Subject inversion in (Proto-)Bantu relative clauses

Fatima Hamlaoui

University of Toronto

Lattimore 513

Inversion constructions are a widespread phenomenon in Bantu languages. Interestingly, some of the languages geographically close to the Bantu ancestral homeland however do not allow any postverbal subjects (e.g. Basaá, A43, Cameroon). In a family in which word order is generally considered flexible, the question arises as to why some languages have a fairly rigid word order and which type of word order characterized Proto-Bantu. Trying to abstract away from the effects of information structure, this talk concentrates on word order in relative clauses, in which SV, VS and SV/VS are attested. Based on a sample of 150 Narrow Bantu languages, we first discuss issues relating to the most frequent word order in the family (VS), the possible motivations for inverted embedded subjects and whether subject inversion in relative clauses and in simple sentences are generally correlated. We propose that Proto-Bantu had both SV and VS relative clauses and that the loss of VS in Bantu A (as represented in our sample) is related to the historical loss of verbal morphology and, in particular, pre-stem object markers. By losing these markers, Bantu A languages would have lost both the word order flexibility typically associated with the head-marking property and the opposition between weak and strong objects, resulting in the loss of association between being after the verb and being strong or focused. This, we claim, would explain why a language like Basaá only shows remnants of an Immediately After the Verb focus position (with wh-pronouns) and focuses subjects in their canonical preverbal position rather than post-verbally. One of the predictions of our proposal, if it is on the right track, is that SV should be preferred among languages with no more pre-stem object markers. Based on a sample of 145 languages (including 51 Northwest Bantu languages), we show that this is indeed the case.