Physiologically motivated auditory models calculate excitation from a compressed and rectified version of auditory filter responses. Unlike the ``power spectrum'' model, they predict that, when the responses of a filter to two sounds have equal power but different peak factors, the sound producing the less peaked response will produce more excitation. This was confirmed by experiments using two sounds, each consisting of harmonics 2--20 of a 100-Hz fundamental, presented at 68 dB SPL/harmonic, and with the harmonics summed in either positive or negative Schroeder phase. Kohlrausch and Sander [J. Acoust. Soc. Am. 97, 1817--1829 (1995)] have shown that the response of an auditory filter centered on 1100 Hz to the positive phase complex is highly modulated, whereas that to the negative phase complex is not. It is reported that the threshold for a 10-ms 1100-Hz signal presented 5 ms after the positive phase stimulus is approximately 10 dB less than after the negative phase stimulus. Also, listeners judged harmonics 9--13 of the positive phase sound to be quieter than the corresponding harmonics of the negative phase sound. It is concluded that the results reflect fast-acting compression in the human auditory system.