The Syntactic Structure of Palauan Resultatives

Embick (2004) has recently challenged the widely held view that English resultative participles (or adjectival passives) are formed in the lexicon while (verbal) passives are built syntactically (a proposal originally formalized in Wasow 1977; see also Bresnan 1982, Levin and Rappaport 1986, and many others). Instead, Embick argues for a syntactic analysis to explain why only certain arguments of a verb may serve as externalized arguments of a corresponding resultative, which must be stipulated on lexical analyses. I enrich the debate by examining the structure of resultatives in Palauan, an Austronesian language with about 15,000 speakers in Micronesia. Assuming a morphological framework with late insertion like Distributed Morphology (Halle and Marantz 1993), I concur with Embick and argue for a syntactic analysis of Palauan resultatives, though I advocate the syntactic structure in (1) for Palauan, which is distinct from that proposed for English resultatives (Embick 2004: 383, ex. 64). The Palauan literature describes resulting state verbs (RSVs) as “derived by taking the verb stem ... and inserting the infix -lor -elafter the stem-initial consonant” (Josephs 1997: 273), exemplified in (2).1 RSVs describe states and differ semantically from passives (which describe events), as shown in (3) (see Dubinsky and Simango 1996: 750, ex. 2 for a similar contrast in Chichewa). And like simple statives, RSVs are incompatible with the aspectual auxiliary mla (irrealis form bla), while passives co-occur freely with mla, also shown in (3). However, RSVs also differ from simple statives in at least two ways. First, RSVs may license manner modifiers describing the event that brings about the state, distinguishing them from simple statives, which may not (see Kratzer 2000: 392, ex. 16 for a similar contrast in German). Second, RSVs are compatible with temporal PPs denoting telic endpoints like in an hour, whereas simple statives are not (see Tenny 1987 for discussion). Both of these differences are shown in (4). Thus far, it seems that RSVs share the properties of resultatives in other languages. These contrasts among Palauan (verbal) passives, resultatives, and (simple) statives could well be explained by virtually any of the lexical or syntactic analyses cited above, but there are other properties of Palauan resultatives that raise questions. First, they may co-occur with PPs expressing what would have been the external argument of a corresponding transitive verb (unlike in English), as in (5) – (7). Second, this optional PP need not contain an agentive DP, as resultatives can be constructed from transitive statives such as merers “hold; contain” in (6) and from morphological causatives such as olsiich er a rengul x “make x proud” (lit. “tighten x’s heart”) in (7) (see also (4a)). Third, the predicate in (7b) is a resultative formed from a phrasal idiom, suggesting (i) that resultatives are phrasal (not lexical), and (ii) that the DP argument of the resultative is base-generated as a complement to the √ root. The emerging generalization is that Palauan resultatives have the distribution of aPs (simple statives) but the internal structure of vPs. Consequently, the structure in (1) contains a √ root that is verbalized to form a vP (much in the same way a passive vP is formed), and this vP then merges with a functional head a to form an adjectival constituent. The resultative a head may be construed as a “stativizer” with -(e)las its exponent (along the lines of Kratzer 2000: 391, ex. 14), which existentially quantifies the event argument of a vP that also contains a target state component, e.g., in (8) [for the interpretation of (2b) with the structure in (1)]. On this analysis, a can in principle merge freely with any vP in the syntax, but the derivation will crash at LF if the event doesn’t have a target state component, e.g., unaccusatives of achievement (*The stars are appeared.) or existence (*My plants are flourished.). Furthermore, if one assumes the Phase Impenetrability Condition, the head of the resultative vP cannot be a phase head (v* in Chomsky’s (2001: 43, fn. 8) terminology), or else the structure associated with the √ root will be Spelled Out before the stativizer a can access it. So resultative a can, in practice, only merge with a semantically and syntactically restricted subset of vPs to form syntactically complex resultatives.