Modeling spectral integration in binaural signal detection

Estimates of critical bandwidths in binaural experiments vary considerably across experimental conditions. For masking experiments using signals of varying bandwidths (i.e., spectral integration), and experiments with a frequency-dependent interaural masker correlation, bandwidth estimates agree with those from monaural experiments. However, if auditory bandwidths are estimated from binaural band-widening masking experiments, the effective bandwidth is usually a factor of 2 to 3 times larger than monaural estimates. One often ignored detail is that this difference between monaural and binaural estimates decreases with decreasing masker level [Hall et al., J. Acoust. Soc. Am. 73, 894–898 (1983)]. Using the binaural model described by Breebaart et al., we have simulated the binaural experiments described above. It can be shown that, without any parameter change in the model, basically all relevant conditions can be modeled accurately, including the band-widening and spectral integration experiments. In agreement with the scheme discussed by Kohlrausch et al., detection in binaural narrow-band-noise conditions is improved by analyzing the internal representation in several adjacent filters. It is concluded that the absence of this detection advantage in monaural random-noise conditions is the primary cause for the so-called ‘‘wider binaural critical band.’’