The detectability of a sinusoidal signal masked by a sinusoid of fixed frequency (920 Hz) was measured in a 2IFC task. The signal and masker were gated with fixed starting phase such that the resultant started at maximum amplitude. For fixed duration of the observation interval (T = 500 msec), there was a slight improvement in performance with increasing frequency difference between signal and masker (Δf = 2.0, 4.0, or 6.0 Hz). With Δf = 4.0 Hz, there was a monotonic improvement in performance with increasing duration (T = 125, 250, or 500 msec); the four observers required an average of 6.2 dB less signal power for constant performance with T = 500 msec as compared with T = 125 msec. For a single beat in the signal interval, there was a monotonic improvement in performance with decreasing frequency difference (Δf = 2.0, 4.0, or 8.0; T = 1kΔf). The average improvement was 3.6 dB over the range in Δf. With Δf = 4.0 Hz, T = 250 msec, the average slope of the Weber function was 0.88. Performance in classical intensity discrimination (Δf = 0.0) was better than in some beat conditions. [Supported by a NIH predoctoral fellowship and by the Air Force Office of Scientific Research.]