We have recently reported that frequency resolution, under conditions of high stimulus uncertainty, is much more accurate for late high‐frequency components than for early, low‐frequency components of word‐length tonal sequences [Watson, Wroton, Kelly, and Benbassat, J. Acoust. Soc. Am.57, 1175–1185 (1975)]. We have now replicated these experiments for the case of temporal resolution. Listeners were tested with high, medium, andminimalstimulus uncertainty until they performed at asymptotic levels of temporal resolution, defined by just‐detectable proportional increments (Δt/t) in the duration of single 40‐msec components of 10‐tone sequences. The duration of late, high‐frequency components is also resolved more accurately than that of early, low‐frequency elements, although the differences are considerably less than for frequency resolution. Asymptotic just‐detectable values of Δt/trange from about 0.75 to 0.15, from high to minimal levels of stimulus uncertainty, suggestive of a shift from detecting increments in tonal sequence duration to detecting increments in the duration of single test components. [Supported by Grant ♯NS‐03856 from NINCDS.]