Is Agreement Really Independent of Case in Choctaw?

In many languages, there is a match between the agreement and Case systems. It is very common to find languages where all and only nominative arguments trigger agreement, and if arguments with other Cases are cross-referenced, it is with pronominal clitics that match those arguments in Case. There is good evidence from the typological literature that agreement cannot be entirely independent of Case. For example, no language with a nominative-accusative Case system has been reported to have ergative agreement (e.g. Anderson 1977, 1985, Comrie 1978, Moravcsik 1978, Wierzbicka 1981), and this typological gap suggests that ergative agreement is dependent on the presence of ergative Case. Although there are languages with the opposite mismatch (e.g. Warlpiri), where an ergative Case system is paired with nominative-accusative agreement, the degree of independence of agreement in such languages turns out to be quite limited. Even in those languages, pronominal clitics always match in Case with the arguments they double and true agreement always cross-references a nominative argument when one is present. The mismatch that occurs is limited to clauses with no nominative argument; some ergative languages allow true agreement with a non-nominative in clauses with no nominative (Woolford 2006b). This all suggests a rather close relationship between agreement and Case. Can we rule out the possibility of a language where agreement is truly independent of Case? Do any such languages exist? One of the best candidates for a language where agreement is truly independent of Case is Choctaw. In Choctaw, “the agreement system and the case marking system function independently, ” according to Davies 1986:8. The Case system of Choctaw is nominative-accusative (Broadwell 2006), but the agreement system is active-stative (Mithun 1991). That is, all subjects get nominative Case, but the more agentive subjects tend to be cross-referenced with Series I forms (using the neutral label from Munro and Gordon 1982), while other subjects tend to be cross-referenced with Series II / III forms, which also cross-reference objects an