The effect of silent-interval duration on the perception of pauses in speech was investigated. The stimuli consisted of five recorded sentences within which the words “lost” and “contact” were manipulated so that their syntactic relation to one another varied in complexity from sentence to sentence. Twelve subjects, working individually, mechanically adjusted the silent-interval duration between the words “lost” and “contact” within each sentence in order to locate (1) the pause detection threshold, (2) optimal fluent pause duration, and (3) minimal hesitation pause duration. Across the five sentences, the mean durations were 6, 186, and 505 msec, for the three pause types, respectively. Statistical analysis, however, showed that silent interval durations for these three pause types differed significantly only when occurring between phrases. This was not true of within-phrase pauses. Thus, contrary to usual assumptions in the literature, these results suggest that for within-phrase pauses, at least, silent-interval duration is an insufficient perceptual cue for differentiation of fluent and hesitation pauses. [Supported by NSF and NIH grants.]