Noise‐masked hearing‐threshold measurements for tone pulses of 800 cps over a wide range of pulse durations and bandwidths of the masking noise show that the slope of the threshold versus bandwidth curve of 3 dB per factor 2 in bandwidth, as predicted by Fletcher [Rev. Mod. Phys. 12, 47 (1940)] occurs only for bandwidths of 10 cps and smaller. This is due to a criterion change for bandwidths between 10 and 100 cps. In this range of bandwidths, the average slope is about half of that expected by Fletcher. In the present experiments, the slope of the threshold versus pulse‐duration curve turned out to be 3 dB per factor 2 in time when white‐masking noise is used, and often even less when bands of masking noise are used. These measurements indicate the existence of a mechanism that adjusts the width of the critical band in such a way that detection of the stimulus occurs more efficiently than it would in the case of a fixed width of the critical band.