Coding of complex temporal and spectral features in the auditory cortex of awake primates

Natural sounds often contain energy over a broad spectral range and consequently overlap in frequency when they occur simultaneously, yet such sounds under normal circumstances can be distinguished perceptually (e.g., the cocktail party effect). Sound components arising from different sources have distinct (incoherent) modulations, and incoherence appears to be one important cue used by the auditory system to segregate sounds into separately perceived acoustic objects. In the primary auditory cortex of awake primates, many neurons responsive to amplitude or frequency modulated tones also demonstrate sensitivity to temporal stimulus coherence (measured by relative modulation phase), most commonly exhibiting maximally suppressed firing for the coherent condition. Coherence sensitivity usually exists in a portion of the inhibitory flank, indicating that only some inputs away from the characteristic frequency (CF) interact with inputs at CF on the same time scale. These results reveal that auditory cortex neurons receive off-CF inputs both temporally matched and unmatched to the CF input and respond in a fashion that could be utilized by the auditory system to segregate natural sounds containing similar spectral components (such as vocalizations from multiple conspecifics) based upon stimulus coherence.; Many auditory cortex neurons, especially in secondary areas, respond poorly to tones or very narrowband stimuli but can respond to filtered noise of relatively wide band-widths, although characterization with such stimuli incompletely describes neuronal responsiveness. Without a basic spectral characterization, more complex properties such as temporal coherence sensitivity cannot be ascertained. Parametric, static-spectrum, wide-band stimuli were designed to fill the gap between tones and noise, providing a common stimulus type to probe both primary and secondary auditory cortical fields for spectral preferences. Linear models of the neurons' rate functions provide poor predictions of behavior in response to arbitrary stimuli, implying frequency-dependent nonlinearities in cortical rate functions. Stimuli tailored to each neuron's rate function reveal specific nonlinearities resulting in low or high spectral contrast or spectral density preferences. Low-contrast preferences in particular were unanticipated. Overall, spiking rates in response to properly designed wideband stimuli are usually high and sustained, indicating that this method can identify near-optimal acoustic stimuli for most auditory cortex neurons.