A fundamental issue in linguistics and cognitive science concerns the dynamics that structure spoken language. In this talk, I outline ongoing work into discovering some of the principles behind speech production using dynamical computational models of articulatory control and speech planning.

The talk is structured in two parts. Part 1 focus on articulatory gestures during speech. I outline a method for discovering new models of articulatory gestures directly from kinematic recordings and then demonstrate an application of such a model to theoretical issues around synchronic vowel variation in English.

Part 2 focuses on models of the cognitive processes underpinning articulatory planning and short-time memory dynamics. I first outline an experimental study of phonetic accommodation and how a dynamic neural field model provides a realistic way of modelling short-term phonetic learning. I conclude with a proof-of-concept model of how many short-term interactions between speakers can lead to longer-term patterns of diachronic change.