Reconstructing the nonconfigurational verbal system
This section presents and then reconstructs the *t-V-ce system, which has no apparent VP constituent. The *t-V-ce verbal system occurs in at least five Cariban languages (Gildea 1997b) and potentially in several more (Gildea 1998 [ch. 13]); it is characterized by the verb bearing some modern reflex of the prefix *t- and the suffix*-ce, functioning in the modern system as an ambifix, usually indicating past tense and/or completive/perfective aspect. There is no verbal person marking, but if an auxiliary occurs, it agrees with the absolutive; the absolutive nominals are unmarked and the A bears an ergative suffix. There is no evidence for any constituency relationship between any nominal arguments and the verb, although in Carib of Surinam, the verb must precede the auxiliary. While all orders of nominal arguments and verb can be elicited, and all are attested in text, the most frequent orders are OVA and VS (Tavares 1999 for Wayana, Meira 1999 for Tiriyó). The following examples from Wayana texts (Tavares p.c.) are illustrative:
Hoff (1995:358) reports: “The participle [t-V-se– SG] must precede the auxiliary verb. Otherwise, constituent order is free…”. In illustrative examples taken from texts (20)–(21), we see two cases of intransitive V with no S, plus transitive verbs with the very unusual orders (for Carib) VOA and OAV:
Koehn and Koehn (1986) do not specifically address word order for the cognate construction in Apalaí, but their examples show A, S and O in both preverbal and postverbal position. Meira (1999:536,552–3) reports similar word order freedom in the Tiriyó cognate construction, and in my elicitation field notes I have recorded parallel freedom for Kaxuyana and Akuriyó. This freedom differs markedly from the patterns seen until now, and thus must reconstruct to a distinct source.
In every Cariban language described to date, a stative/passive participle (very like a verbal adjective) can be derived from a verb by means of an ambifix which is a modern reflex of Proto-Carib *t-V-ce (Gildea 1998:140–51). The semantics of this participle vary between potential (22) and past (23); the morphosyntactic properties are perfectly consistent in that it is never possessed, and thus never forms a tight syntactic bond with a noun, either as an attributive nominal modifier or as the complement of a copula. In some languages the participle has no means of expressing the notional A of the verb (e.g. Hixkary ana, Derbyshire 1985:95), but in others, an oblique agent phrase can co-occur (e.g. Tiriyó, Meira 1999:334).
When this participial form is reanalyzed into a main clause verb, it cannot bring with it any constituent structure to be reanalyzed as a VP. The t-V-ce verbal system is transparently derived etymologically from a source construction in which the t-V-ce participial form was the complement of a copular clause; this then entered into the well-known evolutionary path of participle to passive to ergative (Figure 7). The most detailed look at the entire evolutionary scenario is presented in Gildea (1997b), with additional Cariban languages added to the database in Gildea (1998 [Chapter 13]).
The intransitive participle simply follows the path of the French Passé Composé, going directly to a completive/perfective reading without being subjected to the interplay of voice like the transitive participle.
The lack of configurational relationships in the t-V-ce verbal system follows directly from the lack of such relationships in the source: the A phrase arrives on the scene as an oblique adjunct, formulated by analogy to the agent phrase adjuncts found in other nonfinite forms and thereby sharing in their lack of constituency with the nonfinite verb form. In all the languages where this reanalysis has become prominent, the lack of a syntactically restricted order of S and Copula means the O/S of the new ergative clause type similarly has no syntactically restricted order vis-à-vis the verb or auxiliary.20 Thus, the word order preferences indicated in Carib (Hoff 1995), Wayana (Tavares 1999) and Tiriyó (Meira 1999) are nothing more than preferences, presumably intended to indicate something about the pragmatic status of the various parts of the sentence.
