Anaphoric relations in Jakaltek

This paper will provide an LFG-based analysis of anaphoric relations in possessive and prepositional constructions in Jakaltek (an ergative Mayan language). In doing so it not only offers a solution to some previously unaccounted for language-internal incongruities, but also shows Jakaltek’s realisation of anaphoric relations of this type to be of typological and theoretical significance. Essentially, I show that the general anaphoric co-indexing system of Jakaltek is based on structure sharing and the relative distribution of noun classifiers. First I provide a general description of Jakaltek anaphoric relations in constructions involving possessives or prepositional pronouns and the constraints to which they are subject. Then I present an analysis of the conditions governing these relations. With the exception of clause bound anaphoric constructions, concrete nouns in Jakaltek are obligatorily accompanied by a noun classifier (henceforth NC), attributing the noun to one of twenty-one semantic classes. When occurring without a noun, a subset of these NC function as personal pronouns. Possessed NPs, formed by a possessive marker prefixed to the noun, must also take a NC, always in agreement with the possessor. Abstract nouns have no noun classifier. As observed in Craig (1977) and Day (1973), within the clause, the co-indexing of arguments is signalled by the absence of the otherwise obligatory NC on the second element in the anaphoric relation, as in (1) and (2). If the NC is present, no bound reading is possible (3).