The hypothesis is advanced that a time‐invariant pattern of frequency response can be discriminated in the presence of a variable pattern of frequency‐response irregularity only when the listening interval is of sufficient duration to allow the variable part of the total response pattern to be averaged out. Qualitative evidence seems to support the hypothesis, but quantitative evaluation of the needed integration times is still wanting. If the effect is real, however, it may have an important bearing on the interpretation of the conventionalA ‐ Btests used for the assessment of sound reproducing systems.