Retour vers liste

Détail de la contribution

Auteur: Byron AHN

Universality and Subject-Oriented Reflexivity

Abstract/Résumé: Subjects are well-known to play a crucial role in binding anaphoric reflexives. Some morphological markers of reflexivity are entirely restricted to clauses where the subject is the binder — Local Subject-Oriented Reflexivity (LSOR): e.g. Italian si (Rizzi 1986), Japanese zibunzisin (Katada 1991), Russian sebe (Timberlake 1979), Shona zvi (Storoshenko 2009). LSOR is predicted if reflexive elements (LF-)move to just below the (surface) subject (Chomsky 1986, Reuland 2011). This analysis raises a puzzle: why should *subjects* be relevant for binding at all? Moreover, there is a second independent puzzle: only some subjects are possible binders. Specifically, subjects of passives are incompatible with LSOR (Rizzi 1986, Sportiche 2010). To derive both of these puzzles, I propose that UG makes available two mechanisms for binding anaphoric reflexives. One mechanism is anaphoric binding in a more traditional sense, and the other (the focus of this paper) derives LSOR with a unique grammatical voice, Reflexive, (apart from Active or Passive) instantiated by a functional head: REFL Voice. The REFL Voice head sits outside the thematic domain (just like other grammatical voices; e.g. PASS, Harley 2012) and is endowed with an EPP feature that attracts reflexives. The reflexive interpretation is the result of the semantic function encoded by REFL, (1), which co-identifies two arguments. (It does not result from a valency-reducing function; even if such an operation were provided by UG, such a reflexivization function could not distinguish subject-binding and object-binding in a three-place predicate.) Therefore those arguments must occur in positions where they can compose with this REFL – for the subject, this is the normal phase edge, and for the reflexive this is VoiceP: (1) ⟦REFL⟧ = λPλxλyλe.IDENT(x,y) & P(e) (2) [ Taro [ zibunzisin REFL [ Taro zibunzisin keru ]]] PhaseP VoiceP θ-Domain Thus LSOR is a consequence of independent factors: semantic composition, height of VoiceP and movement of the subject through the phase edge. Since LSOR is derived by REFL (a unique Voice head), passive subjects are not the kind of subject that can bind in LSOR, since REFL and PASS Voice heads are in complementary distribution. This solves both puzzles. Even English provides additional evidence for this analysis of LSOR. Though its LSOR and non-LSOR reflexives are (segmentally) homophonous, a distinction between the two remains in their prosodic behaviors, and this distinction is derived by (1) and (2) (Ahn 2011). This indicates that subject-orientation is a universally core property of reflexivization, across all languages. Even when such a distinction is not entirely (morphologically) apparent, closer investigation can elucidate it. Subject-orientation in binding simply falls out from the general architecture of Language.