Sensitivity of auditory cortical neurons to locations of signals and competing noise sources.

The present study examined cortical parallels to psychophysical signal detection and sound localization in the presence of background noise. The activity of single units or of small clusters of units was recorded in cortical area A2 of chloralose-anesthetized cats. Signals were 80-ms click trains that varied in location in the horizontal plane around the animal. Maskers were continuous broadband noises. In the focal masker condition, a single masker source was tested at various azimuths. In the diffuse masker condition, uncorrelated noise was presented from two speakers at +/-90 degrees lateral to the animal. For about 2/3 of units ("type A"), the presence of the masker generally reduced neural sensitivity to signals, and the effects of the masker depended on the relative locations of signal and masker sources. For the remaining 1/3 of units ("type B"), the masker reduced spike rates at low signal levels but often augmented spike rates at higher signal levels. Increases in spike rates of type B units were most common for signal sources in front of the ear contralateral to the recording site but tended to be independent of masker source location. For type A units, masker effects could be modeled as a shift toward higher levels of spike-rate- and spike-latency-versus-level functions. For a focal masker, the shift size decreased with increasing separation of signal and masker. That result resembled psychophysical spatial unmasking, i.e., improved signal detection by spatial separation of the signal from the noise source. For the diffuse masker condition, the shift size generally was constant across signal locations. For type A units, we examined the effects of maskers on cortical signaling of sound-source location, using an artificial-neural-network (ANN) algorithm. First, an ANN was trained to estimate the signal location in the quiet condition by recognizing the spike patterns of single units. Then we tested ANN responses for spike patterns recorded under various masker conditions. Addition of a masker generally altered spike patterns and disrupted ANN identification of signal location. That disruption was smaller, however, for signal and masker configurations in which the masker did not severely reduce units' spike rates. That result compared well with the psychophysical observation that listeners maintain good localization performance as long as signals are clearly audible.