Figure 9.

Differences in decoding performance between ON and OFF populations are eliminated when populations are equated for the firing-rate range SNR measure specific to the CM stimulus. A, The rejection-sampling procedure selects populations of ON and OFF cells for which the SNR measure on the dark-object CM stimulus is matched. B, Performance of the matched populations of A in the dark-object CM task decoding task. Unlike the measures tested in Figure 7, matching for the SNR measure eliminates decoding performance differences between ON and OFF populations. C, The SNR measure for the light-object CM stimulus exhibited less disparity between ON and OFF populations. D, Performance of the matched populations of C in the light-object CM task decoding task were similar to the randomly chosen populations, with no significant difference between ON and OFF cells. Asterisks (significance) and error bars were calculated as in Figure 3C.