M. A. Burock
L. H. Carney
Dept. of Biomed. Eng., 44 Cummington St., Boston Univ., Boston, MA 02215
Models for level discrimination based on neural counts require ``pooling'' of auditory-nerve (AN) fiber responses to explain human performance. Neither the mechanism nor site of this ``pooling'' is known. This study concerns the transformation of AN responses in the anteroventral cochlear nucleus (AVCN). An AN model with statistics consistent with physiological data [mean-to-variance ratio ((alpha))(approximately equal to)2] provided inputs to coincidence-detecting models for AVCN cells. The (alpha) of coincidence-detecting cell models is lower than the (alpha) of AN fibers. Responses of most gerbil AVCN neurons to (frozen) wideband noise had (alpha) less than 2, consistent with a coincidence-detection mechanism. The lower (alpha) implies these cells would not be as effective as AN fibers for encoding level in terms of rate. To optimally ``pool'' these responses would require either summation of a greater number of elements, or steeper rate-level curves (to compensate for lower (alpha)) and thus more elements to span a wide dynamic range. Neural coincidence detection may contribute to processing spectral information (monaurally) and localization information (binaurally). This mechanism does not provide an advantage, and perhaps introduces a disadvantage, for processing intensity information in the form of discharge rate. [Work supported by NIH, Whitaker Foundation.]