• Keine Ergebnisse gefunden

Teresa Parodi

2. On the phenomenon of clitic doubling

Among pronominal elements clitics present curious research problems, such as the cooccurrence of a clitic with a coreferent NP in clitic doubling (CD) and clitic-left dislocation (CLLD). This phenomenon raises questions both about the status of the clitic and of the NP.

One of the many questions clitics raise is whether they should be analysed as arguments or as agreement markers and, as such, half-way between pronouns and inflectional elements.

Traditionally in studies such as Kayne 1975 and Strozer 1976 among others, they have been treated as arguments although their behaviour differs from that of „normal arguments“ in various aspects: they have a fixed place in the sentence (preverbal with finite verbs, postverbal with nonfinite ones), while the placement of arguments is freer. As opposed to

1 The paper is a revised version of a talk given at the workshop “Syntactic and semantic aspects of specificity in Romance languages” in Konstanz in October 2002. I would like to thank the participants for helpful discussion.

The study is partly based on joint work with Ianthi Tsimpli for Greek and with Daniela Karadzovska for Macedonian. Both of them deserve thanks for long and fruitful cooperation.

arguments, clitics cannot combine freely with one another. Of special interest is the cooccurrence in a sentence of a clitic and a coreferent NP, difficult to explain if the clitic is the pronominal realisation of the corresponding argument.

Two lines of analysis are offered for the cooccurrence of clitic and NP, a movement and a base-generation analysis. The movement analysis is basically the line taken by Kayne 1975 and Strozer 1976; it assumes that clitics move from an argument position to the one in which they surface. The next question is where the coreferent NP is. Placing it in the trace left by the clitic would give raise to ungrammaticality. The other option is that the NP is extraposed, in a A´-position. However, assigning the same argument to both the clitic and the NP would violate the theta-criterion (Chomsky 1981).2

An alternative are base-generation analyses which take clitics to represent a separate category, basically as markers of an agreement relation with the verb, and not to form canonical object positions (Kaiser 1992, Kaiser & Meisel 1992, Suñer 1988, Franco 1993, 2000, Cummins & Roberge 1994, Torrego 1995, Sportiche 1996, Parodi 1998, for Greek Tsimpli 1996, Mise ka Tomi 2000 for Macedonian, among others). The type of category (a head, a feature, etc.) varies according to different authors. For Torrego (1995) and Uriagereka (1995) clitics are D-heads and the doubled element is in SpecDP. For Sportiche (1996) clitics head their own projections, in this case a Clitic-Voice projection.

2. 1 Spanish

In the current study I will follow Parodi (1998) and Franco (2000) who see clitics as instances of object agreement in a continuum which includes inflectional affixes on one end and pronouns on the other. As opposed to subject agreement, we do not observe agreement with the object across the board: object agreement follows certain conditions related to case and specificity. If we take Spanish, agreement with the dative is unrestricted, while there are stricter conditions for the accusative. These conditions are shown to be related to specificity and definiteness.

In Spanish if the element coreferent with the clitic is a pronoun, the presence of the clitic is obligatory, both in the dative and the accusative and in both varieties under consideration.

(1) a. accusative la veo a ella / *veo a ella (I) her see K her / *(I) see K her

‘I see her’

b. dative le doy la carta a él / *doy la carta a él (I) him give the letter K him / (I)give the letter K him

‘I give the letter to him’

2 Further counterarguments are the extraction properties shown by these constituents, as pointed out by Jaeggli (1986), Suñer (1988) and Franco (2000), such as subjacency effects (Jaeggli 1986) or the “unorthodox” starting point for wh-movement if the NP is in an A´-position (Suñer 1988).

Teresa Parodi 105 The cooccurrence of a clitic with a dative NP is optional in both dialects but, at least in Rio de la Plata Spanish, strongly preferred (see 2). The situation does not change when the NP is definite, as in (2a) and (2b) or indefinite as in (2c) and (2d). Its status as + or –specific (cf. 2a, 2c vs 2b, 2d) does not make any difference either.

(2) a. le doy la carta al vecino (I) him give the letter K the neighbour b. al primer vecino que aparezca

K the first neighbour who might turn up c. a un vecino que usa bastón

K a neighbour who uses a walking stick d. a un vecino

K a neighbour

Clitic doubling with an accusative NP is disallowed in European Spanish. It is, however, possible in Río de la Plata Spanish, provided the NP is definite and specific as in (3a), but not if the definite NP is non-specific (3b), or if the NP is indefinite, whether + or –specific (3c-d).3

(3) a. la veo a la mujer (RPlata ≠ European *) +definite

3accfs see.1s K the woman +specific

‘I see the woman’

b. *la busco a la mujer que sepa turco +def, -spec 3accfs search.1s K the woman who might know Turkish

‘I am looking for the woman who might know Turkish’

3 If the -definite NP has a generic reading, then the cooccurrence of clitic and NP is possible, both in doubling and in clitic left-dislocation. Cf. the example below.

te la roban enseguida una bicicleta cara

2dat.s 3acc.f.s steal.3plat once a bycicle expensive (fem.)

´They steal it (fem.) at once an expensive bycicle (fem.)´

c. *la veo a una mujer -def, +spec 3accfs see.1sK a woman

‘I see a woman’

d. *la busco a una mujer que sepa turco -def, -spec 3accfs search.1s K a woman who might know Turkish

‘I am looking for a woman who might know Turkish’

The degree of acceptability does not change according to animacy in either dialect, i.e.

animacy is not one of the conditions ruling doubling. Compare the examples under (4) with inanimate NPs in the accusative (4a-b) with the animate counterparts in (2).

(4) a. la veo la carta (RPlata ≠ European *) b. *la veo una carta

The conditions which guide the cooccurrence of a clitic and an NP are seen in Parodi (1998) as the interaction of a case and a specificity hierarchy. The first one is based on the mapping of thematic roles on syntactic functions following Larson (1988) as in (5) below:

(5) agent >recipient/goal >theme/patient

nominative >dative >accusative (indirect object)(direct object)

With respect to specificity I follow Comrie (1981), who establishes an “animacy hierarchy”

similar to that in (6) below. A literal reading of “animacy” is in this case inaccurate. In this scale animacy, definiteness and specificity all represent ways of identifying an entity. As Franco (2000:169) puts it, animacy is here a cover term that includes the notions of saliency and definiteness / referentiality.

(6) 1+2 pronoun > 3 pronoun>

+animate NP > -animate NP +definite NP > -definite NP +specific NP > -specific NP

One could ask at this stage whether the fact that semantics is playing a role speaks against the view of clitics as agreement markers. The answer is no, given that Comrie (1981) shows that many agreement relations are driven by an animacy hierarchy which holds cross-linguistically. Swahili and Macedonian, the latter very close to Río de la Plata Spanish in this respect, also exclude agreement with the lowest elements of the Animacy Hierarchy, i.e.

indefinite inanimate nouns.

We end up with some sort of parametric account of agreement depending on which elements of the scale are selected by different languages.

Teresa Parodi 107 2. 2 Clitic doubling in Greek

The conditions described above also apply to Greek. Greek is a language with rich overt morphology. NPs are marked for gender, number, case and definiteness, the latter by means of articles. Greek has unrestricted use of clitic doubling in the genitive, but again restricted to definites in the accusative, as shown in the examples under (7).

(7) a. accusative, +definite

I Eléni ta diábase ta biblía pou tis ícha dósi the Helen 3accnpl read.past the books that 3genfs had given.1sg

‘Helen read the books I gave her’

b. accusative, -definite

*O Kóstas to agórase chthes ena avtokíneto the Costas 3MS-acc bought yesterday a car

‘Costas bought a car yesterday’

2.3 Clitic doubling in Macedonian

Macedonian is a South Slavic language which makes extensive use of clitic doubling.

Macedonian has a definite article which is affixal, attached to the first element of an NP and not necessarily to the head noun (cf. 8). Articles in Macedonian carry deictic and agreement features (as opposed to English where it does not carry either). They are also marked for gender (masculine and feminine), for number (singular/plural) and for case on the pronouns (on the full and on the reduced form, i.e. the clitic). In Macedonian there is concord within the nominal. Furthermore Macedonian marks speaker/hearer orientation on the article (cf. 9).

(8) ja vidov kniga-ta I saw book-the

‘I saw the book’

(9) -ov, -va, -vo, -ve, -va (near the speaker) -ot, -ta, -to, -te, -ta (near the hearer)

-on, -na, -no, -ne, -na (far from both speaker and hearer)

Macedonian has unrestricted use of clitic doubling with the dative but restricted to definites in the accusative. These conditions are fairly similar to those in Río de la Plata Spanish. One difference is that in Macedonian clitic doubling is practically obligatory if possible, while it is slightly more restricted in Río de la Plata Spanish.

(10) Macedonian

a. accusative, +definite

Mira ja donese tetrakta-ta Mira it.acc brought notebook-the

‘Mira brought the notebook’

b. accusative, demonstrative

gi procitav vera ovie napisi 3accpl read.1s yesterday these articles

‘I read these articles yesterday’

c. accusative, proper name gi vidov Tome i Vera 3accpl saw Tome and Vera

‘I saw Tome and Vera’

d. dative, +definite

toj im isprati pismo na roditeli-te he 3datpl sent a letter to his parents