A stochastic model of linguistic category dynamics

Production and perception of speech are notions that are distinct, yet intimately coupled. One way to model production and perception is by exemplar models, where we store examples of speech events (exemplars) for these two purposes. Production uses these exemplars to form our utterances; however, due to various influences, what is intended is often not what is ultimately spoken. We can model this situation by introducing a stochastic ele- ment to the exemplar prior to verbalization. Perception uses the lexicon for comparison, so upon hearing a sound we can classify it based on its "closeness" to lexical entries. In the following pages we will use this model to analyze how sounds evolve and remain unique. Recreating work by Wedel will show us how the lexical categories maintain their differentiation. We will also consider a simplification of this model that will facilitate analytical and numerical analysis.