The ability to extract information from relatively slow amplitude changes, the envelope, appears to be a crucial aspect of auditory communication. The general question we address is how can this ability best be understood anddescribed. One approach, the “modulation filterbank” model, postulates that there are filters that are selectively tuned to envelope frequency. This, essentially, is a translation of the well-documented notion of auditory spectral filtering into the domain of modulation frequency. We present data that suggest that a more useful approach may be to consider time-domain descriptions of envelope processing. More specifically, we investigated the modulation masking produced by complex periodic modulation maskers on detection of sinusoidal AM. The carrier was broadband noise and the modulation frequency of the signal coincided with the fundamental frequency of the masker. The data indicate that low duty-cycle pulse-train maskers can be much less effective in masking than a sinusoidal masker even when the amplitude of the fundamental of the masker is held constant. This seems incompatible with a simple modulation filterbank model but is easily understood, at least qualitatively, in terms of the temporal structure of the envelopes: the detectability of the signal is determined by the local modulation depth, which can be much greater in the troughs of a pulse-train masker. It also appears that a model that incorporates such a temporally-based approach can describe the “tuning” seen in modulation masking without invoking modulation filters.