Many languages have restrictions on the co-occurrence of laryngeally marked segments (such as voiced obstruents, aspirates, glottalized consonants, etc.). Current theories of sound change ascribe the origin of these restrictions either to speaker-oriented articulatory forces (grammaticalization of articulatory simplification) or to listener-oriented perceptual forces (grammaticalization of misperception). In this presentation, I will argue for a third possible source for these co-occurrence restrictions, based on a newly developing restriction in Afrikaans. I will argue that co-occurrence restrictions can also arise via a lexical route. Through the gradual lexical accumulation of sound changes, a pattern consistent with a co-occurrence restriction can accidentally arise in the lexicon of some language. Once the pattern has been lexically established, language users can then elevate the pattern to a grammatical principle via a statistical learning mechanism. I will first establish the existence of the voicing co-occurrence in Afrikaans relying on the three kinds of evidence: (i) Evidence for the pattern in the Afrikaans lexicon. (ii) The results of a wug-test with Afrikaans speakers. (iii) Evidence from non-standard minority varieties of Afrikaans in which the restriction has been established more firmly than in Standard Afrikaans. I will then trace the developments of the Afrikaans lexicon from Dutch, showing that the lexical pattern in Afrikaans is an accidental side-effect of a series of unrelated sound changes that applied in the development from Dutch to Afrikaans.

Andries W. Coetzee is Associate Professor and Associate Chair of the Department of Linguistics at the University of Michigan.