Responses of auditory cortical neurons to pairs of sounds: correlates of fusion and localization.

When two brief sounds arrive at a listener's ears nearly simultaneously from different directions, localization of the sounds is described by "the precedence effect." At inter-stimulus delays (ISDs) <5 ms, listeners typically report hearing not two sounds but a single fused sound. The reported location of the fused image depends on the ISD. At ISDs of 1-4 ms, listeners point near the leading source (localization dominance). As the ISD is decreased from 0.8 to 0 ms, the fused image shifts toward a location midway between the two sources (summing localization). When an inter-stimulus level difference (ISLD) is imposed, judgements shift toward the more intense source. Spatial hearing, including the precedence effect, is thought to depend on the auditory cortex. Therefore we tested the hypothesis that the activity of cortical neurons signals the perceived location of fused pairs of sounds. We recorded the unit responses of cortical neurons in areas A1 and A2 of anesthetized cats. Single broadband clicks were presented from various frontal locations. Paired clicks were presented with various ISDs and ISLDs from two loudspeakers located 50 degrees to the left and right of midline. Units typically responded to single clicks or paired clicks with a single burst of spikes. Artificial neural networks were trained to recognize the spike patterns elicited by single clicks from various locations. The trained networks were then used to identify the locations signaled by unit responses to paired clicks. At ISDs of 1-4 ms, unit responses typically signaled locations near that of the leading source in agreement with localization dominance. Nonetheless the responses generally exhibited a substantial undershoot; this finding, too, accorded with psychophysical measurements. As the ISD was decreased from ~0.4 to 0 ms, network estimates typically shifted from the leading location toward the midline in agreement with summing localization. Furthermore a superposed ISLD shifted network estimates toward the more intense source, reaching an asymptote at an ISLD of 15-20 dB. To allow quantitative comparison of our physiological findings to psychophysical results, we performed human psychophysical experiments and made acoustical measurements from the ears of cats and humans. After accounting for the difference in head size between cats and humans, the responses of cortical units usually agreed with the responses of human listeners, although a sizable minority of units defied psychophysical expectations.