The Unresolved Problem of the Order of the Nominative Subject
Like O in clauses with SAP A, the order of the Set I nominative subject is not conveniently linked to reconstructible morphology. For this paper, the important fact about all the order studies in the family (reviewed in Section 1) is that they document a great deal of synchronic variation in the order of the subject: in every single language, pre-VP and post-VP orders are both allowed and zero subjects are common. This indicates that we are dealing with a syntactically free argument. Unlike the case of the syntactically free SAP O pronouns with 3A (Section 4.2), we have no morphological clues to a previous constituency status for the subject argument and as such, we are forced to reconstruct a syntactically free subject at least as far back as Proto-Carib.
This brings up a slightly different issue in reconstruction of syntax: to the extent that basic word order is taken to be constituent structure, we can go no farther. But if we accept the notion that some syntactic notion of basic word order can be determined by either pragmatic markedness interpretation, or by discourse frequency of given orders, we might still be able to determine a basic (i.e. most frequent) order for the subject vis-à-vis the OV unit, and in that case the question might arise as to whether we could reconstruct such a “basic” order to Proto-Carib. My answer is no, that a basic order which is derived from frequency counts cannot be reconstructed for any proto-language because there will never be either a speaker of the proto-language to give pragmatic markedness judgements nor texts in the proto-language that allow one to count order frequency.
Derbyshire (1985) suggests that order in subordinate clauses preserves facts about order in older main clauses. However, as I have noted before (Gildea 1992:56–8; 1998:47–49), while dependent clauses are more conservative than main clauses, what they conserve in Proto-Carib is clearly not main clause morphosyntax, but dependent clause morphosyntax — that is, reconstructions show convincingly that main clause morphosyntax in Proto-Carib is almost completely unrelated to subordinate clause morphosyntax (cf. also sections 2–3). In any event, preposed versus postposed order of the oblique A vis-à-vis the Abs V unit in nominalized subordinate clauses is every bit as open to question as order of the A in main clauses (cf. discussion above in Section 2.1.1).
The one remote possibility I can see for reconstructing a statistically predominant basic order to Proto-Carib would be to start at the bottom, with a solid classification of the Cariban family and text counts in every single relevant Cariban language. We might then discover that all members of given branches of the family share similar order frequencies, which might allow us to recon struct statistically predominant orders to the proto-branch levels, and so on back to Proto-Carib itself. The problem with testing this idea is that we still have no reliable internal classification of the family (cf. Gildea 1998:3–11), and even if we did, we have reliable frequency studies of only Carib, De’kwana, Hixkaryana, Panare, and Wayana. Nonetheless, we may soon be able to do such studies in at least two small branches, the Taranoan group (consisting of Akurió, Karihona, and Tiriyó, cf. Meira 1998) and the Parukotoan group (consisting of Hixkaryana, Kaxuyana, and Waiwai, cf. Gildea 1998:92–94), and as other clear branches are identified, the possibilities will expand.
