The unaccusativity puzzle: Explorations of the syntax-lexicon interface (review)

This collection of twelve papers on unaccusativity originates from a workshop on unaccusativity held in 1998 in Berlin. It includes contributions that were not presented at the workshop. The topics of discussion vary from lexical-semantic versus syntactic approaches to the phenomenon, unaccusativity diagnostics, thematic roles, and the importance of derivational morphology and phrasal constituency in relation to unaccusative predicates and to the possible triggers of unaccusatives in first- and second-language acquisition. All authors share the view that unaccusatives/ergatives are SpecVPs—that is, they have VP-internal arguments as subjects. The VP-external position, Spec vP, is reserved for unergative, agentive/causative subjects. Much, however, depends on how one analyzes the differentiae specificae among reflexives of different sorts, middles, anticausatives, and unaccusatives. What one needs to consider is that not all languages show morphologically, and even behaviorally, what the differences are. English, for one, does not signal middles or anticausatives by verbal morphology. Its behavioral inventory is limited: PP attribution to subject DPs is far from being a reliable diagnostic. And it has no distinguishing Aux signal to tell unaccusatives apart from intransitives, contrary to Italian, Dutch, and German (Abraham 1994).

The contributions revolve around three main topics: unaccusativity and reflexivization, unaccusativity and perfectivity, and internal and external (composite) unaccusatives. For unaccusativity and reflexivization, G ennaroC hierchia(‘A semantics for unaccusatives and its syntactic consequences’, 22–59) and D avidE mbick(‘Unaccusative syntax and verbal alternations’, 137–58) share a syncretistic view: If the morphology does not tell otherwise, functions such as reflexives, middles, and unaccusatives may have an identical, that is, vP-less syntax. Embick goes a decisive step further by assuming that such syncretism may arise through categorial underspecification. In other words, what appears in identical PF form may be different when spelled out structurally below word level, X ⁰—doubtlessly an important insight. It cannot be emphasized enough how crucial this notion of category underspecification is for a theory of, for example, diachronic grammaticalization and language evolution (Chomsky 1995:11). Such underspecification made syntax possible by combining words into multiple unit expressions, with in principle infinite recursion. It has been argued (Gil 2006) that early human language was isolating/analytic with no distinct lexical and grammatical categories. This follows if the initial evolutionary mutation was in fact Merge and if words are underlyingly un(der)specified roots that need syntactic configurations to function as verbs or nouns (cf. Halle & Marantz 1993, Embick 2004).

Chierchia finds himself in fundamental opposition to T anyaR einhartand T alS iloni(‘Against an unaccusative analysis of reflexives’, 159–80) and M arkusS teinbach(‘Unaccusatives and anticausatives in German’, 181–206). Reinhart and Siloni show for Romance and Hebrew that reflexives are true objects, which inevitably have a Spec vP-subject, thereby disrupting any constructional equivalence between eV (internal arguments as subjects) and reflexives (external arguments as subjects). It is tough to follow the authors in their assumption that, for one, eV principally derive from tV by reduction, that is, expletivization (166): Unaccusatives like die(causative kill), German sterben(causative töten), French mourir(causative tuer), Russian umirat’(causative ubivat’/ rezat’) never derive from their semantically (but not morphologically) derivable causative counterparts. And it appears to be at the core of the Perlmutter-Burzio assumption (see Perlmutter 1978, Burzio 1981, 1986) that unaccusatives have no direct transitive counterpart. Second, the argument against reflexive reduction for what is reduced already, as unaccusatives or passives, is empirically well founded. Reinhart and Siloni adduce Dutch, which disallows the impersonal passive of reflexive structures and, likewise, unaccusatives (169). Their refutation of any constructional parallelism is plausible. There is no plausible link between the agent condition and the reflexive coreference constraint to diathetic alternation.

There are seven papers that base their unaccusative analysis on aspectual considerations: A ngeliek vanH out’s ‘Unaccusativity as...