(A) Normalized activity of all pixels in the imaging plane following CS+ presentation on the third day of each odor set (A3 and B3, days 5 and 6 of training). (B) Fraction of neurons coding for cues, licks, rewards, and their combinations in A3 and B3 (days 5 and 6). (C) Mean (± SEM, across mice) correlation between activity predicted by odor set A3 models and its training data (A3, cross-validated) or activity in B3, for true (black) and trial shuffled (gray) activity. Thin lines are individual mice. , for main effect of odor set, , for main effect of shuffle, , for interaction, mice, two-way ANOVA. (D) Mean (± SEM, across mice) unique cue, lick, and reward variance for cells classified as coding cues, licks, both, or neither for odor set A. For each category, odor set A unique variance preference was maintained for odor set B () except for both cells, for which lick and reward variance were not different in odor set B (, Bonferroni-corrected, mice). (E) Distribution of best model fits across all cue cells, with colors from Figure 3C. Dashed line is chance proportion when assuming even distribution. (F) Left: normalized activity of every value cell, sorted by mean firing 0–1.5s following odor set A CS+ onset. Right: mean normalized activity of all value cells, grouped by whether peak cue activity (0–2.5 s) was above (top) or below (bottom) baseline in held out trials. Number of neurons noted for each plot. (G) As in (E), for value-like cells. (H) Mean (± SEM, across neurons) activity of cue cells tracked across A1, A2, and A3 with preferential CS+ firing, defined on half of A3 trials and plotted for the other half of A3 trials and all of A1 and A2 trials. (I) For neurons in (H), correlation between a neuron’s activity in one session and its own activity in the subsequent session, quantified as a percentile out of correlations with the activity of all other neurons on the subsequent day. Plotted as the median for each subject ( with CS+ preferring cue cells) and the mean (± SEM) across these values. Real data was more correlated than shuffled data ( A1:A2, A2:A3, Wilcoxon signed-rank test).