Binaural localization: influence of stimulus frequency and the linkage to covert peak areas.

The influence of selectively filtering a broadband stimulus on binaural localization was investigated. First, head-related-transfer-functions were obtained by placing a miniature microphone at the entrance of the ear canal and presenting broadband noise bursts from each of 104 loudspeakers arrayed in the listener's left hemifield. The microphone's output was transformed into frequency spectra using a Fast Fourier Transform. The microphone and loudspeaker characteristics were accounted for by repeating the procedure with the microphone suspended in space. The in-ear data were divided by the in-space data thereby providing an account of the pinna's interaction with the incident sound wave. Extracted from these data were the covert peak areas (CPAs) associated with different frequency segments. A CPA was defined as the spatial location of those loudspeakers, which when generating the stimulus, produced a sound pressure level at the ear canal entrance within 1 dB of the maximum level recorded for a particular frequency segment. A series of localization tests was conducted using a bandstop stimulus--one in which differently-centered 2.0-kHz wide frequency segments were filtered from a broadband noise. We predicted that when a given frequency segment was filtered, binaural listeners would less often report a sound as originating from the CPA associated with that segment compared to their performances when the sound was unfiltered. This prediction was substantiated by the data (P < 0.0001). While localization accuracy was decreased for the filtered stimuli, the decrement was significantly greater (p < 0.01) for sounds originating in the CPA. We interpreted the results to mean that monaural spectral cues contribute significantly to the accuracy of binaural localization and that the basis of the contribution is the spatial referents of stimulus frequencies.