Syntactic criteria in Split-S Systems

Syntactic approaches to split intransitivity suppose distinct representations for the arguments of intransitive verbs. For instance, in the Relational Grammar framework, SA (‘unergative’) verbs have both an initial and a terminal 1 (subject), while SO (‘unaccusative’) verbs have an initial 2 (direct object) and a terminal 1; this is known as the ‘Unaccusative Hypothesis’. Though Perlmutter 1978 admitted a possible semantic basis for class membership, subsequent works (e.g. Harris 1982, Rosen 1984) have denied the existence of any consistent semantic rationale for it. Being unergative or unaccusative is thus an idiosyncratic characteristic of each verb, which must be specified in the lexicon.

Government-Binding approaches have basically adopted the same idea, though with a different formalism: they suppose that SA (‘unergative’) verbs only occur in D-structures in which the only argument NP is outside of the VP node, whereas SO (‘ergative’) verbs can only fit in D-structures in which the only argument NP is inside the VP node (Burzio 1981). Again, the specific D-structure in which a certain intransitive verb can occur is semantically unpredictable, and must be specified for each verb in the lexicon. Therefore, for both theories, the class membership of a specific intransitive verb is one of its idiosyncratic characteristics, much like its phonetic shape, which must be independently memorized by the speakers.

Since, as we have seen in the previous section, semantic factors do not seem to correlate with class membership in Tiriyó, it may be preferable to ‘give up’ and consider that each Tiriyó verb has as part of its lexical features the specification ‘+SA’ or ‘-SA’. In the absence of any semantic criteria to identify a verb as being SA or SO, this seems to be the simplest analysis.

 However, such an analysis would presuppose that there is nothing else to be said about the two intransitive verb categories: they would seem to be just as arbitrary as e.g. verb conjugations in Romance languages. Interestingly enough, there is something else to be said about this system in Cariban languages.