Despite a long history of debate on how the speech production and perception systems are linked and a vast number of experimental studies indicating an intimate link between perception and production, formal proposals of this link have been conspicuously lacking. We provide a computationally explicit, dynamical model of the process of phonological planning. In this model, the properties of a perceived utterance automatically serve as input to the ongoing planning of an intended utterance. Using tools from non-linear dynamics, we formalize how incoming inputs from perception influence the ongoing choice of phonological parameters to be used in production. Our model provides an account of response time modulations reported in independent experimental work, and makes additional concrete predictions that can be tested experimentally. Our model provides a foundation for better understanding the cognition of speech perception, speech production, and the interaction between the two.