Abstract:

A source identification experiment measures the ability of a listener to identify the location of a sound source. The experimental results are in the form of an rms error and a variability. The experiment can be described in terms of a statistical model in which the listener's responses are determined by the width and bias of an internal distribution. Model calculations were performed to aid the design of source identification experiments. Calculations in the limit of small spans (<<180 deg) showed that the experimental variability provides a good estimate of the width of the internal distribution if both the bias and the source spacing are not greater than the width. The results are insensitive to the number of sources in the array. Specific models were developed for dealing with large biases, a frequent experimental occurrence. Calculations in the limit of large spans showed that experiments that ignore front-to-back reversals may lead to misleading results. Experimental tests of the models, using four human listeners in experiments having 3, 6, 12, or 24 sources, generally corroborated the models for both easy and difficult localization conditions. [Work supported by the NIDCD, DC00181.]