A corpus-based analysis of P indexing in Ruuli (Bantu, JE103)

Abstract

Verbs in Bantu languages usually carry an obligatory subject (or S/A) prefix, whereas the presence of a transitive object (or P) prefixes depends on various language-specific factors. A number of such factors is well described in a range of studies mainly based on elicited data. To examine their interplay in naturalistic texts, we conducted a corpus-based case study of object prefixes (or P indexing in the terminology used in this article) in the Bantu language Ruuli (JE103). The corpus of over 15 000 words was annotated for variables such as animacy, identifiability and textual givenness. The statistically relevant factors for triggering P indexing were identified using conditional inference trees. Unsurprisingly, the results show that the strongest predictor for P indexing in Ruuli is word order. Just as P indexing itself, we assume that word order is a differential pattern expressing the argument’s semantic and pragmatic properties. Taking only the latter into account, the analyses reveal that firstly, P indexing seems to be strongly predictable by textual givenness. Secondly, if the referent is given, the probability that it gets indexed is significantly higher if it is human.