A series of experiments compared the excitation produced in an auditory filter centered on 1100 Hz by two complexes, both of which consisted of harmonics 2–20 of a 100-Hz fundamental. When the components had a level of 69 dB SPL each, summing them in positive Schroeder phase produced substantially less forward masking of an 1100-Hz signal than when the components were summed in negative Schroeder phase. This difference decreased with decreases in overall masker level. Listeners also reported that the components of the positive-phase masker close to 1100 Hz were quieter than the corresponding components in the negative-phase masker. The data are explained using Kohlrausch and Sander’s [J. Acoust. Soc. Am.97, 1817–1829 (1995)] finding that the response of an 1100-Hz auditory filter to the positive-phase complex shows marked peaks and dips, whereas that to the negative-phase complex does not. It is argued that the peaks in the response to the positive-phase masker are attenuated by fast-acting compression in the auditory system, thereby reducing the excitation produced by that sound. It is also argued that, compared to the power functions commonly used to model “excess masking” and the growth of loudness, the present data reflect greater compression at high levels but less compression at low levels.