Cluster formation in speech disfluencies of 15 preschool stuttering children and an equal number of nonstuttering control subjects was studied and compared. Clustering was defined as the occurrence of two or more disfluencies on the same or adjacent words. A computer program, based on the Monte Carlo method, was used to generate sets of expected numbers of single disfluencies and disfluencies in various cluster sizes according to simulation models. Results indicate that disfluencies occurred in clusters more often than could have been expected by chance for both groups. The proportions of clustered disfluencies, as well as the size of clusters, in the speech of the stutterers were significantly greater than those of the nonstuttering subjects.