Modeling the emergence of vowel harmony through iterated learning

Fred Mailhot

Abstract: This chapter provides support for the view of sound change as being listener-driven, specifically with respect to vowel harmony as the result of phonologized vowel-to-vowel coarticulation. It is shown, by means of a computational model, that coarticulatorily-perturbed outputs, coupled with biased perception, result in a pressure towards harmonization when iterated across multiple generations of transmission and acquisition. The model also preliminarily addresses the actuation problem of Weinreich et al. (1968), demonstrating that the march toward harmony need not be inexorable in a diachronic functional model.