Retour vers liste

Détail de la contribution


Labels, phases and VSO order

Abstract/Résumé: Chomsky (2013) notes that phrase-phrase merge creates a labeling paradox. If the label of a phrase is determined by the closest head it dominates, then α cannot be labeled in (1) because both X and Y are equidistant from it. (1) [α XP YP] Chomsky argues that the paradox disappears if either XP = YP or if either XP or YP are moved. In the latter case, the copy of the moved phrase is a segment of a discontinuous element and is invisible to the labeling algorithm LA. For example, if YP is moved then α receives the label of the closest (visible) head, X. Granting head movement, LA can succeed if either X or Y move. For example, if Y is moved, then the lower copy of Y is invisible to the LA. The closest head to α would be X and α is labeled X. If XP in (1) is the subject SUBJ and YP is vP, then raising of SUBJ would lead to α's acquiring the label v, a correct result. But if v raises and SUBJ doesn't, α would be labeled by (the head of) SUBJ. This can be prevented by moving SUBJ or its head. Since LA applies at the phase level, it can apply apply after SUBJ moves but before v does. α is then labeled v and subsequent movement of v cannot undo this. This leads to: SUBJ may not remain in vP if v raises. I then turn to the specific case of VSO lgs and argue from the distribution of the Berber construct state that SUBJ raises out of vP and merges with a head below the position targeted by v. In Berber, the morphological shape of a noun is determined by the syntactic context in which it appears, (Guerssel 1992). Subjects and complements of prepositions appear in the construct state (CS). Otherwise, nouns display the free state. What is puzzling here is that the syntactic environments for CS do not appear to constitute a natural classe. Generalizing Kayne (2004), assume that P is externally-merged in the functional space above vP and attracts a nominal to the specifier of a lower, 'twinned' head W. This is the environment for CS on complements of P. Incorporating Guerssel's (1992) suggestion that Berber subjects are preceded by a null P, I argue that CS is simply the shape of a noun merged with W. Subjects must display CS because they must vacate vP, to permit proper labeling of α. In Kabyle Berber, OBJ can be clitic-doubled. In this context, it displays CS. Suppose that a clitic-doubled OBJ is internally-merged to v. To permit proper labeling, OBJ, like nSUBJ must raise and merge with W. Here, the P licensing a W which attracts OBJ is also null. In many languages, however, clitic-doubled OBJ must be preceded by an (expletive) P – Kayne's Generalization. One should, in the spirit of this analysis, consider the role of e.g., Spanish "a" as an overt selector of W, to be the grammar's response to the labeling paradox which arises when OBJ merges with vP.