Recently, [J. Acoust. Soc. Am. 103, 2578–2589 (1998)] and [J. Acoust. Soc. Am. 103, 2573–2577 (1998)] independently reported that greater masking of interaurally phase-reversed (Sp) tones was produced by diotic low-noise noise than by diotic Gaussian noise. Based on quantitative analyses, Eddins and Barber suggested that their results could not be accounted for by assuming that listeners’ judgments were based on constant-criterion changes in the normalized interaural correlation produced by adding the Sp signal to the diotic masker. In particular, they showed that a model like the one previously employed by [J. Acoust. Soc. Am. 100, 3774–3784 (1996)] predicted an ordering of thresholds between the conditions of interest that was opposite to that observed. Bernstein and Trahiotis computed the normalized interaural correlation subsequent to half-wave, square-law rectification and low-pass filtering, the parameters of which were chosen to mimic peripheral auditory processing. In this report, it is demonstrated that augmenting the model by adding a physiologically valid stage of "envelope compression" prior to rectification and low-pass filtering provides a remedy. The new model not only accounts for the data obtained by Eddins and Barber (and the similar data obtained by Hall et al.), but also does not diminish the highly successful account of the comprehensive set of data that gave rise to the original form of the model. Therefore, models based on the computation of the normalized interaural correlation appear to remain valid because they can account, both quantitatively and qualitatively, for a wide variety of binaural detection and discrimination data. © 1999 Acoustical Society of America.