A fixed amplitude sinusoid that was frequency modulated by a sawtooth or triangle wave was used to mask a 20-msec pure-tone signal. The linear portions of the modulating waveforms were 150 msec in duration and swept the masker frequency between 1.5 and 0.75 kHz. A frequency contour of the masking was obtained when the temporal position of the signal was fixed relative to the sawtooth or triangle and the signal frequency was varied. A temporal contour of the masking was measured when the signal frequency was fixed and the temporal position of the signal was varied. Temporal and frequency selectivity indicated by differences of 30-40 dB of masking were observed in both kinds of contours. A linear model consisting of a bandpass filter followed by a power integrator provided a fair approximation to principal features of the masking contours when the bandwidth of the filter was 100-200 Hz and the time constant of the integrator was 10 msec. Similar resonant systems are found in spectrum channel vocoders and in instruments that produce speech spectrograms. The model tends to underestimate the masking at the higher frequencies and does not exhibit the type of deep, narrow-band notches that are found in frequency contours of the masking.