Cariban languages and word order typology
Modern Cariban languages present many order variations, of which OVS and SOV appear to be the most common. Determining basic order by a variety of methods, the following proposals have been made (D&P refers to Derbyshire and Pullum 1981):
These studies of word order have relied on three different types of methodology: syntactic properties (i.e., constituency) of the various orders, gross frequency counts of the various orders, and pragmatic or stylistic effects associated with individual instances of the various orders. As examples of the first methodology, Derbyshire (1977, 1985) argues fromco-occurrence restrictions that a preverbal subject occupies a syntactic “topic” slot rather than having this as a basic order (a test which has not, to my knowledge, been replicated in any other Cariban language to date); Gildea (1993a) suggests that the subject in innovative Panare clause types is rigidly post auxiliary (which entails also a post-VP order), and Payne (1994) also places the Panare subject following the VP. Hoff (1995) for Carib, Tavares (1999) for Wayana, and Meira (1999) for Tiriyó conclude that there is no evidence for a syntactically restricted subject position (either NP VP or VP NP) in any clause type. Several studies have concluded from morphosyntactic and prosodic phenomena that OV constitutes a valid VP constituent (cf. Section 4.1 below). There are numerous examples of the second methodology, including Hoff (1978) for Carib, Derbyshire (1986) for Hixkaryana, Hall (1988) for De’kwana, Payne (1994) for Panare and Tavares (1999) for Wayana. Only three of these, Hoff (1978) for Carib, Derbyshire (1986) for Hixkaryana, and Payne (1994) for Panare, pass effectively to the third methodology, that of explaining the frequency counts by offering some discussion of the stylistic effects associated with individual tokens of variation from the most frequent orders.
Numerous claims have also been made about historical change in word order in the Cariban family. Derbyshire (1981, 1985, 1987) argues that the OVS order in Hixkaryana, Makushi, and Taulipang, as well as the mixed SOV/OVS order in Apalaí, are innovations from a Proto-Carib basic order of SOV, instan tiated synchronically in Carib of Surinam and Waiwai. In a review of order patterns in Carib, De’kwana, Apalaí, Hixkaryana, and Panare, Payne (1994:582–3) points out that the main order variants are SOV, OVS, and VSO, with the first two variants common in most languages and the final rare in all but Panare, where it is in fact the most frequent order. In a number of languages, SVO is also a common option, but in no language does it appear to be basic for this clause type. As Payne notes, the languages seem to arrange themselves in a cline, with preferred orders as follows:
The simplest scenario would thus appear to begin with SOV (Carib, De’kwana), then via the reanalysis of an afterthought construction (Derbyshire 1981, 1985), SOV > OVS (in progress in Apalaí, completed in Hixkaryana); the afterthought construction would then enter a second cycle, this time acting on the O, yielding OVS > VSO (in progress in Panare, Payne 1994). These changes apparently triggered no other changes in constituent order, so the typological correlates of the innovative OVS order ended up identical to those of the prior (and typo logically common) SOV order (viz. ‘they have postpositions, genitive precedes the head noun, they are dominantly suffixing, and auxiliary elements are suffixed to or follow the verb’ — Payne 1994:582).
In addition to the change from SOV to OVS, Derbyshire posited an earlier ergative case-marking system, retained in all clause types in Makushi, but only in the more conservative subordinate clauses in Hixkaryana and Carib. Whereas in 1981 Derbyshire was reluctant to connect these two changes, and in Derby shire (1985:104) he explicitly suggests that “the breakdown in the case-marking system may follow the change in word order”, in Derbyshire (1987:316–7) he was willing to assert that “a diachronic change from SOV to some other basic order has resulted in a partial change from ergative to accusative (or neutral) case-marking and/or agreement systems”.
However, looking at a much larger database of Cariban languages, I have shown (Gildea 1998) that the opposition is not simply between ergative and non ergative case-marking, but that there are (at least) seven distinct verbal systems attested in modern Cariban languages, and that each of these seven has a distinct etymology. These systems differ in terms of case-marking (two show ergative case-marking, the rest have no case-marking), verbal person marking (one has a complicated series of personal prefies marking both A and O, one shows absolutive proclitics and ergative encilitcs, one shows only absolutive proclitics, one shows only accusative proclitics, one shows only ergative proclitics, and two show noperson-marking) and occurrence/agreement properties of auxiliaries (one shows no auxiliaries, one shows auxiliaries that do not vary for person, three show auxiliaries that agree with nominative, one shows auxiliaries that agree with absolutive, one shows auxiliaries that agree with accusative). These properties and others (Gildea 1998 [ch. 2]) combine to form constellations of morphosyntactic properties that simply cannot be reconstructed to common sources. In particular, it is clear that ergative case-marking and absolutive verbal person-marking do not reconstruct to Proto-Carib main clauses, but rather have their sources in Proto-Carib nominalized and adverbialized subordinate clauses; thus, all of the ergative verbal systems in main clauses in Cariban languages are innovative. Any connection between the putative change from Proto-Cariban SOV to Hixkaryana OVS therefore cannot be connected to the innovation of main clause ergativity, since Hixkaryana does not show this innovation.
While it was not the focus of my attention in Gildea 1998, it is nonetheless the case that at least six of these seven verbal systems have somewhat different VP constituents,5 and that — along with the component morphology of each system — these constituents reconstruct to their nonverbal sources (cf. Bybee 1988; Aristar 1991; Claudi 1994). My goal in the remainder of this paper is to present a more global picture of the innovative configurational properties in the family (Section 2–3), then to have a look at reconstructing a configurational VP for the more recalcitrant Proto-Carib verbal system (Section 4).
