Signal detectability was measured in three temporal conditions as a function of the bandwidth and configuration of simultaneous maskers that either did or did not spectrally overlap the signal. The 20-ms signal was 250 Hz wide and was centered at 2500 Hz (fs). Although there were marked individual differences, performance was typically poorer when signal onset came 1 ms rather than 250 ms after the onset of a 420-ms masker, and poorest when signal onset came 1 ms after the onset of a 23-ms masker. The results support the idea that two separate across-channel processes contribute to temporal changes in signal detectability. One process contributes to the improvement observed as signal onset is delayed from masker onset, and its influence is reduced by the presence of masking components at fs only when the masker extends exclusively below fs. The other process is associated with the improvement observed as masker offset is delayed from signal offset, and its influence is reduced by the presence of masking components at fs when the masker extends exclusively above, or both below and above fs. Both of these processes are primarily activated by frequencies ranging from 0.6 to 0.8fs and 1.2 to 1.4fs. The data also demonstrate that the measured critical bandwidth narrows as signal onset is delayed from masker onset.