Figure 5.

Effect of Varying the Mean and Variance of the Stimulus Distribution on Filter Shape and Time Course

(A) Normalized filters derived for one IC neuron from stimulus distributions with different means. Insets show the same filters before mean subtraction.

(B) Filters derived from stimulus distributions with different variances for a different neuron.

(C) Ratio of positive to negative area (P/N ratio) for zero mean versus nonzero mean distributions. Inset shows mean P/N ratio for distributions with different means.

(D) P/N ratio for high- (SD = 20 dB) versus low- (SD = 10 dB) variance distributions. Inset shows mean P/N ratio as a function of stimulus variance.

(E) Filters derived for another neuron to show the effect of stimulus distributions with different means on latency (defined as the position of the negative peak).

(F) Filters derived for a fourth neuron to show the effect of stimulus distributions with different variances on latency.

(G) Filter latencies for zero mean versus nonzero mean distributions. Inset shows average filter latencies for distributions with different means.

(H) Filter latencies for high- (SD = 20 dB) versus low- (SD = 10 dB) variance distributions. Inset shows mean filter latency as a function of variance. Although we used spike trains sampled at 5 kHz to analyze the time course of the filters, we often measured identical latencies for two or more neurons, resulting in many of the data points shown in the scatter plots of (G) and (H) occluding each other. Error bars are ± SEM.