Fig. 2: Population recording from the V1 of an example mouse.

a Spike raster for the recorded population of 31 neurons in five subsequent multimodal trials in an example animal. Both stimulus presentation period (gray background) and intertrial intervals (white background) are shown. Letters labels indicate neurons on (b). b Sensitivity of four example neurons (rows) to different task-relevant variables (columns). Across-trial average firing rates for two values of the particular variable (dark and light lines, shadings around lines denote 2 s.e.m., values are indicated in the legend). Sensitivity to a particular task variable (difference in means normalized by the standard deviation and averaged across time points) are shown both for the stimulus presentation period and for the intertrial interval (numbers at the top of panels). c Task-relevant modulation of population activity as shown by Principal Component Analysis (PCA) performed on the stimulus presentation period. Time course of population activity projected on the first three principal components (three rows) averaged across different values of the four task-relevant variables (four columns). Colors are matched with those on (b). d Relative signal variance of population activity associated with the four different task-relevant variables (four columns). Variance is normalized by the total variance of the population responses. Relative variances are shown cumulatively for activities projected on a growing subset of PCs. Inset shows signal and total variances on separate vertical scales, the numerator (dark blue), and total variance (red), the denominator, of the relative variance shown in the main panel. Empty circles correspond to the total variance. A baseline relative variance is established by using a random orthogonal projection matrix (mean, and s.e.m. of 20 random projections, gray dashed lines and bands, respectively).