Fig. 2: Mice adapt their previous choice weight to different temporal regularities across multiple days.

a History kernels comprising the past seven successful choice weights of the probabilistic choice model (“Methods”; see Supplementary Fig. 1 for the full set of regression weights and parameter recovery analysis). While mice are biased by several past choices, only the previous (1-back) choice weight differs across environments. Error bars in all panels depict SEMs. Dots parallel to x-axis indicate weights significantly different from baseline, two-sided permutation test based on shuffled trial history, Bonferroni-corrected, p < 0.007. Sample size was n = 10 mice in panels a, b and h. b 1-back successful choice weight across environments for each mouse (gray lines) and group average (black). One-sided t-tests; repeating vs alternating: t(9) = 3.11, p = 0.006; repeating vs neutral: t(9) = 1.97, p = 0.04; alternating vs neutral: t(9) = −2.76, p = 0.01. c 1-back successful choice weights estimated on the first, second, and third day of alternating sessions following a neutral session (n = 6 mice). d Same as in (c), but for repeating sessions (n = 6 mice). e Choice history kernels for neutral sessions conditioned on the temporal regularity experienced on the preceding day (solid/circle: repeating; dashed/square: alternating; n = 9 mice in panels e, f and g). f 1-back successful choice weight of neutral sessions preceded by repeating (circle) or alternating (square) sessions in each mouse (gray lines) and across the population (green line). Stars denote results of one and two-sided t-tests (see main text). g Difference in choice probabilities conditioned on the previous trial’s successful choice in neutral sessions preceded by repeating (solid/circle) or alternating (dashed/square) sessions. The differential impact of the previous regularity is most pronounced when current contrast is low. h Difference in choice probabilities conditioned on the previous trial’s successful response split according to whether the previous trial’s stimulus contrast was high (black) or low (gray). Mice are more likely to repeat the previous choice when it was based on a low rather than high contrast stimulus. *p < 0.05, **p < 0.01. Source data are provided as a Source Data file.