Computational approach to stuttering

Though many causes have been considered from mental or anatomical viewpoint, the mechanism of stuttering is not known yet. One candidate for the stuttering occurrence is a dysfunction of basal ganglia. However the detailed relation between the basal ganglia dysfunction and the stuttering is not understood. In this study, based on anatomical, physiological and clinical evidences, we propose a functional model of the basal ganglia in the utterance movement, and try to explain the stuttering. Traditionally, the basal ganglia is considered as having primarily motor functions. However, recent findings pointed out unexpected roles of the basal ganglia that range from movement control to attention, visual perception and learning. According to an observation of arm movement in Huntington disease, the basal ganglia is suggested to have a feedback control function to compensate an error during the movement. Moreover it is suggested that the function has something to do with language dysfunction, for example Gilles de la Tourette syndrome or stuttering. But we don’t know the concrete mechanism of how the dysfunction of the basal ganglia feedback control causes the utterance dysfunction. When we assume the basal ganglia to have the error feedback control function, next question is what kind of control the basal ganglia performs. The origin of the motor movement, which the basal ganglia works to compensate, is thought being biological noise such as white noise. If so, the basal ganglia has to solve a stochastic control problem. Then we propose a functional model of the corticobasal ganglia loop as a stochastic control system. In the model we assume that the control value is obtained through the feedback of error. This means that control value, that is equivalent to the output of the basal ganglia, is proportional to the error amount. One way to obtain this type of solution is the regulator theory of optimal control framework. Based on our model, we try to explain the phenomenon of stuttering, especially type SR (sound and syllable repetition). From a computational simulation, we conclude that type SR stuttering may be explained by some difficulties in signal transfer caused by error feedback dysfunction of the basal ganglia. Based on our model, our prediction on a fMRI study becomes as follows. By producing the stuttering artificially by DAF, the activation of the basal ganglia should be measured. If a correlation between the degree of stuttering and the activation of the basal ganglia is obtained, it implies that the basal ganglia activates when the error is produced. The result is supported by our hypothesis that the control value is obtained by feedback form of the error. Through these comprehensive studies, we would like to link the computational approaches to fMRI brain study.