Genitive constructions in Niger-Congo

International audience There is a clear areal pattern within Niger-Congo regarding the mutual ordering of a genitive modifier (Gen) and the nominal expression it modifies (N). In a large central zone stretching from Liberia to Benin the order is overwhelmingly Gen N, whereas we find the inverse N Gen ordering at the Eastern and Western periphery. Although this distribution suggests that Gen N order is an innovation, we propose instead that the N Gen ordering found at the peripheries is innovative and that Proto Niger-Congo is more likely to have had Gen N order, in as far as this is reconstructable. Our claim is based on two kinds of indications. The first is the existence of head-final compounds in Proto Bantu, such as *gʊ̀ndà ‘forest’ (< ‘house’ dá ‘of tree’ gʊ̀n), which are arguably retentions from a pre-Proto Bantu stage. The second is the existence of a synchronically attested pattern of change that can easily explain the recurrent emergence of N Gen orders out of an original Gen N construction. What remains to be explained is why such changes have not taken place in a vast central domain of the Niger-Congo languages.