Professional Documents
Culture Documents
Kayne 1996
Kayne 1996
General Editor
E.F. KONRAD KOERNER
(University of Ottawa)
Volume 139
Edited by
JAMES R. BLACK
Memorial University of Newfoundland
VIRGINIA MOTAPANYANE
University of New Brunswick
JAMES R. BLACK
Department of Linguistics
Memorial University of Newfoundland
St. John's, NF, Al 3X9, Canada
VIRGINIA MOTAPANYANE
Humanities
University of New Brunswick - Saint John
Saint John, NB, E2L 4L5, Canada
June 1996
MICROPARAMETRIC SYNTAX:
SOME INTRODUCTORY REMARKS
RICHARD S. KAYNE
Graduate Center, City University of New York
l
Cf. Chomsky (1980, 1981) and Kayne (1981).
RICHARD S. KAYNE
2
A related proposal concerning double object constructions is given in Kayne (1983). The ab
sence of double object constructions in French and Italian is not expected from the perspective
of Collins & Thráinsson (1993), whose analysis lacks the crucial abstract preposition.
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS xi
There is another correlation in Romance that involves the order between clitic
and infinitive: a Romance language/dialect will allow the counterpart of English
*John doesn 't know if to go to the movies tonight (with controlled PRO in the
presence of if), if and only if it has the order infinitive-clitic.7 This correlation,
particularly striking if one looks at little studied (from a syntactic point of view)
null subject languages that have clitic-infinitive order (Sardinian, Occitan,
Gardenese), appears to favor over other approaches an approach to PRO based
on a modified version of Chomsky's (1986) analysis, and to support (a certain
revision of) principles A and of his binding theory.
Italian, although it almost invariably shows infinitive-clitic order, has an ap
parently anomalous pre-infinitival clitic in infinitival imperatives. This anomaly
can be made sense of by taking advantage of a correlation that holds quite
strongly across the dialects of Italy between pre-infinitival clitics in these im
peratives and clitic climbing of the familiar sort. The analysis called for involves
an abstract auxiliary to which to clitics can raise in the manner of clitic climb
ing. 8
The study of Italian dialects also radically changes one's perception of
Romance auxiliary selection. The picture and theory of auxiliary selection that
one arrives at by studying the most widely spoken Romance languages must be
considerably modified and enriched if one is to account for the remarkable di
versity found in the dialects.9
Comparative work on the syntax of a large number of closely related lan
guages can be thought of as a new research tool, one that is capable of providing
results of an unusually fine-grained and particularly solid character.10 If it were
possible to experiment on languages, a syntactician would construct an experi
ment of the following type: take a language, alter a single one of its observable
syntactic properties, examine the result to see what, if any, other property has
changed as a consequence of the original manipulation. If one has, interpret that
result as indicating that it and the original property that was altered are linked to
one another by some abstract parameter.
Although such experiments cannot be performed, I think that by examining
pairs (and larger sets) of ever more closely related languages, one can begin to
7
Cf. Kayne (1991).
8
Cf. Kayne (1992).
9
Cf. Kayne (1993).
10
As illustrated in particular by the various articles in this volume, most of which treat
Romance or Germanic languages. Johns' article on Inuttut and Cheng, Huang & Tang's article
on Chinese demonstrate the importance of microparametric syntax for other language families.
I make no attempt in these introductory remarks at a systematic survey of microparametric
work.
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS xiii
approximate the results of such an experiment. To the extent that one can find
languages that are syntactically extremely similar to one another, yet clearly dis
tinguishable and readily examinable, one can hope to reach a point such that the
number of observable differences is so small that one can virtually see one pro
perty covarying with another.11
In addition to facilitating the accurate individuation of parameters and of the
principles of Universal Grammar required to interact with them, the technique of
examining a large number of very closely related languages promises to provide
a broad understanding of parameters at their finest-grained (microparameters),
i.e., to provide a handle on the question: what are the minimal units of syntactic
variation?
Consider, for example, the phenomenon of past participle agreement. Study
of the better known Romance languages shows past participle agreement with
the subject in SPEC-IP in passives to be general, perhaps exceptionless. In active
sentences, on the other hand, past participle agreement is found in French and
Italian, and to some extent in Catalan, but not in Spanish or Portuguese. On the
basis of these, one might postulate a parameter one setting of which allows past
participle agreement in actives.
Yet consideration of lesser known Romance languages indicates that there
must be parameters with finer-grained effects, to account for facts such as the
following: some of these languages allow past participle agreement in both WH-
and clitic constructions, some only in the latter. No Romance language/dialect,
as far as I can see, has past participle agreement with WH-phrases but not with
object clitics.12
Some allow past participle agreement with all direct object clitics, but some,
while having it with third person clitics, prohibit it with first and second person
clitics. Some allow past participle agreement with the partitive clitic, others do
not. Of those that allow clitics to follow past participles in the auxiliary-past par
ticiple construction, some allow past participle agreement only when the clitic
has moved up to the auxiliary, while others are freer. As far as I know, all
Romance languages with past participle agreement in actives allow such agree
ment in at least some reflexive clitic constructions when the auxiliary is the
equivalent of be. When there is a reflexive clitic in addition to an accusative
clitic, some require agreement with the accusative, others do not. When the
auxiliary is have, some prohibit agreement with the reflexive clitic (or its an
tecedent) completely, while others do not.
11
In the extreme case, one may find an isolated property distinguishing two very close di
alects—cf. Henry's article in this volume.
12
Cf. Kayne (1989b).
XIV RICHARD S. KAYNE
Our understanding of all of the above points will benefit from a closer look
at even more of these languages/dialects. In many of these cases, it is not clear
yet what the exact form of the relevant parameters will be, nor whether they will
concern agreement per se, or whether they will more centrally involve properties
of the pronouns/clitics, or of the auxiliaries, or of the participles, or some com
bination of these.
It seems reasonable to expect work in microparametric syntax to play a privi
leged role in the future in answering the more general question concerning the
form that syntactic parameters may take.13 Chomsky's recent work, for exam
ple, suggests the possibility that all syntactic variation might be expressible in
terms of strong/weak features on various functional heads; 14 microparametric
work will enable us to test this kind of hypothesis in a particularly interesting
way.
In the preceding discussion, I have assumed that the enormous amount of
syntactic variation that can be observed even within the set of Romance lan
guages/dialects lends itself to insightful characterization in terms of the notion
'parameter' as it has developed over the past fifteen or twenty years.
Alternatively put, we can take the study of microparametric variation to provide
an ideal testing ground for the very hypothesis that syntactic variation can be re
duced to a finite set of parameters (interacting with a set of universal principles).
Related to this is the question of how many irreducible syntactic parameters
there really are. Again, work in microparametric syntax should be invaluable,
and should begin to give us some sense of a lower bound for the number of pa
rameters (which in turn will bear on questions of learnability/acquisition). It is
also clear that the study of minimal syntactic variation is bound to provide crucial
evidence bearing on questions of diachronic syntax (which involves the study of
the minimally different stages in the evolution of the syntax of a language).
The question of the number of syntactic parameters leads in turn to the ques
tion of the number of syntactically distinct languages/dialects. To begin with, I
take it for granted that there is no syntactically significant distinction to be drawn
between 'language' and 'dialect' and no justification for neglecting the latter.
Now it is often estimated that the number of languages presently in existence is
4000-5000.15 Such estimates must evaluate the contribution of Italy as one.
Yet Renzi and Vanelli (1983) showed that in Northern Italy alone one can
13
The term 'micro-comparative' was used by Hellan & Christensen (1986). For discussion of
(microparameters vs.) macroparameters, cf. Baker (1996).
14
Cf. Chomsky (1995). In a general way, this seems compatible with the implications of the
approach of Kayne (1994) to word order variation.
15
For some discussion, cf. Comrie (1987:2-5) and Crystal (1987:284-285).
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS XV
21
For an example of theoretically important variation within English particle constructions,
cf. Emonds (1976:83-86); for recent discussion, cf. den Dikken (1995).
MICROPARAMETRIC SYNTAX: SOME INTRODUCTORY REMARKS XVii
to the 33rd power is about 8.5 billion). Although we do not yet have a clear idea
of the number of irreducible syntactic parameters, it seems likely that the number
will turn out to be greater than 33. At the same time, although again there is a lot
yet to be understood, it seems plausible that the child is capable of setting at least
that many syntactic parameters.
If the number of independent parameters is somewhat larger, say 50, then
the corresponding number of syntactically distinct grammars is somewhat more
than one thousand trillion. If the parameters are 100 in manageable number, then
the corresponding number of grammars is, innocuously, over one million trillion
trillion (i.e., greater than 10 raised to the 30th power).
Alongside these numbers, the increase in numerical coverage due to work in
microparametric syntax in recent years is quantitatively modest. Yet it may not
be premature to speak of the beginnings of a qualitative improvement in our un
derstanding of syntactic variation. Microparametric syntax is a powerful tool,
whose growth is perhaps to be compared with the development of the earliest
microscopes, that allows us to probe questions concerning the most primitive
units of syntactic variation. And since the invariant principles of UG can hardly
be understood in isolation from syntactic variation, this tool promises to provide
invaluable evidence that will shape our understanding of those principles them
selves.
REFERENCES
Baker, Mark C. 1996. The Polysynthesis Parameter. Oxford & New York: Oxford
Univ. Press.
Benincà, Paola 1994. La variazione sintattica. Studi di dialettologia romanza.
Bologna: Il Mulino.
Chomsky, Noam. 1980. 'On Binding". Linguistic Inquiry 11.1-46.
. 1981. Lectures on Government and Binding. Dordrecht: Foris.
. 1986. Knowledge of Language. New York: Praeger.
. 1995. The Minimalist Program. Cambridge: MIT Press.
Collins, Chris & Höskuldur Thráinsson. 1993. 'Object Shift in Double Object
Constructions and the Theory of Case". Papers on Case & Agreement II. MIT
Working Papers in Linguistics 19, ed. by C. Phillips, 131-174. Cambridge, Mass:
Dept. of Linguistics, MIT.
Comrie, Bernard, ed. 1987. The World's Major Languages. New York & Oxford:
Oxford University Press.
Crystal, David, ed. 1987. The Cambridge Encyclopedia of Language. Cambridge:
Cambridge University Press.
Dikken, Marcel den. 1995. Particles. On the Syntax of Verb-Particle, Triadic, and
Causative Constructions. New York & Oxford: Oxford University Press.
Emonds, Joseph E. 1976. A Transformational Approach to English Syntax. Root,
Structure-Preserving and Local Transformations. New York: Academic Press.
Hellan, Lars & Kirsti K. Christensen, eds. 1986. Topics in Scandinavian Syntax.
XVIII RICHARD S. KAYNE
Dordrecht: D. Reidel.
Henry, Alison. 1995. Belfast English and Standard English. Dialect Variation and
Parameter Setting. New York & Oxford: Oxford University Press.
Johansson, Stig. 1979. "American and British English Grammar: An Elicitation
Experiment". English Studies 60,2.195-215.
Kayne, Richard S. 1981. "On Certain Differences between French and English".
Linguistic Inquiry 12.349-371.
. 1983. "Le datif en français et en anglais". Analyses grammaticales du
français. Études publiées à ľ occasion du 50e anniversaire de Carl Vikner.
Revue Romane. [Numéro spécial 24], ed. by M. Herslund, O. Mordrup & F.
Sorensen, 86-98. English version in R. S. Kayne, 1984, Connectedness and
Binary Branching. Dordrecht: Foris.
. 1985. "L'accord du participe passé en français et en italien". Modèles
Linguistiques VIL 73-90.
. 1989a. "Null Subjects and Clitic Climbing". The Null Subject Parameter,
ed. by Osvaldo Jaeggli & Ken Safir, 239-261. Dordrecht: D. Reidel.
. 1989b. "Facets of Romance Past Participle Agreement". Dialect
Variation and the Theory of Grammar, ed. by Paola Benincà, 85-103.
Dordrecht: Foris.
. 1991. "Romance Clitics, Verb Movement and PRO". Linguistic Inquiry
22.647-686.
------. 1992. "Italian Negative Infinitival Imperatives and Clitic Climbing".
Hommages à Nicolas Ruwet, ed. by L. Tasmowski & Anne Zribi-Hertz, 300-312.
Ghent: Communication & Cognition.
. 1993. "Toward a Modular Theory of Auxiliary Selection". Studia
Linguistica 47.3-31.
. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press.
Klima, Edward S. 1964. "Relatedness between Grammatical Systems". Language
40.1-20. Reprinted in Modern Studies in English. Readings in Transformational
Grammar, ed. by D.A. Reibel & S.A. Schane, 227-246. Englewood Cliffs, New
Jersey: Prentice-Hall.
Merat, Farokh 1974. Une comparaison grammaticale et lexicale de l'anglais bri
tannique et américain enseigné aux étrangers. Thèse de Doctorat de 3e Cycle,
Université de Paris VII.
Poletto, Cecilia 1995. "Split AGR and Subject Clitics in the Northern Italian
Dialects". Paper presented at the 18th GLOW [Generative Linguists of the Old
World] Colloquium, GLOW Newsletter 34.46-47.
Renzi, Lorenzo & Laura Vanelli. 1983. "I pronomi soggetto in alcune varietà ro
manze". Scritti linguistici in onore di Giovan Battista Pellegrini, 25-35. Pisa:
Pacini.
Trudgill, Peter & Jack K. Chambers, eds. 1991. Dialects of English. Studies in
Grammatical Variation. London: Longman.
& Jean Hannah. 1994. International English. A Guide to Varieties of
Standard English, 3rd edition. London: Edward Arnold.
Zandwoort, Reinard Willem. 1965. A Handbook of English Grammar, 3rd edition.
London: Longmans.
UNE ANALYSE MICROPARAMÉTRIQUE DES
MOYENS DANS LES LANGUES ROMANES*
J.-MARC AUTHIER & LISA REED
Université ď Ottawa & Pennsylvania State University
Abstract
* Nous tenons à exprimer notre gratitude aux personnes qui nous ont si généreusement aidé à
rassembler les nouvelles données dont notre analyse cherche à rendre compte: pour les données
du français canadien nous remercions Louise Deslauriers de Timmins en Ontario, Tanya
Goguen de Moncton au Nouveau-Brunswick, André Lapierre d'Ottawa, Sylvie Micheline
Mainville de Timmins en Ontario et Michelle Patry d'Ottawa; pour les données de l'italien
nous remercions Leonard Sbrocchi de Roseto Valfortore en Italie; et pour les données de l'espa
gnol nous remercions Maria-Luisa Rivero de Madrid. Nous tenons de plus à remercier de leurs
commentaires l'audience du colloque de 1994 de l'Association linguistique des provinces atlan-
tiques, les professeurs Julia Herschensohn et Paul Postal, ainsi que ľévaluateur anonyme sélec
tionné par les rédacteurs. Les recherches ayant mené à la rédaction de cet article ont été subven
tionnées en partie par une bourse post-doctorale (numéro 756-92-0036) attribuée à Lisa Reed
par le Conseil de la recherche en sciences humaines du Canada.
2 J.-MARC AUTHIER & LISA REED
La construction moyenne dans les langues romanes a fait l'objet d'un débat
soutenu au sein de la grammaire generative (voir par exemple Gross 1968,
Obenauer 1970, Ruwet 1972, Rizzi 1976, Chomsky 1981, Williams 1981,
Belletti 1982, Zubizarreta 1982, Keyser & Roeper 1984, Burzio 1986 [1981],
Wehrli 1986, Roberts 1987 [1985], Zubizarreta 1987 [1982], Cinque 1988,
Fellbaum & Zribi-Hertz 1989, Lyons 1989 et Fagan 1992). De ce débat a
émergé un certain accord parmi les chercheurs quant aux propriétés qui carac
térisent cette construction. Parmi ces propriétés se trouvent la présence obliga
toire du morphème SE/SI ainsi que l'apparition d'un objet thématique en position
de sujet de surface. Ces caractéristiques sont apparentes dans les exemples ci-
dessous tirés du français, de l'italien et de l'espagnol:
Il est aussi tenu pour acquis qu'une troisième caractéristique des moyens ré
side dans le fait que ce genre de construction va de pair avec un contexte géné
rique, c'est à dire que les phrases contenant un verbe à la voix moyenne dénotent
un état de faits qui était attribuable, est attribuable ou sera attribuable à n'importe
quel individu ou ensemble d'individus. Les effets de cette contrainte générique
sont perceptibles dans les paires d'exemples comme (2a,b), deux exemples du
français continental qui se distinguent par la présence versus l'absence d'un ad
verbe de temps ponctuel et par la présence versus l'absence d'un temps géné
rique. 1
1
Comme l'ont fait remarquer Belletti (1982) et Cinque (1988), le morphème SE/SI en
espagnol et en italien a au moins deux lectures possibles: la lecture moyenne et la lecture
impersonnelle. Le fait que les exemples de l'italien et de l'espagnol qui sont en apparence
parallèles à (2b) soient grammaticaux pourrait donc être lié à la présence de cette lecture
impersonnelle. Nous avons donc illustré la contrainte générique à l'aide d'exemples tirés du
français continental pour éviter ces complications.
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 3
Enfin, les verbes à la voix moyenne sont tenus d'avoir pour quatrième carac
téristique de ne pas tolérer la présence d'une phrase prépositionnelle en par
semblable à celle qui accompagnent souvent les verbes à la voix passive. Cette
caractéristique se base sur ľagrammaticalité de phrases telles que (3a,c), qui re
flètent les jugements de grammaticalité rapportés dans la plupart des études sur le
sujet (voir par exemple Ruwet 1972:110, Belletti 1982:7, et Aissen & Perlmutter
1983:368). Il est cependant à signaler que l'absence de phrases en par dans les
moyens est une contrainte qui n'est pas sans exceptions puisque des formes plus
anciennes du français utilisé en Europe permettaient la coexistence de la voix
moyenne et des phrases en par (cf. Martinon 1927, Stefanini 1962, Ruwet
1972, parmi d'autres). De plus, Cinque (1988:529, note 11), qui cite Lepschy
(1986, note 4), fait état d'une telle coexistence dans ce qu'il appelle les 'registres
rhétoriques' de l'italien moderne. Néanmoins, il est actuellement tenu pour vrai
que les registres parlés de la forme moderne des langues romanes ne permettent
pas l'utilisation des phraseserdans les constructions moyennes.
2
Certains chercheurs ont en fait remis en question bien avant nous le bien fondé de considérer
la contrainte générique comme étant une propriété caractéristique des moyens en se basant sur
deux types de données. Ces données ont cependant été jugées n'être que des contre-exemples ap
parents à cette contrainte, ce qui est aussi notre avis. Le premier type de données
'problématiques' se retrouve dans les travaux de Ruwet (1972:89,95), Zribi-Hertz (1982:349),
Zubizarreta (1987:150), Lyons (1989:177) et Fagan (1992:159). Ces données exhibent un
SE/SI moyen dans ce qui semble être une phrase non-générique (voir par exemple (ia), un exem
ple tiré de Zribi-Hertz (1982:349)). Or, Ruwet (1972:95), Fellbaum (1986:4), et Lyons
(1989:178) questionnent la valeur de contre-exemple de telles phrases du fait qu'elles sont à un
temps non-ponctuel et donc entièrement compatibles avec la définition de généricité que ľ on
suppose être requise dans les moyens (par exemple qu'à un moment du futur, un agent potentiel
4 J.-MARC AUTHIER & LISA REED
(5) a. Hier, des rubans noirs se sont portés *(par les étudiantes).
b. Ces promesses se sont faites *(par au moins deux politiciens).
quelconque sera en mesure de commettre l'action en question). Pour étayer cette hypothèse on
peut avancer ľobservation de Ruwet qui note que si ľon ajoute à ce genre de phrase un adverbe
de temps ponctuel qui force la lecture événementielle de la phrase (voir (ib)), on obtient un ré
sultat agrammatical.
(i) a. Ce roman se lira bientôt à Moscou.
b. * Demain à huit heures et quart, ce roman se lira à Moscou.
Les phrases en (iia) et (iib) ci-dessous, tirées de Zubizarreta (1987:150) et de Cinque (1988:542)
respectivement, illustrent le deuxième type d'exemple problématique pour la contrainte géné
rique. Il est clair que ces exemples sont cette fois-ci non-génériques en cela qu'ils contiennent
un temps ponctuel. Cependant, l'hypothèse que ce sont de ce fait des contre-exemples à la con
trainte générique en vigueur dans les moyens est loin d'être acceptée par tous. Lyons
(1989:179), par exemple, a suggéré en se basant sur les recherches sur les verbes moyens et
neutres/ergatifs menées par Keyser & Roeper (1984:394), que les exemples comme (iia) pour
raient bien refléter une utilisation neutre de SE et Cinque (1988:542) a défendu l'hypothèse que
les exemples comme (iib) reflètent la lecture impersonnelle de SI. Le fait que les exemples par
allèles à (iib) soient agrammaticaux en français, une langue dans laquelle on ne trouve pas d'u
tilisation productive du SE impersonnel, suggère que l'hypothèse de Cinque est en effet correcte.
Quant à l'hypothèse de Lyons, le lecteur trouvera des données qui l'étayent dans la note 5.
(ii) a. La question est en train de se discuter dans la salle du conseil,
b. Oggi, a Beirut, si è ucciso un innocente,
aujourd'hui à Beirut SI est tué un innocent
"Aujourd'hui, un innocent a été tué à Beirut."
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 5
Dans cet article, nous montrerons que ces nouvelles données du français ca
nadien et de l'espagnol de Madrid permettent en fait de mieux comprendre la
contrainte aspectuelle (ou générique) qui se manifeste dans les constructions
moyennes de la plupart des langues romanes.3 De plus, nous mettrons au point
une analyse microparamétrique des moyens dans les langues romanes qui rendra
compte de leur propriétés quelque soit le dialecte considéré. En bref, nous pro
poserons que les constructions moyennes contiennent un argument externe qui
est actif du point de vue de la syntaxe et qui est projeté sous la forme de la caté
gorie vide PRO dans une position interne au VP (comme l'a suggéré Roberts
1987). Des différences mineures en termes de -sélection seront alors invoquées
pour rendre compte de la variation dialectale observée en ce qui concerne les
phrases en par et la contrainte générique dans les moyens.
3 Nous n'avons considéré, pour les besoins de cet article, que le français, l'italien et l'espagnol.
Cependant, à en juger par ce que dit Cinque (1988:571-574) des autres langues romanes, nos
hypothèses semblent pouvoir s'étendre à ces langues, à l'exception sans doute du trentino, un
dialecte de l'italien qui, d'après Zubizarreta (1982:150, notes), n'a pas de SI moyen.
6 L-MARC AUTHIER & LISA REED
Nous présenterons dans les sections 3 et 4 de cet article cinq arguments qui
étayent une analyse modifiée de la structure proposée par Roberts en (8a). Mais
pour bien saisir la portée de notre discussion, il est nécessaire de commencer par
passer en revue certains tests qui nous permettront de mieux cerner les propriétés
syntaxiques de l'argument externe dans les constructions moyennes.
Nous avons vu que les chercheurs en sont venus à deux conclusions diver
gentes en ce qui concerne le statut syntaxique des arguments externes dans la
8 J.-MARC AUTHIER & LISA REED
construction moyenne des langues romanes. Certains sont d'avis que cet argu
ment est actif dans la syntaxe; d'autres qu'il ne l'est pas. Cependant, les re
cherches entreprises pour en arriver à une analyse de deux autres phénomènes en
anglais—la voix passive et les neutres/ergatifs—ont mené à la mise au point de
quatre tests qui ont pour but d'établir si un argument externe est actif ou inactif
du point de vue de la syntaxe. Par exemple, Chomsky (1981:143, note 60) attri
bue à Manzini (1980) l'observation que seuls les arguments externes qui sont
actifs en syntaxe peuvent contrôler un PRO dans les propositions de but à l'in
finitif, une observation qui a été utilisée depuis par Chomsky (1986:119),
Jaeggli (1986a:611) et Roberts (1987:70). De plus, Chomsky (1986:118),
Jaeggli (1986a:611) et Roberts (1987:70) ont montré que seuls les arguments
externes qui sont actifs en syntaxe peuvent être associés à des adverbes modifi
ant un agent et Chomsky (1986:118) et Jaeggli (1986a:611) ont remarqué que
seuls les arguments externes actifs en syntaxe peuvent être projetés sous la
forme de phrases en par. Enfin Stroik (1992:132, note 7) a découvert que seuls
les arguments externes actifs en syntaxe peuvent fonctionner comme des objets
de prédication.4 Pour bien comprendre la nature de ces quatre tests, examinons
tout d'abord les exemples français en (9a) et (9b), qui illustrent un contraste qui
distingue les passifs des neutres/ergatifs.
(9) a. Les porte-avions de l'ennemi ont été coulés pour PRO met
tre fin à une longue bataille navale.
4
Chomsky (1986:121), Jaeggli (1986a:614) et Roberts (1987:87) avaient en fait proposé que
le rôle thématique externe absorbé par le morphème du passif ne pouvait pas fonctionner
comme un objet de prédication afin de rendre compte de contrastes comme le suivant:
(i) A soldier visited the museum in uniform.
"Un soldat a visité le musée en uniforme."
(ii) The museum was visited *(by a soldier) in uniform.
"Le musée a été visité *(par un soldat) en uniforme."
Il existe cependant des exemples tels (iii)-(v), tirés de Stroik (1992:132, note 7), qui semblent
contredire cette hypothèse. Chomsky (1986:211, note 61) cite Roeper (1984) pour avoir noté
des exemples de ce type et il attribue à Luigi Rizzi l'hypothèse que le contraste qui oppose les
exemples en (i)-(ii) à ceux en (iii)-(v) est lié à la nature adverbiale plutôt qu'adjectivale des mo-
difieurs en (iii)-(v). Si cette hypothèse est correcte, les exemples en (iii)-(v) devraient être con
sidérés comme illustrant le test lié aux adverbes modifiant un agent,
(iii) This bank-job wasn't done alone.
"Ce holdup n'a pas été fait seul."
(iv) I can tell that this letter was written in a good mood.
"Je sais que cette lettre à été écrite de bonne humeur."
(v) That painting was painted blindfolded.
"Ce tableau a été peint les yeux fermés."
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 9
Le fait que le passif en (9a) mais pas le neutre/ergatif en (9b) tolère la pré
sence d'une proposition infinitive de but démontre que seules les phrases conte
nant un argument externe actif en syntaxe, par exemple les phrases passives,
permettent le contrôle du PRO de la proposition de but. Cependant, nous nous
devons de noter qu'il est impossible d'illustrer le fonctionnement de ce test ainsi
que celui des trois tests suivants au moyen de l'équivalent italien ou espagnol
des constructions de l'anglais et du français contenant des verbes comme to
melt/fondre, to sink/couler. Examinons par exemple les versions italiennes et es
pagnoles des exemples du français en (9a,b), qui apparaissent ci-dessous en
(10a,b) et (1 la,b) respectivement:
(12) a. Les policiers sont convaincus que ces pièces d'or ont été
fondues intentionnellement,
b. * Ces pièces d'or ont fondu intentionnellement.
(14) a. Ces pièces d'or ont été fondues par les voleurs,
b. * Ces pièces d'or ont fondu par les voleurs.
Afin d'étayer son hypothèse que les moyens dans les langues romanes conti
ennent un argument externe actif en syntaxe, Roberts (1987:275) a appliqué le
test lié aux propositions infinitives de but ainsi que le test des adverbes modifiant
un agent à des phrases contenant un verbe conjuge à la voix moyenne. Les résul
tats de ces tests indiquent qu'au prime abord, il y a en effet dans ces construc
tions un argument externe présent dans la composante syntaxique de la gram
maire. Les phrases qui le montrent apparaissent ci-dessous. De plus, nous avons
ajouté aux données de Roberts les exemples en (15c), (16c) et (17c) qui illus
trent l'application du test lié au fonctionnement de l'argument exteme comme
objet de prédication et les résultats de ce test concordent avec ceux des deux
autres.
Cependant Cinque (1988, § 4.1 et 4.2) a fait valoir qu'on ne peut pas se fier
à de tels résultats puisqu'on pourrait attribuer la grammaticalité de ces phrases
(comme il le fait d'ailleurs) à une lecture impersonnelle arguméntale du mor
phème SE/SI. Pour éliminer ce problème, Cinque propose l'utilisation de verbes
à l'infinitif avec le morphème SE/SI puisque de tels verbes apparaissent dans des
phrases où le Cas nominatif ne peut pas être assigné et que le SE/SI impersonnel
ajustement besoin d'un tel Cas. Bien que Cinque (1988:562) ait suggéré que les
phrases infinitives avec un SE/SI moyen ne se comportent pas vis à vis des tests
12 L-MARC AUTHIER & LISA REED
comme des phrases contenant un argument exteme actif en syntaxe, nous avons
pu construire des phrases en français, en italien et en espagnol qui semblent
prouver que les conclusions de Cinque étaient un peu hâtives.
(19) a. Visto che questi dimostranti sono così odiati come male ar
mati, avranno il vantaggio di potersi disperdere per PRO
soddisfare un elettorato avido di ordine e di sicurezza.
"Vu que ces manifestants sont aussi méprisés qu'ils sont
mal armés, ils auront l'avantage de pouvoir se disperser
pour satisfaire un electorat avide d'ordre et de sécurité."
b. I contratti preparati dal signor Leblanc hanno la reputazione
di doversi leggere attentamente.
"Les contrats rédigés par M. Leblanc ont la réputation de
devoir se lire avec précaution."
e. Questi musei militari hanno lo svantaggio di non potersi
visitare che in divisa.
"Ces musées militaires ont le désavantage de ne pouvoir se
visiter qu'en uniforme."
(20) a. Dado que los manifestantes son tan despreciados como están
mal armados, tendrán la ventaja de poder dispersarse para
PRO satisfacer a un electorado ávido de orden y seguridad.
"Vu que ces manifestants sont aussi méprisés qu'ils sont
mal armés, ils auront l'avantage de pouvoir se disperser
pour satisfaire un electorat avide d'ordre et de sécurité."
b. Dado que estos manifestantes son tan despreciados como
están mal armados, tendrán la ventaja de poder dispersarse
con entusiasmo.
"Vu que ces manifestants sont aussi méprisés qu'ils sont
mal armés, ils auront l'avantage de pouvoir se disperser
avec enthousiasme."
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 13
Dans la section précédente, nous avons examiné les résultats de quatre tests
qui militent contre toute analyse des moyens qui se base sur des structures du
type en (7a,b). Cependant, il nous reste encore à choisir entre deux hypothèses
possibles: l'hypothèse de l'argument explicite défendue par Belletti (1982) et
Roberts (1987), une hypothèse qui se base sur la structure en (8a), et l'hypo
thèse de l'argument implicite proposée par Chomsky (1981) et Keyser & Roeper
(1984), qui se base sur la structure en (8b). Ces deux types d'analyse sont com
patibles avec les données présentées à la section 3 puisque les quatre tests utili
sés ne peuvent distinguer qu'entre les arguments externes qui sont présents en
syntaxe et ceux qui ne le sont pas et qu'en conséquence les arguments externes
implicites et explicites se comportent de manière identique par rapport à ces tests.
5
L'application de ces tests à la classe de contre-exemples apparents mentionnée dans la note
numéro 2 indique que Lyons (1989:178) avait raison d'analyser de telles phrases comme reflé
tant l'utilisation neutre du morphème SE.
(i) a. ?* La question est en train de se discuter dans la salle du conseil pour
faire peur au gouvernement fédéral.
b. * La question est en train de se discuter dans la salle du conseil en hurlant.
c. * La question est en train de se discuter en uniforme au Q.G. du bataillon.
14 L-MARC AUTHIER & LISA REED
Il nous reste donc à apporter une réponse à la question suivante: des deux struc
tures en (8a) et (8b), laquelle est la bonne?
Nous nous proposons de résoudre cette question à ľaide d'un test mis en
place par Jaeggli (1986a:616-617), un test lié à l'observation qu'il existe deux
types de Contrôle: le Contrôle argumentai et le Contrôle thématique. Alors que le
Contrôle thématique se fait à partir d'arguments implicites, le Contrôle argumen
tai ne peut se faire qu'à partir d'arguments explicites qui occupent une position
arguméntale dans la structure. En se basant sur des contrastes tels celui de
(21a,b), Jaeggli (1986a) a proposé que le Contrôle d'un PRO sujet d'une pro
position infinitive au passif est un exemple de Contrôle argumentai. C'est à dire
que seul (21a) contient un argument explicite (John) qui peut contrôler le PRO
alors que le morphème du passif en anglais ne peut ce faire, étant donné sa na
ture d'argument implicite.6
Donc, étant donné que le Contrôle d'un PRO sujet d'une infinitive à la voix
passive est sensible au statut implicite ou explicite de l'argument externe de la
proposition principale, il devrait être possible d'utiliser ce test pour déterminer le
statut (implicite ou explicite) de l'argument externe dans les constructions
moyennes. Or, pour autant que nous le sachions, personne n'a remarqué que les
constructions moyennes dans les langues romanes permettent le contrôle d'un
6
Les jugements de grammaticalité rapportés par nos locuteurs natifs de l'italien et de l'espag
nol indiquent que les constructions passives dans ces deux langues diffèrent de celles de l'an
glais et du français en cela qu'elles tolèrent le Contrôle à l'intérieur de propositions infinitives
au passif. Ce contraste est illustré ici par (i) versus (ii):
(i) a. * These lies were told PRO to be admired by others.
b. * Ces mensonges ont été racontés pour PRO être admiré par les gens,
(ii) a. Ieri, alle due, queste mensogne sono state raccontate per essere ammirati dalla gente,
hier à 2 h. ces mensonges sont été racontés pour être admiré des gens
"Hier à 2 h., quelqu'un a raconté ces mensonges pour se faire admirer par les gens."
b. Ayer a las ocho esas mentiras fueron contadas para ser admirado,
hier à huit heures ces mensonges ont été racontés pour être admiré
"Hier à huit heures, quelqu'un a raconté ces mensonges pour se faire admirer."
Il semble donc que, contrairement à ce que Jaeggli (1986a) supposait, les constructions à la
voix passive dans les langues romanes n'ont pas exactement les mêmes propriétés. En particu
lier, le statut grammatical des phrases en (ii) indique que le passif en italien et en espagnol con
tient un argument externe explicite.
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 15
PRO contenu dans une proposition infinitive à la voix passive, ce qui indique
qu'elles contiennent un argument nul explicite.
Comme nous l'avons déjà mentionné au début de cet article, les chercheurs
qui se sont penchés sur le problème des moyens ont pris pour acquis que ce type
de construction est incompatible avec un argument externe exprimé au moyen
d'une phrase en par. Cela a donné lieu à des analyses de la voix moyenne qui
comportent des mécanismes qui ont pour effet de bloquer l'apparition d'une
phrase en par. Or, nous avons vu que les phrases en par sont en fait licites dans
certaines variétés du français canadien et de l'espagnol continental, ce qui a pour
conséquence de nous amener à modifier de tels mécanismes.
Belletti et Roberts ont proposé des analyses qui empêchent l'apparition d'une
phrase en par dans les moyens de deux manières différentes, comme l'indiquent
les entrées lexicales ci-dessous.
16 J.-MARC AUTHIER & LISA REED
sorbeur de Cas accusatif, mais nous nous dissocions de leurs hypothèses res
pectives en proposant que ce morphème joue aussi le rôle d'absorbeur de rôle
thématique externe et qu'il -sélectionne exclusivement un DP.
7
Pour des raisons qui nous échappent, les locuteurs de l'italien et de l'espagnol que nous
avons consultés jugent inacceptables les moyens contenant un sujet dérivé à la première ou à la
deuxième personne. Ces jugements, illustrés en (i) et (ii) ci-dessous contrastent avec ceux se
rapportant au français en (28a,b) dans le texte,
(i) a. ?* Dicono che mi leggo bene.
"On dit que je me lis bien."
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 19
Donc, étant donné que PRO ne peut pas être identifié par le morphème des
moyens, son identification doit se faire par le truchement d'un opérateur non-
sélectif. C'est en fait ce processus d'identification par un opérateur non-sélectif
qui explique les faits se rapportant à la contrainte générique. En effet, comme l'a
montré Lewis (1975:7), la sémantique des opérateurs non-sélectifs (une classe
d'opérateurs qui comprend les adverbes de quantification en général, parfois et
jamais ainsi que l'adverbe de quantification nul correspondant à en général) est
telle que le domaine de quantification de ces opérateurs est un ensemble de cas,
c'est à dire un ensemble d'événements ou d'intervalles de temps. Par exemple,
dire qu'un état de choses a cours en général, parfois, ou n'a jamais cours, c'est
vérifier si cet état de choses a cours ou n'a pas cours à travers un grand nombre
d'événements. C'est cette particularité sémantique des opérateurs non-sélectifs
que nous croyons être responsable de la contrainte générique en vigueur dans les
moyens qui ne contiennent pas de phrase en par. Dans ces cas-là le SE/SI moyen
-sélectionne un DP qui, en raison de la Théorie des Cas et de la propriété du
SE/SI moyen d'absorber le Cas accusatif, apparaît sous la forme d'un PRO. Or,
comme ce PRO ne peut être identifié que par un opérateur non-sélectif et que la
sémantique de ce genre d'opérateur en restreint la distribution à des phrases qui
permettent la quantification sur les cas, à savoir les phrases génériques, il s'en
suit que toute phrase moyenne sans phraseerdoit être générique.
Enfin, il est important de noter que notre explication de la contrainte géné
rique dans les moyens en termes de -sélection et des conditions d'identification
imposées à PRO prédit qu'en théorie, une langue ou un dialecte qui possède un
morphème SE/SI moyen pourrait choisir pour ce morphème les mêmes propriétés
de -sélection que celles données par Jaeggli (1986a) pour le morphème du
passif (c'est à dire que le SE/SI moyen absorberait le Cas accusatif et le rôle thé
matique externe mais ne -sélectionnerait qu'une phrase en par et ceci faculta
tivement.)8 A première vue, le SE moyen de l'espagnol madrilène exemplifié en
(6b) paraît justement être un tel cas puisque les exemples du type Ayer a las ocho
esas promesas se hicieron "Hier à huit heures, ces promesses se sont faites" sont
grammaticaux. Cependant, ces données ne sont pas concluantes puisque la
grammaticalité de telles phrases pourrait, si l'on suit l'analyse de Cinque (1988),
être attribuée à la lecture impersonnelle arguméntale du morphème SE.
En conclusion, nous avons, dans cet article, introduit des différences micro
paramétriques liées à la voix moyenne dans les langues romanes et nous avons
défendu trois grandes hypothèses. Tout d'abord, nous avons proposé que les
constructions moyennes dans ces langues ont une structure très proche de celle
en (8a), la seule différence étant que le PRO interne au VP n'est pas déplacé dans
cette position comme le suggère Roberts (1987), mais y est inséré dans la base.
De plus, nous avons suggéré que le morphème des moyens dans les langues
romanes est associé aux entrées lexicales en (27a,b) en nous basant sur le fait
que seules ces entrées lexicales permettent de rendre compte de la variation mi
croparamétrique qui caractérise la voix moyenne dans ce groupe de langues.
Enfin, les données nouvelles que nous avons examinées ont montré que deux
des propriétés qui avaient été qualifiées de caractéristiques des moyens par le
passé en fait ne méritent pas cette qualification. Nous avons donc défendu l'hy
pothèse que ces deux propriétés (l'absence d'une phrase en par et la contrainte
générique), que l'on retrouve dans la majorité des langues romanes, sont tout
simplement les conséquences qui découlent du choix des propriétés de C-sélec-
tion attribuées au morphème des moyens ainsi que celles qui découlent des
conditions d'identification imposées par la grammaire universelle à la catégorie
vide PRO. 9
8
On doit tout de même garder à l'esprit les données de la note numéro 6 concernant l'italien et
l'espagnol qui semblent indiquer que l'analyse du passif de Jaeggli (1986a) pour les langues
romanes est incomplète.
9
Le lecteur attentif ne manquera pas de se poser la question de savoir si oui ou non nos hypo
thèses concernant les moyens dans les langues romanes s'étendent à d'autres langues comme
par exemple l'anglais. Etendre nos hypothèses à l'anglais est certainement du domaine du pos
sible et le fait que les moyens de l'anglais soient soumis à une contrainte générique et ne soient
pas compatibles avec une phrase en by suggère à première vue qu'une telle extension soit en
fait désirable. Par exemple, on pourrait émettre l'hypothèse que les moyens de l'anglais conti
ennent un morphème SE/SI nul (cette idée ayant été proposée par Keyser & Roeper) et associer
ce morphème aux propriétés de -sélection en (27b). On pourrait aussi adopter la récente hypo-
LA CONSTRUCTION MOYENNE DES LANGUES ROMANES 21
BIBLIOGRAPHIE
, Peter & Maaike Schoorlemmer. 1995. "Middles and Nonmovement".
Linguistic Inquiry 26.173-197.
Aissen, Judith & David Perlmutter. 1983. "Clause Reduction in Spanish". Studies
in Relational Grammar 1, red. par David Perlmutter, 360-403. Chicago:
University of Chicago Press.
Authier, J.-Marc. 1988. The Syntax of Uns eleetive Binding. Thèse de doctorat,
Univ. of Southern California, Los Angeles, Cal.
. 1992a. "A Parametric Account of V-Governed Arbitrary Null
Arguments". Natural Language and Linguistic Theory 10.345-374.
. 1992b. "Is French a Null Subject Language in the DP"? Probus 4.1-16.
Belletti, Adriana. 1982. "Morphological Passive and PRO-Drop: The Impersonal
Construction in Italian". Journal of Linguistic Research 2.1-34.
Burzio, Luigi. 1986. Italian Syntax: A Government-Binding Approach. Dordrecht:
D. Reidel. (Version revue et corrigé d'une thèse de doctorat, Intransitive Verbs
and Italian Auxiliaries, MIT, 1981.)
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
. 1986. Knowledge of Language: Its Nature, Origin, and Use. New York:
Praeger.
Cinque, Guglielmo. 1988. "On si Constructions and the Theory of crb".
Linguistic Inquiry 4.521-581.
thèse de Stroik (1992) qui suggère que les moyens de l'anglais comportent un PRO qui reçoit le
rôle thématique externe puisque comme ľa montré Authier (1988), ce PRO peut être lié à un
opérateur non-sélectif, tout comme PRO. L'une ou l'autre de ces deux hypothèses rendrait
compte des deux caractéristiques que nous venons de mentionner. Cependant, il existe des faits
qui incitent à la prudence lorsque l'on songe à uniformiser le traitement des moyens des langues
romanes et de l'anglais. Par exemple, il est bien connu que la voix moyenne a un caractère
beaucoup moins productif en anglais que dans les langues romanes. (Il suffit pour s'en conva
incre d'essayer de traduire en anglais les exemples au moyen présentés dans cet article.) De
plus, Jaeggli (1986a:611) et Roberts (1987:189,190) ont montré que les tests classiques qui
permettent de dépister les arguments externes qui sont actifs en syntaxe ont des résultats posi
tifs dans les langues romanes mais des résultats négatifs en anglais. Enfin notre nouveau test
lié au Contrôle à l'intérieur d'une proposition infinitive au passif appliqué aux moyens semble
indiquer la présence d'un argument externe explicite dans les langues romanes mais pas en an
glais. (Cf. le statut grammatical des phrases en (22)-(24) versus le statut nongrammatical de
phrases comme *Bridges blow up easily to be awarded medals) Hâtons-nous d'ajouter, cepen
dant, que ces faits ne devraient pas forcément nous amener à rejeter l'hypothèse que les moyens
de l'anglais sont entièrement différents de ceux des langues romanes. En effet, Stroik (1992) a
attribué les résultats négatifs pour l'anglais de certains des tests que nous avons appliqués aux
moyens à des facteurs indépendants (mais voir la réplique de & Schoorlemmer 1995).
Nous concluons donc que l'étude des moyens de l'anglais mérite des recherches plus approfon
dies. (Le lecteur intéressé peut se référer entre autres à Fellbaum 1986, à Roberts 1987, à Fagan
1988, 1992, à Massam 1992, à Stroik 1992, et à & Schoorlemmer 1995 qui ont étu
dié en détail les propriétés particulières des moyens de l'anglais.)
22 J.-MARC AUTHIER & LISA REED
Introduction
1
I make use of the following labelling conventions. Coindexation indicates membership in
the same chain. The category S is an abbrevation for AGRSP.
26 PHILIP BRANIGAN
The contrast between (1c) and (1d) is possible in English because the com
plemenüser can be omitted or included freely in complements to the appropriate
class of verbs. The optional presence of a complemenüser is not a necessary-
feature of language. Other languages may require or prohibit overt complemen-
tisers in parallel contexts. Within Germanic, Danish (2) and German (3) provide
a cross-linguistic minimal pair in this respect.
(The V/2 structure of (4) is shown by the position of the finite verb smager
in front of the negator ikke.) As the data shows, Danish and German differ in
the restrictions they impose on complementisers in V/2 complements. Danish re
quires the at complemenüser in such clauses. German does not allow a comple
menüser to appear (den Besten 1983).
Significantiy, this difference between the two languages appears to go hand-
in-hand with the possibility of a phrase undergoing WH-movement out of the
V/2 complement (Vikner 1991).
The 'topic5 category in a V/2 clause need not be the subject. Complements
and adjuncts may also raise to SPECC and legitimate verb raising in the same
manner. And the same contrast between Danish and German is found when ob
jects or adjuncts are extracted from an embedded V/2 SPECC: (7). (Example
(7d) is taken from Vikner (1991).)
Again, the fronted position of the verb shows that WH-movement is preceded by
movement of the complement or adjunct phrase to SPEC.C.
I assume that the derivation is fixed in this respect (in both languages): there
are two distinct -bar movement operations involved in WH-extraction from a
41 assume this to be the case even when subjects are extracted, contra Travis (1984); Zwart
(1993). For arguments against the dual position analysis of verb-second, see Schwartz &
Vikner (1989); Branigan (1995).
TREATING THAT-TRACE VARIATION 29
V/2 complement. The first is topicalisation, which takes the WH-phrase from its
position inside the clause to the SPECC topic position. This movement results—
for whatever reason—in movement of the finite verb to Topicalisation pro
duces a chain with its head in an -bar position and its tail in an A-position (or
possibly an adjunct, 'broadly L-related' position). This chain is not uniformly A-
bar. The second movement then raises the WH-phrase from SPEC to a higher
SPEC position, possibly through a sequence of intermediate SPECC positions.
This produces a second chain, all of the members of which occupy SPEC.C. This
second chain is uniformly A-bar.
I assume as well that the overt complementiser found in embedded V/2
complements in Danish heads a second CP category, which takes the V/2 clause
as its complement (Vikner 1991; Iatridou & Kroch 1992). Thus the structures of
the complement clauses in (4) and (5) are as shown in (8a) and (8b), respec
tively.
One further issue remains to be clarified before the extraction constraints can
be addressed. The CP-recursion structure found in Danish embedded V/2 is
somewhat problematic from the point of view of a restrictive theory of selection.
A V/2 clause is found only in a selected position—such clauses cannot
serve as subjects or adjuncts, nor are they allowed to be complements to other
than a restricted class of predicates (bridge verbs, loosely put). In short, V/2
complements must be selected by an appropriate predicate. Under conventional
assumptions, a head can select properties only on the head of its complements.
This means that the matrix verb should be unable to select a V/2 clause in the
structure (8a). As the verb must in fact do so, we may suppose that the structure
of (8a) is modified in the LF component. Selection will be possible if covert
movement raises the V- complex to the next head, where it can be selected di
rectly at LF.3 Then the LF representation of (8a) will be (9). This -raising op
eration makes sense if we suppose, with Law (1991), that some complementis-
ers are 'expletive' elements which must be eliminated from the LF representa-
3
An anonymous reviewer points out that the matrix verb might select the V/2 complement in
directly, by selecting a complementiser which has the particular property of selecting a V/2
complement. In the absence of any reason to suppose such a complementiser, I prefer the text
analysis. Although indirect selection is certainly a coherent notion, by resorting to such selec
tion 'magic', we mask the relationship between the matrix verb and its complement type,
rather than explaining it.
PHILIP BRANIGAN
tion. Then movement of the lower V- complex is necessary in order to satisfy
Full Interpretation, because only such movement will eliminate the expletive
from the LF representation.
(9) V[CpV-C-C[CPXPtc[s...t...]]]
Now the parameters of the problem are clear. While extraction from SPEC- in a
V/2 complement is possible, as occurs in German, extraction from a lower
SPEC- is impossible in CP-recursion constructions, such as those found in
Danish.
The obvious place to look for a solution is the bounding theory. Let us there
fore quickly dispose of this possibility. Suppose a mechanism similar to those
employed in Chomsky's (1986) theory of barriers. In such a theory, a comple
ment is a barrier unless it is L-marked. In the case of the CP-recursion structures
which concern us here, the consequence seems to be that the second CP should
be a barrier. If the upper CP were L-marked (by the matrix verb) and the lower
CP were not, then movement from the lower SPEC- would cross a CP barrier.
However, any theory of this sort must stipulate, as Chomsky did with the IP cat
egory, that adjunction to the appropriate barrier is not allowed. In the absence of
any justification for this stipulation, a barriers-based approach does not hold
much promise.
The alternative is a constraint on representations, like the ECP. In the frame
work I have adopted, there is no notion of government, so there can be no no
tion of proper government. But it is possible to formulate a Minimalist counter
part to the ECP, by relying on the notion of a checking domain. We require some
principle which constrains the formation of a uniform -bar chain, that is, of a
chain the head and tail of which occupy distinct SPEC- positions. We cannot
exclude such chains entirely, as they are well-formed in the German structures,
where there is no CP-recursion in embedded V/2. Let us suppose then that a uni
form -bar chain must not terminate in the specifier of a CP, the head of which
has been raised. This can be captured in a general LF constraint on chains, which
I refer to as RES(ECP) (the 'residue' of the ECP).
Let us take this formulation as a first approximation only, which will be re
fined as we proceed. Conceptually, the RES(ECP) is quite closely related to the
TREATING THAT-TRACE VARIATION 31
ECP. Both impose an extra licensing condition on traces. The ECP does so di
rectly, requiring proper government in place of simple government. The
RES(ECP) also imposes an extra constraint on traces, but it does so indirectly, by
referring to members of uniform chains.
Now let us compare the effects of the RES (ECP) on the grammatical German
example (5) and the ungrammatical Danish example (4). In the German case,
there is a single CP projection in the complement clause. The tail of the chain
formed by WH-movement is in SPEC.C, and this is part of the checking domain
of the complementiser. So RES(ECP) is satisfied for the tail of the uniform A-bar
chain. As for the head of the -bar chain, it too occupies a SPECC with no CP
recursion, so we may assume that RES(ECP) is satisfied here, too. In the Danish
case, the effect of RES (ECP) is quite different. While the tail of the uniform A-bar
chain again occupies a SPECC position, it is not a part of the checking domain of
any complementiser at LF. The checking domain of any head is determined
derivationally (Chomsky 1993), so that it may change as movement creates new
X°-chains. While the original checking domain of the lower includes its speci
fier, movement of that head to a higher position produces a new checking do
main for the new binary Xo chain, which does not include its original specifier.
As noted above, the V- complex must raise to the higher in the LF component
in order for selection of the V/2 complement to be possible. So LF movement of
the V- complex to the higher will leave the tail of the uniform -chain in vio
lation of RES(ECP). We must also be sure that normal A-bar chains do not run
afoul of the RES (ECP). This is quite straightforward. In A-bar chains which do
not originate in SPECC, the tail normally does not occupy an A-bar position.
Complements to categories other than are either L-related or immobile. L-re-
lated categories can never serve as the tails of uniform A-bar chains. The same is
true of subjects of small clauses or ECM complements, which occupy a
SPEC-AGR position.
So the RES(ECP) has the right empirical effects for distinguishing the gram
matical German examples from the ungrammatical Danish ones. The next step
must be to see how it can best be formulated in principled, minimalist terms. The
definition (10) requires that chain members be in the checking domain of some
complementiser. We might seek to eliminate the reference to a single syntactic
category in favor of a more general constraint. Let us suppose that the prop
erty which distinguishes from other categories is the fact that the domain of
is not L-related. In other words, everything in its domain is in an A-bar position.
Then we can relate the A-bar properties of the chain constrained by the RES (ECP)
to this property of The revised RES(ECP) in (11) expresses this idea.
32 PHILIP BRANIGAN
(11) RES(ECP)
Any category α belonging to a uniform chain Σ must be in the
checking domain of a head γ, such that the features which render Σ
uniform are determined for α by the corresponding properties of γ.
The objects of weigh and bite cannot be raised to subject position, even
though they are regular NPs and NP objects are usually raised in passives. This
follows from RES(ECP), because the objects do not occupy a Ө-position in either
case. Movement of the object to SPEC-AGRS would create a uniform -chain, all
of whose elements are not Ө-marked. The tail of this chain is not in the checking
domain of any head, though, so it does not satisfy RES(ECP).
Another effect of the RES(ECP) is that topicalisation of S will be blocked, ac
counting for the ungrammaticality of (13b).
(13) a. ? Peter promised to tell Mark that it would snow and snow¡, he
told Mark that it would ti.
b. * Peter promised to tell Mark that it would snow and [s it would
snow ] i , [s he told Mark that ti.]
TREATING THAT-TRACE VARIATION 33
The literature on this phenomenon, notably Sobin (1987), Rizzi (1990) and
Sobin (1991), proposes certain alterations to the formulation of the ECP which
are designed to provide a means for proper government to take place in (lc), for
those speakers who find this sentence grammatical. In both cases, the mecha
nism involves strengthening the that complementiser by a procedure which Com
dexes it with the subject, so that it counts as a proper governor. The same pro
cedure is said to take place with the that complementiser found in subject relative
clauses like (lb).
34 PHILIP BRANIGAN
In essence, the strategy adopted is to patch the ECP to deal with the awkward
data. The strategy is not in itself unreasonable—we should hardly expect to
achieve exactly the right formulation of a principle of UG on the first try. But re
cent developments in the form of the theory as a whole—in particular,
Chomsky's (1993) minimalist program—make this strategy for characterising
dialectal variation less attractive, for the simple reason that the ECP is not a for-
mulable notion in the minimalist model. The ECP requires that a restricted form
of government obtain for all traces. In the minimalist model, the notion of gov
ernment is not defined, as all relations between head and phrasal category can be
defined in the more fundamental local -bar relations. Without government,
there is no proper government. Without proper government, there can be no
ECP. Then there is little point in reconciling the properties of dialect with the
ECP. What is needed is a new account of the ungrammaticality of (lc) in dialect
A.
It remains to be seen how RES (ECP) can account for that-trace effects in
English, where the relevant contexts are not V/2 complements. Conventional
wisdom has it that subjects occupy an A-posiüon within the clause—SPEC-AGRS,
in the minimalist model. If this is right, then there is in fact no way to derive the
English facts from RES(ECP). But I maintain that there is sufficient evidence that
conventional wisdom is wrong in this respect, as argued in Branigan (1992,
1996) and Harley (1995). The most direct evidence comes from the locative in
version construction, in which a PP is preposed to the position normally occu
pied by subjects, leaving the subject elsewhere. Example (15) illustrates.
If the locative PP occupied SPEC-AGRS, then the fact that the subject agrees
with the finite verb is inexplicable.4 If the locative PP occupies a different posi
tion, though, then covert movement of the subject to SPEC-AGRS allows the usual
nominative Case and agreement checking to take place. I conclude that the pre-
verbal position usually occupied by subjects in finite clauses is not SPEC-AGRS,
but rather a higher position. I now propose that this position is SPEC-, so that
the structure of a finite complement clause may involve 4CP recursion', as in
(16). 5
4
But cf. Hoekstra & Mulder (1990) and Watanabe (1993) for alternative proposals concerning
agreement in locative inversion.
5
Other literature which points in the same direction includes Culicover (1991) and Shlonsky
(1992). Culicover's Topic projection and Shonsky's AGRC projection can be taken as roughly
equivalent to my projection, for the purposes of this paper. The RES (ECP) is indifferent to
TREATING THAT-TRACE VARIATION 35
(16) John said [CP that [Œ Peter [s t loves green tea ]]]
The that-trace effect in the -dialects now follows from RES(ECP). Consider ex
ample (lc), once more.
(lc) (*) Which appk i do the experts agree that ti had been the tastiest?
(18) [CP whOi did [s John say [CP C -that [ ti tc [s ti had ...] ]]]]
The uniform -bar chain in (18) violates RES(ECP) because its tail is not in
the checking domain of any complementiser. The same explanation holds for
(19), where the locative PP occupies the position normally used by subjects, be
fore WH-movement raises it higher.
the label used on the category for which the subject is the specifier. What matters is that the
subject must occupy an -bar specifier position.
6
See Branigan (1995) for arguments that C-to-C movement takes place only in the overt syn
tax.
36 PHILIP BRANIGAN
(19) * Behind which doori did Tom say [cp C-that [CP ti t c [s sleep three
large dogs ti ]]]
When the that complementiser is not present, the primary complementiser re
mains in place, so that RES(ECP) is satisfied in (20).
(20) a. Which eli do the experts agree [CPti[s ti had been the
tastiest ]]
b. Behind which doori did Tom say [ cp ti C [s sleep three
large dogs ti ]]
As C71 is not raised, the tail of the uniform -bar chain is in its checking domain
at LF, thereby satisfying the RES(ECP).
The remainder of the (1) sentences (in dialect A) fall readily into place.
Consider first example (la).
The derivation of (la) matches that of (lc). The subject is extracted from
SPECC by WH-movement, giving rise to a strongly uniform chain. And the pri
mary complementiser must again adjoin to the secondary —in this case, the if
complementiser. The RES(ECP) is then violated by the LF representation.
The subject relative in (lb) is unproblematic, as long as the structure of the
relative clause is that shown in (21).7
In this case, the primary complementiser takes the form that. The primary
has phonetic content in a relative clause of this type, because its specifier has
none. Examples like (22) are excluded as a 'doubly-filled Comp' (DFC) effect
(Chomsky & Lasnik 1977).
As DFC effects are common elsewhere in the grammar, this need not be a
matter for concern. What is significant in the structure of (21) is the absence of
any uniform -bar chains in the relative clause. The sole -bar chain terminates
7
Branigan (1992) shows that 'vacuous movement effects' can be explained on the basis of a
structure for subject relatives similar to that proposed in the text.
TREATING -TRACE VARIATION 37
(23) which appleSj do the experts agree [CP ti that [s ti are the tastiest ]]
3. Conclusion
The data in (1) pose problems for an ECP account of constraints on subject
extraction. The ECP is too coarsely grained to deal with the actual variability of
the data, especially if the structures assigned to identical strings are invariant
across dialects. By relaxing the latter assumption, and by reformulating the ECP
as RES(ECP), a minimalist constraint on chains, the theory is made more flexi
ble, and the necessary fine distinctions in the data can then be drawn. The pic
ture of English that-trace variation which results from this analysis includes a
limited, and familiar, brand of variation. Dialect A and Dialect differ only in
the form of the primary complementiser in finite complement clauses, as sum
marised in (25).
Dialect A Dialect
non-relative relative non-relative relative
primary : ø that or ø that or ø that or ø
secondary C: that or if that or ø that or if that or ø
REFERENCES
Besten, Hans den. 1983. "On the Interaction of Root Transformations and Lexical
Deletive Rules". On the Formal Syntax of the Westgermania, ed. by Werner
Abraham, 47-131. Amsterdam & Philadelphia: John Benjamins.
Borer, Hagit. 1984. Parametric Syntax: Case Studies in Semitic and Romance
Languages. Dordrecht: Foris.
Branigan, Philip. 1992. Subjects and Complementizers. MIT dissertation.
. 1996. "Verb-second and the -bar Syntax of Subjects". Studia
Linguistica 50,1.51-79.
Browning, Marguerite. 1987. Null Operator Constructions. MIT dissertation.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
———. 1986. Barriers. Cambridge, Mass.: MIT Press.
. 1993. "A Minimalist Program for Linguistic Theory". The View from
Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. by
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press.
———. 1994. "Bare Phrase Structure". MIT Occasional Papers in Linguistics 5.
[Published as Chomsky, Noam. 1995. "Bare Phrase Structure". Government
and Binding Theory and the Minimalist Program, ed. by Gert Webelhuth, 381-
439. Oxford and Cambridge, Mass.: Blackwell]
— — . & Howard Lasnik. 1977. "Filters and Control". Linguistic Inquiry
8.425-504.
. 1993. "Principles and Parameters Theory". Syntax: Ein internationales
Handbuch zeitgenössicher Forschung = An International Handbook of
Contemporary Research, ed. by Joachim Jacobs, Arnim von Stechow, Wolfgang
Sternefeld & Theo Vennemann, 506-569. Berlin: Walter de Gruyter.
Culicover, Peter W. 1991. "Topicalization, Inversion, and Complementizers in
English", ms., Ohio State University.
Harley, Heidi. 1995. Subjects, Events and Licensing. MIT dissertation.
Hoekstra, Teun, & Rene Mulder. 1990. "Unergatives as Copular Verbs:
Locational and Existential Predication". The Linguistic Review 7.1-79.
Iatridou, Sabine, & Anthony Kroch. 1992. "The Licensing of CP Recursion and
its Relevance to the Germanic Verb-Second Phenomenon". Working Papers in
Scandinavian Syntax 50.1-25.
Larson, Richard. 1988. "On the Double Object Construction". Linguistic Inquiry
19.335-391.
Law, Paul, 1991. Effects of Head-Movement on Theories of Subjacency and Proper
Government. MIT dissertation.
Rizzi, Luigi. 1990. Relativized Minimality. Cambridge: MIT Press.
Schwartz, Bonnie & Sten Vikner. 1989. "All Verb-second Clauses are CPs".
Working Papers in Scandinavian Syntax 43.27-49
Shlonsky, Ur. 1992. "The Representation of Agreement in Comp and Subject
Clitics in West Flemish". Geneva Generative Papers 0,0.27-38.
Sobin, Nicholas. 1987. "The Variable Status of Comp-trace Phenomena".
Natural Language and Linguistic Theory 5.33-60.
. 1991. "Agreement in CP". Lingua 84.43-54.
TREATING THAT-TRACE VARIATION 39
Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. MIT disserta
tion.
Vikner, Sten. 1991. Verb Movement and the Licensing of NP-Positions in the
Germanic Languages. Université de Genève dissertation.
Watanabe, Akira. 1993. AGR-based Case theory. MIT dissertation.
Zwart, . Jan Wouter. 1993. Dutch Syntax: A Minimalist Approach. Univ. of
Groningen dissertation.
NEGATIVE PARTICLE QUESTIONS:
A DIALECTAL COMPARISON*
LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
University of California, Irvine & Academia Sinica, Nanking, Taipei
1. Introduction
There are a number of ways to form yes-no questions in Chinese, though
not every dialect employs all the choices. In this paper, we discuss a particular
yes-no construction which uses negation markers to form yes-no questions, as
in (1)-(3).
(1) Mandarin
hufei kan-wan-le nei-ben shu meiyou
Hufei read-finish-PERF that-cL book not-have
"Has Hufei finished reading that book?"
(2) Cantonese
wufei lei-zo mei
Wufei come-PERF not-yet
"Has Wufei come yet?"
(3) Taiwanese
i u tsiak beng bo
he have eat rice not-have
"Did he eat?"
* We thank the audiences of the APL A meeting on Mircroparametric Syntax and Dialect
Variation, the Workshop on Theoretical East Asian Linguistics at UC Irvine, the Stanford
Workshop on Historical Chinese Linguistics, and audiences at the universities of Connecticut,
Toronto, Illinois at Urbana-Champaign and Arizona. In particular, we would like to thank
Audrey Li, Richard Kayne, Chao-Fen Sun, Rint Sybesma, and Pei-Chuan Wei for their com
ments and suggestions. We also thank Cheng-Sheng Liu, Xiaoguang Li, Ruo-Mei Hsieh, Sze-
Wing Tang, Teresa Griffith, Pauline Huynh and Deng Wei for their native speaker judgements.
42 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
(4) Mandarin
hufei meiyou -wan nei-ben shu
Hufei not-have read-finish that-cL book
"Hufei did not finish reading that book."
(5) Cantonese
wufei mei lei
Wufei not-yet come
"Wufei has not come yet."
(6) Taiwanese
i bo tsiak beng
he not-have eat rice
"He did not eat."
(7) a. ta chang qu bu
he often go not
"Does he go often?"
b. ta yijing kan-wan shu meiyou
he already read-finish book not-have
"Did he already finish reading the book?"
1
There are other types of questions on a par with VP-not-V, such as VP-not-VP and V-not-VP.
We will only discuss VP-not-V questions. The arguments can be easily extended to the other
types.
44 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
As we can see in (11), A-not-A and VP-not-V questions can co-occur with the
question particle ne, though they cannot appear with ma.2 However, NPQs can
not co-occur with either ma or ne, as in (12).
2
Though A-not-A questions take ne as a question particle, they are still interpreted as yes-no
questions. This may seem strange at first glance. However, given Huang's (1991) proposal
which treats the formation of A-not-A questions on a par with typical WH-questions, the fact
that ne is used is not unexpected.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 45
(12) a * ta qu bu ne/ma
he go not WH/Y-N
"Is he going?"
b. * ta you qian meiyou ne/ma
he have money not-have WH/Y-N
"Did he have money?"
2. Negation Forms
The crucial element in NPQs is the negation marker. To understand the for
mation of NPQs, we must first consider the properties of negation in Chinese.
Every dialect in Chinese has more than one negation form. The negation form
varies depending on the aspectual markings on the verb or the verb type itself. In
other words, there is a matching or agreement requirement which holds between
the negation marker and the aspect/verb form. (In the following discussion, we
will not consider the negation marker in imperatives.)
2.1 Mandarín
Mandarin has two negation markers: bu and meiyou (see Wang 1965, Chao
1968 and Li & Thompson 1981). is used with bare verbs and modals.
Meiyou is used with various aspects and with accomplishment verbs.3 In the ex-
3
There are some apparent counterexamples to this description of the usage ofbuw and meiyou.
As we can see in (i), bu seems able to appear with the aspectual marker -zhe "PROG":
(i) ta bu bao-zhe zhen-tou shui-jiao
he not hold-PROG pillow sleep
"He does not sleep by holding a pillow."
46 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
amples below, we see that meiyou cannot appear with the modal hui (13b),
while bu cannot appear with the perfective aspect -le (14b) or the experiential as
pect -guo (14e). (14a) shows that meiyou is interpreted as perfective without the
presence of the perfective aspect -le (and in fact, meiyou cannot co-occur with
-le (14c), see Wang (1965) for an account of this restriction).
(13) a. ta bu lai
he not come
"He is not coming."
b. hufei bu/*meiyou hui qu
Hufei not/not-have will go
"Hufei will not go."
Note that both bu and meiyou can be used in A-not-A questions.4 As we will
see in subsequent sections, in Cantonese and Taiwanese, not all negation mark
ers can appear in A-not-A questions, further supporting our claim that NPQs can
not be derived from A-not-A questions.
2.2 Cantonese
Cantonese has three negation forms: m, mou, and mei (see Cheung 1972
and Yue-Hashimoto 1993).5 M is used with bare verbs and modals (on a par
with bu in Mandarin) and cannot be used with aspectual markers (18a-c). is
4
In Beijing Mandarin, instead of (17b), it is possible to say (i):
(i) qiaofeng qu-mei-qu
Qiaofeng go-not-go
"Did Qiaofeng go?"
5
It should be noted that Mandarin also has a negation marker, wei "not yet" which corresponds
to mei "not-yet" in Cantonese. However, the negation form wei "not-yet" has to co-occur with
the adverbials shang "yet" or hai "yet".
(i) a. ta shang/hai wei lai.
he yet not come
"He has not come yet."
b. * ta wei lai
he not come
"He has not come yet."
The contrast in (i) shows that wei is no longer a free form in Mandarin.
48 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
used with various aspects and accomplishment verbs and, like meiyou in
Mandarin, it cannot co-occur with the perfective aspect marker and its mere
presence is interpreted as perfective. Mei is similar to mou except that the former
has an added meaning of "not yet".
6
It should be noted that mou "not-have" can be used in NPQs only when the verb is the pos
sessive verb jau "to have", as shown in (i):
(i) keoi jau tsin mou
he have money not-have
"Does he have money?"
This indicates that there is a contrast between mou being the suppletive form of NEG plus the
aspectual jau "perfective" and mou being the suppletive form of NEG plus the verb jau "to
have". We will leave this issue open.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 49
However, both m and mou can appear in A-not-A questions while mei cannot.
2.3 Taiwanese
Taiwanese has four monosyllabic negation markers, m, bo, be, and buei
(see P. Li 1971, Teng 1992 and T.-C. Tang 1993). M is the neutral negation, bo
the perfective negation, be the future negation and buei is the negation marker
indicating "not-yet".
(23) a. i m lai
he not come
"He is not coming."
50 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
b. і be lai
he not-FUT come
"He will not come."
* i m/bo/buei e lai
he not/not-have/not-yet will come
"He will not come."
d. i m/*bo/*buei/be gaN chu-ki
he not/not-have/not-yet/not-FUT dare out-go
"He dare not/will not dare go out."
The examples in (23a-d) show that bo "not-have" and buei "not-yet" cannot
appear with modals (such as e "will" and gaN "dare"). M can appear with typical
modals except e "will" (probably due to the fact that be "not-future" is the sup
pletive form of NEG and e "will").
(24) a. i bo ki hak-hao
he not-have go school
"He did not go to school."
b. * i m u ki hak-hao
he not have go school
"He didn't go to school."
(26) a. li ki m
you go not
"Are you going?"
b. i u tsiak beng bo
he have eat rice not-have
"Did he eat?"
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 51
c. i e ki be
he will go not-FUT
"Will he go?"
d. і ki buei
he go not-yet
"Has he gone?"
A-not-A questions in Taiwanese are restricted to certain verbs such as si "to be"
and only the negation marker m.
(28) a. * і lai-bo-lai
he come-not-have-come
"Did he come?"
b. * і lai-be-lai
he come-not-FUT-come
"Will he come?"
c. * і lai-buei-lai
he come-not-yet-come
"Has he come yet?"
Hence, Taiwanese offers further evidence for our claim that NPQs are not de
rived from A-not-A questions.
3. Classical Chinese
Q-PARTICLE]. He shows that there were sentences of the VP-NEG form followed
by a question particle, as in (29), where the negation is fou followed by the
question particle hu. It should be noted that Classical Chinese has more than a
dozen negation markers. However, only bu,fou, wei, fei, and wu can occur in
NPQs.
The data in Classical Chinese above show the historical development of negation
markers as question particles. We will see that this sheds light on the formation
of NPQs in various dialects of Chinese.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 53
We have so far presented data showing that the three dialects under discus
sion differ with respect to the number of negation forms they have as well as
which one(s) can be used in NPQs. It is also clear that there is agreement be
tween the negation form and the aspect/verb. We now present data showing that
such agreement is maintained in NPQs in Mandarin while in Cantonese and
Taiwanese, the agreement collapses in NPQs.
Consider first examples of NPQs in Mandarin.
(34) a. * ta qu-le bu
Һ gO-PERF not
"Did he go?"
b. * ta qu-guo bu
Һ gO-EXP not
"Has he gone?"
c. ta qu bu
he go not
"Is he going?"
(36) a. ta hui/yinggai/neng qu bu
he will/should/can go not
"Will/should/can he go?"
b. * ta hui/yinggai/neng qu meiyou
he will/should/can go not-have
"Will/should/can he go?"
these two aspects, as shown in (35).7 (36) further shows that in NPQs, bu can
appear with modals while meiyou cannot. Hence, in Mandarin, the agreement
that we have seen in Section 2.1 between negation and verb/aspect is maintained
in NPQs as well.
In Cantonese, however, the agreement between negation and aspect/verb
does not seem to hold in NPQs. As noted earlier, mei "not-yet" is the only nega
tion form that can be used in NPQs. Thus, if agreement were to hold in NPQs in
Cantonese, we would expect that NPQs cannot contain modals, as mei cannot
appear with modals, as we have seen in (18b). It tums out however that though
mei carries the interpretation of "not-yet", it can still appear with modals (37), as
well as the typical perfective and experiential aspects in NPQs (38).8
The sentences in (38) and (37) show that even though Cantonese exhibits
agreement between negation and aspect/verb in typical negation environments,
such agreement is not observed in NPQs.
Taiwanese is similar to Cantonese in that there is no strict match
ing/agreement requirement in NPQs. We have shown earlier that Taiwanese is
similar to Cantonese and Mandarin in that there is agreement between negation
7
There are different proposals which address the question of why -le cannot appear with
mei(you) in regular negation contexts. We assume here that whatever the constraint is, it is not
a semantic incompatibility and that the structural description that leads to the non-co-occur
rence in this case is no longer met when the negation is in the C° position.
8 There is a yes-no question particle me in Cantonese which differs from mei in that it is not a
negation marker and does not have any restriction on the verb form. Thus, it is quite similar to
the ma question particle in Mandarin.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 55
The examples in (39) and (40) contrast with the ones in (23)-(25). In (23), we
see that m, bo, and buei cannot appear with the modal e "will". However, in the
NPQ in (39a), all the negation markers can appear with e "will". Similarly, in
(24b), we see that m cannot appear with the perfective aspect. In contrast, in the
NPQ in (40a) we see that m can be used even though the perfective marker is
present.
In short, the formation of NPQs differs in these dialects. Mandarin NPQs re
tain the same kind of agreement observed in typical negation environments while
Cantonese and Taiwanese NPQs do not.
5. Analysis
Two apparent questions arise given the above data in Classical Chinese and
in the three synchronic dialects of Chinese:
Let us now turn to the second question, the question of dialectal difference
with respect to agreement. Consider first the dialects which lack agreement in
NPQs (i.e., Cantonese and Taiwanese). We consider these dialects to resemble
Classical Chinese in the formation of NPQs. In particular, as we have noted ear-
9
It should be noted that in Shanghai, the sentence final negative particle has a 'literal' incor
porated question particle, as shown below:
(i) yi φ le a (ii) yi φ le va (iii) yi ve qi
Һ gO PERF Q Һ gO PERF not-Q he not gO
"Did he go?" "Did he go?" "He is not going."
As we can see in (iii), the typical negation form is ve and the one used in NPQ is the combina
tion of ve and the question particle a. We need to examine Shanghai more closely to see the
pattern of NPQs. Furthermore, it is pointed out to us (Sybesma, p.c.) that in a Northern dialect
of Mandarin, NPQs with bu can co-occur with ma, though the ordering of the Q-particle and the
negation marker appears to differ from what we find in Classical Chinese:
(iv) ni qu ma bu
you go Q not
"Are you going?"
Further tests are needed to see whether these are genuine NPQs (see Appendix for some basic
tests).
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 57
These questions are related to the nature of negation as well as to the nature
of the NEG-particle in these dialects. Before we turn to these questions, we will
first consider some supporting evidence for the movement vs. base-generation
distinction.
58 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
We have so far examined simplex NPQs, which illustrate a basic dialectal dif
ference in terms of agreement. Below we present data involving verbs which
take clausal complements. In particular, we consider sentences in which the
agreement requirement of the matrix verb differs from that of the embedded
verb. We show that such cases provide further evidence for a movement analy
sis in the formation of NPQs in Mandarin Chinese. Consider first a very simple
case, where the matrix and embedded verbs share the same agreement require
ment:
(42) ta yiwei ni qu bu
he think you go not
(a) "Does he think or not think that you are going?"
(b) "Does he think that you are going or not going?"
In (45), the NEG-particle is bu and only the matrix verb satisfies the agree
ment requirement since the embedded one has the experiential marker attached to
it. As expected, the question does not have an embedded reading. On the other
hand, the NEG-particle in (46) is meiyou and only the embedded predicate can
agree with it since the matrix has the modal hui "will". And again, as expected,
the question does not have a matrix reading.
If the agreement constraint is some sort of non-local constraint on the C° and
the verb/aspect, it is possible to account for (45) but not for (46). For (45), it is
still possible to maintain that the NEG-particle cannot be base-generated in the
embedded C° due to the constraint. Thus, the only possibility is for it to be base-
generated in the matrix. However, (46) presents a problem for such an analysis.
It should be noted that even though the allowed reading in (46) is an embedded
reading, it is still a matrix question. Thus, though meiyou is allowed to be base-
generated in the embedded clause since it can occur with the experiential aspect
marker -guo, it has to move to the matrix C°. That is, the NEG-particle will
eventually end up in the matrix C°, even though the modal hui "will" is present
in the matrix. The non-local agreement constraint which rules out sentences such
as (44) will also rule out (46).
The above data show that if NEG-particles in Mandarin were to be base-gen
erated in C° positions, sentences such as (46) cannot be accounted for. On the
other hand, given a movement analysis, the grammaticality and the readings of
the sentences in (42), and (44)-(46) naturally follow. Let us consider the sen
tences one by one. In the grammatical and ambiguous (42), the NEG-particle bu
can be either generated in the embedded NEG or matrix NEG. In the embedded
case, the NEG-particle moves to the matrix C° via the embedded C°. In (44), the
NEG-particle meiyou cannot be generated in either the embedded NEG or the ma
trix NEG due to the selectionai restriction between the verb/aspect and the NEG
and thus the sentence is ruled out. On the other hand, in (45), even though the
NEG-particle bu cannot be generated in the embedded NEG due to the experiential
aspect marker -guo, it can be generated in the matrix NEG and subsequently
moves to the matrix C°. Turning now to the problematic case for the non-move
ment analysis, in (46) we see that the NEG-particle meiyou can indeed be base-
generated in the embedded NEG. As in the ambiguous case in (42), meiyou first
moves to the embedded C° and it subsequently moves to the matrix C°. The
movement from the embedded C° to the matrix C° does not involve the matrix
predicate, nor does it involve the matrix NEG. Hence, even though the
verb/aspect type of the matrix in (46) does not appear to agree with the NEG-par
ticle which ends up in the matrix C°, the sentence is still grammatical, with the
embedded reading.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 61
A question that arises given such mixed cases is whether or not Cantonese
and Taiwanese data differ from the Mandarin data presented above. Given the
fact that Cantonese and Taiwanese do not display agreement in NPQs, it is ex
pected that even in mixed cases, ambiguous readings are allowed since NEG-par-
ticles are base-generated in C°s and can undergo movement. This prediction is
borne out, as (47)-(54) show.10
Taiwanese
(47) i giosi li (e) ki m
he think you will go not
(a) "Does he think or not think that you are leaving?"
(b) "Does he think that you are leaving or not leaving?"
(48) і e giosi li yinggai ki bo
he will think you should go not-have
(a) "Will he think or not think that you can leave?"
(b) "Will he think that you can leave or cannot leave?"
Cantonese
(51) keoi yiwai ni zau mei
he think you leave not-yet
(a) "Does he think or not think that you are leaving?"
(b) "Does he think that you are leaving or not leaving?"
(52) keoi wui yiwai ni hoyi zau mei
he will think you can leave not-yet
(a) "Will he think or not think that you can leave?"
(b) "Will he think that you can leave or cannot leave?"
10
It should be noted that even though the sentences are ambiguous, in some cases, there is a
preferred reading.
62 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
Mandarin
(55) ta xiang-zhidao ni lai-le meiyou
he wonder you come-PERF not-have
"He wonders whether you came."
(56) * ta xiang-zhidao ni qu bu
he wonder you go not
"He wonders whether you are going."
The contrast between meiyou and bu in their ability to indicate embedded yes-no
questions is illustrated in (55) and (56). The verb xiang-zhidao "to wonder"
requires an embedded interrogative and the ungrammaticality of (56) is due to
the fact that bu cannot indicate an embedded question. This property of bu is on
a par with the typical yes-no question particle ma, as we see in (57) and (58):
It thus appears that bu is similar to ma in that both have a 'matrix property'. The
matrix property of ma has been attributed to the speaker-oriented interpretation
of ma. We suggest that bu is on a par with ma in its speaker-oriented property.
Taiwanese m also cannot indicate embedded questions, in contrast with the
other negation markers, as shown in (59).
ers are both negative markers and question particles because not every negation
marker can be used in NPQs (for example, in Cantonese, only mei can be used in
NPQs). These 'dual status' negation markers are thus like some negation markers
in Classical Chinese in that they can function as a negation marker or as a ques
tion particle.
Let us turn now to the movement dialect, Mandarin. Following Chomsky
(1995), we assume that overt movement is triggered by unchecked features.
Hence, in Mandarin, movement of the negation markers is to check some formal
feature in C°. Note that the negation markers in Mandarin cannot be inserted as
question particles in C° to mark yes-no questions. In other words, negation
markers in Mandarin do not have a dual status, even though they do appear in
C°. We propose that Mandarin has a phonologically null C° with the formal fea
tures [Q, NEG]. This C° can be considered to be another residue of the historical
development of negation as question particles. That is, instead of having a full-
fledged negation marker functioning as a question particle, Mandarin has a C°
with [Q, NEG] features, with just a 'trace' of negation in it. This [NEG] feature
however has to be checked off in order for the sentence to be properly inter
preted as a yes-no question. Thus, negation markers bu and meiyou can and
must move to C° to check the [NEG] feature.
The picture that has emerged here provides answers to the questions in (60).
In particular, the answers all relate to the properties of or NEG. With respect to
Mandarin, its negation markers cannot be base-generated in NPQs because they
are 'pure' negation markers rather than those with a dual status. And what
'allows' negation markers to move in Mandarin is the particular feature [NEG] in
the C°. The answer to the question raised in (60c) may have to do with
Economy. Given the fact that Cantonese and Taiwanese also have negation
markers, why is it the case that they cannot undergo movement, just as negation
markers do in Mandarin? There are indeed two different possibilities:
(a) Cantonese / Taiwanese also has a C° with a [Q, NEG] feature;
(b) Cantonese / Taiwanese does not have such a C°.
Consider possibility (b) first. If these two dialects do not have such a C°, there is
then no motivation for the negation markers to undergo movement. On the other
hand, if we have possibility (b), the trigger for movement is present, and the
question then is why movement is lacking. Note however that these dialects
have negation markers as question particles which can be directly merged into
the computation. Assuming that Merge is less costly than Move (see Chomsky
1995), the Merge option and thus the insertion of these negation markers always
rules out the movement possibility.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 65
5.4 Extension
Note that in Mandarin, mei can be used alone in both NPQs and A-not-A
questions. The difference between mei and meiyou in A-not-A questions is
shown in (61) (see also foonote 4). We assume here that mei is a reduced form
of meiyou, which has a verbal element vow "to have" in it.
(61) a. ta lai-mei-lai
he come-not-have
"Did he come?"
b. * ta lai-meiyou-lai
he come-not-have-come
"Did he come?"
Cantonese and Taiwanese as well as the dialectal split may appear to be mysteri
ous. However, we suggest that the analysis proposed in this paper together with
Huang's (1991) analysis of A-not-A questions provide an answer to the above
distribution.
Huang (1991) proposes that the formation of A-not-A questions involves a
PF-insertion of a negator. In other words, the negator in A-not-A questions does
not enter into the computation. At PF, the insertion of a negator is to occupy the
'not' slot in A-not-A questions. It is thus reasonable to assume that only 'pure'
negators can be inserted. The complementary distribution in Cantonese and
Taiwanese is thus explained. In both dialects, the negators that can be used in
NPQs have dual status. They thus do not qualify as 'pure' negators. On the other
hand, in Mandarin, negators differ from the ones in Cantonese/Taiwanese in that
they are simple NEG elements and thus they can also be inserted in A-not-A ques
tions. Note that meiyou cannot be inserted in A-not-A questions because meiyou
is a composite form consisting of both the negator mei and the verb you "to
have". Thus PF insertion will only see mei listed as a NEG.
The analyis of NPQs presented above thus provides a rather simple account
of the complementarity noted as well as the dialectal split in terms of A-not-A
questions.
6. Conclusion
We have argued that the main dialectal difference between Mandarin and
Cantonese/Taiwanese is due to the NEG°-to-C0 movement in the former and the
lack of it in the latter. We have seen a basic difference in terms of agreement
patterns in these two types of dialects. Sentences involving embedding further
support our claim that in Mandarin, there is NEG°-to-C° movement.
The dialectal variation we have seen may be traced back to historical devel
opment of negation markers as question particles. It appears that
Cantonese/Taiwanese maintains the grammaticalization of negation markers as
question particles and thus these negation markers can be simply inserted in the
C° position to form a yes-no question. On the other hand, Mandarin negation
markers are no longer question particles. The only 'trace' of the grammaticaliza
tion of negation markers as question particles that remains in this dialect can be
seen in the C° feature [Q, NEG], which triggers the movement of NEG in NPQs.
Lastly, we would like to point out that NPQs are not just found in Chinese
dialects. It appears that NPQs exist in languages such as Cambodian, Thai and
Vietnamese:
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 67
Appendix
Comparison of NPQs with other kinds of yes-no questions
We examine here in detail the properties of NPQs and show that they are in
terpreted as other types of yes-no questions (such as Mandarin ma questions,
tag-questions, haishi-questions, A-not-A questions, VP-not-V questions, VP-VP
questions), but that they differ from other types of yes-no questions. We will
discuss Mandarin and Taiwanese, the former a representative of the movement
type in NPQs, the latter a representative of the base-generation type in NPQs.
Mandarin yes-no questions
Before we compare NPQs with the other types of yes-no questions, we first
provide a brief overview of some yes-no question types in Mandarin.
(67) Μα-question
ta lai-le ma
he come-PERF Q
"Did he come?"
(68) Tag-question
ta hui lai, bu shi ma
he will come not be Q
"He is coming, isn't he?"
Μα-questions are characterized by the sentence final particle ma. As we can see,
ma is also needed in tag-questions.
(69) Haishi-qaestion
ta lai haishi bu lai
he come or not come
"Is he coming or is he not coming?"
(70) A-not-A question
ta xihuan-bu֊xihuan ni
he like-not-like you
"Does he like you?"
(71) VP-not-V question
ta xihuan ni-bu-xihuan
he like you-not-like
"Does he like you?"
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 69
(73) a. ta qu bu
he go not
"Is he going?"
b. * ta bu qu bu
he not go not
"Isn't he going?"
c. ta ku-le meiyou
he cry-PERF not-have
"Did he cry?"
d. * ta meiyou ku meiyou
he not-have cry not-have
"Didn't he cry?"
The same restriction can be found in A-not-A questions and VP-not-V questions,
as in (74) and (75).
11
Some yes-no questions are irrelevant in this sub-section, such as ma-questions, tag-ques
tions andhaidhi'-questions.Themaz-questionsare irrelvant because the questions cannot be con
sidered to consist of a yes and a no part. As for tag-questions and haishi-questions, there is no
requirement on the co-occurrence of yes and no parts. VP-VP questions require that both VPs
consist of non-negated verb forms.
12
The NPQs with the negation marker m in Taiwanese are an exception. See example (86) for
details.
70 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
(74) a. ta qu-bu-qu
he go-not-go
"Is he going?"
b. * ta bu qu-qu
he not go-go
"Isn't he going?"
c. ta you-meiyou qu
he have-not-have go
"Did he go?"
d. * ta meiyou qu qu
he not-have go go
"Didn't he go?"
This property alone may indicate that NPQs seem to be related to A-not-A
questions or VP-not-V questions. However, we saw earlier that NPQs cannot be
derived from A-not-A or VP-not-V questions: preverbal adjuncts and ne, which
may appear with these question types, may not appear in NPQs.
Co-occurrence with aspect markers
Each Chinese dialect has a number of aspectual markers. As already pointed
out in section 2.1., aspect markers can appear in NPQs, as shown in (76).
b. ta qu-le meiyou
he go-PERF not-have
"Has he gone?"
(78) Speaker A:
ni haoxiang shou-le yidian
you seem thin-PERF a little
"You seem to have lost some weight."
72 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
Speaker B:
shi ma? ni kan wo shou-le ma?
be Q yOU See I thin-PERF Q
wo ziji dao bu juede
I self on:the:contrary not feel
"Is that so? Do you think I've lost weight?
I haven't noticed it myself."
Speaker B:
* shi-bu-shi? ??ni kan wo shou-le meiyoul
be-not-be you see I thin-PERF not-have
wo ziji dao bu juede
I self on:the:contrary not feel
In contrast, adverbials like daodi "on earth" can only occur in NPQs.13
13
In addition to NPQs, daodi can also appear in haishi-questions, A-not-A questions and VP-
not-V questions. It however cannot occur in VP-VP questions.
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 73
b. daodi ta hui qu bu
really he will go not
"Is he really going?"
c. daodi ta lai-le meiyou
really he come-PERF not-have
"Did he really come?"
Taiwanese yes-no questions
Compared to Mandarin, Taiwanese has ma-type questions, tag-questions
and haishi-type questions, but not A-not-A questions (except in cases with the
copula si "be"), VP-not-V questions, and VP-VP questions. However, Taiwanese
has other kinds of yes-no questions like gam-questions and VP-a-VP ques
tions. 14 ՛ 15
14
We will not discuss all kinds of Taiwanese yes-no questions here.
15
For a discussion of the pragmatic function of various Taiwanese sentence final question par
ticles, see Chen (1993).
74 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
Comparisons
Let us now turn to the comparison between Taiwanese NPQs and these other
types of yes-no questions.
Positive and negative
Except for m, in NPQs the verb has to be in a non-negation form, as shown
in (86).
(86) a. i ki/ m ki m
he go/not go not
"Is he going?/Is he not going?"
b. і u/*bo ki bo
he have/not-have go not-have
"Did he go?"
і e/*be ki be
he will/not-FUT go not-FUT
"Will he go?"
Since Taiwanese has neither A-not-A nor VP-not-V questions, no comparison can
be made with such sentences. This constraint is irrelevant for /no-questions,
sibo-questions, asi-questions and gam-questions. The hio- and gam-questions
are irrelevant because they cannot be treated as consisting of a yes and no part.
As for sibo-questions and asi-questions, there is no requirement on the co-oc
currence of yes and no parts. VP-a-VP questions require both VPs to consist of
negation or non-negation verb forms.
Non-temporal preverbal adjuncts
As with Mandarin NPQs, in Taiwanese it is possible for NPQs to have pre-
verbal adjuncts, as in (87).
(87) . і tiaNüaN ki m
he often go not
"Does he go often?"
b. i yiting e ki be
he certainly will go not-FUT
"Will he certainly go?"
CHINESE NEGATIVE PARTICLE QUESTIONS: A DIALECTAL COMPARISON 75
Note that, except for VP-a-VP questions, other types of Taiwanese yes-no ques
tions can also take non-temporal/locative preverbal adjuncts.
(90) . і ki-gue bo
he go-EXP not-have
"Has he ever been there?"
b. i jiak-a buei
he eat-PERF not-yet
"Has he eaten?"
However, aspect markers cannot appear in VP-a-VP questions, though they can
occur in the other types of Taiwanese yes-no questions.
(92) a. * i ki m hio
he go not Q
"Does he go?"
b. * i u ki bo hio
he have go not-have Q
'Oid he go?"
c. * i e ki be hio
he will go not-FUT Q
"Will he go?"
The comparisons discussed above have clearly shown that while NPQs and
certain other questions are semantically yes-no questions, they are syntactically
very distinct from one another.
REFERENCES
Chao, Yuen-Ren. 1968. A Grammar of Spoken Chinese. Berkeley and Los
Angeles: University of California Press.
Chen, Janet C.-W. 1993. "Taiwanese Sentence-final Question Particles".
Proceedings of the First Symposium on Languages in Taiwan, C4, 01-30. Taipei:
National Taiwan University.
Cheng, Lisa L.-S. 1991. On the Typology ofW/h-questions.PhD dissertation, MIT.
Cheung, Samuel H.-N. 1972. Cantonese as Spoken in Hong Kong. Hong Kong:
The Chinese University of Hong Kong.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Huang, C.-T. James. 1991. "Modularity and Chinese A-not-A questions".
Interdisciplinary Approaches to Language: Essays in Honor of S.-Y. Kuroda,
ed. by Carol Georgopoulos & Roberta Ishihara, 305-332. Dordrecht: Kluwer.
Li, Charles & Sandra A. Thompson. 1981. Mandarin Chinese: A Functional
Reference Grammar. Berkeley: University of California Press.
Li, Paul. 1971. "Two Negative Markers in Taiwanese". Bulletin of the Institute of
the History of Philology 43.201-220. Institute of History and Philology,
Academia Sinica, Taiwan.
Noss, Richard . 1964. Thai Reference Grammar. Washington D.C.: Foreign
Service Institute, Department of State.
78 LISA L.-S. CHENG, C.-T. JAMES HUANG & C.-C. JANE TANG
Tang, T.-C. 1989. Studies on Chinese Morphology and Syntax 2. Taipei: Student
Book Co.
. 1993. "Mingnanhua Foudingci de Yuji Neihan yu Jufa Biaoxian (A
Semantic and Syntactic Study of Negation in Taiwanese)". Publications of the
National Science Council, Part C: Humanities and Social Sciences. 3,2.224-243
Taipei, Taiwan.
Teng, Shou Xin. 1992. "Diversification and Unification of Negation in
Taiwanese". Chinese Languages and Linguistics 1. 609-629.
Wang, William. 1965. "Two Aspect Markers in Mandarin". Language 41,3.457-
470
Yue-Hashimoto, Anne. 1988. "A Preliminary Investigation into the Subclassifica
tion Problem of the Yue Dialects". Computational Analysis of Asian and
African Languages 30.7-38. Tokyo.
.1992. "Comparative Dialectal Grammar: Problems and Prospects". Paper
presented at the Fourth North American Conference on Chinese Linguistics,
University of Michigan, May 1992.
.1993. Comparative Chinese Dialectal Grammar: Handbook for
Investigators. Paris: Ecole des Hautes Études en Sciences Sociales, Centre de
recherches linguistiques sur ľAise Orientale.
Zhang, Min. 1990. A Typological Study of Yes-no Questions in Chinese Dialects:
in Diachronic Perspective. PhD dissertation, Peking University.
IMPERATIVE INVERSION
IN BELFAST ENGLISH
ALISON HENRY
University of Ulster at Jordanstown
Introduction
Chomsky (1995) admits some optionality, but the concept of 'economy' still raises the
question of why movement takes place at all if it is optional.
80 ALISON HENRY
cess (for example verb raising) obligatorily while in the other it is optional, sits
uneasily with current approaches.
From the perspective of recent work in linguistic theory, then, we would ex
pect differences between language varieties to be of the following types:
The question arises as to whether such an approach can in fact account for
thefine-graineddifferences we find between dialects, where there appear to be a
very large number of possible small-scale differences between language vari
eties. Two types of study outside the mainstream of work on linguistic theory
suggest that there may be problems. Work on sociolinguístics (see Labov 1994
for an overview) has argued, on the basis of studies of the output of a variety of
speakers, that language is characterised by a high degree of variability, subject to
both linguistic and extra-linguistic constraints. This view of language, where
speakers are seen to select between sociolinguístic variants made available by the
grammar, is essentially at variance with a model in which optionality is excluded
in principle.
There have also been findings in the study of historical syntax which have
called into question a straightforward parameter-setting model. Kroch (1994) ar
gues that historical change takes place via competing grammars which differ in
relation to one parameter setting, with the use of one grammar replacing the
other through a gradual shift. Again, this approach also seems to call into ques
tion a model in which there is no optionahty, and parameters are categorically set
in one direction or another.
In this paper, we will explore dialect differences in relation to overt-subject
imperatives between standard English and Belfast English, and within Belfast
English, in order to find out whether the kind of differences we find can in fact
be accommodated within current parameter theory.
It will be shown that the dialect variation found can to some extent be ac
commodated within the highly restrictive model of syntax envisaged by the
Minimalist program of syntactic theory (Chomsky 1994 ) : in particular, it can be
accounted for in terms of whether movement to a given functional projection is
IMPERATIVE INVERSION IN BELFAST ENGLISH 81
overt or covert. However, there are two respects in which the Belfast English
data presents problems for aspects of the theory. First, to account for the vari
ability found within the grammars of individual speakers, we need to admit a
degree of optionality into the syntax, allowing functional categories to be op
tionally strong. Secondly, to account for the fact that verb-raising can be specific
to certain constructions, we need to be able to specify the strength or weakness
of individual functional elements (e.g., the imperative morpheme) and not just a
functional category (e.g., C) in general. The theory thus on the one hand pre
dicts the type of variation which is found (overt versus covert movement) but
does not appear to allow sufficient flexibility in relation to the optionality of
movement and the level at which the strength or weakness of functional elements
may be specified.
Like standard English imperatives, Belfast English imperatives can have null
subjects:
(1) Go away.
However, imperatives with overt subjects differ from standard English in that
they can have postverbal subjects. Thus, in addition to the standard English
version with a preverbal subject, as in:
In such sentences, the verb appears to have moved out of the verb phrase to a
position in front of the subject.
The order main verb-subject is restricted to imperatives; in other sentence
types the word order is as in standard English. Like standard English, Belfast
English does not allow the raising of main verbs, other than be and have, out of
the Verb Phrase.
(4) a. * Yesterday met Mary the students.
b. Yesterday Mary met the students.
82 ALISON HENRY
Henry (1992) argues that in this structure for lowers to to, but one might be
tempted to seek an analysis of this which linked it to the imperative inversion
construction—though it is less obvious how the link might be construed than is
the case with inversion in embedded questions. However, such an effort would
again be misleading: the for-to construction does not necessarily co-occur in in-
IMPERATIVE INVERSION IN BELFAST ENGLISH 83
In this dialect verb-subject order can occur with all verbs. It seems that the
verb has moved out of the Verb Phrase to a position above the subject.
Object shift is obligatory for weak pronouns where the verb raises; all speakers
who have verb raising also have this object shift.
And do-support is not available in imperatives even where required to fulfill the
NEG-criterion, nor is emphatic do available.
Moreover, don't appears with verbs which in general do not allow do support,
that is, be and auxiliaries.
This marker cannot co-occur with don't, the negative imperative marker.
(31) a. * Don't gon you make your mummy a cup of tea.
b. * Gon don't you make your mummy a cup of tea.
In this dialect, it seems that the subject can only occur postverbally where it
originates in a post-verbal position; the subject does not appear to be forced to
raise out of VP. Where there is an auxiliary, the subject occurs, not after the
auxiliary, but after the lexical verb.
Henry (1995) attributes the availability of this construction to the fact that subject
raising to SPEC-AGRS is not obligatory in Belfast English, a fact that is reflected
in the optionality of subject-verb concord.
However, further study of speakers outside the inner Belfast area has shown
a dissociation between the availability of 'singular concord' and that of non-
raising of subjects in imperatives; there are speakers who allow non-raising of
subjects in imperatives, but do not have singular concord. Thus, the optionality
of subject raising seems again to be exclusive to imperatives, a property of the
strength of the NP-feature of some functional head in the imperative construc
tion. It is only in this construction that postverbal subjects are possible; they do
not occur generally in Belfast English.
This variety of Belfast English, then, differs from standard English in that,
in imperatives, the subject raises at LF rather than before Spell-out, exactly the
type of difference we would expect to see under the Minimalist program.
However note that, like verb raising in the unrestricted inversion dialect, this is a
property of a single construction—perhaps of the imperative morpheme—and it
is also an optional property. Alongside imperatives which do not have subject
raising, the same speakers use imperatives in which the subject does raise.
We can tell that the preverbal DPs in these sentences are true subjects rather than
vocatives (which of course also occur with imperatives, as in (47), with an into
nation break), not only because an intonation break is not necessary, but also
because pronouns can refer back to third-person subjects. Beukema &
Coopmans (1989) point out that this is only possible with true subjects; where
the DP is a vocative, non-third-person pronouns cannot be co-referential with it.
(47) Everybodyi go to thehj *i places.
Finally, let us note that there are some speakers, mainly outside the inner-
city areas of Belfast, who show a pattern of inverted imperatives where the sub
ject occurs after the main verb in sentences with auxiliaries, even where the verb
is not unaccusative or passive. In Henry (1995) I argued that speakers in Belfast
did not in fact use such structures, but it turns out that they are used in some
other areas of Ireland which have inverted imperatives.
IMPERATIVE INVERSION IN BELFAST ENGLISH 91
For speakers of this dialect, this order is obligatory when the verb raises. In
all other respects, this variety is like the unrestricted inversion dialect; it permits
NEG criterion sentences and object shift, and disallows negation with inversion.
This dialect appears to have, like Spanish, movement of the main verb along
with the auxiliary to C.
Here again we see a dialect exploiting a possibility offered by UG. However, the
selection of this possibility does not appear to depend on other selections in the
grammar; the dialect of speakers who select this option does not differ from that
of the unrestricted inversion dialect in any other way.
In relation to imperatives, there appear in fact to be four different dialects in
the Belfast area: one which like standard English does not allow inversion; one
which allows verb movement to in imperatives, and in which subject raising is
optional; a variant of this in which, where an auxiliary raises, the main verb
raises with it; and one which has subject raising as optional but does not allow
verb raising. Thus, even within a small area, there are several competing di
alects.
5. Conclusion
REFERENCES
Beukema, Fritz & Peter Coopmans. 1989. "A GB Perspective on the Imperative in
English". Journal of Linguistics 25.417-436.
Chomsky, Noam. 1986. Knowledge of Language. New York: Praeger.
. 1993. "A Minimalist Program for Linguistic Theory". The View from
Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. by
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge, Mass.: MIT Press.
. 1994 "Bare Phrase Structure". MIT Occasional Papers in Linguistics 5.
[Published as Chomsky, Noam. 1995. "Bare Phrase Structure". Government
and Binding Theory and the Minimalist Program, ed. by Gert Webelhuth, 381-
439. Oxford and Cambridge, Mass.: Blackwell]
. 1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
IMPERATIVE INVERSION IN BELFAST ENGLISH 93
Introduction
We discuss two parameters which account for much of the variation in the
form of the noun phrase among the Scandinavian languages and dialects, in par
ticular possessive constructions, and more particularly, constructions with a
postnominal possessor. We will claim that one of the two parameters is a so-
called major parameter, that is a parameter which (a) concerns the feature values
of a functional category with a general distribution, (b) which is therefore resis
tant to change, and (c) therefore resistant to dialectal variation. The other pa
rameter is a so-called minor parameter, that is to say a parameter which (a) con
cerns the feature values of a functional category with a restricted distribution, (b)
which is therefore vulnerable to change, and (c) therefore subject to dialectal
variation.
constructions. Other parameters are minor parameters, in the sense that they af
fect few constructions, in the most extreme case only one.
The value of a minor parameter is much more likely to change from one gen
eration to the next than the value of a major parameter. All it takes to change the
value of a minor parameter is that the construction where the parameter has its
effects, and which hence provides the crucial trigger experience for fixing its
value, is pushed out of use by another construction, spontaneously created or
taken over from, say, a neighbouring dialect or language. In other words, a mi
nor change in the linguistic environment may be enough to change the value of a
minor parameter from one generation to the next. To change the value of a major
parameter, on the other hand, it is not enough that just one construction is
changed or replaced: since the value for a major parameter is selected on the ba
sis of information from several constructions, there have to be several more or
less simultaneous changes in the linguistic environment, affecting several of the
relevant constructions, to create the conditions for a change in the setting of the
parameter.
Since the value of a minor parameter is easily changed from one generation
to the next, it is likely to be subject to dialectal variation. A major parameter, on
the other hand, is less likely to exhibit dialectal variation. That is to say, what we
call dialectal variation typically concerns minor parameters.
One of the leading ideas within P&P theory is that all syntactic variation is
due to variation in the feature values of certain functional categories. In this per
spective a major parameter is one which concerns features of a functional cate
gory which is involved in many constructions, while a minor parameter is one
which concerns features of a functional category which may occur in only one
construction, in the most extreme case. We may summarized the properties of
major and minor parameters as follows:
a MINOR PARAMETER:
• concerns feature values of a functional category
with a restricted distribution,
• is (therefore) vulnerable to change, and
• (therefore) likely to show dialectal variation.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 97
Within the Scandinavian languages, there are basically two forms of post-
nominal possessor constructions: one with a bare head noun, the other with a
head noun bearing the definite suffix ('N.D' = noun with definite suffix).
The construction exemplified in (2a), which we will call the N-POSS con
struction, is found in Icelandic and Old Scandinavian, but in no variety of
Mainland Scandinavian. The N.D-POSS construction of (2b) is common to
Icelandic, Norwegian, Northern Swedish, and, with a PP possessor, Farnese.
We will address mainly two issues. First, we will explain why the N-POSS con
struction (2a) is restricted to Icelandic and Old Scandinavian. We claim that this
is an effect of a major parameter in Scandinavian syntax, namely strong versus
1
Dialect variation may be restricted by various social factors, such as a strong normative tradi
tion, enforced in primary education and by various forms of 'language control'.
98 ANDERS HOLMBERG & GÖREL SANDSTRÖM
weak Case: Icelandic and Old Scandinavian have strong Case, while the other
Scandinavian languages all have weak Case. Second, we will discuss the N.D-
POSS construction (2b). This construction has been discussed in the literature on
the basis of facts from Icelandic, Norwegian, and some varieties of Northern
Swedish. In all of these languages/dialects the possessor can only be a pronoun,
or possibly a proper name in construction with a pronoun (examples will be
provided below). When we look at a wider range of Northern Swedish dialects,
however, it emerges that the variation in the categorial and morphological fea
tures of the possessor in this construction is much richer than is generally as
sumed. We will propose an analysis which is consistent with this variation. The
claim is that the definite form of the noun is moved to D, the highest functional
head in the nominal projection, and the possessor is moved to the specifier posi
tion of an AGR head situated between D and NP.2 The variation is due to a minor
parameter in the feature-content of the DP-internal AGR, where, however, the
features are subject to a universal hierarchy of argument features which is well
known from the typological literature.
(a) there is also some striking variation which cuts across this division, and (b)
there is a remarkable amount of variation within one of the main groups, namely
Mainland Scandinavian. In fact, it appears that there is at least as much word or
der and morphological variation in possessive constructions within Northern
Swedish alone as there is among all the Scandinavian standard languages. In (3)
are exemplified three possessive constructions found in the Scandinavian lan
guages. The table in (4) illustrates how they are distributed among the
Scandinavian languages.5
5
Mainly for ease of presentation we have left out Faroese from the table. Faroese has + in all
three rows. However, there are special restrictions on all the forms: N-POSS occurs only with
family relations, N.DEF-POSS only with a PP possessor, and POSS-N only with proper names.
We will not deal with Faroese in this paper, but see Holmberg (1994). The table represents an
idealization also in that not all Norwegian or Northern Swedish dialects have POSS-N as well as
N.DEF-POSS, but some do, including both standard varieties of Norwegian.
100 ANDERS HOLMBERG & GÖREL SANDSTRÖM
Why is (2a), repeated here as (5a), not well formed in any MSc dialect, in
cluding those dialects which allow postnominal possessors in some construc
tions?
6
The claim made in these works, apart from Holmberg (1994), is that AGRS is actually absent
in MSc, reflected in the absence of any subject-verb agreement morphology. As mentioned
above in fn. 3, Faroese can be described as being truly in-between the two main groups, having
strong AGRS (like Icelandic) but weak Case (like MSc); see Holmberg & Platzack (1995),
Holmberg (1994).
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 101
These properties are exemplified in (8) and (9) (W = the so-called weak adjec
tival inflection).
less true (as far as we know) that no MSc dialect has a construction such as (9a)
as the unmarked alternative, the way Icelandic does.7
We assume the following principle, given here in two versions. Following
Delsing (1993) we call it 'the Argument Rule'.
7
Some MSc varieties have it as a marked, semantically restricted alternative; see Delsing
(1993:118). More interestingly, many Northern Swedish dialects lack a free definite article, but
nevertheless do not allow the construction (9a). Instead the adjective is incorporated in the defi
nite noun.
(i) Jag såg nybilen.
Í saw new.car.D
"I saw the new car."
According to Sandström and Holmberg (1994) the definite noun is moved to D (called in that
work) across the attributive adjective, followed by adjective incorporation into the definite
noun, It is not inconceivable that the lack of a free definite article is one of the factors behind
the preference for the N.DEF-POSS construction over POSS-N in Northern Swedish. It is not,
however, the case that all dialects which have N.DEF-POSS also have Adjective Incorporation.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 103
functional head determining the scope, and more generally, the syntactic role of
the feature, namely D.8 Both Case and definiteness are syntactically properties of
DP, although they may be morphologically realized on N, or some other head
other than D within DP. Checking is here regarded as a way to ensure that the
morphological feature and the syntactic feature have matching values. If the
syntactic feature is strong, in Chomsky's (1993) sense, it must be checked be
fore spell-out, either by insertion of an appropriately specified category in one of
the checking positions (cf. 1 lb), or by movement of an appropriately specified
lexical category. If the feature is weak, checking can wait until LF. We assume
that the need for checking is mutual between the morphological and the syntactic
feature. Let us say that the morphological feature cannot be interpreted at LF un
less it is checked. Following Chomsky (1993) we assume the syntactic feature
cannot be phonologically interpreted, and therefore must be checked off before
spell-out if it is strong (= 'visible'), assuming checking to result in deletion of
the syntactic feature. A weak (= 'invisible') feature is innocuous in PF, and
therefore need not be checked off before LF. Also following Chomsky (1993),
we assume the Greed principle:
(12) GREED
Movement of a category α is triggered only if the move
ment helps to satisfy the needs of α itself.
The strong feature required by the Argument Rule (10b) may be Case,
clearly a property of DP, hence its head D. Alternatively it is definiteness, also
clearly a property of D/DP. It follows from the assumptions made above that
strong Case in D requires that the language have either some kind of free Case
morpheme(s) which can be inserted in D, checking off the strong Case feature,
or bound Case morphology on N or some other nominal category within DP,
triggering movement of this category to D or SPEC-D. Correspondingly, strong
definiteness in D requires that the language have either free definiteness mor
phemes (articles) which can be inserted in D, checking off the strong definite
ness feature, or bound definiteness morphology on N or some other nominal cat-
8
This presupposes a lexicalist theory of inflectional morphology, according to which lexical
heads are inserted with (at least some) inflections already attached, as proposed by Chomsky
(1993).
104 ANDERS HOLMBERG & GÖREL SANDSTRÖM
(14)
9
See Pollock (1994) for a similar view of checking.
10
The converse does not hold: a language may have, for example, Case-morphology on N, and
yet have weak Case in D, checked only in LF. According to Holmberg (1994) Faroese is such a
language.
11
N-to-D movement may be blocked by an adjective intervening between D and N: see Delsing
(1993: 116-134), Holmberg (1993), Kester (1993), Sandström and Holmberg (1994),
Santelmann (1993). In that case many varieties of Swedish, including Standard Swedish, have a
free article inserted in D, effecting the check-off of the strong definiteness feature. (Not all of
these varieties allow a postnominal possessor.)
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 105
We conclude that the reason why Icelandic unlike MSc can do without an in
definite article in (8), or a free definite article in (9) is that Icelandic has strong
Case. The strong Case feature satisfies the Argument Rule at LF, and the case
morphology on the noun triggers movement of the noun to D, checking off the
strong Case feature before spell-out. MSc, on the other hand, has to rely on
strong definiteness coupled with free or bound definiteness morphology. For the
same reason (15a) below is ungrammatical in MSc, while the corresponding
form is grammatical in Icelandic.12 (15b), on the other hand, is grammatical in
MSc as well as in Icelandic, subject to dialectal variation.
Since MSc does not have strong Case, it has to rely on definiteness mor
phology triggering noun movement to D, eliminating the strong feature.
Therefore the noun in the postnominal possessor construction has to be definite,
as in (15b).
Insertion of a free definite article in (15a) would satisfy the Argument Rule,
but the result is nevertheless ungrammatical.
(16) * denbil(hans)
the car his
Note that (16) is ill-formed with or without the possessor: the definite form
of an unmodified DP can only be [DP bilen ]. We conjecture that this is an effect
Here the corresponding possessorless form [DP en bil ] is fine. The Argument
Rule and strong feature checking is (potentially) satisfied in (17), so the problem
seems to be the possessor. We return to (16) and (17) below, after having es
tablished the more precise structure of DP in Swedish. 15
Characteristic of the construction (2b), repeated here, is that there are con
straints on the categorial features of the possessor, subject to dialectal variation.
13
It is not obvious in which sense the free definite article is more specified than the bound
definite article, apart from containing one more phonetic segment. Tarald Taraldsen (p.c.) sug
gests that the free definite article may actually be bimorphemic, consisting of two parts:
[d [en]]. If so, it would indeed be structurally more complex and contain more feature specifica
tions than the bound article.
14
The construction with a postnominal PP (en bil åt läraren) is well formed as long as the PP
has a goal or benefactive reading: 'a car for the teacher'. In at least some dialects, including
standard varieties of Norwegian, it is ill formed if the PP is a possessor; see Taraldsen (1990).
See also footnotes 22 and 23, below.
15
An anonymous reviewer points out that formulation of the Argument Rule in terms of a
strong feature implies that there are languages which have a weak feature in D, given the usual
employment of the strong/weak dichotomy as a formalization of parametric variation. Insofar
as there are languages which make do without either articles or case-morphology, or any other
form of overt marking of argumenthood (Chinese is a possible candidate) this could be for
mally expressed as a weak feature in D. An interesting possibility is that predicative noun
phrases have a weak feature in D, perhaps universally.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 107
(19) pronoun < pronoun + name < bare name or family term < definite DP
16
The pronoun + name construction, as in (18b), is usually regarded as instantiating not a
pronoun with a proper name in some kind of appositional relation, but rather as a form of arti
cle (a 'preproprial article'), homonymous with a pronoun, with a proper name as complement:
see Taraldsen (1990), Sigurösson (1993), and especially Delsing (1993). This is clearly not
correct for all NSw dialects, as will be shown below, hence possibly not correct for any dialect.
17
Certain Icelandic dialects, too, allow a wider range of categories as possessors, according to
Sigurösson (1993). The more precise properties of the possessors in these dialects have not
been investigated.
18
Västerbotten and Norrbotten make up roughly the northernmost third of Sweden.
Österbotten is a region in Finland where Swedish is widely spoken, primarily along the coast,
across the Baltic Sea from Västerbotten.
108 ANDERS HOLMBERG & GÖREL SANDSTRÖM
19
We also have a few examples of postnominal family terms and other DPs constructed with a
pronoun. One informant consistently constructed all postnominal arguments, names, family
terms and other definite DPs, with a pronoun.
20
The geographical name within parentheses denotes the village or region where the informant
(or one of the informants) who produced the example in question grew up. We do not know, at
this stage in our research, to what extent the informant's judgments are representative of the di
alect spoken in this region. For all we know they may represent only the informant's idiolect.
We will still refer to these varieties as 'the dialect of Älvsbyn', etc., for convenience.
21
The postnominal pronoun+name construction was discussed briefly in footnote 15 above.
(20a) shows that the pronoun is not necessarily a preproprial article in this construction, since
in (20a) the pronoun co-occurs with a preproprial article (usually (e)n for masculine, a for femi
nine). Five different dialects in our data combine a pronoun and a preproprial article in this
way. The morpheme n in (20c) is the dative singular of the feminine preproprial article. For
some reason, the corresponding masculine () seems never to be used as a possessive marker.
22
An indefinite DP may occur embedded in a postnominal possessor PP in various dialects.
We are indeterminate as regards the status of postnominal possessor PPs. On the one hand,
Taraldsen (1990) has presented arguments which appear to show that possessor PPs behave just
like postnominal possessor DPs, which in our terms would mean that they are subject to the
same movement-and-checking requirements as possessor DPs (see the text below). On the other
hand it is tempting to view postnominal possessor PPs as an analogue of other well known
cases where insertion of a 'dummy preposition' is an alternative to Case-triggered DP move
ment.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 109
German allows pronouns, family terms, and proper names, but no other DPs
in prenominai position:
Russian, too, allows pronouns, family terms, and proper names, but no
other possessors in prenominai position:
(24) Piteå:
boken hans, boken mormors, boken Jannes
* boken lärarens, lärarens bok
book.D his, book.D grandmother's, book.D Janne's
book.D teacher.D's, teacher.D's book
(25) Älvsbyn:
boken min, boken pappas, boken hans Janne
* boken lärarens, boken åt läraren
book.D my, book.D dad's, book.D his Janne
book.D the.teacher's, book.D to teacher.D
24
In Italian, too, there is direct evidence of a prenominai possessor position between D and
NP, since the possessive pronoun co-occurs with an article: il mio libro, lit. "the my book" .
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 111
(26)
25
See, however, Sigurðsson (1993) for some observations regarding the interplay of noun
semantics and possessor constructions in Icelandic.
26
The fact that the prenominai possessor cannot co-occur with an article in German, indicates
that the possessor is in SPEC-DP (das Buch, Peters Buch, *das Peters Buch).
112 ANDERS HOLMBERG & GÖREL SANDSTRÖM
27
Delsing (1993: 173-175) argues that the N.D-POSS construction does not involve N-move-
ment to D, but instead would be derived by movement of a DP headed by the definite noun to
SPEC-DP, and movement of the possessor (he only considers constructions with a pronominal
possessor) to D. The structure would be
(i)
This solves some descriptive problems which beset the N-to-D hypothesis, but creates other
problems. For one thing, it is far from obvious how this theory will account for the cases
where the possessor is not a simple pronoun, hence plausibly a head, but a complex DP, as for
instance in (6b, c, d).
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 113
ently more definite than names, which are more definite than definite DPs, which
are (obviously) more definite than indefinite DPs.
6. Prenominai possessors
Let us assume an analysis along the lines of Fiva (1985) for Norwegian and
Abney (1987) for English: The genitive -s is base-generated as a functional head
which takes NP as complement. More specifically we propose that -s is base-
generated in AGRG. That is to say, -s is a phonetically visible instantiation of
AGRG. The genitive -s blocks movement of the definite noun to D, hence the
noun has the bare form (but see below for a counterexample). Hence the only
way the strong feature in D, required by the Argument Rule, can be checked off,
is by movement of -s to D, the possessor ending up in SPEC-DP, by movement
from inside NP. Assuming Greed, this presupposes that -s as well as the posses
sor have some feature or features which require checking in D and SPEC-DP, re
spectively. Assuming strong definiteness in D, -s must have a definiteness fea
ture. The affixation of -s would be a late phonetic process suffixing -s to the fi
nal word of the possessor DP; see Fiva (1985).28
In Standard Swedish possessive -s is not restricted in any way, but accepts
any kind of DP as specifier (unlike for instance those varieties of English where
genitive -s takes only [+animate] DPs as its specifier: the boy's book, *the beer's
colour). The reason why Icelandic does not have prenominai possessors is,
28
This analysis of (27) is consistent with the fact that a noun phrase containing a prenominai
possessor can function as a predicate, as in
(i) Per är Jannes son. (Swedish)
Per is Janne's son
As discussed by Holmberg (1993) and Mandelbaum (1994), this sentence can have a predicative
reading equivalent with, say, John is a student. According to Holmberg (1993) and
Mandelbaum (1994) the predicative reading presupposes that the possessor does not occupy the
highest SPEC-position in the nominal projection. That is to say, in this construction -s and the
possessor have not moved to D/SPEC-DP, but remain in AGRG/SPEC·AGRG. Not being an ar
gument, the construction is not subject to the Argument Rule.
114 ANDERS HOLMBERG & GÖREL SANDSTRÖM
quite simply, that Icelandic does not have any overt AGRG, i.e., no morpheme
corresponding to MSc and English genitive -s.
Consider again the dialect exemplified in (24): as long as the possessor was
a pronoun, proper name, or family term, the 'abstract AGRG strategy' could be
employed, with definite noun movement through AGRG to D, followed by
Possessor Shift to SPEC-AGRG. However, for any other possessor category this
option is closed, by hypothesis because abstract AGRG does not have the features
required to check DPs other than pronouns, names, and family terms. For such
DPs, the dialect employs the 'overt AGRG strategy', inserting genitive -s, an un
restricted DP-checking head, in AGRG. The possessor DP moves to the SPEC of
-s.
In Standard Swedish as well as in Danish and Norwegian the head noun
must have the bare form in the presence of a prenominai possessor, as in (27).
However, in many varieties of NSw the head noun can be definite. Thus (28) is
sharply ungrammatical in Standard Swedish, Danish and Norwegian, but per
fectly well formed in many Northern Swedish dialects.29
It seems to be the case in all, or nearly all, dialects where (28) occurs, that
the prenominai possessor cannot be a pronoun. This suggests that the construc
tion requires overt AGRG, i.e., genitive -s in AGRG. 30 Why this should be so, we
do not know. The construction is problematic in the framework articulated here.
In particular, how is checking of the definite noun ensured, without violation of
the Head Movement Constraint? We leave this construction for future research,
however.
This all implies that postnominal instances of genitive -s, as in (20b) or (24)
do not instantiate overt AGRG, but something closer to a genuine Case inflection,
presumably base-generated with the noun, checked in overt syntax. Note that in
several dialects, in fact most dialects in our data, postnominal lexical DPs have a
special morphology. Particularly common is the construction with a pronoun
plus lexical DP, but, as noted in passing, other possessive forms occur as well,
for instance (20c, d).
29
It is common also in Swedish dialects in South Finland.
30
Strikingly, one informant accepted a third person pronoun in prenominai position, but with
a 'double' genitive form, an -s added to the standard possessive form.
(i) hanses bilen
his's car
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 115
We can now conclude that the reason, or one of the reasons, why (17) is un
grammatical is that the possessor has not moved to SPEC AGRG to check its fea
tures. Then what about (29a, b)?
One property which Object Shift and Possessor Shift share is that they apply
only to pronouns in some varieties of Scandinavian, although not the same va
rieties. Thus, in MSc (including NSw) Object Shift is restricted to (weak) pro
nouns, while in Icelandic any definite object DP can shift:
The fact that only pronouns undergo Object Shift in MSc while all DPs do in
Icelandic has been explained as a consequence of the fact that only pronouns
have morphological case in MSc, while all DPs do in Icelandic (Holmberg &
Platzack 1995). The underlying assumption is that Object Shift only affects DPs
with overt Case. This hypothesis is confirmed by the Possessor Shift construc
tion, since the shifted possessor is either a pronoun, a DP headed by a pronoun,
or a DP with some form of overt possessive morphology, the only exception
being the apparently inflectionless possessor in the dialect represented in (20d).
At first sight, Icelandic seems to be a problem for this theory, though. The the
ory predicts that Icelandic should have Possessor Shift of a bare lexical DP,
since Icelandic has morphological Case, and strong Case, in the sense discussed
above. But as mentioned, Standard Icelandic allows only a pronoun or a pro
noun-plus-name as possessor in the N.D-POSS construction. This is not neces
sarily a problem, though. First, we have been informed by Halidór A.
Sigurðsson (p.c.)» that forms such as bókin Jóns, although not acceptable in
Standard Icelandic, are not infrequently encountered in spoken Icelandic.
Second, we conjecture that the fact that Icelandic has the N-POSS option for lexi
cal DP possessors means that the N.D-POSS construction with a lexical DP pos
sessor is blocked, as an effect of economy.31 This presupposes that Case is
cheaper than definiteness, and/or that (bound) case morphology is cheaper than
(bound) definiteness morphology (as discussed in section 3, what makes the N-
POSS construction work is strong Case coupled with bound case morphology).
At present we do not have any independent support for this hypothesis, except
the observation that bound case morphology is more common (perhaps even
31
There is a form of complementarity between the N-POSS and the N.DEF-POSS construction
in Standard Icelandic: N-POSS is used with lexical possessors but normally not pronouns, while
N.DEF-POSS is used with pronominal possessors only (including pronoun+name possessors);
see Sigurösson (1993). This suggests that, for some reason which we do not understand, the N-
POSS construction does not work when the possessor is a pronoun. In this case only, the
NDEF-POSS construction is allowed, in Icelandic.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 117
much more common) among the languages of the world than bound definiteness
morphology.
Another property shared by Object Shift and Possessor Shift is that they
both presuppose movement of the lexical head, i.e., they both conform to
'Holmberg's Generalization'. Consider (32) and (33): in (32), Object Shift has
applied without verb movement (there is no verb movement in embedded clauses
in MSc). The result is ungrammatical. In (33) Possessor Shift has applied with
out definite noun movement. The result is ungrammatical.
Above we proposed that the reason why the Possessor Shift requires noun
movement is that abstract AGRG must be included in a visible head chain to be
visible itself for grammatical processes, and thus able to check the features of a
possessor. This hypothesis can easily be extended to Object Shift, given the
analysis in (31): abstract AGRO needs to be included in a visible head chain to be
come visible, and thus able to check the features of an object.32 ,33
32
See Chomsky (1993) for a different account of why Object Shift requires verb movement.
Following Branigan (1992) Chomsky assumes that the reason why for instance (31) is ruled
out is that movement of the object across the subject position in SPEC.VP violates the
Shortest Movement condition.
(i) [.. Jegi [ A G R 0 P ham ... ikke [Vp e¡ [v- så e ]]]]
According to Chomsky (1993), the reason why verb movement helps to circumvent Shortest
Movement is that verb movement extends the domain of the verb, in a sense, so that
SPECAGROP and SPECVP count as equidistant from the object position, and therefore object
movement directly to SPECAGROP is technically as short as movement to SPECVP. It is not
obvious how to extend this explanation to Possessor Shift. We could postulate that SPEC-NP
of a definite noun contains an abstract category corresponding to the subject in SPECVP. This
category might be the abstract demonstrative pronoun postulated in Holmberg (1993). This
category would block possessor movement, unless the noun moves as well, making SPEC-NP
and SPEC-AGRGP equidistant from COMPN. However, as mentioned in the text, we do have
possessor movement to SPEC.AGRG, even across a definite noun (in some dialects), namely
when AGRG is overt.
33
Insofar as definite noun movement to D is obligatory, triggered by the needs of the definite
noun, (33) will be ungrammatical already because it contains an unmoved definite noun. In that
case it will not tell us anything about Holmberg's Generalization. Note, however, that in con
struction with an attributive adjective, the definite noun does not move to D (at least not
118 ANDERS HOLMBERG & GÖREL SANDSTRÖM
8. Conclusions
overtly). This is the case when a free definite article must be inserted: den nya boken 'the new
book.DEF'. Movement of a possessor to D is ruled out, as shown by (32), and as explained by
the theory presented here in terms of the visibility of AGRG.
SCANDINAVIAN POSSESSIVE CONSTRUCTIONS 119
We showed that the construction is, indeed, subject to rich dialectal varia
tion, especially in the Northern Swedish region, where, at first glance, it looks
as if every village has its own NP grammar. We showed that there is a pattern in
the variation though, the feature composition of AGRG being subject to a
(possibly) universal argument feature hierarchy. We also showed that the N.D-
POSS construction provides strong evidence of a functional head checking pos
sessors between D and NP. Finally we pointed out some striking similarities
between the N.D-POSS construction and Object Shift, another controversial
Scandinavian construction.34
REFERENCES
Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. PhD dis
sertation, MIT.
Branigan, Philip. 1992. Subjects and Complementizers. PhD dissertation, MIT.
Chomsky, Noam. 1991. "Some Notes on Economy of Derivation and
Representation". Principles and Parameters in Comparative Grammar, ed. by
Robert Freidin, 416-454. Cambridge, Mass.: MIT Press.
____. 1993. "A Minimalist Program for Linguistic Theory". The View from
Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. by
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge Mass.: MIT Press.
Comrie, Bernard. 1989. Language Universals and Linguistic Typology. Oxford:
Blackwell.
Delsing, Lars-Olof. 1993. The Internal Structure of Noun Phrases in the
Scandinavian Languages. Department of Scandinavian Languages, University
of Lund.
Fiva, Toril. 1985. "NP-internal Chains in Norwegian". Nordic Journal of
Linguistics 8.25-47.
Holmberg, Anders. 1993. 'On the Structure of Predicate NP". Studia Linguistica
47.126-138.
____. 1994. "Morphological Parameters in Syntax: the Case of Faroese".
Reports from the Department of General Linguistics, University of Umeå, Report
35.21-62.
34
We wish to thank our informants Mona Marklund, Ann-Louise Marklund, Carita
Lundmark, Ingegerd Jonsson, Majken Widmark, Pelle Wållberg, Peder Alex, Johan Svenlin,
Britta Lundgren, Björn Nilsson, Belinda Backström, Malena Mattbäck, Daniel Gustavsson,
Carin Agerhäll, Gertrud Wallin, Anders Persson, Göran Boman. Thanks also to Tarald
Taraldsen, Phil Branigan, the audiences at the Atlantic Provinces Linguistic Association
Conference on Microparametric Syntax in Saint John, N.B., October 1944, the Scandinavian
Syntax Workshop in Oslo, January 1995, and the seminar of the Phonetics and Linguistics
Department in Bergen, February 1995, and last but not least, two anonymous referees, for their
comments on previous versions of this paper.
120 ANDERS HOLMBERG & GÖREL SANDSTRÖM
___. & Christer Platzack. 1991. "On the Role of Inflection in Scandinavian
Syntax". Issues in Germanic Syntax, ed. by Werner Abraham, 93-118. Berlin:
Mouton de Gruyter.
____&____. 1995. The Role of Inflection in Scandinavian Syntax. New
York and Oxford: Oxford University Press.
Kester, Ellen-Petra. 1993. "The Inflectional Properties of Scandinavian
Adjectives". Studia Linguistica 47.139-153.
Källskog, Margareta. 1992. Attityd, interferens, genitivsyntax. Studier i Nutida
Överkalixmål. [Attitude, Interference, Genitive Syntax. Studies in the Contempo
rary Överkalix Dialect] Publications of the Institute of Dialect and Folklore
Research, Uppsala, Ser. A:18.
Mandelbaum, Deborah. 1994. Syntactic Conditions on Saturation. PhD disserta
tion, City University of New York.
Platzack, Christer. 1987. "The Scandinavian Languages and the Null-Subject
Parameter". Natural Language and Linguistic Theory 5.377-402.
Pollock, Jean-Yves. 1994. "Checking Theory and Bare Verbs". Paths Towards
Universal Grammar: Studies in Honor of Richard S. Kayne, ed. by G. Cinque, J.
Koster, J.-Y. Pollock, L. Rizzi & R. Zanuttini, 293-310. Washington, D.C.:
Georgetown University Press.
Rohrbacher, Bernard. 1994. The Germanic VO Languages and the Full Paradigm.
PhD dissertation, University of Massachusetts at Amherst.
Sandström, Görel & Anders Holmberg. 1994. "Adjective Incorporation and the
Syntax of the Scandinavian Noun Phrase". Reports from the Department of
General Linguistics, University of Umeå, Report 35.81-97.
Santelmann, Lynn. 1993. "The Distribution of Double Determiners in Swedish:
Den support in D°". Studia Linguistica 47.154-176.
Sigurösson, Halldór Á. 1993. "The Structure of the Icelandic NP". Studia
Linguistica 47.177-197.
Taraldsen, . Tarald. 1990. "D-projections and N-projections in Norwegian".
Grammar in Progress, ed. by Jóan Mascaro & Marina Nespor, 419-431.
Dordrecht: Foris.
Vikner, Sten. 1994. Verb Movement and Expletive Subjects in the Germanic
Languages. Oxford and New York: Oxford University Press.
THE OCCASIONAL ABSENCE OF ANAPHORIC
AGREEMENT IN LABRADOR INUTTUT*
ALANA JOHNS
Memorial University of Newfoundland
Introduction
Thanks to the organizers and participants of the Annual Meeting of the Atlantic Provinces
Linguistic Association held at the University of New Brunswick in Saint John in October
1994, where this paper was first presented at the special session on Microparametric Syntax
and Dialect Variation. Thanks also to the University of Toronto, where I worked on this paper
during my sabbatical leave, and to an anonymous reviewer for comments. This research was
made possible by funding provided by the Social Sciences and Humanities Research Council of
Canada (research grant number 410-94-570). Finally, thanks to all the speakers of
Qairnirmiutut and Labrador Inuttut whom I have consulted on this topic, especially Sally
Ikuutaq, Sybella Tuglavina and Harriet Lyall. All errors are mine.
122 ALANA JOHNS
It is beyond the scope of this paper to provide an explanation for the entire
phenomenon of anaphoric agreement in the Inuit language (see Bittner 1994;
Sadock 1994). Instead, the purpose is to show (a) that there is dialectal variation
1
The nature of this variance is not agreed upon by all generative grammarians. For instance
Chomsky (1995) seems to assume that all grammatical properties are found in all languages,
differences resulting from whether or not a particular feature of a lexical item is strong, giving
one set of properties or weak, giving another set of properties. Chomsky (1995:176) allows
that an agreement element may be "inert or perhaps missing". In my view, grammatical prop
erties form a universal set, but not all properties are observed in all languages. Variance results
from the set selected, and the interaction of this set with other sets.
2
That inflectional morphology reflects parameters does not mean that all parameters are mani
fested morphologically. For an example of parameterization within a language with little or no
morphology, see Cheng, Huang & Tang (1996) [this volume].
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 123
in the facts of anaphoric agreement in the Inuit language, and (b) that this varia
tion is predicted within the parametric approach developed in Johns (1993) and
Johns (1995). Before describing the facts pertaining to dialectal variation, it is
necessary to review some of the general facts concerning anaphoric agreement in
the language.
Anaphoric agreement in Inuktitut is an explicit indication by means of
agreement morphology on either a possessed NP or a verb that an argument is
referentially co-indexed with a higher one. This contrast is illustrated in
examples (1) and (2), from the Qairnirmiut dialect of the Inuit language
(Qairnirmiutut or Q), spoken around Baker Lake, Northwest Territories.3
3
Abbreviations used in this article are the following: s for singular; Ρ for plural; D for dual; R
for reflexive; INTR for intransitive; TR for transitive; REL for relative (ergative) case; (ABS) for
absolutive case (null); SECOND for secondary case; PART for participial mood; INDIC for indica
tive mood; OPT for optative mood; NEG for negative, and 1, 2, for first, second and third per
son. The notation 3S/1s indicates third person singular agent and first person singular
theme/patient.
124 ALANA JOHNS
In (3) and (4) we see examples from the Qairnirmiut dialect where the pos
sessed NP "his son" is in secondary case -nik, and the agent of the action "the
man" is in absolutive case. Only the latter agrees with the verb. As before, we
see that disjoint agreement between the possessor and the potential co-referent is
indicated by the possessive affix -a, while anaphoric agreement is indicated this
time by the suffix -mi.
The presence of the morpheme -nil-mi indicates obligatory co-reference be
tween an argument and its antecedent (see Finer 1985 for the seminal analysis of
switch reference phenomena under binding theory). In this property, anaphoric
agreement in the Inuit language resembles reflexive morphemes found other lan
guages, e.g., Scandinavian (Taraldsen 1996). Now consider examples of
anaphoric agreement on verbs.
4
English could have The man sees his own son, but the addition of own adds a marked inter
pretation that is absent from the example in (2).
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 125
In (5) and (6), we see that the verb in an adverbial clause may also display
anaphoric agreement, taking an argument of the matrix clause as antecedent. The
presence of third person agreement on the adverbial verb in (5), indicates that the
person who is happy is disjoint in reference (*i) to the agent of the kissing
(Mary), and must therefore have an independent index. Under this interpreta
tion, the one who is happy may be the patient of the kissing (John), or equally,
it may refer to someone other than the two NPs in the matrix clause.5 These
possibilities are indicated by the indices j or k
In example (6), however, the presence of the anaphoric agreement -mi on the
verb indicates obligatory co-indexing between the theme of this verb "to be
happy" and the agent of the kissing (Mary). No independent indexing is possi
ble, as indicated by *j,*k.
Examples (5) and (6) demonstrate two additional properties of anaphoric
agreement. The first is that co-indexing is not sufficient to license anaphoric
agreement. As stated above, one of the possible interpretations in example (5) is
that the one being happy is identical in reference to the one being kissed (John).
Even though these two arguments are co-indexed under this interpretation, the
morphological instantiation of co-indexing, i.e., anaphoric agreement, may not
appear. If anaphoric agreement were present in this construction, as in (6), it can
only indicate co-reference with the agent. In other words, anaphoric agreement
in an adverbial clause cannot take an absolutive NP as antecedent when a relative
case-marked NP agreeing with the verb exists. Nevertheless co-indexing of the
absolutive may take place without anaphoric agreement, and I will consider this
to be a separate process (see Grodzinsky & Reinhart 1993 for discussion and
references).
The second property that these examples demonstrate is that the third person
agreement is the unmarked (default) case. The fact that third person agreement
is found on the verb when the interpretation co-indexes the absolutive NP in (5)
shows that third person is not intrinsically marked for disjoint reference. In other
words, third person agreement is found either (a) when the licensing conditions
for possible anaphoric agreement are met (in which case, its presence signals
disjoint agreement), or (b) when the licensing conditions for possible anaphoric
agreement are not met (in which case, its presence is neutral as to co-reference).
In this respect, third person is pronominal, while anaphoric agreement is an
anaphor (Finer 1985).
When we compare the examples in (1) and (2) with parallel constructions in
the Labrador dialect of the Inuit language (referred to as Labrador Inuttut), we
note the following difference.6
In (7) we see that where the possessor of the noun innik "son" has a differ
ent index from the agent of the action angutik "the man," the possessive agree
ment is the third person -nga, cognate of the -a that we saw in the Qairnirmiut
example (1); however, when the possessor of the noun has the same index as
the agent of the action, the agreement is still third person, and not the expected
anaphoric agreement affix -ni that we saw in (2). In fact, the presence of such an
affix makes the Labrador example unintelligible, as shown by the ungrammati-
cality of (8).
6
In the Labrador Inuit Standardized Spelling System stands for q; e for ii; for uu; and â for
aa.
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 127
In (9) and (10), which are parallel to the Qairnirmiut examples in (3) and (4),
we see that Labrador Inuttut clearly can license anaphoric agreement in oblique
case-marked NPs. Similar to the previous examples in (5) and (6), it can also li
cense anaphoric agreement on a verb in an adverbial clause, as shown in (11)
and (12).
What the above examples indicate is that agreement in Labrador Inuttut can
not be anaphoric in a particular syntactic context—on an absolutive NP.
Obviously syntax is involved to some degree. The morpheme is not lacking
from the language, and the restriction must be described in terms of a specific
case position. In the following sections I will show that this fact in Labrador
Inuttut not only is predicted by the parameter for Labrador Inuttut proposed in
Johns (1993 & 1995), but is one of many (seemingly unrelated) facts that fall
together under this parameter.
7
This tree could be changed to have all heads on the left of the complements, thereby ensuring
that all movement is leftward in keeping with Kayne (1994a). As far as I know, nothing in the
analysis discussed here hinges on this.
128 ALANAJOHNS
What (13b) indicates is that the transitive clause in (13a) contains two inflec
tional projections. The lower projection AGRPN contains the verb in the form of a
participial complement, an inflectional head AGRN and a specifier which gets rela
tive case. Note that AGRN marks the person of the agent plus NUM. The latter is
the number category of the participial. Third person theme/patient (unlike first
and second person) is a property of NUM.8 Above this projection is another
phrase AGRPV. Within this phrase, the entire AGRPN is the complement. The head
is AGRV, which marks the person of the theme/patient. The specifier of AGRPV
gets absolutive case. The participial becomes a full predicate (verb) through
movement upwards to the inflectional heads. First it moves up to join AGRN, and
then the two of them move up to join AGRV. Although this layering of inflection
cannot be seen in (13a), it is clear in (14) where the theme/patient is first person.
Here we see that the inflection attached to the verb is in the order predicted
by the clause structure in (13b). The third person agent inflection is internal to
the first person theme/patient inflection, due to the structure layering in (13b).
8
The idea that NUM can carry third person properties is based on ideas cited and developed in
Ritter (1995). Further work, beyond the scope of this paper, needs to be done on the structure
of AGRN, investigating, for example, whether NUM needs its own projection.
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 129
In (16) we see that the participial moves up to AGRN, and subsequently the
participial+AGRPN moves to AGRV. In accordance with (15), the specifier (anguti-
up "the man") of AGRP N 'follows' its head by adjoining to AGRP V , a position in
which the Nominal Specifier Condition is met.
9
As pointed out by an anonymous reviewer, not all specifiers are subject to this restriction.
130 ALANA JOHNS
Not only participial mood morphemes, but indicative mood morphemes can
appear in this structure in Qairnirmiutut. The participial mood morphemes are
characterized by the presence of initial j/t, and the indicative mood morphemes
are characterized by the presence of initial v/p10, as in (17).
(18) a. taku-jara Q
see-TR-PART-3s/s
"the one who I see" or "I see him/her/it"
b. taku-vara Q
see-TRINDIC-3s/s
"I see him/her/it" (*"the one who I see")
10
The continuant forms /j/ and /v/ are found after stems ending in vowels, while the stop
forms /t/ and /p/ are found after stems ending in consonants.
11In fact the indicative mood is historically related to C(omp), involving properties including
narrative introduction, questions, etc. How it has become a mood of assertion is a question of
some interest.
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 131
As we saw in section 2.1, both the participial mood and the indicative mood
morphemes can appear in AGRV in Qairnirmiutut, this being obligatory in the case
of the indicative mood. Johns (1993 & 1995), shows that in Labrador Inuttut,12
there is a lexical restriction which does not allow the participial mood to move to
or project13 AGRV, stated in the form given in (20).
The lexical restriction in (20) effectively ensures that the only mood that will
appear in AGRV is the indicative mood in Labrador Inuttut. Although the restric
tion in (20) is a very simple and local restriction, it is equivalent to a parameter,
in that numerous dialect-particular facts of Labrador Inuttut are predicted by this
single property. Where it differs from some versions of parameters, e.g., Baker
(1996) is that it is not a property of the grammar, but of an individual lexical
item—the participial mood morpheme.
Briefly the effects of the Labrador Parameter are schematized in (21).
12 It should be noted that the Labrador region is not homogenous linguistically, and that the
parameter refers to only one speech group.
13 I leave aside an exact formulation of how a category projects, but refer the reader to recent
discussion along these lines in Grimshaw (to appear), Ghomeshi & Massam (1994), and
Chomsky (1995).
132 ALANA JOHNS
b. taku-vaga taku-jaga L
See-TR-INDIC.lS/3 S See-TR-PART- IS/3 S
"I see him" (surprise) "I see him"
As predicted by the Labrador Parameter, first and second person agent are
not subject to mood restrictions. Because their features are in AGRN, the predicate
can, but need not, move to AGRV via the indicative mood.
Qairnirmiutut, not having the Labrador Parameter, displays no mood restric
tions with first or second person theme/patient, as shown in (23).14
14
The example in (23a) was colourfully translated with the English expression 'Yikes!' by
one speaker of Q.
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 133
(24) a. * taku-nngit-paanga Q
See-NEG-TR-INDIC3S/lS
b. taku-nngi-taanga Q
See-NEG-TR-PART-3S/lS
"She/he/it doesn't see me"
In (24a) we see that the negative morpheme -nngit- cannot be followed by
the indicative mood. Instead, after the negative, we find the participial in
Qairnirmiut declarative clauses, as in (24b). This generalization applies irrespec
tive of person.
Turning to Labrador Inuttut, we encounter a problem, in that the generaliza
tion that the indicative mood cannot follow the negative morpheme will intersect
and overlap with the Labrador Parameter, which states that the indicative mood
is required to project a level of structure wherein a subset of person features can
be expressed. Where the person features of AGRV are not required, e.g., third
person theme/patient, as in (22b), we observe the non-projecting participial
mood following the negative, as in (25).
(25) taku-nngi-taga L
See-NEG-TR-PART-lS/3S
"I don't see him/her/it"
In contrast, where the extra level of structure is required, e.g., first and sec
ond person theme or patient, neither the participial mood nor the indicative mood
is permitted following a negative, as shown in (26).
15
Further research needs to be done on examples such as (24a). Johns (1995) describes them
as ungrammatical, but subsequent research has shown that the combination of nega-
tive+indicative is possible in highly marked situations (see also Lowe 1988). Suffice it to say
that they are not grammatical for the reading given in (24a).
134 ALANA JOHNS
The example in (26a) is ruled out by the Labrador Parameter, since the par
ticipial mood cannot project AGRV in order to accommodate first/second person
features. On the other hand, the indicative mood cannot 'rescue' the structure, as
shown in (26b), since the pan-arctic generalization restricts the appearance of an
indicative mood morpheme following a negative. As a result of this overlap,
Labrador Inuttut is forced to use another mood after the negative when first or
second person theme/patient is involved, as shown in (27).
In the preceding section we have seen that the Labrador Parameter described
in (20) accounts for a number of dialect-particular facts of Labrador Inuttut. In
this section, we will see that the Labrador Parameter is also responsible for the
lack of anaphoric agreement in absolutive case that was discussed in section 1.1.
Recall that in section 2, the adjunction of the NP in relative case to AGRPV
was triggered by the movement (or projection) of the predicate from head of
AGRN to head of AGRV position. In Labrador Inuttut, as we have just seen, the
predicate in the participial mood is never permitted to project or move to AGRV
(no matter what the person combination). This fact also predicts that specifier
adjunction cannot be triggered.
Consider the example in (28). Because the condition specified in (15) above
is not met (i.e., the head AGRN does not move), the specifier of AGRPN remains in
situ and does not move up to adjoin to AGRPV (where it would c-command the
specifier of AGRP V ). 17 As a result, the Labrador Parameter predicts that the abso-
16
Negative forms followed by -la- are not traditionally termed optative, but are considered to
be special negative forms (see Smith 1977). Here I use the term optative to suggest that the
-la- is related to the same form found in first and second person optative. This morpheme is
probably more accurately termed irrealis or subjunctive.
17
Alternatively, under a projection account, the AGRV projection is not there, and the absolu
tive is the subject of a nominal clause with no AGRV (see Johns 1987).
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 135
We see then that anaphoric agreement is not possible on absolutive case NPs,
exactly the configuration where the relative does not c-command the absolutive
in Labrador Inuttut. As mentioned above in section 1, we can assume that third
person agreement is pronominal, and can therefore be co-indexed through a pro
cess of co-reference (see Grodzinsky & Reinhart 1993), so that (29a) is ambigu
ous between a co-indexed and a non-co-indexed interpretation (as in English).18
18
As mentioned above, word order is not a fixed property in Inuktitut; however, for some, but
not all, speakers of Labrador Inuttut the word order for the examples such as that in (29a) is of
ten given as [NPABS NP R E L Predicate] or "OSV" in Labrador Inuttut, rather than the unmarked
order [NPREL NP ABS Predicate] or "SOV" observed in other dialects. Word order preferences are
subtle facts in a language with mostly free word order, and further research is needed on this
topic.
136 ALANA JOHNS
At this point we might expect that the relative NP, which remains Iη situ' in
the structure shown in (28) should be able to bear anaphoric agreement, with the
absolutive NP as antecedent. After all, we have just seen that the relative NP can
not move up to c-command the absolutive one. Why then are examples such as
(30) not found?
The fact that such sentences are also not possible leads us to conclude that
anaphoric agreement requires more than a particular structural configuration,
i.e., c-command. Following Finer (1985), Johns (1987) proposes that
anaphoric agreement also requires that there be a syntactic link between the po
tential anaphor and some agreement node of the antecedent. We can make this
notion of syntactic link more precise by stating that a syntactic domain must be
established between an antecedent and an anaphoric feature (i.e., an anaphoric
inflection) and mediated through some head. Syntactic links may be formed
between the antecedent and the potential anaphoric feature either by agreement or
by case assignment. In other words, anaphoric agreement involves more than
structure; a morphological chain must be constructed between the two elements.
Thus the conditions for anaphoric agreement can be characterized as in (31).
We can now see that the condition stated in (31) will explain the set of
agreement facts in Labrador Inuttut. Recall that the core constructions under
consideration are examples such as (29b = 8) and (30), where anaphoric agree
ment cannot hold between a relative case NP and an absolutive case NP. As stated
above, the Labrador Parameter will prevent the potential antecedent (angutik "the
man") in (29b) from adjoining to AGRPV where it could c-command the potential
anaphor (the possessive agreement on innik "son"). At the same time, the
Anaphoric Feature Condition in (31) will prevent the SPEC of AGRN from form
ing a link with an antecedent in SPEC-AGRV, thus ruling out the possibility that
the former could bear anaphoric agreement, as shown by the ungrammaticality
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 137
of (30). This link cannot be formed because the Labrador Parameter in (20) pre
cludes movement or projection of the participial mood with AGRV, thus resulting
either in two non-linked projections, under the movement analysis, or only one
projection (under the projection analysis).
Now let us examine the structures in Labrador Inuttut where anaphoric
agreement is possible. Being grammatical, the examples from (10) and (12)
above, repeated below in (32), must have a syntactic link which is lacking in the
examples (29b) and (30).
19
The oblique cases are -mik, -mut, -mit, -mi, -kkut, and -tut. Respectively, they indicate pa
tient, goal, source, location, path, and similarity.
138 ALANA JOHNS
ous third person -nga is found. Assuming that the Anaphoric Feature Condition
in (31) is correct, there must be some explanation for this fact. Either the an
tecedent does not c-command the possessed NP, or there is a link missing from
the syntactic chain through AGRV which links the NP in relative case to the pos
sessed NP in absolutive case. Recall that the Nominal Specifier Condition in (15)
stipulates that a specifier must follow a nominal head with which it has formed a
syntactic link. In addition, the Anaphoric Feature Condition requires that there
be some syntactic means by which the potential anaphoric agreement can link up
with the NP in relative case. The question now is whether or not there is
morphological evidence that the indicative mood in Labrador Inuttut does not
achieve this linkage, thereby accounting for the lack of anaphoric agreement. Is
there some principled factor that allows us to distinguish between Qairnirmiutut
and Labrador Inuttut in this respect? It turns out that Labrador Inuttut is distinct
from all other dialects of Inuktitut in that the indicative mood agreement does not
reflect a syntactic relation with the agent NP. Consider the Labrador data in (34).
(34) a. taku-vauk L
see-TR-INDIC-3/3s
"He/she, they (2), they (many) see him/her/it"
b. taku-vâgik L
See-TR-INDIC-3/3D
"He/she, they (2), they (many) see those two"
taku-vait L
see-TR-INDIC-3/3
"He/she, they (2), they (many) see them (many)"
As the examples in (34) show, the agreement markers in the Labrador in
dicative mood convey only the number of the theme/patient. In this regard, ex
amples (34a, b, and c) are singular, dual and plural respectively. These markers
do not convey number agreement with the agent NP in relative case. 20 What this
indicates is that the conditions of the Anaphoric Feature Condition in (31) are not
met; the possessive AGR on the NP in absolutive case cannot form a syntactic link
through AGRV to the NP in relative case. Anaphoric agreement cannot occur.
Compare similar examples in the indicative mood from a northern Quebec
dialect (NQ), shown in (35).
20
In fact the participial mood in Labrador Inuttut also lacks agent agreement. I leave it for fu
ture research to determine what rôle, if any, this has in the more general phenomenon of the
Labrador Parameter.
140 ALANA JOHNS
What (35) shows is that, even though the number of the agent is not ex
pressed by the inflection when the theme/patient number is more than one (e.g.,
35 d, e), the inflection reflects the number (singular, dual or plural) of the agent
when the theme/patient is singular. From this we can conclude that AGRV is
linked up to both the NP in relative case and the NP in absolutive case in this di
alect. Not surprisingly then, we find examples such as that in (36) taken from
Dorais (1988).
In this account then, the Anaphoric Feature Condition correlates the unique
ness of Labrador Inuttut having -vauk for 3s/3s and no third person agent
agreement in the transitive indicative paradigm with the lack of anaphoric agree
ment on possessed absolutive NPs. In all other dialects that can use the -v- as a
declarative mood, the 3s/3s form is -vaa . In addition, all other dialects inflect
the predicate for the number of the agent in relative case.21
In summary, we have seen that there are two properties whose existence
disallows anaphoric agreement in possessed absolutive NPs in Labrador Inuttut.
One property is the Labrador Parameter which prevents a participial mood from
moving to or projecting an AGRV node. As a consequence of this parameter, there
is no possibility of a link between the possessed NP in absolutive case and the
NP in relative case via AGRV. The second property is that the indicative mood
paradigms do not agree with third person agents. Without agreement, there is
again no possibility of the absolutive and the relative NPs forming a link that
would satisfy the Anaphoric Feature Condition.22
21
Certain issues remain to be explained, such as the role of number in this dialect difference,
i.e., is it a by-product or source of the difference? As well, the fact that the -vauk form can be
found in some other dialects in the transitive interrogative mood needs to be explained.
22
There may some common basis underlying this conspiracy (see footnote 20).
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 141
As we have seen, the absence of anaphoric agreement in some but not all
constructions in Labrador Inuttut is explained under an analysis whereby
anaphoric agreement must be both c-commanded by an antecedent, and syntacti
cally linked to the agreement of the antecedent (as proposed in Johns 1987).
Under this view, the basis of anaphoric agreement is overt morpheme to mor
pheme linking. In the case of the participial mood morpheme in Labrador
Inuttut, the requisite conditions are not met, since the Labrador Parameter pre
vents the participial mood morpheme from forming a link (via movement or
projection to AGRV) between the NPs in absolutive and relative case. In the case
of the indicative mood, the mood morpheme appears in AGRV, but lack of
agreement with the relative NP precludes a link being established with the abso
lutive NP.
As stated at the beginning of this paper, it is not the goal here to provide full
and detailed analysis of anaphoric agreement in general in Inuktitut.
Nevertheless, it is useful to review some of the facts and conclusions of this pa
per with previous analyses. For example, an analysis of anaphoric agreement
that is based purely on c-command of a 'syntacticized' argument structure, such
as that in Manning (1994) (see also Grimshaw 1990), will have difficulty ex
plaining why the 'transitive' form in Labrador Inuttut cannot have anaphoric
agreement, but the 'antipassive' can, as shown in examples (8) and (10), re
peated here in (37).
REFERENCES
Baker, Mark. 1996. The Polysynthesis Parameter. Oxford University Press.
Bittner, Maria. 1994. Case, Scope, and Binding. Dordrecht: Kluwer Academic.
___ & Ken Hale. 1996. "Ergativity: Towards a Theory of a Heterogeneous
Class". Linguistic Inquiry21,4.
Bok-Bennema, Reineke. 1991. Case and Agreement in Inuit. Studies in Generative
Grammar 38. Berlin & New York: Foris Publications.
Cheng, Lisa L.-S., C.-T. James Huang & C.-C. Jane Tang. 1996. "Negative
Particle Questions: A Dialectal Comparison". Microparametric Syntax and
Dialect Variation, ed. by James R. Black & Virginia Motapanyane, 41-78.
Amsterdam and Philadelphia: John Benjamins Ltd. [this volume]
Chomsky, Noam. 1991. "Some Notes on Economy of Derivation and
Representation." Principles and Parameters in Comparative Grammar,ed. by
Robert Freidin, 416-454. Cambridge: MIT Press.
___. 1993. "A Minimalist Program for Linguistic Theory." The View from
Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. by
Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge: MIT Press.
___. 1995. The Minimalist Program. Cambridge: MIT Press.
Dorais, Louis-Jacques. 1988. Tukilik: An Inuktitut Grammar for All. [Distributed
by Association Inuksiutiit Katimajiit Inc., Université Laval, Québec]
23
Argument structure may indeed play a role in anaphoric agreement. My point is only that it
cannot be the sole factor.
LABRADOR INUTTUT: ABSENCE OF ANAPHORIC AGREEMENT 143
1. Introduction
lr
This paper is the result of a complete collaboration between the authors; the order of names is
alphabetical.
2
Although HI constructions occur in both CF and QF, they stand out as a productive option
only in QF. Villiard (1984:18) notices the absence of this construction in CF to the extent that
some native speakers do not recognize it; however, most speakers accept it, and show a passive
knowledge.
3
from Villiard 1984:18
4
from Vinet 1985:411
146 FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
The parallel behavior in (1) to (3) led Villiard & Vinet 1983, Villiard 1984,
Vinet 1985 to an analysis in which HIs have the same underlying structure in
both CF and QF. Along these lines (but in the framework of the Minimalist
Program), this paper will argue in § 2-5 that the properties in (1) to (3) follow
from the interaction between the features of the Complementizer (C) and the
placement of the respective clause (Complementizer Phrase—henceforth CP) in
Topic. These two factors have an impact on the internal structure of HIs, which
qualify as 'absolute clauses'; that is, HI constructions behave like independent
(vs. subordinate) clauses and yield effects of finiteness, such as lexical subjects
in Nominative6.
In § 6, the paper will focus on facts of variation, which concern the referen
tial content of the null subject. More precisely, HIs present arbitrary null subjects
in CF, as in (Ic) above (see also Vinet 1985). On the other hand, the null subject
in QF may be co-referent with the subject of the matrix clause, as in (4a), disjoint
in reference, as in (4b), arbitrary, as in (4c), or may even occur with atmo
spheric verbs, as in (4d).
The contrast between CF (lc) and QF (4a-d) revolves therefore around the
interpretation of null subjects: QF presents some flexibility in recuperating the
referential content of the null subject; it also includes internal identification, as in
(4b). Several questions arise from these observations: if we consider that HIs
have the same underlying structure in CF and QF, how does the structure allow
for different conditions on null subjects? Do HIs in QF meet the criteria for pro-
drop finite clauses? What property of the language would justify a pro-drop op
tion in HIs?
The analysis proposed here argues that HI configurations meet the criteria for
finite clauses. Hence, licensing and identification of null subjects (pro) at the in
ternal level is possible if the language has opted for the Null Subject Parameter;
on the other hand, the conditions on pro identification at the internal level are in
sufficient in non-pro-drop languages, such as Standard CF. In this sense, the
paper claims that QF behaves like a pro- drop language, in contrast to Standard
CF, and to a higher degree than Colloquial CF (as discussed in Roberge & Vinet
1989, Auger 1994).
The sentences in (1), discussed in Villiard (1984) and Vinet (1985), show
that HI clauses surface only in Topic position. Further investigation of these
clauses will lead to the claim that HI clauses are base generated in Topic position,
not moved to this position. Tests confirming this analysis contrast HIs with non-
8
p.c. Françoise Mougeon, corpus Mougeon-Béniak
9
from Villiard 1984:32
10
from Dulong 1952:151
148 FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
(5) a. * C'est [gagner le gros lot] queje saurais quoi faire avec.
it is win-INF the big ticket that I know-coND what do-INF with
b. C'est [vivre dans cette ville]i que je ne veux plus ti.
it is live-iNF in this town that I not want any more
"It is living in this town that I don't want any more."
The contrastive behavior of the HIs and the fronted non-finite clauses which
originate in a position subcategorized by the verb indicates that only the latter
undergo a form of WH-movement. Fronting through WH-movement licenses
clefting (5b), parasitic gaps (6b), and disallows parallel WH-chains, as in (7b).
Since clauses do not move, we consider them base generated in Topic.
11
Positional tests are conducted in QF, since HIs have a higher productivity in this language
and native speakers have clear judgments.
HYPOTHETICAL INFINITI VES IN CONTINENTAL & QUEBEC FRENCH 149
Furthermore, the types of chains that obtain in (5) to (7) are also different.
We adopt the typology of A'-chains proposed in Cinque (1990), who distin
guishes between: (i) chains headed by a constituent base generated in an A'-po
sition; and (ii) chains headed by a constituent which moved to an A'-position.
The latter, but not the former, has quantificational properties. Along these lines,
HI clauses form non-quantified A'-chains, which include resumptive pronouns,
such as ce, ça "that" in (1c), (3b) for CF, and (8a) for QF. Note that resumptive
pronouns are optional in QF, as shown in (8b).
To sum up this section, the tests in (5) to (8) show that HI clauses are base
generated in Topic position, and head non-quantified A'-chains, which may in
clude resumptive pronouns.
Second, dislocation to Topic position within the HI, as in (10), shows that
the hierarchy of projections by-passes the IP level
150 FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
(11) Tu sais, ma Lauri, ben des fois, j'ai l'air d'un sans coeur,
mais [d'íe voir comme ça] tomber en gondole, j'braillerais!
you know my Lauri many of times I have the air of one
without heart but of you see-INF like that fall off a gondola I
bawl-coND
"You know, my Lauri, I often behave like a rough man, but
if I were to see you like that, falling apart, I would cry." 13
The sentences in (9) to (11) would receive a natural account if we consider that
HIs project to CP.
The definition of HIs as CP clauses would also explain their free alternation
with si-conditionals, which project to CP. The alternation is illustrated in (12):
The contrast between Italian (13a) and French (13b) follows from the fact
that the infinitive verb intervenes between se andpro in overt syntax, canceUng
the illicit government relation in Italian, whereas the same type of V-to-I move
ment does not apply in clitic-infinitive languages like French. Si is incompatible
with non-finite clauses in French, since it interferes with control on the non-lexi
cal subject15. Most HIs display control, in free alternation with lexical subjects;
therefore, it is expected that si be excluded from these constructions.
The analysis sketched above confirms that a C-head with [hypothetical] fea
tures dominates both finite and non-finite conditional clauses; the former exhibit
si, the latter present an empty head, which move from V-to-l-to-C in LF. Since
the clause with the latter configuration occurs in Topic, the hypothetical features
of must interact with the non-finiteness of the embedded inflectional phrase
(IP), and lead to a configuration which changes the into a finite clause.
15
In this paper we adopt the hypothesis initially proposed in Borer (1989) that all the prono
minal empty categories correspond to pro; the licensing conditions on pro differ in finite and
non-finite clauses.
152 FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
16
Tests such as (14b) have led Vinet (1985) to the conclusion that HIs project to IP, and lack
the necessary CP host.
HYPOTHETICAL INFINITIVES IN CONTINENTAL & QUÉBEC FRENCH 153
The observations in (14) and (15) show that the hypothetical features on
must imply [+qu] features. Hence the prediction that (i) C enters a local relation
with SPECCP, and marks it with [+qu] features; and (ii) lack of si triggers verb
movement to C, necessary to fulfill the checking on the hypothetical feature. The
mapping of these two operations depends on the properties of the embedded IP.
Vinet attributes the ungrammaticality of (17) to the clash between the unreal
ized tense in the infinitive and the past tense of the matrix clause; only a condi
tional verb would render (17) comprehensible.
A closer look will reveal, however, that the clash does not concern the tense
value in the matrix, as long as those tenses are included in conditional
paradigms:
The continuous past (18a) or the present tense (18b) can appear in the ma
trix, on a par with conditional forms, and the hypothetical interpretation of the
infinitive is maintained.
The tense and mood variations in (18) cast doubts on the realization of
anaphoric binding between the tense in the matrix and in the HI clause: the latter
does not copy, feature-by-feature, the mood/temporal specifications in the ma
trix. Moreover, anaphoric binding should be excluded for technical reasons: HIs
occupy dislocated positions, and they head chains on an optional basis—for ex
ample, the sentences in (18) do not exhibit chain formation. Lack of chains situ
ates the respective clauses outside the binding domain of the matrix clause: in
this configuration, one cannot suppose that binding involves some internal
empty category and becomes a property of the chain. An account of the realiza
tion of mood and tense values is therefore crucial for the understanding of the
internal structure of HI clauses.
HYPOTHETICAL INFINITIVES IN CONTINENTAL & QUÉBEC FRENCH 155
Kayne makes the observation that infinitive inflections cannot appear in root
clauses, unless they are licensed in a way which ensures them the required
finiteness. The exact mechanism proposed in this analysis is triggered by the
presence of the negation: the negative marker non licenses an empty modal ele
ment, specified for positive tense value, as part of the inflectional chain; the in
finitive raises to the level of the empty modal head and acquires the tense fea
tures that enable it to appear in a root context. Arguments toward this analysis
come from Italian dialects where there is a lexical counterpart to the modal ele
ment, i.e. stà in Paduan (19b)—the element stà is restricted to negative impera
tives, as shown by the contrast with (19c):
ture of C and consider it to be the licensing factor for positive temporal features
on L
We mustfirstobserve that hypothetical clauses (finite or non-finite) display a
variety of inflectional forms. Thus, indicatives, conditionals and infinitives al
ternate as grammatical tools for conveying the hypothetical reading: the resulting
reading depends, as already mentioned, on the presence of si or on the availabil
ity of I-to-C movement, in a specific context (i.e.,Topic-position in QF).
In hypothetical infinitives, the implication of temporality in the [hypothetical]
feature appears in the near obligatory conditional forms in the matrix clauses:
The transfer of modal and temporal values of the matrix verb is not restricted
to conditional matrix verbs. As the indicative may also express the hypothetical
feature, we expect to find hypothetical infinitives when the matrix verb is indica
tive, as shown in (5b) and (22):
However, a matrix indicative verb, without the hypothetical value, turns the
sentence ungrammatical, as shown in (23):
Similarly, a conditional matrix verb which does not have the hypothetical inter
pretation rules out the sentence:
Note that in (24) the ungrammaticality is not due to the fact that the condi
tional verb is in a subordinate clause since sentence (25), with the tensed hypo
thetical conditional in a subordinate clause, is grammatical:
The facts in (20) to (25) show that a transfer of temporal and modal values
takes place in hypothetical infinitive constructions, restricting the choice of in
flectional forms and the past/present interpretation. We relate this restriction to
158 FRANCE MARTINEAU & VIRGINIA MOT AP ANY ANE
(26)
The examples in (27) occur in QF, and seem to contradict our analysis,
wherein has both [+hypothetical] and [+qu] features: SPEC.CP in these exam
ples displays the element de usually compatible with [-qu] complementizers.
We will argue that the examples in (27) represent a second type of HI con
struction, in which hypothetical has [-qu] features. We follow the analysis in
Kayne (1991), where the complementizer de receives the definition of a
Specifier. Placement of de in SPEC.CP blocks WH-movement to this position on a
par with the presence of an empty Operator, as suggested in (26). On the other
hand, as a non-operator, de does not have the capacity to license abstract
modals, which are crucial for the status of subjects in HI clauses.
The following observations will indicate that a de-HI does not meet the crite
ria of an 'absolute clause'; thus, de-HIs must be typologically and functionally
different from HIs with empty CPs. First, clitic chain formation is obligatory in
de-Hls, as in (28a); the chain is represented in (28b).
The empty category falls in the binding domain of the matrix IP. Binding and
control become properties of the chain, and are implemented through
'connectivity'. Note that HIs with empty CPs do not enter the chain in (28b),
unless optionally; therefore, they are not subject to connectivity.
Further examples in (28c,d) confirm that lack of resumptive pronouns rules
outde-HIs;the sentences become grammatical if we delete de.
17
from Villiard 1984: 32
18
from Villiard 1984: 32
160 FRANCE MARTTNEAU & VIRGINIA MOTAPANYANE
The facts in (27) to (30) indicate that a de-HI behaves like a dislocated clause
which is chain related to a position in the matrix clause. The chain mediates the
processes of binding and control, and yields the 'connectivity' effects proper to
adjunct clauses. Therefore,de-HIs do not meet the criteria of 'absolute clauses',
which are finite, quasi-independent clauses.
To sum up sections 4 and 5, HI clauses project to CP, and are equivalent to
si-conditional clauses. The C-head carries the hypothetical feature, which allows
for the interpretation of HIs irrespective of the mood specifications in the matrix
(the latter may be conditional or indicative). The hypothetical feature pairs with
[+/-qu] features: [-qu] features allow for insertion of lexical elements in CPs
(i.e., de) and confer on HIs the properties of adjunct clauses; on the other hand,
[+qu] features license an Operator in SPECCP, which, in turn, licenses an ab
stract modal; the abstract modal confers finite features on HIs , and allows them
to function as 'absolute clauses'.
Sections 1 to 5 have argued that HI clauses have the same underlying struc
ture in CF and QF. Nevertheless, variation arises within this common structure
with respect to the licensing of null subjects. Vinet (1985) shows that null sub
jects receive an obligatory arbitrary interpretation in CF, as in (31a); arbitrary
null subjects alternate with lexical subjects, as in (31b):
19
from Villiard 1984:38. Villiard mentions that judgments vary for the sentences (32): some
speakers accept them very readily, others less so.
162 FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
The properties in (i) to (iii) indicate that HI clauses behave like finite clauses
in pro-drop languages. This behaviour is compatible with the hypothesis put
forth in Roberge & Vinet (1989) and Auger (1994, 1995a,b) that QF presents the
characteristics of a pro-drop language in finite clauses.
Roberge & Vinet (1989) and more recently Auger (1994, 1995a, b) have
proposed an analysis of QF as a pro-drop language22. Roberge & Vinet suggest
that languages where subject clitics must be expressed, as it is generally the case
in QF, are pro-drop. They differ from languages such as Italian, which allow the
absence of subject clitics, in the way pro is licensed in subject position. In
Italian, pro receives its semantic content from the verbal ending; in French, pro
is licensed by the subject clitic, as in (36):
20
Martel 1971: 602 [Montréal 25h 18:23/source: projet de l'Estrie]
21
Martel 1971: 658 [Montréal 26f 4:13/source: projet de l'Estrie]
22
As shown further in this paper, the pro-drop character of QF is best argued for on the basis
of subject clitic doubling, a phenomenon which also occurs in Colloquial CF. Since Standard
CF does not exhibit subject clitic doubling, we prefer not to extend the pro-drop analysis to
this variety of French, without further evidence.
HYPOTHETICAL INFINITIVES IN CONTINENTAL & QUÉBEC FRENCH 163
For Auger, who adopts a similar analysis, the subject clitic, as pronominal
affix, is not an argument which saturates the subcategorizing frame of the verb.
Thus, an overt lexical subject may co-occur with the subject clitic, as in (37):
Roberge & Vinet (1989) also argue for the presence of pro in expletive ex
pressions, where the lexical expletive subject pronoun has been deleted, as in QF
sentences in (39)24.
23
from Auger 1995a, 108-71:036
24
Roberge & Vinet (1989) leave aside cases where the lexical expletive is obligatory in em
bedded clauses (ib), although its deletion is possible in independent clauses (ia):
(i) a. Pleut un peu.
rains a little/ "It rains a little."
b. * Je sais bien que pleut un peu.
I know well that rains a little
For them, (ia) is the result of a possible phonological deletion.
25
from Roberge & Vinet 1989:110-11
164FRANCEMARTINEAU&VIRGINIAMOTAPANYANE
They propose that a syntactic rule may delete an expletive subject when an
object marker (me in (39)) or another element which could licence the expletive
subject is present. Explicitly, they propose the following principle:
(42)
According to (42), the abstract modal shares a positive value with the com
plex [AGR+T], as opposed to Τ only. Further effects on the subject position de
pend on the intrinsic features on Τ—that is, whether Τ has strong or weak [N]
features in the language.
As mentioned for (41a), Τ has weak [N] features in Standard CF, and re
quires the presence of a lexical D, such as iL The function of il, which is to
complement and transfer [AGR+T] features to pro, must be implemented in overt
syntax: il lacks lexical content and becomes invisible at LF. Therefore, if a DP
with the D-head il is inserted in SPEC-MP in (42), the structure would be ruled
out. Verb movement to the modal and further to takes place at LF in clauses,
but il is not visible to LF rules, and the assignment of substantive features to the
166 FRANCE MARTINEAU &VIRGINIAMOTAPANYANE
null pronominal cannot be implemented. Thus, the analysis in Corver & Delfitto
(1993), applied to clauses in French, accounts for non-occurrence of clitic
pronouns in these constructions (35b), although lexical subjects and strong pro
nouns are allowed (35a).
The same analysis indicates, on the other hand, that licensing and identifica
tion of null pronorninals in (4), (32-35) must take place according to the pattern
in (41b). In this sense, the QF versions of HI clauses exhibit the properties of
pro-drop languages such as Spanish, and exclude subject clitics. Indeed, the
grammaticality of those examples can be accounted for only if we assume that
(41b) is the configuration for the null pronominal, and it is inserted in the subject
position in (42). The D-head is non-lexical, so it pairs with a T-head with strong
[N] features. The complex [AGR+T], with strong [N] features, raises to the ab
stract modal, which confers finiteness to it; thus, the complex [AGR+T] transfers
the required features to D, which in turn transfer these features to the pronominal
in SPEC-DP. This process is licit at LF.
To conclude this section, HI clauses in CF present the properties of Standard
non-pro-drop Continental French, and thus contrast with Hypothetical
Infinitives in Québec French with respect to the licensing of null subjects: (41a)
applies in Standard CF, (41b) applies in QF.
REFERENCES
Auger, Julie. 1994. Pronominal Clitics in Quebec Colloquial French: A
Morphological Analysis. Ph.D dissertation, University of Pennsylvania.
__. 1995a. "A Morphological Analysis of Quebec Colloquial French
Pronominal Clitics". CLS 31-II: Papers from the Parasession on Clitics, ed. by
A. Dainora et al., 32-49. Chicago: Chicago Linguistic Society.
___. 1995b. "Les clitiques pronominaux en français parlé informel: une ap
proche morphologique". Revue québécoise de linguistique 24,1.21-60.
Borer, Hagit. 1989. "Anaphoric AGR". The Null Subject Parameter, ed. by O.
Jaeggli & K. Safir, 69-109. Dordrecht: Kluwer Academic Publishers.
Chomsky, Noam. 1993. "A Minimalist Program for Linguistic Theory". The
View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger,
ed. by Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge: MIT Press.
Cinque, Guglielmo. 1990. Types of A'-Dependencies. Cambridge: MIT Press.
Corver, Norbert & Denis Delfitto. 1993. Feature Asymmetry and the Nature of
Pronoun Movement. University of Utrecht: OTS Working Papers [OTS-WP-TL-
93-013].
Dulong, Gaston. 1952. "La langue franco-canadienne". Pédagogie Orientation
6.148-155.
Junker, Marie-Odile & France Martineau. 1992. "The Structure of Infinitives".
Probus 4,3.1-27.
Kayne, Richard. 1991. "Romance Clitics, Verb Movement and PRO". Linguistic
Inquiry 22.647-686.
___. 1992. "Italian Negative Infinitival Imperatives and Clitic Climbing".
Hommage à Nicolas Ruwet, ed. by L. Tasmowski & A. Zribi-Hertz, 300-312.
Ghent: Communication and Cognition.
Martel, Pierre. 1971. Corpus de VEstrie (includes Corpus de l'Estrie, de Québec,
de Montréal, de Saguenay Lac Saint-Jean). Sherbrooke, Québec: Université de
Sherbrooke.
Martineau, France. 1990. La montée du clitique en moyen français: une étude de
la syntaxe des constructions infinitives. PhD dissertation, Université d'Ottawa.
Pollock, Jean-Yves. 1989. "Verb Movement, Universal Grammar and the Structure
of IP". Linguistic Inquiry 20.365-424.
Rizzi, Luigi. 1982. Issues in Italian Syntax. Foris: Dordrecht.
__. 1986. "On the Status of Subject Clitics in Romance". Studies in
Romance Linguistics, ed. by O. Jaeggli & Silva-Corvalan, 391-420.
Dordrecht: Foris.
___. 1990. Relativized Minimality. Cambridge: MIT Press.
Roberge, Yves & Marie-Thérèse Vinet. 1989. La Variation dialectale en gram
maire universelle. Montréal: Presses de l'Université de Montréal.
Terzi, Arhonto. 1992. PRO in Finite Clauses. A Study of the Inflectional Heads of
the Balkan Languages. PhD dissertation, City University of New York.
Thibault, Pierrette. 1984. Corpus Montréal 84: Projet de recherche sur le français
parlé à Montréal, Université de Montréal. See also Thibault, Pierrette & D.
168 FRANCE MARTINEAU & VIRGINIA MOTAPANYANE
Abstract
that there is more variation amongst English dialects at the morphological and
syntactic levels than has previously been supposed
1. Introduction
The following study is based in part upon data from my Grammar of the
traditional urban vernacular of the Farnworth area (1981), the fieldwork for
which was carried out in the 1970s (especially 1972-1974), as well as on further
data from fieldwork that took place in the 1980s within the Bolton Metropolitan
Area.1 Bolton itself is situated circa 12 miles north-west of Manchester. It is part
of a very densely populated area in what was formerly south-east Lancashire
(the South-East Lancashire Conurbation),2 and there is currently a wide variety
of industries there. In the past, however, the traditional industries were textiles
and mining. A consideration of historical, geographical, social, economic, ad
ministrative, and cultural factors shows the Bolton Metropolitan Area to be a pe
culiarly homogeneous unit under the urban field of influence of Bolton (see
Shorrocks 1981:1-27, § 0, especially pp. 15-21, § 0.5; Freeman, Rogers &
Kinvig 1966:218-219; Saxelby 1971:114-115; and Smith 1969:73). This unity
is reflected in the speech of the area, which is relatively homogeneous in charac
ter, and distinctive vis-à-vis that of other areas (see further Shorrocks 1981:21-
27,75-77, and 681). It is important to realize that the major cotton towns around
Manchester are not in any sense mere suburban satellites of Manchester, but that
each constitutes a relatively independent entity. Bolton is both the largest, and
least dependent on Manchester (see Freeman, Rogers & Kinvig 1966:218-219,
and 222). The data for the study are chiefly derived from tape recordings of
'free', i.e., unscripted, conversation made with informants representative of the
traditional vernacular of the area.
1
1 am grateful to the Social Sciences and Humanities Research Council of Canada for Standard
Research Grant #410-92-1137, which has enabled me to continue investigating these second
person singular forms as part of a wider study of the dialects of the North-West of England.
2
In 1972 the Local Government Act was passed, based upon the recommendations of the
Maud Commission, and the county boundaries were redrawn in 1974 in accordance therewith.
Bolton and the surrounding area then became a part of the newly-created Greater Manchester
County. For details of the history, geography, and socio-economic make-up of Greater
Manchester County, see Frangopulo (1977).
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 171
3
In dialect literature, of which there has been and still is a great deal in this area, the second
person singular pronoun is spelled theaw or theau in the subjective case, and thee in the objec
tive case; the plural pronoun is spelled yo. Theaw is always singular; but the singular-plural
distinction is complicated by traces of yo as a polite singular, and also by the fact that un
stressed yo and the unstressed form of the standard pronoun you—which has both singular and
plural reference—have the same pronunciation in the Bolton Metropolitan Area, viz., /jө/.
The reader should note that [Θ] does not have the value currently assigned to it in the latest re
vision of the alphabet of the International Phonetic Association (see Journal of the
International Phonetic Association 23/1 (1993):32-34 and unpaginated centre-page chart), but
that it rather represents a vowel that is typically rounder, fronter, and tenser than the /ə/ of
British Received Pronunciation.
172 GRAHAM SHORROCKS
The first verb, CON or CAN, is presented in greater detail. The presentation
of the remaining verbs follows the same basic pattern. Variants within round
brackets are much less frequent in occurrence.
CON/CAN "can". From a historical point of view, the dialect under discussion
is of a northern west-Midland type, so that /o/ often appears before a nasal
where other dialects of English have /a/. Even within residual traditional ver
nacular, both vowels are encountered in this lexeme. The declarative forms are
included for purposes of comparison, viz., non-past affirmative declarative: /ðε:
kon, kan/ "thou can". Non-past affirmative interrogative forms incorporating the
pronoun: /kontto, kont; 'kantte, (kant)/ "can thou?".4 The former of each inter
rogative pair here is the older—a fact borne out by observation, dialect literature,
and informants' judgments alike. The even-more-contracted form /kont/ occurs
morefrequentlythan its more residual counterpartkɔonttө/.The general import
of these remarks will apply with equal force to the other verbs, below. Non-past
negative declarative: /ðε: kɔænt, komt,(kæmt)/"thou can't". Non-past negative
interrogative: /'kɔtnt, 'katnt, (kɔənt,koint,kaeint)/"can thou not?/can't thou?".
The first two forms, which are both traditional and more frequent, are presum
ably to be derived from modal + cliticized pronoun + cliticized negative particle.
This derivation would entail a hypothetical interstage */kɔntnt, 'kantnt/, phoneti
cally "can thou not?". Final consonant clusters tend to be
glottalized, and medial clusters to be geminated or glottalized, in this dialect (see
further Shorrocks 1988). Subsequent simplification of the cluster has resulted in
the loss of the final consonant of the base.5 Preterite affirmative declarative: /ðε:
kuð/ "thou could". Preterite affirmative interrogative:
4
/tt/ represents a geminated consonant. For reasons of space, I will not debate here the appro
priateness or inappropriateness of this phonemicization, though I have suggested under could
thou? that a contrast is at stake in that particular case.
5
Trudgill (1990:87) mentions a number of interrogative forms from areas considerably to the
south of the one under discussion here:
" Traditional working-class Bristol dialect [...] has cassr't? = "can't you?" from canst
thee not?, and dissn't? = "didn't you?" from didst thee not? The Potteries area of
northern Staffordshire has ast? = "have you?" from hast thou?, and thee coost = "you
could" from couldest,"
The form cassn't? also shows loss of/n/. Trudgill (87) additionally cites asta? "have you?",
dusta? "do you?", and tha cossna "you can't" from western Derbyshire after a folksong in
Scollins & Titford (n.d.). However, he does not attempt any systematic presentation or analy
sis.
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 173
lar development for "should thou not?", and presupposes an earlier */'∫udtnt/.
/'∫utntө/ might be due to informant error—possibly somewhat analogical in na
ture, or a different derivational route: /'∫udnt/ "shouldn't" + cliticized pronoun +
devoicing of/d/ under the influence of heavy glottalization of the medial conso
nant cluster. There is too little evidence to resolve this matter here.
WILL Non-past affirmative declarative: /ðε: wil/ "thou will". Non-past affirma
tive interrogative (cf. forms with a stressed pronoun: /wíl ðε/ "will thou?"; /wilt
dei/ "wilt thou?" also occurs): /"wilttө, wil(t)t/ "will/wilt thou?". Non-past nega
tive declarative: /δε: weint, weint/ "thou won't". Non-past negative interroga
tive: /'wөtnt, 'witnt/ "will/wilt thou not?". It is probable that the lateral (like the
nasal of CON/CAN and MUN, above) has been lost through simplification of the
glottalized consonant cluster. The vowel /ө/ is explicable in terms of rounding
and lowering of /I/ in the historical environment /w-(l)/. Preterite affirmative
declarative: /δει wud/ "thou would". Preterite affirmative interrogative: /'wutte,
'wudtө, wu(t)t, (wudt)/ "would thou?". Preterite negative declarative: /δει
'wudnt, wunt, ('wudn)/ "thou wouldn't". Preterite negative interrogative:
/'wutnt, 'wudnt/ "would thou not?".
BE Non-past affirmative declarative: ðε: æ(r)t, at/ "thou art". Non-past affir
mative interrogative: /attө, a(t)t/ "art thou?". In allegro speech we also meet
such highly-contracted sequences as /wiə(r)t/ "where art thou ...?"; /wot 'le
):(r)nin/ "what art thou learning?"; /eit 'fiilin/ "how art thou feeling?". Non-past
negative declarative: /ðε: æ(r)nt/ "thou aren't". Non-past negative interrogative:
/*atnt, æi(r)nt/ "art thou not?, aren't thou?". Preterite affirmative declarative: /δει
wθ(r), wθ(r)/ "thou were".6 Preterite affirmative interrogative: /"wθttθ, wθ(t)t/
"were thou?". Preterite negative declarative: /ðε: wө:(r)nt/ "thou weren't".
Preterite negative interrogative: /wө(R)nttө, 'wetnt, wθ(r)nt/. The second vari
ant has the morphological derivation "were thou not?", and the first and third
"weren't thou?". Whilst the former is the predominant pattern in the dialect, the
latter is probably more frequent in this particular tense of this particular verb—a
state of affairs that may well relate to the fact that /'wetnt/ is homophonous with
the commoner variant of "will thou not?" The homophony is unsurprising, as
historical /r/ is optionally retained after /ө/, but has been lost without exception
in the environment /θ/ + fortis consonant.
HAVE Non-past affirmative declarative: /ðε: az/ "thou has". Non-past affirma
tive interrogative: /aste, ast/ "has thou?", with the latter often /as/ before a con
sonant. Though the third form is highly contracted, the pronoun remains de
tectable in the fortis /s/ resulting from an earlier assimilation. In allegro speech
we meet such highly-contracted forms as /wots bakt/ "what has thou backed?"
Preterite forms in this dialect have been regularized to were throughout the paradigm.
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 175
[horse racing]; /wɔəss dun/ "what has thou done?". Non-past negative declara
tive: /ðει 'aznt/ "thou hasn't". Non-past negative interrogative: /as(ə)nt/ "has
thou not". Again, the pronoun is perceptible in the devoicing of a lenis conso
nant, the /z/ of /az/ "has". Preterite affirmative declarative: /δει ad/ "thou had".
Preterite affirmative interrogative: /'atto, 'adtө, a(t)t, (adt)/ "had thou?". Note
how the first and third variants are homophonous with the forms of the non-past
affirmative interrogative of BE! Preterite negative declarative: /ðει 'adnt/ "thou
hadn't". Preterite negative interrogative: /'adnt, 'atnt/. Note how the latter, "had
thou not?", though the regular development, is homophonous with the chief
variant of the non-past negative interrogative of BE. This fact may explain why it
seems to occur less frequently than /adnt/ "hadn't thou?/had thou not?".
DARE This verb is sometimes a modal auxiliary. It is a generally difficult verb,
at least from the standpoint of the analyst, with preterite and non-past forms not
clearly distinguished. I have recorded: /dust, dest/ "does thou dare?" The loss of
historical /r/ in these particular phonological environments has resulted in a form
/dust/ that is homophonous with a variant of the non-past affirmative interroga
tive of DO.
DO Non-past affirmative declarative: /δει duz/ "thou does". Non-past affirma
tive interrogative: /dustө, dust/ "does thou?". The final /t/ of the latter is quite
often elided before a consonant, though again the pronoun is still perceptible in
the devoicing of the fricative that has resulted from an earlier assimilation. In al
legro speech we meet such highly-contracted forms as /wots mi:n/ "what does
thou mean?". Non-past negative declarative: /ðε: 'duznt/ "thou doesn't". Non-
past negative interrogative: /'dus(ө)nt/ "does thou not?". Preterite affirmative
declarative: /δει did/ "thou did". Preterite affirmative interrogative: /dittө, dI(t)t,
(didt)/ "did thou?". Preterite negative declarative: /ðει 'didnt/ "thou didn't".
Preterite negative interrogative: ditnt, 'didnt/ "did thou not?, (didn't thou?)".
Can t I have one?1 The same preference carries over into 'modified' speech that
is in many, or even most, other respects relatively standardized. Thus, con
tracted negative interrogative forms are mostly found in tags. (ii) Only the
anomalous finite BE has these second person interrogative enclitic forms when
functioning as a full (lexical) verb; otherwise the forms are restricted to auxil
iaries. HAVE is followed by getten "got" (past participle of get) when possession
is indicated affirmatively; and in negative questions the dialect uses HAVE
(usually + getten) + noanlnone rather than the usual English English standard
form HAVE + contracted negative + any. (iii) In tags after negative propositions,
affirmative contracted forms do not necessarily occur in this dialect. This is so
important a contrast with other varieties of English that it warrants further com
ment.8
Lester (1971:161) is representative of grammars of standard varieties of
English generally in furnishing the following examples of question tags:
7
Trudgill (1984:33) notes:
Speakers of Standard English English in the south of England tend to use, in their
speech, contracted negatives of the type:
I haven't done it
I won't do it
In the north of England, the alternative contraction is, in some areas, more common:
I've not done it
I' ll not do it.
Trudgill & Hannah (1994:97) have drawn attention to the same tendency in Scottish English:
There is a tendency not to contract the negative element not in Scot Eng, especially
in yes-no questions. If an auxiliary is present in a negated sentence, the auxiliary
usually contracts.
ScotEng EngEng
Is he not coming? Isn't he coming?
Did you not see it? Didn' t you see it?
He' 11 not go He won't go
You've not seen it You haven't seen it.
and then commenting: "In short, given the main sentence we can always predict
exactly what the tag will be." (164).9 My impression is that linguists and dialec-
tologists alike assume that other varieties are in accord with the standard as far as
the syntactic subsystem of interrogative tags is concerned. In the dialect of the
Bolton Metropolitan Area, however, this is certainly not the case. There are two
types of negative tag, the one of which obligatorily follows an affirmative
proposition and the other of which optionally follows a negative proposition. In
the Bolton Metropolitan Area equivalents of sentences (1) and (3), above, a
negative tag is used, the tag taking the form:
i.e., a contracted negative form is used in the tag. In (2) and (4), we get:
tive proposition must take the form verb + subject pronoun + /not/ "not". Thus
it is kepi entirely distinct from the contracted negative tags that obligatorily fol
low affirmative propositions. For the Bolton Metropolitan Area dialect, then, we
cannot "predict exactly what the tag will be" after a negative proposition, but
rather that it will take one of two forms.
5. Historical background
it would be distinguished by its intonation contour from the negative tag that optionally fol
lows a negative proposition.
13
I have changed the thorn symbol of the original to [Θ]. Use of reduced /tə/ in affirmative in
terrogatives is noted in various parts of the North, as in Wright's Windhill (West Riding of
Yorkshire) monograph (1892) and Brilioth's Lorton (Cumberland) study [1913]. Brilioth com
ments as follows,rightlydrawing attention to the possibility of partial assimilation:
§ 408. 2nd Person. The weak form to of the 2nd pers. sing, is only used interroga
tively in unaccented position; the t of this form has probably arisen through assimi
lation with the t-ending in verb forms like āt art, aut ought, dušt durst, and partial
asimilation with the s-ending of the 3rd pers. sing., which ending is often extended
to all three persons of the pres. sing. Professor Wright (A Grammar of the Dialect of
Windhill, p. 118) ascribes this transition of θ into t of the form ta, tö solely to the
unaccented position of the pronoun, basing his theory on the fact that there are only
three verbal forms ending in t (at art, out oughtest, d s t durst), but he has not
taken into consideration that partial assimilation with the s-ending of the 2nd pers.
sing, may also have been at work: the effort required in pronouncing the sibilant fol
lowed by the dental spirant in combinations like wants-өə, dis- is undoubtedly
much greater than in the case of s + the dental stop in wants-t, dis-tô. (104-
105, with the original thorn symbol changed to [Θ].)
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 179
example artow "art thou", has tow "hast thou" (21/311), wiltow, woltou
"wilt thou", wenest tǔ "thinkest thou" (3/25), wilte "wilt thou" (12/528).14
Jordan (1974:149-150, § 151) comments specifically on the weakening of
the vowel of the pronoun: "With enclisis appears weakened in seiste, woste
"do you say, know" RG, wilte Hav. 528, 1135, thenkeste Hav. 578 (Klaeber
1926:321)."
We may note a not dissimilar development in the history of the German lan
guage during the Middle High German period, with enclitic formations evident
in second person singular interrogatives under conditions of weak stress, as
similation, and a blending of verb and second person singular pronoun that was
very likely a contributing factor in the re-analysis of the verb inflection from -s
to -st:
Das von der Wortfügung der natürlichen Rede her häufig dem Verbum
nachgestellte Personalpronomen pflegen die Schreiber angesichts seiner
satzrhythmischen Untertonigkeit gern enklitisch dem Verbum anzuhängen:
vgl. bistu oder in »Christus und die Samariterin« 27 mahttu; nimistu. In
solchen Verschmelzungsformen ist du wegen der assimilatorischen
Angleichung seines Anlauts an den voraufgehenden stimmlosen
Konsonanten nicht mehr ohne weiteres erkennbar; so begreift sich, daß es in
falscher Formauflösung als zum Verbum gehörig aufgefaßt wurde.(Tschirch
1989:30)
Early Modern English has affirmative interrogative forms that are clearly di
rect descendants of the Middle English forms cited above. For instance, in
Hamlet, Act 5, Scene 1, we find the lines: "Woot weep? Woo't fight? Woo't
fast? Woo't tear thyself? | Woo't drink up eisel, eat a crocodile?" (Shakespeare
[1600]:265-266). Some versions of the play follow up these lines with "Dost
come here to whine?" The editor's Note to line 265 in the edition cited here
reads: "Woo't wilt (a colloquial form of the second person singular of will (OED
v.l Α 3δ)). The form seems to have been associated with challenges and the like
in Shakespeare's mind. Compare 2 Henry IV 2.1.54-5; Anthony 4.2.7." These
remarks agree with the interpretation in Onions (1986:322).
Cliticizaüon of the negative particle is a later development. Britton (1992:38)
observes that the negative enclitic /nt/ is a contraction of the negative particle
under reduced stress which was contemporaneous with the process of cliticisa-
tion and which was first recorded in English in the second half of the seven
teenth century. Subsequent phonological changes affecting some of the cliti-
cised verb forms in Standard English have made the /nt/ redundant in certain
instances as a marker of negation because in a number of forms the contrast
6. Earlier scholarship
The standard glossary of the Lancashire dialects is Nodal & Milner (1875-
1882). Here I note only:16 arto? "art thou?"; artono?, artrito? "art thou not?";
asto? "hast thou?"—again with supporting quotations from early literature. The
first of the more localized glossaries is Cunliffe (1886), dealing with Rochdale
and Rossendale to the east and north-east of Bolton, and again only a few, af
firmative forms are recorded: arto? "art thou?"; hasto? "hast thou?"; wilto "wilt
thou?"; and the adverb wilto-sholto "whether or not" [the gloss itself is regional;
"whether orno"].
Ellis (1889) contains only one second person interrogative in the
Comparative Specimen, and none in the Dialect Test. The style of the relevant
sentence is formal, and no instructions are given about number or stress. There
are four renderings of the Comparative Specimen for the Variety that includes
what is now the Bolton Metropolitan Area, and the plural pronoun is used in
three of these. (See further p. 334 [1,766].)
In an early south Lancashire glossary (1901), dealing with the area between
and to the east of Bolton and Manchester, Taylor gave a "Conjugation of the
Verb, to have" (n.p.), which includes the following second person singular in-
17
In common with many dialect writers and glossarists, Taylor spells have and other words
with the h of the standard, even though /h/ is not a phoneme of the dialects of south
Lancashire. His use of the apostrophe is inconsistent, as will be evident from a comparison of
the forms cited.
18
Strictly it is historical /ө/ that became /t/, /ð developed in atonic syllables in the fourteenth
century after the sandhi development whereby /Θ/ became /t/. See further Jordan (1976:187, §
207).
182 GRAHAM SHORROCKS
19
This is a terribly vague designation, since pre-1974 Lancashire was geographically exten
sive, had a big population, and the boundary between the Northern and north-Midland dialects
ran right through it.
20
Schilling's Grammar is generally thought to be of limited reliability. The forms cited here
are from the following pages: § 108, p. 113; § 124, p. 123; p. 124; p. 125; p. 126; p. 127; p.
129; p. 130. I have added some glosses, since Schilling's paradigmatic layout has been lost
here; also the diacritic to the shall-förm, which otherwise seems anomalous; and have omitted a
negative interrogative form of DO, where there seems to be a fairly obvious misprint. I have
replaced the thorn symbol of the original by [ө]. Schilling says that the enclitic interrogative is
preferred in Oldham (p. 123, under CAN), and draws attention in a note to the contracted second
and third person singular negative interrogative forms of CAN.
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 183
7. Conclusions
This study has shown that there are second person singular interrogative
forms in the dialect of the Bolton Metropolitan Area that are highly distinctive
both phonologically and morphologically in comparison to interrogative forms in
most other varieties of English spoken today. Extremely contracted enclitic
forms have developed through the processes of reduction, assimilation and
simplification under conditions of weak stress. The contractions have sometimes
been of such an order that certain second person singular interrogative forms of
different verbs are now homophonous, even within such a small and sharply
delimited subset of forms. This homophony may account for an apparent prefer
ence for /'adnt, 'wө:(r)nttө/ over more usual variants. Negative forms are more
often than not of the type verb + cliticized pronoun + cliticized negative particle.
This is because the cliticized forms of the singular pronoun developed in Middle
English times, whereas cliticization of the negative particle was a later develop
ment. There are also important syntactic differences between this dialect and the
standard in the preference for the construction verb + pronoun + negative parti
cle rather than verb + cliticized negative particle + pronoun; and in the existence
of two types of negative tag, the one of which obligatorily follows an affirmative
statement and the other of which optionally occurs after negative statements. Our
review of the previous scholarly literature reveals that certain forms such as
/'katnt, 'kɔtnt, 'mөtnt, 'wөtnt, 'witnt/ have not been recorded before, even im-
pressionistically, and that there has been no systematic treatment of second per
son singular interrogative forms.
The syntactic subsystem treated here has implications for our methodology
and fieldwork: at the linguistic level of syntax, many phenomena cannot be reli
ably elicited by questionnaire techniques. Whole sentences, themselves ques
tions rather than responses, are needed to elicit the interrogative patterns.
Furthermore, the syntactic subsystem described here manifests itself typically in
conversation rather than narrative, so that a free exchange of views between in
formant and fieldworker or between two (or perhaps more) informants should
be encouraged. These interrogative forms emerge fully only if the informant is
free to question the fieldworker (or another informant), and to do this in a rela
tively informal atmosphere—otherwise the plural form or the standard pronoun
might be used as a polite singular. Interviews in which thefieldworkerasks all
the questions, and the informant simply provides answers, are inappropriate
where the interrogative is concerned. Whilst dialectologists have generally rec
ognized that questionnaires do not elicit syntax effectively, they have been slow
to advocate the methodological corollary: namely, that monographs be written
based upon prolongedfieldworkin a locality and a large corpus of tape-recorded
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 185
REFERENCES
Brilioth, Börje. [1913.] A Grammar of the Dialect of Lorton (Cumberland).
Historical and Descriptive, with an Appendix on the Scandinavian Element and
Dialect Specimens and a Glossary. Publications of the Philological Soc. 1.
London: Oxford Univ. Press.
Britton, Derek. 1992. "Secondary Contraction and Deletion of Negative Enclitics
in English Dialects". Zeitschrift für Dialektologie und Linguistik 59,1.38-49.
Cunliffe, Henry, comp. 1886. A Glossary of Rochdale-with-Rossendale Words and
Phrases. Manchester & London: John Heywood.
Ellis, Alexander J[ohn]. 1869-1889.On Early English Pronunciation, with
Especial Reference to Shakspere and Chaucer, Containing an Investigation of
the Correspondence of Writing with Speech in England, from the Anglosaxon
Period to. the Existing Received and Dialectal Forms, with a Systematic Notation
of Spoken Sounds by Means of the Ordinary Printing Types. Including a Re-
Arrangement of Prof. F.J. Child's Memoirs on the Language of Chaucer and
Gower, Reprints of the Rare Tracts by Salesbury on English, 1547, and Welsh,
1567, and by Barcley on French, 1521, Abstracts of Schmeller's Treatise on
Bavarian Dialects, and Winkler's Low German and Friesian Dialecticon, and
Prince L.L. Bonaparte's Vowel and Consonant Lists. 5 vols. Publications of the
Chaucer Society. 2nd Ser. 1, 4, 5, 11, 25. Publications of the Early English Text
Society. Extra Ser. 2, 7, 14, 23, 56. And without numbers in the publications of
the Philological Society. London: 1869-1889. Pag. cont; vol. 5 also has sepa
rate pag. Title: Parts I-III, title stops at Barcley on French, 1521. Imprint: Parts I-
IV: Asher & Co. for the Philological Society; Trübner & Co. for the Chaucer
Society and the Early English Text Society. Part V: Trübner & Co. for the
Philological Society, the Early English Text Society, and the Chaucer Society,
1889. Part V. Existing Dialectal as Compared with West Saxon Pronunciation.
186 GRAHAM SHORROCKS
With Two Maps of the Dialect Districts. [The subtitle is used much more fre
quently: The Existing Phonology of English Dialects Compared with That of
West Saxon Speech. Forming Part V. of Early English Pronunciation.] Repr.
New York: Greenwood Press, 1968.
Frangopulo, N[icholas] J[oseph]. 1977. Tradition in Action. The Historical
Evolution of Greater Manchester County. East Ardsley (Wakefield): EP
Publishing Limited.
Freeman, T[homas] W[alter], H.B. Rogers & R.H. Kinvig. [1966.] Lancashire,
Cheshire and the Isle of Man. Regions of the British Isles. [London: Thomas]
Nelson [& Sons Ltd.].
Hargreaves, Alexander. 1904. A Grammar of the Dialect of Adlington
(Lancashire). Anglistische Forschungen 13. Heidelberg: Carl Winter. Repr.
Amsterdam: Swets & Zeitlinger NV, 1967.
Hunt, Jean. [Listed in some bibliographies under her married name, Jean Wade.]
1959. A Grammar of the Dialect of Heywood, Lancashire. MA thesis, Dept. of
English, Univ. of Leeds. Publ, in microfiche, East Ardsley (Wakefield): Micro
Methods Limited (later SR Publishers Ltd./EP Microform).
International Phonetic Association's Alphabet. 1993 revision. Journal of the
International Phonetic Association 23,1.32-34 and unpaginated centre-page
chart.
Jordan, Richard. 1974. Handbook of Middle English Grammar: Phonology.
Transl. & rev. Eugene Joseph Crook. Janua Linguarum. Series Practica 218. The
Hague & Paris: Mouton.
Lester, Mark. 1971. Introductory Transformational Grammar of English. New
York: Holt, Rinehart & Winston.
Mossé, Fernand. 1952. A Handbook of Middle English. Transl. James A. Walker.
Baltimore & London: The Johns Hopkins Univ. Press. (9th printing, 1987.)
Nodal, John H[oward] & George Milner. 1875-1882. A Glossary of the
Lancashire Dialect. Publications of the Manchester Literary Club. Publ, in 2
Parts. A Glossary of the Lancashire Dialect. Part I, A-Ε. Publications of the
Manchester Literary Club. Manchester: Alexander Ireland & Co.; London:
Trübner & Co. for the Manchester Literary Club, 1875. Part II, F-Z.
Manchester: Abel Heywood & Son for the Manchester Literary Club, 1882.
Repr. in one vol. in a limited ed. of 100 copies. Bath: Olivers, 1972. Also publ,
as Part /.— A to E. Part IL— F to Ζ. Publications of the English Dialect Society
10 & 35. Series Original Glossaries. Vol. 14. London: Triibner & Co. for the
English Dialect Society, 1875-1882. Repr. Vaduz: Kraus Reprint Ltd., 1965.
Onions, C.T. A Shakespeare Glossary. Enl. & rev. Robert D. Eagleson. Oxford:
Clarendon Press, 1986.
Orton, Harold & Wilfrid J. Halliday, eds. 1962-1963. Survey of English Dialects.
(B). The Basic Material. Volume I. The Six Northern Counties and the Isle of
Man. 3 Parts. Leeds: E.J. Arnold & Son Limited for the Univ. of Leeds.
Picton, J.A. 1865. "The South Lancashire Dialect". Proceedings of the Literary
and Philosophical Society of Liverpool 54 (1864-1865):17-56. Extracted by
permission & repr. as Notes on the South Lancashire Dialect. Liverpool:
Privately printed.
BOLTON VERNACULAR: SECOND PERSON SING. INTERROGATIVE 187
Saxelby, C[harles] H[arold], ed. 1971. Bolton Survey. General ed. C.H. Saxelby.
Subject eds.: Geography, Alice Dyson; History, W,E. Brown; Natural History and
Pre-History, A. Hazlewood. Pub, under the auspices of the Bolton branches of
the Geographical and Historical Associations and Bolton Field Naturalists
Society. [Bolton: Printed by Tillotsons (Bolton) Limited, Bolton & London, for
the Bolton Survey Committee,] 1953. Repr. with a new introduction by W.E.
Brown. County History Reprints. Wakefield: SR Publishers Ltd.
Schilling, Karl G. 1906. A Grammar of the Dialect of Oldham (Lancashire).
Doctoral diss. Univ. of Gießen, 1906. Publ. Darmstadt: G. Otto's Hof»
Buchdruckerei.
Scollins, R. & J. Titford. N.d. Ey Up, Mi Duck. Notes to record produced by Ram
Productions, Ilkeston, Derbyshire.
Shakespeare, William. [1600.] Hamlet. Ed. G.R. Hibbard. The Oxford
Shakespeare. Oxford & New York: Oxford Univ. Press, 1987.
Shorrocks, Graham. 1981. A Grammar of the Dialect of Farnworth and District
(Greater Manchester County, Formerly Lancashire). 2 vols. Ph.D. diss. Univ. of
Sheffield, 1980-1981. Ann Arbor: University Microfilms international; London:
Information Publications. 81-70, 023.
___. 1985. "Aspects of Affirmation and Negation in the Dialect of Farnworth
and District (Greater Manchester County, Formerly Lancashire)". Journal of the
Lancashire Dialect Society 34.20-29.
___. 1988. "Glottalization and Gemination in an English Urban Dialect". The
Canadian Journal of Linguistics/La Revue canadienne de Linguistique 33,1.59-
64.
— _ . In press. 'Fieid Methods and Non-standard Grammar". Methods VIII.
Proceedings of the International Conference on Dialectology ¡Colloque interna
tional de dialectologie, 2-7 August, 1993, University of Victoria, B.C. Ed. James
Arthurs. [With a summary in French.]
Smith, David M. 1969. The North West. Industrial Britain. Newton Abbot: David &
Charles.
Survey of English Dialects. See under Orton & Halliday, eds. (1962-1963).
Taylor, Francis Edward, comp. 1901. The Folk-Speech of South Lancashire: A
Glossary of Words Which Are, or have Been During the Last Hundred Years, in
Common Use in That Portion of the County Palatine Situate Between Bolton
and Manchester, Including Dialect Words, Children's Words, Local
Mispronunciations, Colloquialisms, and Local Slang, with an Appendix of
Quaint Sayings. Manchester & London: John Heywood.
Tschirch, Fritz. 1989. Geschichte der deutschen Sprache. Zweiter Teil:
Entwicklung und Wandlungen der deutschen Sprachgestalt vom Hochmittelalter
bis zur Gegenwart. Grundlagen der Germanistik 9. 3te, ergänzte und überar
beitete Auflage bearbeitet von Werner Besch. Berlin: Erich Schmidt Verlag.
Trudgill, P[eter]. 1984. "Standard English in England". Language in the British
Isles,ed. by Peter Trudgill, 32-44. Cambridge: Cambridge Univ. Press.
—-. 1990. The Dialects of England. Oxford, & Cambridge, Ma.: Basil
Blackwell.
188 GRAHAM SHORROCKS
1. Introduction
In this article, I will study two seemingly unrelated topics in Icelandic and
Faroese syntax—the properties of subject /verb agreement and the distribution of
reflexives and pronouns in the two languages. It will turn out, however, that the
analysis I propose to explain certain properties of subject/verb agreement easily
extends to provide an account of the distributional constraints on reflexives and
pronouns, whence I will ultimately draw the conclusion that the relevant proper
ties in the two domains actually reflect the same set of syntactic principles. These
principles include a general licensing condition on empty arguments and an
economy principle preferring unspecified feature values over specified ones.
I will first present an analysis of subject/ verb agreement and extend this
analysis to object/verb agreement. This analysis is subsequently shown to ex
plain an otherwise unexpected restriction on reflexives in Icelandic. The notion
of Economy of Representation embodied in the analysis is then shown to predict
the complementary distribution of reflexives and non-reflexive pronouns.
Finally, I show that this approach leads to a unified analysis of a set of contrasts
between Icelandic and Faroese.
I will begin by considering the question why certain DPs move to SPEC-po-
sitions. In particular, I will use certain properties of subject /verb agreement in
Icelandic and Faroese to argue against the currently fashionable view that
movement to SPEC-positions is in general driven by feature checking. Instead, I
will suggest that such movement more plausibly reflects the way predication is
realized in syntactic form. The appearance of agreement features on the verb will
be taken to be forced by movement rather than the other way around. This as
sumption will be the key to my analysis of subject/verb agreement, and even
tually also of object/verb agreement, in Icelandic and Faroese.
1
Technically, it is of course possible to say that oblique subjects have 'abstract' nominative
Case, on top of whatever Case is reflected in their morphology, but this clearly has no ex
planatory value in the absence of independent motivation.
2
A finite verb always has overt number- and person-agreement with a nominative subject; cf.
2.4. below. Hence, if one were to posit covert agreement for sentences with oblique subjects,
one would have to explain why the agreement features are invisible in just this case. Moreover,
such covert agreement would have to be restricted to person, since the verb may have visible
number-agreement with a nominative object.
192 KNUTTARALD TARALDSEN
and is enforced by Full Interpretation on the assumption that these features re
ceive no interpretation, either at the PF- or the LF-interface. Thus, the relevant
features stand out as vestiges of an autonomous syntactic component in a theory
where syntax is taken to be determined by interface conditions. From this point
of view, the current use of feature checking to fuel movement to SPEC-positions
seems curiously non-minimalist.
(8)
The intuition behind this proposal is that if some head Y licenses a Case K, and
there is no overt DP bearing K, then the existence of a K-marked ec is deduced.
3
See Holmberg (1993:128) for a similar proposal, which, however, does not include an identi
fication condition on the predicate-internal empty argument.
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 193
Otherwise, the presence of the ec must be flagged by copying its person /number
features onto the head of the relevant domain.4
We can now treat subject/verb agreement as a side-effect of movement to
SPECXP in (8). If the DP moved to SPECXP in (8) is not Case-licensed in X', its
trace, the ec in (8), is uniquely identified in the predicate X' only if its person/
number features are copied onto X. Since the person / number features of the
trace must have the same values as those of the moved DP, the effect of specifier
/head agreement will arise, although, on this view, there is no feature checking
relation between X and SPECXP. If the DP in SPECXP is Case-licensed within
X', however, the identification condition does not require the person / number
features of its trace to appear on X, which may then take on default values.
We shall now see how this proposal can provide an account of the subject/
verb agreement pattern in Icelandic and Faroese.
4
In this formulation, my identification requirement for predicate-internal empty arguments
bears a certain resemblance to various disjunctive formulations of the ECP, taking (9b) to cor
respond to lexical proper government, while (9a) is a form of antecedent-government. The pos
sible significance of this relationship remains to be explored.
5 The exact nature of Case-licensing seems largely irrelevant here.
194 KNUT TARALD TARALDSEN
petitors that are equally well-formed with respect to other constraints, the prin
ciple of Economy of Representation will select the one with the least number of
specified feature-values and discard the others:
This follows immediately from the proposal in 2.2 if nominative subjects are not
Case-licensed in a position lower than SPEC·AGRS. Then, the trace left behind in
AGRS', corresponding to the ec in (8), is uniquely identified, according to (9),
only if its person/number features are copied onto AGRS. Hence, we have the ef
fect of obligatory subject/verb agreement with nominative subjects.
6
The term 'non-distinct' is to be understood in such a way that (a) the phonetic matrix is not
taken into consideration and (b) an unspecified occurrence of a feature is non-distinct from any
specified occurrence of the same feature. This has the consequence that two representations A
and where A has X in position Ρ and has Y in P, may sometimes be considered non-distinct
from one another even when X and Y are generally regarded as different lexical items. This
property becomes crucial in 3.2. below. Thus, I disagree with Chomsky (1995), whose notion
of 'numeration' allows two representations to belong to the same competitor set only if they
are built from exactly the same lexical items.
7 Actually, the finite verb overtly inflects only for number in Modern Faroese. I have not ex
plored the possible relevance of this for the text analysis except for the suggestion in footnote
22.
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 195
8 The distinctive property of an object with lexically selected Case is that it retains its Case
when promoted to subject, e.g., in passives.
9
This seems true for the majority of Icelandic speakers, although there are quite a few speakers
who do not accept infinitival clauses with oblique subjects and nominative objects in ECM-
contexts. According to Sigurðsson (1993: footnote 4), some of these speakers actually have ac
cusative Case on the object in this situation. Possibly, for the speakers who cannot have nom
inative Case on the embedded object in ECM-contexts, the infinitival AGRO cannot host a num
ber-feature in this configuration, as would be required for the trace of a nominative object to
satisfy (9) under the analysis in 2.5.
196 KNUT TARALD TARALDSEN
10
Like most other current accounts, this analysis cannot use Case-theory to rule out sentences
like *There seems a man to be sleeping under the table; cf. Chomsky (1995) for a possible ex
planation not based on Case-licensing.
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 197
This suggests that the restriction found in Icelandic is linked to the fact that ob
ject has nominative Case in sentences with oblique subjects in Icelandic, but has
accusative Case in Faroese, a minimal difference between the two languages,
which will be exploited again at various points below.
The conjecture that the ill-formedness of (15) is to be traced back to the di
rect object's having nominative Case is also supported by the observation that
whenever the direct object itself has oblique Case, it may perfectly well denote a
human being even in Icelandic:
At the same time, we learn from (19)-(20), that the relevant distinction is not
simply accusative vs nominative, but rather nominative vs all other Cases. This
now becomes entirely natural if the nominative differs from all other Cases by
being the default value assigned to an unspecified Case-feature. We can then
characterize the condition relevant to (15) as follows:11
11
(21) is obviously not more than a descriptive generalization. A more explanatory account
might try to link (21) to the appearance of a preposition in front of direct objects denoting hu
man beings in Spanish and various other varieties of Romance. Such an account should also
explain why verbs carrying the -st suffix (at least historically an affixed reflexive) are systemat
ically exempt from (20) in Icelandic. Finally, we will eventually have to explain why (21) does
not apply to subjects.
198 KNUT TARALD TARALDSEN
As mentioned above, the verb agrees with the direct object in Icelandic sen
tences with oblique subjects; cf. (7) (repeated below as (22)) and (23):12
In Faroese, however, the verb always takes the default 3SG form in such
sentences. This contrast between Icelandic and Faroese with respect to object
agreement would appear to follow from the fact that the object has accusative
Case in sentences with oblique subjects in Faroese. Not even in Icelandic does
the verb agree with an object which does not have nominative Case (because of
selectionai properties of the verb, the infinitive leiðast in (24), overriding default
Case assignment):
12
In general, object agreement is strongly preferred over the default option (always 3SG) in
sentences with nominative objects. Nevertheless, many speakers find the default form relatively
acceptable. The text analysis is based on the assumption that such speakers allow pure number
agreement not to be spelled out at PF, taking verbal agreement with nominative objects to be
obligatory at least at LF.
13 On the basis of the proposals made in 2.2., we may conjecture that all direct arguments of a
verb are linked to their θ-roles via predication at LF. Then, AGRO, like AGRS, must be the head
of a predicate at LF. Word order seems to indicate that raising to S P E C A G R O P generally only
occurs after Spell-out, a property I have no explanation for.
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 199
disallows it. 14 But if the object is nominative, it really has no specified Case-
feature, outside of PF, and its trace in AGRO' is uniquely identified only if its
features are copied onto AGRO.
Notice that object agreement is restricted to agreement with respect to num
ber. Person agreement with the object is not possible:15
This suggests that AGRO, unlike AGRS, can only host number features, a hypoth
esis which will become important in the next section.16
2.6. Summary
14
As a special case, objects assigned 'structural' accusative Case in sentences with non-
oblique subjects will never induce object / verb agreement, assuming that the 'structural' ac
cusative, like lexically selected Case, is licensed within AGRO', albeit by general rule.
15 Some Icelandic speakers do not accept 1st and 2nd person nominative objects at all. This is
not predicted by my analysis, unless, for these speakers, forms like ég "I" vs við "we" have
only person features, being distinguished as, say, 1 person vs 4 person rather than as lSG vs
IPL; cf. Kayne (1989). Another fact for which I have no explanation is that speakers who do
accept 1st and 2nd person nominative objects apparently prefer the default form of the verb over
object / verb agreement in sentences like (25), although the judgment goes in the opposite di
rection with 3rd person nominative objects.
16 Suppose number features can only be represented on AGRO, and that, moreover, a specified
person feature on AGRS needs the support of a specified number feature agreeing with the same
DP. Then, placing a nominative DP in SPEC-AGRSP, by requiring a specified person feature on
AGRS, will force AGRS to attract AGRO and the number feature of AGRO must agree with the
nominative subject rather than the object. Consequently we would predict correctly that the ob
ject cannot be nominative in a sentence with a nominative subject. Licensing of 'structural' ac
cusative Case for the object could then be viewed as a 'last resort' option (distinct from the de
fault assignment of nominative Case at PF, since, crucially, its effect must be LF-visible).
200 KNUTTARALD TARALDSEN
Icelandic and Faroese both have a so-called simple reflexive pronoun, sig in
Icelandic and seg in Faroese, which, unlike the non-reflexive pronouns, is refer-
entially dependent, and needs to be bound by a suitably local antecedent. Below,
I will first examine a surprising gap in the distribution of Icelandic sig, and then
discuss the nature of the complementary distribution of sig/seg and the non-re
flexive pronouns. In each case, I will argue that the Binding Theory does not
explain the pertinent observations, whereas the assumptions underlying the
analysis in Section 2 actually do.
Yet, (26) is ungrammatical, although (27) is fine. A priori, one might of course
claim that oblique subjects are not licit antecedents for sig. But the grammatical
y of (28) (where the embedded verb has a dative subject) immediately shows
this to be false:
The only other difference between Faroese and Icelandic potentially relevant
to this contrast concerns the Case assigned to the object in sentences with
oblique subjects: in Faroese, it is accusative, but in Icelandic, nominative.
Hence, we are led to conclude that (26) is ungrammatical because sig has nomi
native Case. But why should precisely sig be allergic to nominative Case ?
In fact, the analysis developed in the preceding section provides an immediate
answer to this question. By previous assumption, sig ¡seg must eventually raise
to the matrix SPECAGROP in (26)-(29). By the identification condition on the
predicate internal ec, and the definition of unique identification in (9) (repeated
below), the trace of sig /seg must either be Case-licensed within AGRO', or else
have a copy of its features placed under AGRO:
Since AGRO, as seen in 2.5 may host only number features, the latter option is
open only to DPs that in fact have specified number features. Since the nomina
tive Case is simply the PF-value for unspecified Case, we therefore predict that a
nominative DP can successfully raise to SPECAGROP only if it has a specified
number feature.
Now, it so happens that sig /seg is unique within the pronominal system of
Icelandic and Faroese in that it does not have both singular and plural forms. In
fact, even when the antecedent is plural, sig /seg has the morphological shape of
a singular pronoun.17 We may take this to mean that sig /seg is actually unspec
ified for number and is assigned the singular form at PF by default. It follows
17
Compare the following paradigms : By contrast, the 1PL and 2PL forms are:
1SG 2SG sig 1PL 2PL
G mín fl ín sín G okkar ykkar
D mér flér sér D okkur ykkur
A mig flig sig A okkur ykkur
Faroese seg is morphologically related to the 1SG and 2SG forms in the same way.
202 KNUT TARALD TARALDSEN
then from the preceding observations that a nominative sig cannot raise to
18
SPECAGROP, and hence, sig cannot be a nominative object.
The success of this analysis provides independent support for the assump
tions it is based on, viz., the identification requirement associated with predica
tion and the claim that nominative DPs have unspecified Case (outside of PF).19
This observation might suggest that sig is subject to the Binding Theory's
Principle A, while non-reflexive pronouns go by Principle B. However, sig too
seems subject to Principle rather than Principle A. On the one hand, its an
tecedent need not be included in its binding domain, as shown by the well-
known 'long distance' cases. On the other hand, it does not normally even tol-
18
Sig cannot occur as a nominative subject either. Although it is somewhat more difficult to
eliminate competing explanations for this fact, I would like to suggest that here too, we see
the effect of the identification requirement on predicate-internal empty arguments. In particular,
I would suggest that sig not only lacks specified number, but in fact also has no person-fea
ture. Providing adequate motivation for this claim would take us outside the bounds of this ar
ticle, but notice that if sig has no specified number-feature, the claim made in footnote 16 that
a specified person-feature needs the support of a specified number feature entails that sig has no
specified person-feature either.
19 It is potentially embarrassing for the analysis presented here that it does not extend to (i),
since reciprocal hvor annar "each other" would not appear to be unspecified for number :
(i) * Börnunum leiddist hvorum annar
children-the-D got-bored-3SG each-D other-N
However, it appears that the reciprocal facts are somewhat less clear-cut than those involving
sig. Thus, Margret Jónsdóttir (p.c.) (unlike Höskuldur Práinsson, p.c.) considers (ii) as quite
acceptable. In (ii), the each-part of the reciprocal expression, like , also has nominative
Case, and does not agree in Case with its antecedent. Corresponding examples with nominative
(sjálf) sig are strongly rejected:
(ii) Börnunum leiddist hvor annar
children-the-D got-bored-3SG each-N other-N
For the moment, then, the significance of (i) is hard to assess.
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 203
erate an antecedent within its binding domain. In simple clauses like (31), for
instance, sig is impossible unless sjálfur "self is added:
The reason (30) is grammatical without sjálfur is that sig here is part of an em
bedded clause, so that it may be assigned a binding domain which does not in
clude its antecedent.
In this perspective, sig is a pronoun, not an anaphor, in the standard sense
of this term. It is however referentially dependent, possibly because it has no
specified value of the number feature.
Assuming, therefore, that Binding Theory provides no explanation for the
complementarity of sig and non-reflexive pronouns, we are led back to the
question why the non-reflexive pronoun cannot co-refer with the matrix subject
in sentences like (30). xA.gain, the answer is supplied by one of the ingredients
entering into the analysis developed in the preceding section, the principle of
Economy of Representation.
As we saw in 3.1, the contention that sig has no specified number feature is
supported both by morphological considerations and the fact that this assump
tion allows a straightforward account of the absence of nominative sig. The non-
reflexive 3rd person pronouns, on the other hand, both have separate singular
and plural forms and can appear as nominative objects, as we shall see shortly.
Thus, they must have specified number features. By Economy of Representation
then, a non-reflexive pronoun is excluded from any context where it can licitly
be replaced by sig, which has fewer specified features.20 In (30), for instance,
20
When sig inside an infinitival or a subjunctive clause is bound to the subject of a higher
clause, it can be replaced by a non-reflexive pronoun. My analysis requires a structural differ
ence between the two sentences that result from this: the one with sig must have a structural
property allowing sig to link up with a distant antecedent, while the one with a non-reflexive
pronoun must not have this property. Then, the two will not be in the same competitor set.
This view is supported by the following observation (Höskuldur Práinsson, p.c.) : Consider a
sentence of the form indicated in (i), where IP2 and IP3 both are infinitival or subjunctive com
plements allowing for long distance binding of sig, and DP 2 and DP3 are both non-subjects :
(i) DP1 ... [IP2 ...DP2 ... [ I p 3 ... DP3 ...
Suppose both of DP 2 and DP3 are bound by DP 1 . Placing sig or the appropriate form of hann
in the positions of DP 2 and DP3, seems to yield the following acceptability pattern :
204 KNUT TARALD TARALDSEN
h cannot be co-indexed with the matrix subject, although this does not violate
Principle B, since this indexing licenses sig, also a non-anaphoric pronoun, in
contrast with the corresponding indexing in (31).21
If this analysis is correct, it provides independent evidence for our notion of
Economy of Representation as well as the assumption that sig has no specified
number feature, two important components of the analyses presented above. In
the next subsection, I provide a piece of striking evidence that the analysis is in
deed on the right track.
We have seen that sig cannot occur in the nominative Case, according to the
analysis in 3.1 because it would then fail to satisfy the identification requirement
on predicate-internal empty arguments as defined in (9). Since objects have
nominative Case in Icelandic sentences with oblique subjects, provided the verb
does not select an oblique Case for its object, there will therefore be sentences
where the object cannot be sig, even though it would be bound by a suitable an
tecedent, as shown in 3.1. The analysis in 3.2 now predicts that exactly in these
Likewise, the object pronoun cannot be bound by the subject in the Faroese
counterpart to (32):
This follows in our analysis because the object has accusative Case in Faroese,
making seg possible:
These observations show quite conclusively that the possibility for the object
pronoun to be bound by the subject in (32) has nothing to do with the fact that
the subject is oblique, but rather depends on the Case-marking of the object.
Notice, finally, that the effect of nominative Case relevant to (32) is not
simply that of lifting the normal Principle requirements. If nominative pro
nouns were immune to Principle effects, we would not understand the con
trast between (32) and (35), where the object pronoun cannot be bound by the
subject, although it has nominative Case:
If on the other hand Principle holds for nominative pronouns too, the contrast
is expected, since the pronoun has a binding domain (the infinitival clause) not
including its binder in (32) but not in (35).
I conclude that attributing the complementary distribution of sig and non-re
flexive pronouns to the principle of Economy of Representation makes correct
predictions that are not matched by alternative approaches.
3.4. Summary
In this section, we have seen that the distributional properties of sig /seg and
the non-reflexive pronouns in Icelandic and Faroese provide independent moti
vation for the crucial assumptions that the analysis in Section 2 was based on. In
particular, the absence of a nominative object sig was derived from the identifi
cation requirement on predicate-internal empty arguments and the assumption
that the nominative is the PF-spell-out of unspecified Case. The complementarity
of sig /seg and the non-reflexive 3rd person pronouns, on the other hand, was
taken to follow from the same notion of Economy of Representation that was
used in Section 2, a hypothesis that was perhaps most convincingly supported
by the facts discussed in 3.3.
Putting together the analyses in Sections 2 and 3 then, we have also reached
the surprising conclusion that the unavailability of nominative object sig and the
obligatoriness of verbal agreement with nominative subjects and objects are the
same fact at an appropriate level of abstraction, both reflecting the identification
condition on predicate-internal empty arguments and the status of the nominative
as a default Case. The complementarity of sig /seg and the non-reflexive pro
nouns, on the other hand, follows from the same principle of Economy of
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 207
4. Icelandic vs Faroese
I pointed out in 2.4 that Faroese differs from Icelandic with respect to the
Case assigned to the direct object in sentences with oblique subjects. Whenever
the Case is not lexically selected by the main verb, it is nominative in Icelandic,
but accusative in Faroese. The second difference we observed was that Icelandic
líka, but not its Faroese counterpart, rejects an object denoting a human being:
We then saw that a finite verb shows number agreement with the object in sen
tences with oblique subjects in Icelandic, but not in Faroese:
The next contrast concerns the possibility of using the simple reflexive sig
/seg in the object position of a verb with an oblique subject, when the verb does
not select an oblique Case for its object. In Faroese this is perfectly possible,
and in Icelandic, it is not:
Thus, we have isolated five different contrasts between Icelandic and Faroese.
One might take all the five contrasts listed above to follow from the first of
them, the contrast with respect to the Case-marking of the direct object in sen
tences with oblique subjects. From this, the remaining four contrasts follow,
within the framework of assumptions adopted in Sections 2-3. To recapitulate,
the contrast illustrated by *(15) vs (17) is a consequence of the difference with
respect to Case-marking together with the assumption that the nominative is a
default Case, and the condition in (21):
must then make sure that the 'last resort' mechanism does not become available
for (26), even though its failure to apply actually causes (26) to be ill-formed.
This can be done by taking 'last resort' mechanisms to come into play only if
none of the members in a reference set is grammatical when more highly favored
options are taken. Assuming as before that (26) and (32) are in the same refer
ence set, the possibility for the object in (32) to be licensed without 'structural'
accusative Case makes 'last resort' Case-marking illicit in (26) as well. It is un
clear, however, how to extend this approach to (15), which is perhaps unlikely
to be in the same reference set as (16). Hence, the preceding remarks must be
left as suggestions for future research.22
4.3. Summary
In this section, I have listed five contrasts between Icelandic and Faroese,
and shown how they can be related to one another on the analysis developed
earlier. We have seen that there are at least two possibilities with respect to
which of the five contrasts should be considered basic. In either case, all five
contrasts are deduced from a single source in this analysis.
5. Conclusion
22
Deriving the relevant contrasts between Icelandic and Faroese from the properties of AGRO
would seem particularly attractive, if the stipulated impossibility of associating a number fea
ture with AGRO in Faroese could be linked with the lack of person features on AGRS in Faroese
(see footnote 7).
REFLEXIVES, PRONOUNS & AGREEMENT: ICELANDIC & FAROESE 211
REFERENCES
Chomsky, Noam. 1993. "A Minimalist Program for Linguistic Theory". The
View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger,
ed. by Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge: MIT Press.
. 1995. "Bare Phrase Structure". Government and Binding Theory and
the Minimalist Program, ed. by Gert Webelhuth, 385-439. Oxford & Cambridge:
Blackwell.
& Howard Lasnik 1993. "Principles and Parameters Theory". Syntax:
Ein internationales Handbuch zeitgenös sicher Forschung = An International
Handbook of Contemporary Research, ed. by Joachim Jacobs, Arnim von
Stechow, Wolfgang Sternefeld & Theo Vennemann, 506-569. Berlin: Walter de
Gruyter.
Hestvik, Arild. 1990. LF-Movement of Pronouns and the Computation of Binding
Domains. PhD dissertation, Brandeis University.
Holmberg, Anders. 1993. "On the Structure of Predicate NP". Studia Linguistica
47.126-138.
& Christer Platzack. 1995. The Role of Inflection in Scandinavian Syntax.
New York & Oxford: Oxford University Press.
Kayne, Richard S. 1989. "Notes on English Agreement". Unpublished ms., City
University of New York.
Sigurðsson, Halldor A. 1989. Verbal Syntax and Case in Icelandic. PhD disserta
tion, Lund University.
. 1993. "Agreement as Head Visible Feature Government". Studia
Linguistica 47.32-56.
ADVERBIAL QUANTIFIERS AND DIALECTAL
VARIATION IN A MINIMALIST FRAMEWORK*
MARIE-THÉRÈSE VINET
Université de Sherbrooke
Introduction
This study was partially funded by the Social Sciences and Humanities Research Council of
Canada [SSHRC grant number 410-93-0838) and FCAR (94ER0401). I would like to thank
the members of the FCAR research group (UQAM-Sherbrooke) for their comments and helpful
suggestions as well as the audiences in Tarragona (17/3/94), at the University of Toronto
(4/3/94), the University of Ottawa (28/9/94), Université du Québec à Montréal (20/10/94) and
the Saint John campus of the University of New Brunswick. Special thanks also to Guglielmo
Cinque, Esther Torrego and Anne Zribi-Hertz. The usual disclaimers apply.
214 MARIE-THERESE VINET
Adverbial quantifiers of the type studied here have received very little atten
tion in the literature. For some recent related studies, cf. Léard & Amyot (1993),
Vinet (1994), Vinet & al. (1994).
Benben can be loosely compared to beaucoup in SF in the sense that both are
quantifiers and when they appear with Negation they both show scope interac
tion effects. Benben in (la) presents scope ambiguities just as beaucoup would
in the same position (cf. Rizzi 1990:18). It can mean "many friends are such that
I did not meet them" or "not many friends are such that I met them". The sen
tences (lb, c), however, are unambiguous and allow only the second reading. It
can be observed, as a result, that benben in a pre-past participle position and
benben in a post-NP position both receive the same unambiguous internal scope
interpretation.
However, there exist many obvious differences in behaviour between the
two Qs (cf. Vinet & al. 1994). First, it can be observed that benben can appear
before a noun and it can be used as a quantifier or an intensifier for a verb, an
adjective or an adverb (cf. ). On the contrary beaucoup is more restricted. It
can quantify over a noun and it can quantify a verb but it is ruled out with most
adjectives and adverbs (cf.(3b)), except for comparative adverbs, as illustrated
in (3c) below:
The contrast in (4b) and (4d) clearly indicates that de is a negative polarity
item when it appears with benben (cf. Horn 1978) and a partitive preposition
when it appears with beaucoup. In other words, de is not in the scope of ben-
ben. Though beaucoup can license an existentially quantified N, benben cannot.
Note that benben is always optional in both positive or negative contexts, as il
lustrated in (5a, b). Nevertheless, in a positive context, benben is followed by
an indefinite des and it can never appear sentence-finally, as (5c) illustrates:
Note that if benben does appear with the indefinite in certain negative envi
ronments, it then has the meaning of the adverb vraiment "really" as in C'était
pas des cours benben parce qu'on faisait rien "They were not really courses be
cause we weren't doing anything" (P. Thibault: corpus sur la langue parlée à
Montréal, Université de Montréal).
Benben, as well as beaucoup, can quantify at a distance (cf. Obenauer 1983,
1984). However, when benben appears in a positive context, acceptability
judgements are not always clear: quantification at a distance (QAD) seems more
acceptable with certain verbs only:1
All the facts presented in this section tend to illustrate that benben is a polar
ity sensitive item whose distribution and interpretation is sensitive to negative
versus affirmative contexts (cf. Progovac 1993). This situation is reminiscent of
a relatively similar phenomenon with the quantifier not much which is normally
in suppletive variation with a lot. As noted by Klima (1964:283), only a lot can
appear in a positive sentence, as exemplified in (8) below:
Other adverbial Qs, like benben, can also appear postnominally in QF, as il
lustrated in (9b):
2
An apparent counterexample which could be raised is the case of en masse "masses/a lot"
which can be used postnominally in a positive context as exemplified below:
(i) a. J'ai en masse d'amis. b. *J'ai d'amis en masse,
. J'ai des amis en masse.
"I have masses / a lot of friends."
218 MARIE-THÉRÈSE VINET
Cellar (1991) mentions that bezef is used more frequently in the negative.
Note that bezef in colloquial French behaves differently from bezzaf in dialectal
Arabic which can be used in a positive context and also sentence-finally as in:
Ahmed mrid bezzaf"Amhed (is) ill a lot". The behavior of bezef seems to follow
more closely the behavior of adjectival-quantifiers in QF and we now turn to the
study of such quantifiers.
2.1 Adjectival-quantifiers
However, when en masse appears postnominally, it does not quantify over the N any more.
The partitive de is then obligatorily transformed into an indefinite determiner des, as in (ic).
ADVERBIAL QUANTIFIERS AND DIALECT VARIATION 219
From this point of view, molto can be said to behave more like an adjective
or a determiner. However, quantification over individuals, when the head noun
is in the singular, is always ruled out:*Hotrovato molto amico "I found a lot of
friend". In the same fashion, quantification over individuals, with adjectival-
quantifiers in QF, never allows a head noun in the singular:*J'ai pas gros gros
de cheval à vendre "I do not have big big of (a) horse to sell".
The invariable nature of adverbs in general explains the impossibility of gen
der and number agreement between the adjectival-quantifier and the head noun:
Kayne (1981) has shown that French [QP[e] de N] structures were headed by an
empty initial QP and that beaucoup could fill this position. However, as illus
trated in (17) below, there is a sharp contrast between the two forms since ben-
ben in (17b,c) is a negative polarity item which must be licensed by a sentential
NEG:
The negation forms which can license benben are very few. There is the
minimal negation quantifier pas, the aspectual quantifier plus (pronounced pu
[py] in QF) and jamais the temporal quantifier:
Cinque (1994) has proposed that adverbs in general are specifiers and not
adjuncts, as in the Emonds / Pollock approach. His main argument is connected
to the relatively rigid order which is found among adverbs. He therefore hypoth
esizes that adverbs occur in the SPEC of various intermediate functional phrases
(FPs) such as NEGP (pas), TP (toujours), ASPP (benben), etc., an hypothesis
which explains the rigid word order among adverbs:
Benben and similar adverbial-quantifiers can appear as the most deeply em
bedded element of the VP because, as negative polarity items, they are focus
sensitive. These all appear with at least two syllables and there is always a
falling intonation on the last syllable when it appears sentence-finally. Adverbs
with one syllable only, such as ben or trop, can usually not bear the greatest
prominence of the sentence (cf. Cinque 1993a), as illustrated in (22) below:
(22) a. J'ai pas ben aimé ça. /*J'ai pas aimé ça ben.
"I didn't (much) like it (much)."
b. Il a pas trop vu la différence.
c. * Il a pas vu la différence trop.
"He didn't (too much) see the difference (too much)."
Note that there must also be a rigid order between the -commander and the
polarity item, as claimed by Baker (1970), Lasnik (1993) and many others. This
explains the unacceptability of the following sentence in (23):
It can therefore be observed that the proper NEG -commanders for benben
usually correspond to negative adverbial forms and not to syntactic morphemes
such as ne I non, a class distinction also noted in Di Sciullo & Tremblay (1993).
I leave aside the preposition sans which also acts as a -commander in sans ben-
ben réfléchir I sans réfléchir benben "without much thinking / without thinking
much".
Sportiche (1988) has proposed that 'floating' Qs are generally partitive and
that tous "all" is treated as a partitive structure, without de insertion. Even
though adverbial Qs like benben are also partitive and require an association with
a plural NP just like tous, benben behaves nonetheless quite differently. First,
the scope of the domain of benben is VP only, contrary to tous where a syntactic
dependency between Q and the NP in SPEC-IP position is possible. This contrast
is illustrated below in (24):
224 MARIE-THÉRÈSE VINET
I suggest that the NEG item is both a trigger and a binder for the polarity of Q
and de. It is also hypothesized that the trigger must be in the appropriate domain
since negative polarity is a clause-bound process. As often noted, volitional
verbs can create a larger domain for negative polarity contrary to others like
epistemic and factive verbs, as illustrated by the very clear contrast in the follow
ing examples:
(26) a. Elle veut pas que j'aye ďamis benben après Fècole.
"She does not want me to have-SUBJ friends much after school.
b. * Elle pense pas que j'ai d'amis benben après ľécole.
c. * Elle regrette pas que j 'aye d'amis benben après l'école.
"She does not think / regret that I have (of) friends much..."
3
Furthermore, note that there can be no minimality effects with benben since (contrary to
beaucoup) it cannot bind by itself de N expressions. Minimality effects, as in (ib), are rather
linked to the presence of the Neg item pas which binds the existentially quantified N, as (ic,d)
illustrate:
(i) a. Combien de films que t'as pas benben aimés?
b. * Combien que t'as pas benben aimé de films?
c. * Combien que t'as pas aimé de films (benben)?
"How many movies did you not like a lot?"
d. Combien que t'as aimé de films?
"How many movies did you like?"
ADVERBIAL QUANTIFIERS AND DIALECT VARIATION 225
NEG licenser cannot bind more than one polarity item at a time in the same do
main, as (27) illustrates:
(27) a. ?* Les enfants ont pas benben ramassé de fraises trop trop.
"The children have not a lot picked up strawberries too
much."
b. * Les enfants ont pas benben ramassé de fraises benben.
4. Conclusion
In sum, the paper proposes that different factors can explain the syntactic
representations of adverbial Qs in QF and in other varieties of French as well.
First, it was indicated that certain adverbial features within a lexical form such as
benben were more dominant than in the case of adjectival-quantifiers. Moreover,
all these quantifiers which can easily appear sentence-finally always bear a
reading that falls within the standard negative polarity environments. These lexi
cal characteristics thus explain why adjectival-quantifiers cannot quantify at a
distance.
It was also demonstrated that the quantifier benben can receive several inter
pretations in the grammar of QF and that its interpretation as a negative polarity
item always depends on a trigger, as also observed by Cheng (1991) for the
various interpretations of WH-words in Mandarin Chinese. On the other hand,
adverbial quantifiers like beaucoup in SF are not interpreted as negative polarity
items depending on a trigger. They are rather independent items and therefore
they are not focus sensitive or polarity sensitive items.
Secondly, since benben and other quantifiers of the same type appear in a
negative polarity environment, it was indicated how they obey principles of UG,
such as locality conditions, c-command and the notion of syntactic domain.
Furthermore, it was shown that such Qs simply appear in various A positions in
the structure, as proposed by Chomsky (1993) and Cinque (1994). All of these
principles allow one to identify invariants in the model, i.e., elements which are
considered part of conceptual necessity for developing any human language.
226 MARIE-THÉRÈSE VINET
REFERENCES
Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. PhD dis
sertation, MIT.
Baker, Leroy . 1970. "Double negatives". Linguistic Inquiry 1,2.169-186.
Baker, Mark. 1988. Incorporation: a Theory of Grammatical Function Changing.
University of Chicago Press.
Cheng, Lisa. 1991. On the Typology of WH-Questions. PhD dissertation, MIT.
Cellar, Jacques. 1991. Dictionnaire du français non conventionnel. Paris:
Hachette.
Chomsky, Noam. 1993. "A Minimalist Program for Linguistic Theory". The
View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger,
ed. by Kenneth Hale & Samuel Jay Keyser, 1-52. Cambridge: MIT Press.
. 1994. "Bare Phrase Structure". MIT Occasional Papers in Linguistics 5,
[Published as Chomsky, Noam. 1995. "Bare Phrase Structure". Government
and Binding Theory and the Minimalist Program, ed. by Gert Webelhuth, 381-
439. Oxford and Cambridge, Mass.: Blackwell]
Cinque, Guglielmo. 1993a. "A Null Theory of Phrase and Compound Stress".
Linguistic Inquiry 24,2.239-298.
. 1993b. 'On the Evidence for Partial N Movement in the Romance DP".
ms., University of Venice.
—. 1994. "Romance Past Participle Movement and Clause Structure". Paper
read at "Table-ronde: linguistique comparée des langues et dialectes maro
cains", Rabat University, Marrakesh, Morocco, February 3-6, 1994.
Di Sciullo, Anna-Maria & Mireille Tremblay. 1993. "Négation et interfaces".
Toronto Working Papers in Linguistics 12.1, ed. by Carrie Dyck, 75-89.
Toronto: Linguistics Graduate Students Union.
Doetjes, Jenny. 1995. "Quantification at a Distance and Event Relatedness".
Linguistics in the Netherlands 1994, ed. by Reineke Bok-Bennema & Crit
Cremers, 13-24. Amsterdam & Philadelphia: John Benjamins Ltd.
Emonds, Joseph. 1985. A Unified Theory of Syntactic Categories. Dordrecht:
Foris.
Fauconnier, Gilles. 1980. Étude de certains aspects logiques et grammaticaux de
la quantification et de l' anaphore en français et en anglais. Presses de l'Uni
versité de Lille.
Horn, Laurence R. 1978. "Remarks on Neg-Raising". Syntax and Semantics 9:
Pragmatics, ed. by Peter Cole, 129-220. New York: Academic Press.
Jakubowicz, Célia. 1994. "On the Morphological Specification of Reflexives:
Implications for Acquisition". Proceedings of NELS 24, 205-219. Graduate
Students Linguistic Association: University of Massachusetts at Amherst.
Kayne, Richard. 1981. "ECP Extensions". Linguistic Inquiry 12,1.93-133.
Klima, Edward S. 1964. "Negation in English". The Structure of Language:
Readings in the Philosophy of Language, ed. by Jerry Fodor & Jerrold Katz,
246-323. Englewood Cliffs, New Jersey: Prentice-Hall.
Lasnik, Howard. 1993. "Lectures on Minimalist Syntax". University of
Connecticut Occasional Papers in Linguistics 1.
ADVERBIAL QUANTIFIERS AND DIALECT VARIATION 227
1. Introduction
On the basis of the word order facts in (1), the Continental West Germanic
languages have been classified as OV-languages in the tradition of generative
syntax (Bach 1962, Koster 1975). The word order in (la) and (lc) was consid
ered to reflect the deep structure ordering of the meaningful elements, the SVO
order of main clauses being derived by movement of the finite verb to the second
position in the sentence (see also den Besten 1983). This analysis presupposes a
phrase structural split dividing the Germanic languages into English and North
Germanic on the one side, and Continental West Germanic on the other. The
split was related to the position of the head in the phrase: North Germanic and
English would have the structure in (2), Continental West Germanic the one in
(3):
Recently, however, it has been argued that phrases in all Germanic lan
guages, including Continental West Germanic, should be construed as in (2)
(Kayne 1994, Zwart 1994). This presupposes that the word order in (la) and
(lc) does not immediately reflect the deep structure ordering of the meaningful
elements. Rather, (la) and (lc) may be derived by movement of the object noun
phrase to the left, as demonstrated by Vanden Wyngaerd (1989). This does not
affect the explanation for the SVO word order in main clauses, which is derived
from the embedded clause word order by the verb movement identified by
Koster (1975) and den Besten (1983).
Immediate evidence for the object noun phrase movement to the left is pre
sented by the phenomenon that the object noun phrase and the verb in OV-con-
structions need not be adjacent:
(5) a. (Hij denkt) dat ik wil dat hij het boek leest
He thinks that I want that he the book reads
"He thinks that I want him to read the book."
b. * (Hij denkt) dat ik dat hij het boek leest wil
c. (Je moet) willen dat hij het boek leest
You must want that he the book reads
"You've got to want him to read the book."
d. * (Je moet) dat hij het boek leest willen
In Zwart (1994), it is argued that the word order in (5a) and (5c) does reflect
the basic ordering of the meaningful elements in the Dutch VP. This is because
the factor triggering noun phrase movement in (1) and (4)—Case assignment by
assumption (see Vanden Wyngaerd 1989)—has no effect on clauses (which are
not assigned Case). This deviates from the traditional analysis of (5), in which
the clause is assumed to move to the right, by a process called extraposition
(Reuland 1981). Extraposition, however, has a number of curious properties,
leading Kayne (1994) to conclude that this movement process actually does not
exist.
In this paper, we will discuss the remaining type of verbal complementation
to be studied in connection with the position of the V in the verb phrase in
Continental West Germanic: complementation by a verbal or infinitival con
stituent. This type of complementation typically gives rise to the construction of
232 JAN-WOUTER ZWART
verbal clusters in Continental West Germanic (cf. Evers 1975). The order of the
verbs in the verbal cluster shows a bewildering variation across Continental
West Germanic dialects. It will be argued that this variation can best be analyzed
by assuming that all phrases involved are structured as in (2).
Having established this, the paper explores the possibilities of participle
placement in Continental West Germanic, based on the analysis of auxiliary
constructions as possessive constructions proposed in Kayne (1993).
The Continental West Germanic dialects show a large variety of verb clus
ters, especially in embedded clauses, where the verb movement which puts the
finite verb in the second position does not apply. The most elementary cases are
those in which an auxiliary verb has a past participle in its complement domain,
or when a modal, causative, or perception verb has an infinitive in its comple
ment domain. In these situations, clusters are created that consist of two verbs
only. More complex clusters arise by iteration of the processes that give rise to
these simple clusters. (The infinitives in the complement domain of raising and
control verbs, generally marked by a prefix cognate of EngUsh to, do not appear
to give rise to cluster formation, and will be kept out of the discussion.)
Examples of simple clusters are given in (6)-(8), from Standard Dutch. The
organization of the clusters is represented numerically on the extreme right,
where a verb β originating in the complement domain of a verb α gets a higher
number than α (so if α = 1, β = 2 or higher):
(7) a. (Ik denk) dat Jan het boek kan lezen 1-2
I think that John the book can read-INF
"I think that John is capable of reading the book."
b. (Ik denk) dat Jan het boek lezen kan 2-1
The 1-2 order in (6a) is prominent in written Dutch, whereas the 2-1 order in
(6b) is prominent in spoken Dutch (Stroop 1970). In (7) and (8), the 1-2 order
is slightly favored in both written and spoken Dutch, but the 2-1 order is not im
possible.
Standard Dutch is presumably a composite of several systems present in the
various dialects of Dutch (Stroop 1970). The freedom of word order in the verb
clusters in (6)-(8) is generally absent from the dialects. Nevertheless, both the 1-
2 order and the 2-1 order are represented in the dialects, so that (6)-(8) is a
proper representation of the word order possibilities in Continental West
Germanic verb clusters in general.
Looking at the dialects of Dutch, it appears that in the Southern dialects, the
participle has a tendency to precede the auxiliary, as in (6b) (Verhasselt
1961:153, Vanacker 1969, but cf. Shepherd 1946:61 on the dialect of
Maastricht, who gives examples of the 1-2 order only). This tendency is less
strong in the dialects spoken in the Eastern part of the Netherlands (Stroop
1970:250), whereas the dialects spoken in the North appear to employ the 2-1
order exclusively. The latter is also true of Frisian, High German, and the di
alects of German, as far as I have been able to ascertain (Bruch 1973:93 men
tions some Luxemburgish vestiges of the 1-2 order, which was a possibility of
Middle High German).
The 2-1 order in (7)-(8), where the finite verb has an infinitive in its com
plement domain, is used exclusively in High German, Frisian, and the dialects
spoken in the North of the Netherlands. However, many German dialects show
the 1-2 order in this case (Bruch 1973:94 on Luxemburgish, Baur 1988:157 and
Lötscher 1978:4 on Swiss German), which is also used very prominently in the
Dutch dialects spoken in Belgium, Limburg, and the dialects spoken in the West
and the South of the Netherlands. In the East of the Netherlands, a mixed situa
tion seems to exist (Stroop 1970:254).
In more complex verb clusters, tendencies tend to become rule. Thus, in
Standard Dutch verb clusters, the general word order becomes 1-2-3 (see (9)),
while in High German and Frisian the general word order becomes 3-2-1 (see
Frisian (10), from Tiersma 1985:139):
(9) a. (Ik denk) dat Jan het boek moet kunnen lezen 1-2-3
I think that John the book must can-INF read-INF
"I think John must be capable of reading the book."
b. * (Ik denk) dat Jan het boek lezen kunnen moet 3-2-1
234 JAN-WOUTER ZWART
(10) a. ... wêrom 'tik de hiele dei sitten bliuwe moatten ha 4-3-2-1
why that I the whole day sit-INF stay-INF must-PART
have-FIN(ITE)
"... why I have had to remain sitting all day."
ե *... wêrom 't ik de hiele dei ha moatten bliuwe sitten 1-2-3-4
But many exceptions exist. In Standard Dutch, for instance, when the sec
ond verb is an auxiliary, the participle in its complement may show up in three
positions:
(11) a. (Ik denk) dat Jan het boek moet hebben gelezen 1-2-3
I think that John the book must have-INF read-PART
"I think John must have read the book."
b. (Ik denk) dat Jan het boek moet gelezen hebben 1-3-2
c. (Ik denk) dat Jan het boek gelezen moet hebben 3-1-2
The general tendency is confirmed, however, in that the order of the modal
and the auxiliary in the complement domain of the modal is fixed:
(11) d. ?? (Ik denk) dat Jan het boek gelezen hebben moet. 3-2-1
e. * (Ik denk) dat Jan het boek hebben gelezen moet. 2-3-1
f. * (Ik denk) dat Jan het boek hebben moet gelezen. 2-1-3
In Frisian, only the 3-2-1 order is allowed in this case, whereas High German
allows both the 3-2-1 order and the 1-3-2 order.
In clusters of three verbs, the first of which is an auxiliary, the verb in its
immediate complement domain often takes the shape of an infinitive, instead of
the expected participial shape. This phenomenon, referred to as the infìnitivus
pro participio or IPP-phenomenon is present in all those dialects of Continental
West Germanic in which the auxiliary precedes its complement in the verbal
cluster (Hoekstra 1994; the IPP effect is also attested in clusters in which the
participle-turned-infinitive and its complement verb both precede the auxiliary
(2-3-1), as in (35); see also Lange 1981, Hoeksema 1988, Haegeman 1995, and
references cited there). (12a) is from Standard Dutch, (12b) from High German:
(12) a. (Ik denk) dat Jan het boek heeft kunnen/*gekund lezen 1-2-3
I think that John the book has can-INF/pART read-INF
"I think John could have read the book."
b. (Ich glaube) daß J. das Buch hat lesen können/*gekonnt 1-3-2
I think that John the book has read-iNF can-INF/PART
VERB CLUSTERS IN WEST GERMANIC DIALECTS 235
(Notice that the modal is in the complement of the auxiliary, in spite of what
the English translation suggests.) German (12b) again deviates from the ex
pected 3-2-1 pattern, the 1-2 order setting up the context for the IPP-effect. In the
3-2-1 order, no IPP-effect occurs (van Dam 1972:146; another order in which
the IPP-effect shows up in verbal clusters in German dialects is the 3-1-2 order,
as in kommen hat können"come-INFhas -INF", again displaying the 1-2 or
der):
(14) a. (Ik dacht) dat Jan het boek gelezen had kunnen hebben 4-1-2-3
I thought that John the book read-PART had can-INF have-INF
"I thought John could have read the book."
b. ... omdat zi'j et wel es daon hebben kund had 4-3-2-1
because she it PARTICLE PARTICLE done-PART
have-iNF could-PART had
"... because she may very well have done it."
(15) a. (Ik wist niet) dat Jan ontslagen was (?geworden) 3-1-2
I knew not that Johnfired-PARTwas become-PART
"I did not know John had been fired."
b. Grad wo է abgfaare gsy bisch... 3-2-1
just when you taken-off-PART ЄЄN-PART are-FIN
"Just after you had left."
236 JAN-WOUTER ZWART
These constructions are curious, in that in all dialects, no matter what the
preferred word order in the verbal cluster, the most deeply embedded participle
has to precede the hierarchically higher participle. Thus, whereas the placement
of the participle is particularly Uberal in Dutch, (16) is completely ungrammatical
(cf. (15a)):
(16) (Ik wist niet) dat Jan was (*geworden) ontslagen 1-2-3
I knew not that John was ЄMЄ-PART fired-PART
"I didn't know John had been fired."
Thus, it does not occur in Frisian and in the Northern Dutch dialects (though it
does in West Frisian, which behaves more like High German in this respect, cf.
Hoekstra 1994), and it shows up in High German only in the 1-3-2 cases
(again, the auxiliary is hierarchically superior to the modal, contrary to what the
English translation suggests):
(19) a. ... dat Jan heeft gelezen [het boek] (cf. 6)
that John has read-PART the book
b. ... dat Jan [het boek]i heeft gelezen ti
These elements, then, will not normally interfere with the verb movements
giving rise to verb clustering.
There are certain restrictions on the movement of the object and the embed
ded predicate in (19) and (20). For example, the movements must give rise to
crossing paths, rather than nesting paths:
(21 ) *... dat Jan [uit]j [het boek]i heeft gelezen ti tj (cf. 20)
Assuming, with Chomsky (1993), that the relevant movements are directed
to the specifier position of a functional projection, (21) can be excluded if the
functional projections have to be ordered in some way.
Such an ordering, however, in principle leaves open the possibility of gen
erating the relevant functional projections in various positions. This is not ex
cluded, as long as the result does not violate the restrictions on the ordering of
the functional projections involved.
More concretely, if the functional projection that hosts the object must pre
cede the functional projection that hosts the embedded predicate, it is not ex
cluded that the latter functional projection is generated somewhere between the
auxiliary and the participle. This would yield the word order in (22):
(22) ... dat Jan [het boek]i heeft [uit]j gelezen ti tj
we will argue that participles move as phrases to a specifier position. But, unlike
the Verb Projection Raising proposal, this does not increase the number of pos
sible movements, as the participle movement is of the well-known type of move
ment of a phrase to a specifier position.)
A further restriction on Verb Projection Raising is that the material breaking
up the cluster (such as uit in (22)) cannot be placed to the right of the verb which
it properly belongs to. Thus, in (22) uit "out (i.e., finished)" is the predicate of a
Small Clause het boek uit "the book out", occurring in the complement of the
verb gelezen "read", In this sense, uit belongs to gelezen (cf. Neeleman 1994,
who analyzes the combination of uit and gelezen as a complex predicate). We
know that gelezen may appear both to the right and to the left of the auxiliary
heeft (cf. (6)). But when gelezen appears to the left of heeft, so must uit:
(23) a. ... dat Jan [het boek]i [uit]j gelezenk heeft tk ti tj
that John the book out read-PART has
b. *... dat Jan [het boek]¡ gelezenk heeft [uit]j tk ti tj
c. *... dat Jan [het boek]i gelezenk [uitį heeft tk ti tj
In Standard Dutch, (23b) and (23c) contrast sharply with the marginally ac
ceptable (22). In Verb Projection Raising dialects like West Flemish and Swiss
German, examples like (23b) and (23c) are never found.
Apparently, the intervening material in Verb Projection Raising constructions
must be construed (in a pretheoretical sense) with the verb it belongs to.
'Construing with' is apparently sensitive to directionality, as is also suggested in
Kayne (1994). Below, we will provide a more technical description of
'construing with' in Verb Projection Raising constructions.
So far, we have seen that the VO-approach makes it unnecessary to refer to
the phrase structural status of the category moved (in the process of construing
of verb clusters). It does not seem to be the case that there is more going on than
object movement, predicate movement, and verb movement.
Let us next consider the question of the direction of adjunction. In the OV-
approach, the variation in (6), reflecting the range of possibilities in two-verb
clusters in Continental West Germanic, is analyzed as involving a choice be
tween adjunction to the right and adjunction to the left. As can be seen, there is
no way of keeping the direction of adjunction constant, even within a single lan
guage (e.g., Standard Dutch and High German).
As a first approximation, we could propose that the order in (6a) results
from there being no movement at all, while the order in (6b) is the result of ad
junction of the participle to the left of the auxiliary:
242 JAN-WOUTER ZWART
(24) a. ... [het boek]j... [VP heeft [ү gelezen tj]] = (6a)
b. ... [het boek] j ... [VP gelezeni-heeft [ү ti tj]] =(6b)
In other words, the variation in (6) is the result of the absence vs. presence
of a particular movement (or, in Chomsky's (1993) framework, of the timing of
the movement of the participle, taking place before Spell-Out in (24b), and after
Spell-Out in (24a)). Similarly, the variation in (7) and (8) could be taken to re
sult from the absence vs. presence of raising and adjunction of the infinitive to
the modal, perceptive, or causative verb.
This approach, which was taken in Zwart (1993), has the immediate advan
tage that no language or construction specific statements need to be made about
the direction of adjunction. It is also in accordance with Kayne's (1994) obser
vation that adjunction seems to always take place on the left.
A similar approach could not be as successful if we were to start from the
OV-structure in (3). In principle, the 2-1 orders in (6a), (7a), and (8a) could be
obtained by abstaining from verb movement (instead of raising of the embedded
verb and adjoining it to the left of the embedding verb):
(25) a. [het boek]j [ү [ү tj ti ] heeft-gelezeni ] = (6a)
b. [het boek]j [ү [VP tj gelezen ] heeft ] = (6b)
But this generates several problems. For one thing, the Verb Projection
Raising order heeft uit gelezen must be derived by raising a (part of a) verb
phrase, as discussed above. For another, it is unclear why the 2-1 cluster is im
penetrable in all Continental West Germanic dialects. Finally, this would lead to
problems if clusters with more than two verbs are taken into account.
Consider, for instance, the three-verb clusters in the sentences (11) from
Standard Dutch. Allowed are the orders 1-2-3, 1-3-2, and 3-1-2. Very marginal
is 3-2-1, and unacceptable are 2-3-1 and 2-1-3. From the VO-perspective, it
looks like the basic 1-2 order is fixed (i.e., the auxiliary always follows the
modal verb), while the 3 verb (the participle) can choose to not move at all,
move part way, or move all the way:
(26) . [VP moet [ү hebben [VP gelezen ]]] =(lla)
b. [ү moet [ү gelezenrhebben [VP ti ]]] =(11b)
c. [ү gelezenrmoet [ү hebben [VP ti ]]] = (11c)
If necessary, the 3-2-1 order of (11d) can be derived by adjoining the com
plex gelezen-hebben in (26b) to the left of the matrix verb moet:
VERB CLUSTERS IN WEST GERMANIC DIALECTS 243
This movement clearly deviates from the ones considered up to now. (11e)
appears to be excluded in Continental West Germanic, although the 2-3-1 order
does occur in several dialects when the 1 verb is an auxiliary (e.g., in West
Flemish zien gebeuren heęft [see happen-INFINITIVE has] "has seen happen"). The
process by which (26e) is derived, then, cannot entirely be excluded, but it does
not seem to apply in constructions in which the 1 verb is a modal verb (see
section 5 and Zwart 1995 for further discussion of this asymmetry between
modal verbs and auxiliaries). Finally, the 2-1-3 order can only be derived by
moving the middle verb out to the left:
joining the participle 3 to the left of the auxiliary 2, followed again by adjunction
of the participle-auxiliary cluster to therightof the modal verb 1 (27b). Finally,
the 3-1-2 order is derived by leaving the participle 3 in its place, while moving
the auxiliary 2 out to theright(27c):
(28) a. ... dat Jan het boek moet hebben uit gelezen
that John the book must-FIN have-INF out read-PART
"... that John must have finished the book."
b. ... dat Jan het boek moet uit hebben gelezen
As (28b) shows, the penetration of the cluster cannot be the result of some
complex uit-gelezen moving and adjoining to the auxiliary as a single participle.
Such an analysis is generally not feasible in Verb Projection Raising construc
tions, since the material breaking up the cluster may very well be a complete
phrase.
Returning to the question of how to derive the various orderings in (11)
from an OV-point of view, the 3-2-1 ordering of (11d) is unproblematic. It can
be derived by abstainingfrommovement entirely:
But of the remaining two orders, the first cannot be blocked without utter
stipulation. Thus, the 2-3-1 order of (11e) could be derived by adjoining the
participle 3 to the auxiliary 2, as in (27a), and by abstaining from further move
ment (27e). It is entirely unclear what blocks this process (apart from the stipu
lation that once the movement process has started, it has to go all the way). The
completely ungrammatical 2-1-3 order of (11f) is derivable by moving the par
ticiple 3 all the way to therightand adjoining it to the modal 1 (27f):
Given the possibility of adjunction to the left (as needed in the derivation of
(27b)), this derivation should in principle be allowed. Yet (11f) is hopelessly
ungrammatical in all dialects of Continental West Germanic.
In conclusion, the various word orders discussed here can be derived in a
simpler and more restrictive way if it is assumed that the verb phrase in
Continental West Germanic is structured as in (2).
The analysis of the Continental West Germanic verb clusters from a VO-
point of view, as presented in section 4, has one unsatisfactory aspect. The phe
nomenon that in certain dialects both the auxiliary-participle order and the par
ticiple-auxiliary order are possible is described in terms of optional movement of
the participle to the left. But the ungrammaticality of (11f) suggests that such
optionality in fact does not exist (see the discussion around (26f)). (As
Chomsky 1993 points out, optionality ought not to be part of a system of gram
mar. However, it cannot be excluded that the optionality in (6) in fact reflects a
limited form of bilingualism among speakers of the relevant dialect—perhaps not
accidentally a standard dialect. (11f), however, strongly suggests that there is
always participle movement in Continental West Germanic.)
If participle movement is optional, movement of the auxiliary to the left of
the modal should not interfere with the placement of the participle. Yet this is
what happens, as the following paradigm shows:
The movement of the object het boek in (32) (in fact, a Small Clause subject
in (32c)) and of the predicate uit in (32c) takes place under the conditions dis
cussed in section 4. The landing site of these elements is generally located to the
left of the BE-OF composite have. (We will return to Verb Projection Raising
below.)
The representation in (32b) has the participle gelezen inside the VP embedded
under OF. But, as we have seen, the participle must also move to the left. When
we zoom in on the construction in (32b), we find that the structure must be
something like (33):
(33) ... [het boek]i [VP-- BE [VP- OF [AGRP [VP gelezen ti ]]]]
(34) a. ... [het boek]i [VP-- BE [VP- OF [AGRP [VP gelezen ti ]]]]
b. ... [het boek] i [VP g e l e z e n BE [VP - OF [ AGR p[ VP tj tį ]]]]
Assuming now that BE +OF (=have) is spelled out in the position of BE, (34)
immediately yields the two possible word orders in (6):
phrase as well, moving around the auxiliary to what must be analyzed as a spec
ifier position. Generalizing this result, it must be the case that auxiliaries license
the verbs in their complement in a specifier position.
(This might lead to the conclusion that what is moved in (34) is actually not
the participle itself, but the VP containing the participle as a whole. This, how
ever leads to a problem in cases where the participle has a clausal complement.
Contrary to what one would expect under the scenario under consideration here,
the clausal complement does not appear to the left of the auxiliary, as pointed out
to me by Daniel Biiring (p.c.). I will not discuss this issue here, leaving the
phrase structure status of the participle in (34) open.)
The observation that participles move to a specifier position is not easily ac
commodated under the assumption that the VP in Continental West Germanic is
structured as in (3). Since specifiers do not appear to the right, participle-final
orders can only be described as the result of head-adjunction under that ap
proach. As a further disadvantage of the OV-approach, it should be mentioned
that there is no way of relating the optionality of participle placement to the pos
sessive structure of auxiliary constructions as argued for in Kayne (1993). To be
more precise, movement to each of the two participle constructions would al
ways yield a participle-auxiliary order, as (36) shows:
(37) a. .., dat Jan het boek gelezen moet hebben = (11c)
that John the book read-PART must-FIN have-INF
"... that John must have read the book."
b. ?... dat Jan het boek moet gelezen kunnen hebben 1-4-2-3
that John the book must-FiN read-PART can-INF have-INF
"... that John must have been able to read the book."
The 1-2-3 and 1-3-2 orders of (11a-b), not repeated here, can now be de
scribed as the immediate result of the participle movement illustrated in (34). The
250 JAN-WOUTER ZWART
participle occupies either the lower or the higher specifier position associated
with the composite have. But this leaves the 3-1-2 order in (37a)/(llc) unac
counted for. The same can be said about the 3-2-1 order in (11d), marginal in
Dutch, but grammatical in several Continental West Germanic dialects (High
German, among others).
In a pretheoretic sense, we could say that the modal verb moet in (37a)
'takes over' the licensing of the participle from the auxiliary hebben. Similarly
for kunnen in (37b). A first approximation therefore could be to assume that in
(37) the participle is licensed in the specifier position associated with the modal
verb.
This analysis cannot work if the modal verb itself needs the specifier posi
tion involved to license the verb in its immediate complement domain (i.e.,
hebben in (37)). However, it is not clear whether the infinitive in the comple
ment domain of the modal is licensed by movement to a specifier position or by
head-adjunction to the modal. Constructions like the ones in (35), showing that
participles are licensed in a specifier position, are typically absent from
Continental West Germanic dialects when the 1 verb is a modal verb (as far as I
have been able to ascertain). (See Zwart 1995 for more discussion of the differ
ence between verb clusters involving modal verbs and verb clusters involving
auxiliaries, in particular in relation to the Stellingwerf dialect.)
Moreover, on the basis of the pair in (38), one would expect to find the pah
in (39) also, if the complement of the modal were to move as a phrase:
But the 2-3-1 order in (39b) is apparently only found where 1 is an auxiliary
instead of a modal. This leads me to conclude that infinitives, if they move, ad
join to a head, instead of moving to a specifier position.
Following a suggestion by Eric Hoekstra, I will assume that a modal may
'take over' from an infinitival auxiliary (i.e., license the participle in its specifier
position) on the basis of the licensing relation that exists between the modal and
the infinitive. Even if the infinitive does not move to the modal overtly, as in
(37), the licensing relation between the two verbs must be said to exist. In terms
VERB CLUSTERS IN WEST GERMANIC DIALECTS 251
of Chomsky (1993), we may assume that the infinitive adjoins to the modal in
the hidden component of syntax LF. In a more representational approach, as ad
vocated by Groat & O'Neil (1994), we may assume that the infinitive has in fact
adjoined to the modal in (37), leaving a copy in its original position behind. The
language may then choose which of the copies of the infinitive to spell out:
Spelling out the higher copy yields the High German order 3-2-1, spelling
out the lower copy yields the Standard Dutch order 3-1-2.
We may now assume that the auxiliary hebben transfers its capacity to li
cense the participle in a specifier position to its sister in the adjunction configu
ration in (40). In this way, the modal can take over from the auxiliary. The op
tionality of this process (attested by the possibility of the various word orders in
(lla-)) remains unaccounted for under this approach.
This analysis of the word order possibilities in (37) carries over to the Dutch
constructions with several infinitivals, where the participle is allowed to appear
to the extreme left of the cluster:
(41) ... dat Jan het boek gelezen zou moeten kunnen hebben 5-1-2-3-4
that John the book read-PART should-FIN must-INF
can-INF have-INF
"... that John should have been able to read the book."
(11) f. *... dat Jan het boek hebben moet gelezen 2-1-3
that John the book have-INF must-FIN read-PART
252 JAN-WOUTER ZWART
So far, we have been studying the range of possible word orders across
Continental West Germanic. In this concluding section, I would like to concen
trate on a particular area where Continental West Germanic dialects are spoken,
to see whether there is a system in the actual patterns of word order that corre
sponds to a particular group of speakers or dialects.
The area under consideration is the territory of the Netherlands, and the data
and generalizations are derived from Stroop (1970). I will limit myself to con-
tructions involving a participle, an auxiliary, and zero or more modal verbs.
Stroop (1970:264) identifies three systems in the verb clusters used among
the speakers of Netherlandic dialects. The relevant examples are given below
(the auxiliary worden is comparable to hebben in relevant respects; in particular,
I assume that worden is a composite of a directional and a stative head, as in the
paraphrase 'come to be'):
(44) System I
a. ... dat hij gehaald werd 2-1
that he fetched-PART became
"... that he was fetched."
254 JAN-WOUTER ZWART
(45) System
a. ... dat hij gehaald werd 2-1
that he fetched-PART became
b. ... dat hij gehaald moest worden 3-1 -2
that he fetched-PART must-FIΝ become-INF
... dat hij gehaald zou moeten worden 4-1 -2-3
that he fetched-PART should-FIΝ must-INF become-INF
(46) System
a. ... dat hij werd gehaald 1-2
that he became fetched-PART
b. ... dat hij moest gehaald worden or worden gehaald 1-3-2 or 1-2-3
that he must-FIN fetched-PART become-INF or become fetched
could optionally be realized as gehaald werd (2-1). The absence of this possibil
ity in Stroop's system ΠΙ remains a subject for further study.
7. Conclusion
1
This paper could not have been written without previous exposure to presentations and
publications on the subject of verb clusters in Continental West Germanic by many Dutch,
Belgian and German colleagues. I would like to mention in this context (without intending to
leave anyone out) Hans den Besten, Hans Broekhuis, Marcel den Dikken, Arnold Evers, Liliane
Haegeman, Eric Hoekstra, Jan Koster, Karen Lattewitz, Eric Reuland, Henk van Riemsdijk,
Jean Rutten, and Manuela Schönenberger. For this paper in particular, comments and data from
Liliane Haegeman, Eric Hoekstra, Karen Lattewitz, and an anonymous reviewer for the APLA
Conference Proceedings have been extremely helpful. I also would like to thank the audiences
at the APLA conference on Microparametric Syntax and Dialect Variation, at the 10th
Comparative Germanic Syntax Workshop in Brussels (January 1995), and at the linguistics
colloquia at the University of Cologne, the University of Frankfurt, and the University of
Groningen. The author is supported by a grant from the Netherlands Organization for Scientific
Research, NWO.
256 JAN-WOUTER ZWART
REFERENCES
Bach, Emmon. 1962. "The Order of Elements in a Transformational Grammar of
German". Language 38.263-269.
Baker, Mark 1988. Incorporation: A Theory of Grammatical Function
Changing. Chicago: The University of Chicago Press.
Baur, Arthur. 1988. Schwyzertüütsch. Winterthur: Gemsberg Verlag.
Benveniste, Emile. 1966. "Être et avoir dans leurs fonctions linguistiques".
Problèmes de linguistique générale 1, ed. by Emile Benveniste, 187-222. Paris:
Gallimard.
Besten, Hans den. 1983. "On the Interaction of Root Transformations and Lexical
Deletive Rules". On the Formal Syntax of the Westgermania, ed. by W.
Abraham, 47-131. Amsterdam: John Benjamins.
& Jerold Edmondson. 1983. "The Verbal Complex in Continental West
Germanic". On the Formal Syntax of the Westgermania, ed. by W. Abraham,
155-216. Amsterdam: John Benjamins.
Bloemhoff, Hein. 1979. "Heranalyse van een Stellingwerver oppervlaktestruc
tuur". Us Wurk 28.31-38.
Bruch, Robert. 1973. Précis populaire de grammaire luxembourgeoise.
Luxembourg: Éditions de la Section de Linguistique de l'Institut gr.-d.
Chomsky, Noam. 1993. "A Minimalist Program for Linguistic Theory". The
View from Building 20. Essays in Linguistics in Honor of Sylvain Bromberger,
ed. by Ken Hale and Samuel J. Keyser, 1-52. Cambridge: MIT Press.
Dam, Jan van. 1972. Syntax der deutschen Sprache. Groningen: Wolters-
Noordhoff.
Dikken, Marcel den. 1994. "Minimalist Verb (Projection) Raising". Minimalism
and Kayne's Asymmetry Hypothesis. Groninger Arbeiten zur germanistischen
Linguistik 37, ed. by C. Jan-Wouter Zwart, 71-88. Department of German,
University of Groningen.
& Eric Hoekstra. 1995. "Parasitic Participles". Paper presented at the 10th
Comparative Germanic Syntax Workshop, Brussels, January 18.
Evers, Arnold. 1975. The Transformational Cycle in Dutch and German. PhD dis
sertation, University of Utrecht.
Freeze, Ray 1992. "Existentials and Other Locatives". Language 68.553-595.
Groat, Erich & John O'Neil. 1994. "Spell-out at the LF-Interface". ms., Harvard
University.
Haegeman, Liliane. 1994. "Verb Raising as Verb Projection Raising: Some
Empirical Problems". Linguistic Inquiry 25.509-522.
. 1995. "IPP Constructions and Verb Movement in West Flemish", ms.,
University of Geneva.
& Henk van Riemsdijk. 1986. "Verb Projection Raising, Scope, and the
Typology of Rules Affecting Verbs". Linguistic Inquiry 17.417-466.
Hoeksema, Jack. 1988. "A Constraint on Governors in the West Germanic Verb
Ouster". Morphology and Modularity, ed. by Martin Everaert, Arnold Evers &
Mieke Trommelen, 147-161. Dordrecht: Foris.
VERB CLUSTERS IN WEST GERMANIC DIALECTS 257
WH .
WH-chains 148 X°-chain 31
WH-criterion 87
WH-extraction 26, 28, 148 Y.
WH-movement 26, 27, 28, 31, 32, 35, yes-no questions 43, 56, 62, 64, 68, 69,
148, 152, 159 73, 76, 77
WH-phrase xiii, 28, 29 yes-no questions in Chinese 41
WH-questions 76 yes-no question types in Mandarin 68
whole-verb inversion dialect 83 Yue-Hashimoto, Anne 43, 47
WILL 174
Williams, Edwin 2, 5, 13 Z.
word order Zhang, Min 51, 52
Germanic 230 Zribi-Hertz, Anne .
Wright, Joseph 181 Zubizarreta, María-Luisa 2, 6
— & Elizabeth Mary Wright 178 Zwart, Jan-Wouter 28η., 230, 236, 242,
250