A different technique of analysis is applied to the experiment suggested by Harvey Fletcher for measuring the width of the critical band. This experiment determines the ability of noise bands of different widths to mask a pure tone centered in the band. The analysis considers two filters in series, one outside and one inside the observer. The width of the second filter (the critical band) can be estimated from measurements of the reduction in the noise power at the detector which is effected by the pair of filters. The width of the critical band is estimated under four different assumptions about the shape of the band. The results provide a context for discussing the reasons that may underlie the widely varying estimates of the critical bandwidth which have been obtained in previous studies.