The clear [OV] Verb Phrase in 3A3O sentences
The evidence for an OV constituent in modern Cariban languages is as follows:
– alternation between a free O nominal and a verbal proclitic *n- (illustrated for 14 languages in Gildea 1998:60–74).
– prosodic evidence for a phrase, usually involving shift in accent in one or both of the words (Mattéi-Muller 1981:64–66; Payne 1993).
– phrase-internal segmental alterations in the verbal prefix for Panare (Mattéi Muller 1981:64–66; Payne 1993)
– placement of second-position clitics/particles, in which the [OV] VP is treated as a single unit, and cannot be separated by a clitic/particle (Hoff 1995; Meira 1999).
For the first and last properties, consider these illustrative examples from Carib of Surinam (Hoff 1995:353): (24a–d) illustrate the alternation between the proclitic n- and the preverbal O, but not A; (25a–b) illustrate that the second position particle hkuru ‘certainly’ can follow the entire VP (25a), but it cannot come between O and V (25b).
While such studies of the interaction between particles and constituency have been published only for Carib (Hoff 1995) and Tiriyó (Meira 1999), until counter examples begin to appear, I hypothesize that cognate Set I VPs will show the same behavior. In the meantime, relying only on the alternation between preverbal O and proclitic, I have found that a clear synchronic VP exists for 14 of 19 languages, and for two others such a VP is at least available as a stylistic option (Gildea 1998:78). The four languages where the alternation does not exist are easily identified as innovative (Gildea 1998:74–78), leaving a clear preponderance of evidence in favor of reconstructing the 3A3O VP to Proto-Carib. This VP is exactly the source of most claims that Cariban languages have a basic OV order.
