Figure 14.

Neural decoding of temporal information in Experiment 2. A, Classification of sample intervals into short and long periods based on activity of all recorded units (n = 372) in Blocks 1, 2, and 3 (B1, B2, and B3, respectively). Same format as in Figure 10C. B, Results of a neuron-dropping analysis for A. C, Neural decoding of elapsed time. The 8 s time interval in the second block was divided into 10 equal-duration bins and the order of the middle eight bins was decoded based on activity of all recorded units within each bin. Same format as in Figure 10G. D, Results of a neuron-dropping analysis for C. E, Precision of temporal discrimination as a function of elapsed time. Each sample interval was divided into 10 equal-duration bins and Mahalanobis distance was calculated for each pair of adjacent bins. The slopes are significantly different from 0 for all curves (p < 0.040). F, Mahalanobis distance for the simulated neural data obtained from linearly (red) or logarithmically (black) changing functions during the longest sample interval (8 s). The slope is significantly different from 0 for the logarithmic (p = 0.003), but not for linear function (p = 0.433). The error bars (SEM) are too small to see. Same analysis procedure and same format as in Figure 9. G, All neurons were grouped into quintiles according to their PC1 loading values and their mean normalized activity (z-score, left) and SD (right) are shown for all six interval durations in 50 ms time bins. The bottom panels show mean normalized activity and mean SD for all neurons. Same format as in Figure 8B.