Representation of temporal features of complex sounds by the discharge patterns of neurons in the owl's inferior colliculus.

The spiking pattern evoked in cells of the owl's inferior colliculus by repeated presentation of the same broadband noise was found to be highly reproducible and synchronized with the temporal features of the noise stimulus. The pattern remained largely unchanged when the stimulus was presented from spatial loci that evoke similar average firing rates. To better understand this patterning, we computed the pre-event stimulus ensemble (PESE)-the average of the stimuli that preceded each spike. Computing the PESE by averaging the pressure waveforms produced a noisy, featureless trace, suggesting that the patterning was not synchronized to a particular waveform in the fine structure. By contrast, computing the PESE by averaging the stimulus envelope revealed an average envelope waveform, the "PESE envelope," typically having a peak preceded by a trough. Increasing the overall stimulus level produced PESE envelopes with higher amplitudes, suggesting a decrease in the jitter of the cell's response. The effect of carrier frequency on the PESE envelope was investigated by obtaining a cell's response to broadband noise and either estimating the PESE envelope for each spectral band or by computing a spectrogram of the stimulus prior to each spike. Either method yielded the cell's PESE spectrogram, a plot of the average amplitude of each carrier-frequency component at various pre-spike times. PESE spectrograms revealed surfaces with peaks and troughs at certain frequencies and pre-spike times. These features are collectively called the spectrotemporal receptive field (STRF). The shape of the STRF showed that in many cases, the carrier frequency can affect the PESE envelope. The modulation transfer function (MTF), which describes a cell's ability to respond to time-varying amplitudes, was estimated with sinusoidally amplitude-modulated (SAM) noises. Comparison of the PESE envelope with the MTF in the time and frequency domains showed that the two were closely matched, suggesting that a cell's response to SAM stimuli is largely predictable from its response to a noise-modulated carrier. The STRF is considered to be a model of the linear component of a system's response to dynamic stimuli. Using the STRF, we estimated the degree to which we could predict a cell's response to an arbitrary broadband noise by comparing the convolution of the STRF and the envelope of the noise with the cell's post-stimulus time histogram to the same noise. The STRF explained 18-46% of the variance of a cell's response to broadband noise.