Fig. 6. General linear encoding model (GLM) reveals PFC subregion-specific patterns of information encoding across task phases of a working memory task.

After removing the poke context regressor, a GLM uncovers the percent of neurons in each region that significantly encode working memory task variables (sample ports, choice ports, and outcome (correct vs incorrect)) at key timepoints. a The region with the highest percentage of significant sample location neurons shifts from MOs to dmPFC as the mice progress through task phases. b Prospective choice port location is not detectable in any region until the choice phase, during which vmPFC shows the highest selectivity. c The trial outcome (correct vs incorrect), is similarly not encoded by any region until the choice is made. Each region encoded the outcome (correct vs incorrect) after the choice was made, with the MOs having the largest proportion of outcome-encoding neurons. Double-colored straight lines in any panel represent statistically significant differences (p-value < .05) between the two respective subregions in that time bin, after correcting for both false discovery rate and family-wise error rate.