With the same experimental technique, two sets of related experiments have been performed. In one, a band of random noise is used to mask a pure tone. In the other experiment, the intensity difference limen for the band of noise has been determined. Thresholds for masking and discrimination were obtained with help of a Békésy audiometer. A large range of bandwidths (5–12 000 cps) and five central frequencies (500, 1000, 2000, 4000, and 8000 cps) have been employed. Both sets of data agree qualitatively with data appearing in the literature. From the discrimination data, it appears that, for small bandwidths, inherent intensity fluctuations of bands of noise determine the maximum intensity discrimination. These fluctuations should influence the masking situation to the same degree. The similarity of masking and discrimination thresholds in the region of small bandwidth lends support to this prediction. Thus, it is argued why masking data should not be judged in terms of a fixed standard (as Fletcher has done) in order to arrive at values for the critical bandwidth. If one judges masking data in terms of the discriminatory power that the ear exhibits for the masking noise employed, one arrives at critical‐bandwidth data much more in line with generally accepted data. The accuracy with which these can be determined is so poor that one should consider masking experiments of this kind as totally unsuited to measure the critical bandwidth.