Learning Abstract Phonological from Auditory Phonetic Categories: An Integrated Model for the Acquisition of Language-Specific Sound Categories

We introduce a two-stage model for the perceptual acquisition of speech sound categories within the framework of Stochastic Optimality Theory and the Gradual Learning Algorithm [1]. During the first stage, learning of language-specific sound categories by infants is driven by distributional evidence in the linguistic input. This auditory-driven learning leads to a warping of the baby’s perceptual space, to discrimination curves, to the perceptual magnet effect, and ultimately to the creation of phonetic categories. In the transition to the second stage, these phonetic categories turn into simple abstract phonological categories. During the second stage, when the lexicon is in place, lexically-driven learning will develop more abstract representations and optimize multi-dimensional perception. The results of our simulations compare well with findings from the infant literature and from cross-language studies.