Masked thresholds for constant and gliding tones were determined by the method of adjustment for durations between 0.5 and 5000 ms in three overlapping frequency regions between 0.25 and 3.3 kHz. The masker was a continuous white noise at 70-dB SPL. Listening was monaural; subjects had normal hearing. Below 10 ms the thresholds for upward glides were lower and those for downward glides higher than the thresholds for constant tones. In the 10--300 ms duration range, which encompasses formant transitions of speech, the highest thresholds are for downward glides and the lowest ones for constant tones. These differences could result from different time courses of neural decay and inhibition for constant tones, upward and downward glides. The differences between upward and downward glides indicate that the phase spectra influence sound detectability. The thresholds for constant tones reach minimum around 1 s. The thresholds for glides continue to decrease at least up to 5 s. The "critical" duration for constant tone integration can result from the overriding of integration effects by adaptation effects, the latter ones being eliminated by changing frequency. The curves for constant-tone threshold between 10- and 1000-ms duration were fitted by a product of exponential and hyperbolic functions.