"Residual masking' was measured in a tonal forward masking paradigm. In one experiment, psychophysical tuning curves and masking patterns were obtained at several frequencies and levels for a fixed masker-probe time delay. In a second experiment, tuning curves and masking patterns were measured as masker-probe time delay was varied. Our results indicate that tuning curves and masking patterns are sharpest at low levels, high frequencies and brief masker-probe time delays. In addition, we observed that masked probe threshold returned to the level of unmasked probe threshold at approximately the same post-masker time regardless of masker level or the probe-to-masker frequency relationship. These findings suggest that frequency, level and time delay all affect the degree of frequency selectivity observed with these measures.