Four experiments were performed in an attempt to refine the measurement of psychophysical tuning curves (PTCs). PTCs were determined using sinusoidal signals and narrow-band noise maskers, in both simultaneous and forward masking. In experiment I a fixed low-level notched noise was gated with the masker in order to restrict off-frequency listening and to eliminate "confusions" between the signal and masker in forward masking. In contrast to previous work, this procedure produced PTCs with similar round tips in both simultaneous and forward masking. The PTCs differed mainly in that the high-frequency skirts were steeper in forward masking. Experiment II compared PTCs with and without an additional notched-noise masker, for a number of signal levels. The notched noise had the effect of broadening the tips of the PTCs, this effect being greater in forward masking than in simultaneous masking. With increasing signal level the high-frequency skirt of the PTC became slightly steeper, and the low-frequency skirt slightly shallower. Experiment III studied the effect of signal duration on PTCs determined in simultaneous masking. Duration had little effect when off-frequency listening was restricted with a fixed notched-noise masker, but had a substantial effect in the absence of such a masker. Experiment IV studied the effect of signal delay in forward masking. When off-frequency listening was restricted and the signal levels adjusted to produce similar masker levels at threshold for each delay, the shape of the PTC was not affected by delay. It is concluded that, when PTCs are determined in the presence of notched noise, suppression is probably responsible for most of the differences between simultaneous and forward masking. Previous work may have overestimated the influence of suppression, because it did not control off-frequency listening and/or provide appropriate detection cues in forward masking.