Consonant mutation is a linguistic phenomenon whereby two or more sets of consonant phonemes alternate systematically within roots (or other morphemes) in a way that is not entirely predictable from the phonological environment. A number of Atlantic (Niger-Congo) languages of West Africa exhibit root-initial consonant mutation to mark noun class as well as various verbal morphosyntactic categories. This study treats the historical origin and development of these consonant mutation systems. I argue that despite their typological similarity, the mutation systems of the various Atlantic languages arose for the most part independently, in contrast with the prevailing assumption in the existing literature that these systems were inherited from a common source.
I begin with a typological survey of consonant mutation in the rest of the world’s languages, providing a brief description of each phenomenon, as well as summarizing the theoretical treatments of these phenomena in the literature. I also examine the historical development of these mutation systems where such proposals are extant in the literature. Historically, consonant mutation— which is rather less rare than has often been claimed— generally arises from the interaction between an initial (or less often final) segment of one morpheme with the immediately adjacent segment of a preceding (or following) morpheme. For example, earlier *a-ta and *ak-ta (in which *a- and *ak- are two prefixes which can appear on the root *ta) might develop into a mutation alternation /a-sa/ ~ /a-ta/.
The Atlantic languages which exhibit mutation are Fula and Sereer, Wolof, Kobiana and Kasanga, the Tenda languages, and Biafada and Pajade— all traditionally part of the proposed Northern Atlantic subgroup within Atlantic. The development of consonant mutation in all of these languages is consistent with the type of diachronic origin sketched above. For each of these languages I reconstruct the forms of the original triggers of mutation, as well as the regular sound changes that operated between these triggers and the initial consonants of lexical roots. In doing so I present reconstructions of the phonological and noun class systems of each proto-language, as well as their relevant verbal/pronominal systems. In addition to the regular sound changes that are the ultimate source of mutation, I identify a number of analogical changes that reshaped these systems of alternation in each language.
Finally, I explore the wider question of what genetic relationships exist between the Northern Atlantic languages, using the lexical, phonological, and morphological evidence presented in the preceding chapters. Having established that mutation in the Northern Atlantic languages cannot be attributed to shared inheritance, the principal remaining sources of evidence involve shared lexical material and similarities in the noun class systems. I argue that there are no convincing developments in the noun class systems of the Northern Atlantic languages that can be taken as shared developments of this proposed subgroup, though a strong argument can be made for the grouping of Wolof with Bainunk-Kobiana-Kasanga, and much more tentatively for grouping Biafada-Pajade with this group. Otherwise, the similarities in the noun class systems of Northern Atlantic languages can be attributed to shared inheritance from Niger-Congo. The lexical evidence for this subgroup is somewhat more convincing, but is still extremely limited. Taking into account the patterns of language contact in the area, it seems entirely possible that the few potential lexical innovations of Northern Atlantic can be attributed to borrowing or shared retention. Thus I argue against the genetic unity of the Northern Atlantic subgroup, though all of these languages can be rather securely situated within Niger-Congo at large.