In experiment 1, masking patterns were obtained with a tonal masker that was sinusoidally amplitude modulated (SAM) at a rate of 8 Hz and a depth (m) of 1.0. The signal was centered at a masker peak or masker valley. Masker frequency (f(m)) was 750, 1350, or 2430 Hz, and signal frequency (f(s)) ranged from 0.8 to 1.62 f(m). Thresholds were generally higher for a signal in a masker peak than in a masker valley. The magnitude of this peak-to- valley (PV) difference was governed by f(s)/f(m), rather than by f(s), and was largest for f(s)>f(m). The PV differences were smallest at the lowest f(m), at least when f(s)>f(m). In experiment 2, growth-of-masking functions were measured (f(m)=M 1350 Hz, f(s)= 1.44f(m)). The masker was modulated at a depth (m) of 1.0, 0.75, or 0.50. These thresholds were compared with those obtained with an unmodulated masker in forward or simultaneous masking. The comparisons suggest that thresholds for a signal at a peak of an 8-Hz SAM masker are due to simultaneous masking, while those in a valley are due primarily to forward masking when m=1.0 or simultaneous masking when m=0.75 or 0.50. For these masker depths, the PV difference first increased but then decreased as masker level increased from 60 to 90 dB SPL. This was a consequence of the slope of the masking function for peak placement changing from a value greater than 2.0 to a value of 1.0 at the highest signal levels (an effect that was also observed with the unmodulated simultaneous masker), a result that may be understood in terms of basilar membrane nonlinearity.