The Full Set II system
The Full Set II system, found in at least five Cariban languages, is characterized by absolutive verb proclitics (1b), (1d), (2b), (2d) and ergative case-marking on explicit A (pro)nouns (1a), (2c–d). In the Pemón subgroup (Pemón, Kapón and Makushi, spoken primarily in Brazil, Guyana and Venezuela, respectively), certain tenses and aspects take a non-agreeing auxiliary (1a), (1c) and the verb frequently bears personal enclitics marking person of A, followed by the ergative case-marker (1b).
The Full Set II system presents a solid VP constituent which takes the absolutive as its internal argument (most carefully argued in Franchetto 1990 for Kuikúro, but with similar evidence easily encountered in the other languages, as discussed below). There is no evidence for a rigid syntactic order of A vis-à-vis the VP, as A occurs frequently either before or after the OV unit (although the most reduced form of A occurs as a postverbal clitic). While no frequency study has been published for any Full Set II system, in the text appended to Abbott (1991), only SV and OV were attested for verbs with explicit S and O arguments, whereas of the 12 clauses with explicit A arguments, 8 preceded the VP and 4 followed.
I have reconstructed the Full Set II system before (Gildea 1997a, 1998 [Chapter 9]), but since the source of the Full Set II VP is also relevant to the Partial Set II system, I briefly reiterate the source here.
Throughout the Cariban family, a possessor noun and its possessed noun form a tight constituent: only certain particles (that modify the possessor NP) are allowed between the two; in some languages, phonological evidence indicates phrasal status (Derbyshire 1985:181 for phrase as “phonological word” in Hixkaryana; Franchetto 1990:409 for suprasegmental marking of the VP-internal absolutive in Kuikúru; Mattéi-Muller 1981:64–66, Payne 1993 for segmental alternations and stress shift in NP, VP and PP in Panare; Gildea 1991, citing Hoff 1968, personal communication, for Carib of Surinam); in several languages, a ‘linker’ morpheme occurs between the two, prefixed to the possessed; and in nearly all of the family, the possessor NP is in alternation with a possessive proclitic on the possessed noun (Gildea 1998:105–116).
Most Cariban subordinate clauses are formed on the basis of nominalized verbs, in which the notional absolutive argument of the verb obligatorily possesses the derived noun; the notional A, if it occurs explicitly, is an oblique adjunct (Derbyshire 1994; Gildea 1994b, 1998 [Chapter 7]). Consider the illustrative examples from Makushi ((3a), Abbott 1991:68) and Carib ((3b), Hoff 1968:121). In (3a), the intransitive verb enna’po ‘return’ is possessed by the second person possessive proclitic, which refers to the notional S, the ‘returner’; the entire possessed derived noun then serves as the O of the transitive main verb. In (3b), the transitive verb eta ‘hear’ is nominalized, then possessed by the notional O ayaura ‘your words’; this [Genitive N] phrase is then the object of the postposition ke ‘Instrumental’, with the notional A coming after the PP as an adjunct.
Thus, if we can identify a construction in which such a nominalized verb might be reanalyzed as part of either Set II system, the Genitive N possessive phrase provides a ready source for the [OV/SV] absolutive VP. And in fact, I have identified six such source constructions (Gildea 1998 [chapters 9–10]), four of which have developed into the Full Set II verbal system (cf. below), and two into the Partial Set II verbal system (cf. Section 2.1.2 below). One of the four will be sufficient here to illustrate the general Full Set II evolutionary pattern.
Figure 1 diagrams what I call the pleonastic pivot source,7 in which the main clause is an equative construction with a pleonastic subject and the predicate is a possessed nominalization, with the possibility of adding an oblique agent phrase if the nominalized verb is transitive. After reanalysis, the pleonastic subject simply drops out of use or is reanalyzed into another function (cf. Gildea 1998:164). If such a change were to occur in English, it would be roughly equivalent to: ‘It (is) my sleeping’ > ‘my sleeping’ = ‘I sleep / am sleeping’ and ‘It (is) my employing by the university’ > ‘my employing by the university’ = ‘The university employs me / is employing me.’
Examples of the source are not so easily encountered in the literature. The clearest pre-reanalysis examples come from Apalaí ((4) Koehn and Koehn 1986:109), the post-reanalysis examples from Makushi being exactly parallel, but with a new epistemic or emotive value now attached to the optional sentence final pronouns ((5) Abbott 1991:106):
The order of the oblique A in Proto-Carib is not clear, as it does not appear to have a fixed pre-reanalysis order in any language described to date. Derbyshire (1981, 1985) takes the preposed order as primary in Hixkaryana, and uses this as the basis for reconstructing SOV order to Proto-Carib. However, in recent personal communication he suggests that it is simply an adjunct postpositional phrase, and that, while like any adjunct it can be placed preceding the derived nominal, more recent work with co-translators has convinced him that the other possibility, post-nominal placement, is in fact more frequent. To my knowledge, statistical studies of the placement of this oblique agent phrase have not been carried out in any Cariban language, and as such, we have no basis for claiming any order as synchronically basic, nor for reconstructing any order as basic in Proto-Carib. My personal suspicion is that the order in Proto-Carib was syntactically free, and thus pragmatically conditioned, as it appears to be today in those Cariban languages which have not reanalyzed it.
In the absence of a syntactically rigid postverbal placement of the ergative A, the ergative enclitic in the Pemón Group Full Set II system requires explanation. I propose that the postverbal position of the postpositional agent phrase correlates with a given, topical information status (cf. Payne 1992:6–11 for discussion of how typologically common such placement appears to be when the order of S/A is not configurationally restricted). Thus, the anaphoric and speech act participant pronouns would normally occur postverbally, and as such would be readily available for encliticization, while leaving full NP agent phrases free for placement either pre- or postverbally.
