Extended Data Fig. 6. Significance of choice decoder weights.

To assess significant weights of the choice decoder, we combined spatially downsampled choice maps from all sessions in each PyN type and subsequently performed a t-test in each pixel to determine which decoder weights are significantly different from zero. The resulting maps show significant pixels for different trial periods in white (t-test, p < 0.05, bonferroni-corrected for 3364 pixels). Significant pixels closely match choice decoder weights (Fig. 6c) with significant regions being largely tied to anterior cortex. In all PyN types, weights in frontal cortex are significant in the stimulus and delay period, thus supporting the main conclusions of the choice kernel analysis.