The influence of masker variability on estimates of monaural and binaural critical bandwidths

If integration bandwidths are derived from band-widening masking experiments, binaural estimates are usually a factor 2 to 3 larger than monaural estimates, at least at high levels. It is proposed that this difference does not reflect ‘‘different critical bandwidths,’’ but that it is a consequence of detection statistics. In monaural experiments, the variability in the masker energy limits detectability of the signal. Therefore with increasing subcritical bandwidths, the S/N ratio at threshold decreases. In binaural NoSπ experiments, the S/N ratio at subcritical bandwidths remains constant because the masker correlation is always 1 without uncertainty. Due to this absence of external variability in the masker correlation, detection must be limited by internal noise. Binaural detection in narrow-band conditions can gain from the spread of excitation across several critical bands, assuming that the internal noises are uncorrelated. The wider binaural critical band is then caused by masking the off-frequency spread through masker components outside the central critical band (for binaural model simulations, see Breebaart et al.). In this study it is shown that this scheme also applies to monaural frozen-noise maskers, which do not have external variability. In agreement with results from binaural experiments, the integration bandwidths increase considerably with masker level.