Successive-cyclic case assignment in Amis: evidence from case-stacking

Tingchun Chen, MIT This paper proposes an analysis of Amis (Formosan, Austronesian; VSO) case alternations, implemented in a dependent case model (Marantz 1991), in which nominals crucially may receive case multiple times. Moreover, I argue that, contra some previous proposals, e.g. Levin (2016), the presence of a nominal in multiple phases is not a prerequisite for multiple case assignments, but focus movement is necessary for overt case-stacking. I first show that in Amis, case patterns of the first (vP) Spell-Out in main clauses parallel those in gerunds: the external argument (EA) receives genitive (GEN) in both. I argue that DPs headed by bare roots illustrate the consequences of this first Spell-Out most clearly. In imperfective main clauses, the second (CP) Spell-Out replaces GEN on the EA with nominative (NOM), masking an earlier case assignment. I argue that case-stacking, licensed only in contrastive focus contexts, supports the existence of these two rounds of case assignment, where realization of the unmarked case is sensitive to the category of the Spell-Out domain (nominal ⇒ GEN, verbal ⇒ NOM). Nominal roots Bare event/state roots, e.g. cefus ‘splash,’ parallel bare entity roots, e.g. lupas ‘peach,’ in distribution and case marking. First, Amis has a number of affixes commonly termed voice, e.g. mi-/pi‘actor voice (AV)’ (miappears in indicatives only). All voices can directly attach to either type of root, e.g. mi-cefus ‘splash,’ mi-lupas ‘pick peaches.’ Second, infixal plural reduplication can also apply to either type of root, yielding either a plurality of events or entities, e.g. cefus ‘splash repeatedly,’ lupas ‘peaches.’ Third, DPs headed by either type of root, as in (1), mark case in the same way: the higher DP (EA or possessor) receives GEN and the lower DP receives accusative (ACC). Moreover, (2a) shows that the transitive internal argument (IA) receives ACC even without an overt EA. This contrasts with the unaccusative IA in (2b), which always receives GEN. Following Baker & Vinokurova (2010), I will assume that a null argument (EA in (2a)) counts for the purpose of case competition. Last, gerunds (3) also show the same case pattern just described. This contrasts with imperfective main clauses (4), where the EA receives NOM instead of GEN. For reasons of space, I will not discuss perfectives, which show differential subject marking. (1) [BareRootDP U U cefus/cudad splash/book nu GEN tawki boss tu ACC epah wine ] ku NOM sakaketer reason.anger aku. GEN.1SG ‘The boss’ splashing water/ the boss’ book about wine is the reason why I am angry.’ (2) a. u U cefus splash *nu/tu *GEN/ACC epah wine ‘splashing wine’ (Transitive IA) b. u U leneng sink nu/*tu GEN/*ACC tamina’ boat ‘sinking (a) boat(s)’ (Unaccusative IA) (3) Faheka surprised kaku NOM.1SG [Gerund (tu)ya (ACC)that pi-faca’ AV-wash nu GEN tawki boss tu ACC riko’ clothes ].