Conférence du Professeur Geoffrey HAIG | Alignment in discourse: the discourse basis of ergativity revisited

In a pioneering piece of research, Du Bois (1986) suggested that cross-linguistically, natural connected discourse exhibits a characteristic pattern in the way syntactic arguments are realized, called Preferred Argument Structure (PAS): While transitive subjects are overwhelmingly expressed via pronouns or zero, intransitive subjects and transitive objects cluster together in that they are significantly more frequently full NPs. Du Bois claimed that PAS in discourse effectively mirrors ergative alignment in morphosyntax (cf. Dixon 1996, Du Bois 2003a, 2003b). In this paper, the claims regarding a universal preference for ergative structures in discourse are re-evaluated, based on data from natural spoken discourse from seven typologically and genetically diverse languages (Haig et al 2011). It is suggested that the ergative bias in discourse is in fact an artefact of the way the data has been analysed (Haspelmath 2006), and in fact languages vary quite significantly in this respect. Furthermore, our data demonstrate that the best predictor for lexical vs. pronominal/zero expression of arguments is not syntactic function, but animacy (cf. Everett 2009). Once animacy is taken into account, natural discourse does in fact reveal characteristic patterns, but they are better seen as split-intransitive rather than ergative.