This study examined how effectively information about amplitude modulation (AM) on two carriers is combined, and whether the detection of AM depends on the relative phase of the AM across carriers. Psychometric functions were measured for detecting 5-Hz sinusoidal AM of carriers with frequencies 1100 and 1925 Hz, with a mean level 65 dB SPL for each carrier. The carriers had a duration of 400 ms with 50-ms raised-cosine ramps on either side of this. A single cycle of 5-Hz sinusoidal AM (200 ms in duration) was imposed on the temporal center of the stimulus, with 100-ms steady-state fringes before and after the modulation. The modulators for the two carriers were either in phase or in antiphase. The modulation of each carrier was equally detectable, as determined in a preliminary experiment. A continuous pink noise background was used to mask the outputs of auditory filters tuned between the two carrier frequencies. There was no effect of relative modulator phase. However, performance was consistently better than predicted from the assumption that information about AM from the two carriers is processed independently and combined optimally. The results are discussed in terms of (1) predictions using Dau’s “modulation filter bank model” [T. Dau et al., inPsychoacoustics, Speech and Hearing Aids, edited by B. Kollmeier (World Scientific, Singapore, 1996), pp. 45–48], and (2) the fact that relative modulator phase does have an effect on the detection offrequencymodulation on two carriers, as found by Furukawa and Moore [J. Acoust. Soc. Am.100, 2299–2311 (1996)].