A consideration of the normalization that is typically included in correlation-based models of binaural detection.

An analysis of binaural detection and new data that elucidate the nature and precision of normalization that must be assumed if binaural detection is accomplished via mechanisms that effectively compute the coefficient of cross correlation is presented. Based on that analysis, it is argued that the precision of normalization required to remove deleterious effects resulting from variations in the levels of the stimuli is so great that it is highly unlikely that normalization, per se, actually occurs as part of binarual processing. Instead, it appears more likely that binaural processing is accomplished via "subtractive" mechanisms, such as the one originally described by Durlach [J. Acoust. Soc. Am. 35, 1206-1218 (1963)]. Within that framework, deleterious effects that could result from variations in the levels of the stimuli simply do not arise.