Comparing causal-noncausal alternation in three West-African families in contact: Atlantic, Mel and Mande

Submitted to Language Dynamics and Change, special issue on Valence orientation in contact: a cross-linguistic perspective. Proceedings of the Workshop on Valence orientation in contact at the 51st Annual Meeting of the Societas Linguistica Europaea (SLE) (August 29-September 1st 2018), Tallinn, Estonia,This paper investigates the coding of causal-noncausal alternation in three families of languages spoken in West Africa, studying divergence and convergence within and between these families along with various methodological problems related to this cross-linguistic comparison. Atlantic, Mande and Mel languages belong to the same Niger-Congo phylum but display quite different typological profiles and have long lasting historical contacts in Senegal and the surrounding areas. In order to evaluate the correlation between typological profile and valence orientation and to identify contact-induced changes, the same 18 verb-pair meanings (Nichols et al. 2004) have been retrieved from the RefLex lexical database for all the documented Atlantic languages, and for the Mande and Mel languages in contact with them. After overcoming a bias in the list of pairs (favoring causativization), the first results confirm the expected correlation between the favored strategies and the typological profiles, namely lability for Mande (isolating languages with a limited set of derivational suffixes and regularly labile verb) vs. derivational strategies for Atlantic and Mel languages (displaying a large inventory of verbal extensions). In addition, the equipollent strategy shows a surprisingly significant score in Atlantic and Mel, and the suppletive one in the three families. The first one is attributed to structural features of these languages, the second one is interpreted as an areal feature. A standard pattern of distribution of the five strategies for coding valence alternation (based on optimized samples of languages for overcoming data shortage) is then calculated for each family and used for measuring the deviation of individual languages from family pattern. Thanks to a fine-grained study taking into account the structural features of individual languages and the geographical contacts, these deviations are finally attributed to internal evolution of individual languages or of sub-groups of languages and, in a few cases, to contact induced changes. The most salient result is the remarkable heterogeneity observed inside Atlantic family for valence orientation, undoubtlessly reflecting the historical depth of the family