the
fact thatnontaxicitems should be spelt out in a high-visibilitylocus. High
visibility is a propertyof a geometricalnature,which in this specific case
is assessed in syntacticterms (topmosthead of phrasaldomain).Underthe
modularapproach(80), it is unproblematicthat,as exemplifiedby (43b) or
(76), OLCsseem insensitive to the contrastbetween functionaland lexical
heads, or that, as illustratedby (78b), their target may be made up of a
syntactichead and a syntacticnonhead.



Our descriptionof Wolof person and locative markershas led us to dis-

tinguish three types of phonologically-deficient(WK) items: nominative
pronouns,which encliticize to any adjacentword within theirphrasaldo-
main and otherwise receive default initial stress; person inflection, spelt
out as a suffix on the finite verb or auxiliary;and OLCs, spelt out as
nontaxic phrasalclitics. We conclude that clitics, in Wolof, do not form
a syntactic class. Clitics only share the phonological property(57a), but
they do not form a linguistic class, not even in phonology - the relevant
phonological distinctionfor Wolof being WK/STR, ratherthan [+clitic].
Despite Cardinalettiand Starke's (1999) predictions, we have found no
empiricalevidence that WK person and locative markersform a class of
syntactically deficient items. Diagram (82) attemptsto capture the rela-
tion between the syntactic, morphological, and phonological properties
revealedby our study:

(82) Wolof person and locative markers:the syntax-phonologyin-


+taxic -taxic
(featuresprojected (phrasal features)
in X-barsyntax)


-structural +structural
case case SYNTAX

dislocated, nominative person accusative

isolated pronouns inflection andlocative
or P-governed pronouns

full-fledged simple word- special MORPHOLOGY

words clitics or level clitics
domain- affixes


strongstress weak stress

The only syntax-phonologycorrelationswhich seem broughtout by (82)

are the facts that strong pronounsdo not appearin structural-caseposi-
tions and that all nontaxicitems have weak-stressphonologicalspell-outs.
Since there is no evidence that Wolof strong-stressand weak-stresspro-
nounscontrastas to theirfeaturecontent(see section 3.3.5), the descriptive
resultspresentedin (82) do not sufficeto establishthatstrongstressis sens-
itive to syntactic case; it might be that nonstructural-casepositions (i.e.,
dislocated,isolated, and P-governedpositions) in Wolof, as in French,in-
dependentlyrequirefull-fledgedprosodicwords,for phonologicalreasons.
Anderson(1992) providesevidence thatthe othercorrelationsuggestedby

(82) (nontaxic/WK)is not a necessary one. Leaving these issues open for
future research,we conclude that the Wolof data presentedin this study
globally supporta modularapproachto pronountypology treatingweak-
stressandcliticizationas phonologicalpropertieslargelyindependentfrom


Anderson, Steven. 1992. A-morphousMorphology, CambridgeUniversity Press, Cam-

Anderson,Steven. 1993. 'Wackemagel'sRevenge: Clitics, Morphology,and the Syntax of
Second Position', Language69(1), 68-98.
Baker,Mark. 1988. Incorporation:A Theoryof GrammaticalFunction Changing,Univer-
sity of Chicago Press, Chicago.
Borer, Hagit (ed.). 1986. The Syntax of Pronominal Clitics, Syntax and Semantics 19,
Academic Press, New York.
Boskovic, Zeljko. 2000. 'Cliticizationand the Syntax-PhonologyInterface',unpublished
manuscript,Universityof Connecticut.
Bouchard, Denis. 1982. On the Content of Empty Categories, Ph.D. dissertation,MIT
[published1984 by Foris, Dordrecht].
Boukhris, Fatima. 1998. Les clitiques en berbere tamazighte: approche minimaliste,
unpublishedPh.D. dissertation,UniversitdMohamedV, Rabat.
Burzio, Luigi. 1986. Italian Syntax,Reidel, Dordrecht.
Cardinaletti,Anna and Michael Starke. 1999. 'The Typology of StructuralDeficiency: A
Case Study of the Three Classes of Pronouns',in H. van Riemsdijk(ed.), Clitics in the
Languagesof Europe,Moutonde Gruyter,Berlin, pp. 145-233.
Chomsky,Noam. 1998. MinimalistInquiries: The Framework,M1TrOccasionalPapersin
Linguistics 15, Departmentof Linguistics,MIT,Cambridge,MA.
Chomsky, Noam. 1999. Derivation by Phase, MIT Occasional Papersin Linguistics 18,
Departmentof Linguistics,MIT, Cambridge,MA.
Church,Eric. 1981. Le systemeverbaldu wolof, DocumentsLinguistiques27, Publications
of the Departmentof Generaland African Linguistics,DakarUniversity,Dakar.
Collins, Chris. 1997. Local Economy,MIT Press, Cambridge,MA.
Dell, Fran,ois and Mohamed Elmedlaoui. 1989. 'Clitic Ordering, Morphology and
Phonology in the Verbal Complex of Imdlawn TashlhiytBerber', Langues orientales
anciennes. Philologie et Linguistique2-3, Peeters,Leuven/Paris.
Diallo, Amadou. 198la. Une phonologie du wolof, Les langues nationalesau Sdnegal78,
Instituteof Applied Linguistics,DakarUniversity,Dakar.
Diallo, Amadou. 1981b. Structures verbales du wolof contemporain,Les langues na-
tionales au Sendgal80, Instituteof Applied Linguistics,DakarUniversity,Dakar.
Diouf, Jean-Leopold. 1982. Introductiona une etude du systeme verbal du wolof. relation
modes,pronomssujets et autres modalites du predicat, Instituteof Applied Linguistics,
Dobrovie-Sorin,Carmen. 1999. 'The Typology of Pronounsand the Distinction between
Syntax and Morphophonology',in H. van Riemsdijk (ed.), Clitics in the Languages of
Europe,Moutonde Gruyter,Berlin, pp. 249-257.

Dunigan, Melynda. 1994. The Clausal Structureof Wolof.A Studyof Focus and Cliticiza-
tion, unpublishedPh.D. dissertation,Universityof NorthCarolina,ChapelHill.
Emonds, Joseph. 1975. 'A TransformationalAnalysis of French Clitics without Positive
OutputConstraints',LinguisticAnalysis 1(1), 3-24.
Fal, Aram, Rosine Santos and Jean-Leonce Doneux. 1990. Dictionnaire wolof-francais,
Fal, Aram. 1991. Alphabetisationin wolof: Guide orthographique,Institute of Applied
Frota, Sonia. 1992. 'Is Focus a Phonological Category in Portuguese?',Proceedings of
ConSole, UtrechtUniversity,pp. 69-86.
Givon, Talmy. 1972. 'HistoricalSyntax and SynchronicMorphology:An Archaeologist's
Field Trip', Proceedingsof the Chicago LinguisticsSociety 7, 394-415.
Giv6n, Talmy.1976. 'Topic,Pronoun,andGrammaticalAgreement',in C. Li (ed.), Subject
and Topic,Academic Press, New York,pp. 149-188.
Greenberg,Joseph. 1966. The Languagesof Africa, Mouton,The Hague.
Halle, Morrisand Alec Marantz.1993. 'DistributedMorphologyand the Pieces of Inflec-
tion', in K. Hale andS. J. Keyser(eds.), The ViewfromBuilding20: Essays in Linguistics
in Honor of SylvainBromberger,MIT Press, Cambridge,MA, pp. 111-176.
Halpern,Aaron. 1995. On the Placement and Morphologyof Clitics, Centerfor the Study
of Languageand Information,Stanford.
Harris,James. 1997. 'Why n'ho is pronounced [li] in Barceloni Catalan:Morphological
Impoverishment,Merger,Fusion and Fission', RecherchesLinguistiquesde Vincennes
26, 61-86.
Harris, James. 1998. 'Spanish Imperatives:Syntax Meets Morphology', Journal of Lin-
guistics 34, 27-52.
Inkelas, SharonandDragaZec (eds.). 1990. ThePhonology-SyntaxConnection,Centerfor
the Study of Languageand Information,Stanford.
Jaeggli, Osvaldo. 1982. Topicsin RomanceSyntax,Foris, Dordrecht.
Ka, Omar. 1978. L'accent en wolof: essai d'analyse phonetique et linguistique, unpub-
lished M.A. thesis, Universityof Dakar,Dakar.
Ka, Omar.1994. WolofPhonologyand Morphology,UniversityPress of America,Lanham,
Ka, Ousman. 1982. La syntaxe du wolof. essai d'analyse distributionnelle,unpublished
Ph.D. dissertation,DakarUniversity,Dakar.
Kaisse, Ellen. 1983. 'The Syntax of Auxiliary Reduction in English', Language 59, 93-
Kaisse, Ellen. 1990. 'Towarda Typology of Postlexical Rules', in S. Inkelas and D.
Zec (eds.), The Phonology-SyntaxConnection,Center for the Study of Language and
Information,Stanford,CA, pp. 127-144.
Kayne, Richard.1975. FrenchSyntax,MIT Press, Cambridge,MA.
Kayne, Richard. 1989. 'Null Subjects and Clitic Climbing', in 0. Jaeggli and K. Safir
(eds.), TheNull SubjectParameter,Kluwer,Dordrecht,pp. 239-261.
Kayne, Richard. 1991. 'Romance Clitics, Verb Movement, and PRO', LinguisticInquiry
22, 647-686.
Kayne, Richard.1994. TheAntisymmetryof Syntax,MIT Press, Cambridge,MA.
Kayne, Richard. 1999. 'A Note on Clitic Doubling in French', unpublishedmanuscript,
New YorkUniversity.

Keenan, Edward. 1996. 'MorphologyIs Structure:A Malagasy Test Case', in M. Pear-

son and I. Paul (eds.), The Structureof Malagasy, Vol. 1, UCLAOccasional Papers in
Linguistics17, Departmentof Linguistics,Universityof California,LA, pp. 92-112.
Kihm, Alain. 1995. 'Le capuchonde mon stylo - mon capuchonde stylo: morphosyntaxe
et semantiquedes complementsdu nom en wolof', in J. Gueron(ed.), Rencontres:etudes
de syntaxeet de morphologie,Publidix,UniversiteParis-lO,Nanterre,pp. 147-170.
Klavans,Judy. 1980. Some Problemsin a Theoryof Clitics, Ph.D. dissertation,University
College London.Published 1982 by IndianaUniversityLinguistics Club, Bloomington,
Klavans, Judith. 1985. 'The Independence of Syntax and Phonology in Cliticization',
Language61, 95-120.
Legendre,Geraldine. 1997. 'OptimalRomanianClitics: A Cross-LinguisticPerspective',
TechnicalReportJHU-CogSci98-9, Departmentof Cognitive Science, Baltimore,MD.
Miller,Philip. 1992. Clitics and Constituentsin Phrase StructureGrammar,Garland,NY
Miller, Philip, Geoffrey Pullum, and Arnold Zwicky. 1997. 'The Principle of Phonology-
free Syntax:FourApparentCounterexamplesin French',Joumal of Linguistics33, 67-
Miller,Philip and Ivan Sag. 1997. 'FrenchClitic MovementwithoutClitics or Movement',
NaturalLanguageand LinguisticTheory15, 573-639.
Nash, Lea and Alain Rouveret. 1999. 'FeatureFission and.the Syntax of ArgumentDPs
and Clitics', paper presented at the Clitic Round Table, Departementde Recherches
Njie, Codu Mbassy. 1982. Description syntaxique du wolof de Gambie, Les Nouvelles
Ouhalla,Jamal. 1989. 'CliticMovementand the ECP:Evidence from BerberandRomance
Languages',Lingua79, 165-215.
Pullum, Geoffrey and Arnold Zwicky. 1988. 'The Syntax-Phonology IInterface', in
F. Newmeyer (ed.), Cambridge Survey of Linguistics, Cambridge University Press,
Cambridge,pp. 255-289.
Rizzi, Luigi. 1997. 'The Fine Structureof the Left Periphery', in L. Haegeman (ed.),
Elements of Grammar:Handbookin GenerativeSyntax, Kluwer, Dordrecht,pp. 281-
Roberge, Yves. 1990. The Syntactic Recoverabilityof Null Arguments,McGill-Queen's
UniversityPress, Montreal.
Robert,Stephane.1991. Approche6nonciativedu systemeverbal.Le cas du wolof, Editions
du CNRS, Paris.
Rouveret,Alain. 1999. 'Clitics, Subjects and Tense in EuropeanPortuguese', in H. van
Riemsdijk (ed.), Clitics in the Languages of Europe, Mouton de Gruyter,Berlin, pp.
Sauzet,Patrick.1996. 'Ordredes mots, ordredans les mots', Languefran!aise 111, 10-37.
Sauzet,Patrick.1998. 'Enamourer,enivrer,enorgueillir:le statutdes prefixes',in P. Sauzet
(ed.), Langues & GrammaireII & III: Phonologie, Linguistics Department,Universite
Paris-8, Saint-Denis,pp. 117-140.
Sauzet, Patrick. 1999. 'Linearite et consonnes latentes', Recherches Linguistiques de
Vincennes28, 59-86.
Sauzet, Patrick.2000. 'Clitiques:commentbouger sans changerde place, ou: si la phono-
logie rempla,ait le mouvement?',paperpresentedat the Langues & Grammaireweekly

Sauvageot, Serge. 1965. Description synchronique d'un dialecte wolof: le parler du

Dyolof, InstitutFrancaisd'AfriqueNoire, Dakar.
Selkirk, Elisabeth. 1984. Phonology and Syntax: The Relation between Sound and
Structure,MIT Press, Cambridge,MA.
Shlonsky,Ur. 1994. 'Semitic Clitics', Geneva GenerativePapers 2, 1-11.
Sportiche, Dominique. 1996. 'Clitic Constructions',in J. Rooryck and L. Zaring (eds.),
Phrase Structureand the Lexicon, IndianaUniversity Linguistics Club, Bloomington,
IN, pp. 213-276.
Wackernagel,Jacob. 1892. 'Uber ein Gezets der indogermanischenWortstellung',Indo-
germanischeForschungen1, 333-436.
Zagona, Karen. 1982. Governmentand Proper Governmentof VerbalProjections,unpub-
lished Ph.D. dissertation,Universityof Washington.
Zribi-Hertz,Anne. 2000. 'Les pronomsforts du fran,ais sont-ils [+animes]?Spdcification
morphologiqueet specification s6mantique',in M. Coene, W. De Mulder,P. Dendale
and Y. D'Hulst (eds.), TraianiAugusti VestigiaPressa Sequamur:Studia Linguisticain
HonoremLilianae Tasmowski,Unipress,Padova,pp. 663-679.
Zribi-Hertz, Anne and Lamine Diagne. 1999, 'Description linguistique et grammaire
universelle: reflexions sur la notion de finitude a partir de la grammairedu wolof',
in A. Sores and C. Marchello-Nizia (eds.), Typologie des langues, universaux lin-
guistiques, special issue of LINX,LinguisticsDepartment,UniversiteParis-IO,Nanterre,
pp. 205-215.
Zribi-Hertz,Anne and Liliane Mbolatianavalona.1999. 'Towardsa ModularTheory of
LinguisticDeficiency: Evidence from MalagasyPersonalPronouns',NaturalLanguage
& Linguistic Theory17, 161-218.
Zwicky, Arnold. 1977. On Clitics, IndianaUniversityLinguisticsClub, Bloomington,IN.
Zwicky, Arnold. 1982. 'Strandedto andPhonologicalPhrasingin English', Linguistics20,
Zwicky, Arnold. 1985. 'Clitics and Particles',Language 61, 283-305.
Zwicky, Arnold. 1986. 'The Unaccented Pronoun Constraintin English', Ohio State
UniversityWorkingPapers in Linguistics32, 100-113.
Zwicky, Arnold. 1990. 'SyntacticRepresentationsand Phonological Shapes', in S. Inkelas
and D. Zec (eds.), ThePhonology-SyntaxConnection,Centerfor the Studyof Language
and Information,Stanford,CA, pp. 379-397.
Zwicky, Arnold and Geoffrey Pullum. 1983. 'Cliticization vs. Inflection: English n't',
Language 59, 502-513.

Received 12 August 1999

Revised 31 August 2001

UPRESA 7023, CNRS/Paris-8

Departementde LinguistiqueGenerale
2 rue de la Libertd
F-93526 Saint-DenisCedex 02