Thresholds for changes in the amplitudes of two-tone complexes have been measured as a function of frequency separation. Thresholds have been measured both with and without a roving intensity level. Thresholds for changes in the spectral shape of two-tone complexes were generally lower than those for changes in the overall intensity. Thresholds for the former changes are lowest at a frequency separation of one semitone. Thresholds for the latter decrease as a function of frequency separation. Two models have been applied to the data. For narrow-bandwidth signals, the EWAIF model [Feth, Percept. Psychophys. 15, 375–378 (1974)] can partly account for the obtained data. For (partially) resolved signals, the two-channel model [Durlach et al., J. Acoust. Soc. Am. 80, 63–72 (1986)] can account for the data. The obtained values of the model parameters show that the auditory system does not process the changes optimally, in the sense that amplitude changes in one component contribute more in the detection process than those in the other component.