When we speak we hear the sound of our own voice. This auditory feedback plays an important role in maintaining the precision of fluent speech. In a series of experiments my colleagues and I have been studying this sensorimotor control of speech using a novel signal processing system that allows us to modify speech feedback in real time. Subjects compensate for changes to their voice and these compensations persist following return to normal feedback. The picture that is emerging from our study of speech is one of an activity that is produced with the aid of rich, predictive motor representations that learn aspects of the task, performance environment and the feedback that is expected when an action is performed. For speech, this means that the sound system of the language, the individual vocal tract morphology, and the expected auditory and somatosensory feedback are contained in a motor representation that is drawn upon when a sound sequence is planned
.