Masked threshold levels for signal detection or pitch ranking (low, middle, or high) were determined in conditions where one of three possible signal frequencies could be presented on a given trial of a three-interval forced-choice task. Thresholds were determined under conditions investigating binaural masking release (the masking-level difference, or MLD) and monaural masking release in modulated noise. It was assumed that part of the masking release in modulated noise was due to a within-channel analysis of information in the masker dips, and that part of the masking release was due to across-frequency analysis of temporal envelope information (comodulation masking release, or CMR). In the MLD experiment, the masker was a low-pass noise with a frequency cutoff of 1000 Hz. In the modulated-noise experiment, a broadband noise was square-wave amplitude modulated at a rate of 25 Hz and then digitally low-passed filtered at 1000 Hz. The results indicated that when the frequency separation between frequency components was relatively wide, masking release for pitch ranking was similar to that for signal detection. However, as the frequency separation between components narrowed, masking release for pitch ranking decreased. The results are consistent with an interpretation that information regarding signal frequency is relatively coarse under the conditions of masking release examined here.