Some tests of the EXPLAN account of serial speech performance

Peter Howell

University College London, UK

EXPLAN is a model of the serial order of speech and how it can break down. According to the model, speech control involves two processes – planning and execution and the timing relation is critical. More complicated words take longer to plan than simpler ones. A number of factors affect execution timing – complexity of the motor program, the number of concurrent events and so on. Crucially, a) the two processes can be operationally distinguished and b) the model posits the involvement of a synchronization detector in the cerebellum. Three topics will be covered: First, the planning and execution processes are simulated and it is shown how these produce serial ordering of words of different complexity (this uses a graphics package and sound output system). Perturbations are then made to the model at planning and execution levels that lead to various forms of fluency breakdown. The next studies represent experimental tests of some specific predictions about EXPLAN. Second, during planning pauses will not occur at random. Support is presented for the prediction that pauses occur at the onset of planning units. Third, features of the execution process that is supposed to be located in the cerebellum are presented. A prediction about a manipulation that can be made to speech (DAF) is investigated. According to the model, DAF is an asynchronous event and this should make it affect the cerebellar processes. There is evidence against this prediction from one imaging study. We replicated the study and found cerebellar activity under DAF. Second, the cerebellum is regarded as operating as a general purpose synchrony detector. We show that if speakers are induced into considering they are performing an asynchronous syllable repetition task, motor and clock variance in the Wing-Kristofferson task is higher than when the same task is performed, but speakers are induced to think they are performing a synchronous task. Both these studies support the general notion of timekeeper involvement in speech tasks.