This study examines the role of temporal cues in spectral masking, such as
beats and intrinsic envelope fluctuations. Predictions from the modulation-
filterbank model developed by Dau ct al. [J. Acoust. Sec. Am. 102, 2906-291
9 (1997)] are compared to average masking patterns from Moore et al. [J. Ac
oust. Sec. Am. 104, 1023-1038 (1998)]. In these experiments, tones and narr
ow-band noises have been used as the signal and the masker, so that all fou
r signal-masker combinations are considered. In addition, model predictions
are compared with new experimental data in conditions of notched-noise mas
king, where the masker consisted of two narrow-band noises whose bandwidth
and frequency separation were varied systematically. The model uses a perip
heral filtering stage with linear and symmetric Gammatone filters, an adapt
ation stage that includes a static compressive nonlinearity for stationary
input stumuli and a higher sensitivity for envelope fluctuation, and a modu
lation filterbank that analyzes the output for each peripheral channel. For
low and medium masker levels, the model accounts very well for the masking
patterns in all signal-masker conditions, as well as for the notched-noise
conditions. In contrast, predictions from a version of the model that acts
like an energy detector account for only some of the notched-noise data, a
nd generally do not account for the shape of the masking patterns. For a hi
gh masker level, the simulations suggest the use of asymmetric filters, wit
h a steeper high-frequency slope than is used in the linear model, consiste
nt with results from previous studies. In addition, several nonlinear effec
ts become apparent at this masker level, which cannot be accounted for by t
he current model. (C) 2000 Acoustical Society of America. [S0001-4966(00)02
707-7].