Fig. 1: Mice bias visual decisions to exploit temporal regularities of stimulus sequences.

a Schematic of the two-alternative visual decision-making task. Head-fixed mice reported the location (left/right) of gratings with varying contrasts by steering a wheel with their forepaws, receiving water reward for correct responses. Adapted from ref. ²¹. https://creativecommons.org/licenses/by/4.0/. b Stimulus sequences of left and right grating presentations followed distinct transition probabilities (left column), interleaved across different days. In the neutral environment, stimulus location was determined randomly (top). In the repeating and alternating environments, stimulus locations were likely repeated (middle) or alternated (bottom) across successive trials. Right column shows example sequences in each environment. Different shades of green and pink denote different stimulus contrasts, varying randomly across trials. c Mice exhibit expert performance, demonstrated by steep psychometric curves with near-perfect performance for easy (high contrast) stimuli (data pooled across environments). Negative and positive contrasts denote stimuli on the left and right sides, and the y-axis denotes the probability of a rightward choice. Black data points show the group average and gray lines indicate individual mice (n = 10 in all panels). The black line indicates the best-fitting probabilistic choice model (see “Methods”). Error bars in all panels depict SEMs. d Psychometric curves conditioned on the previous successful choice direction “left” (green) or “right” (pink). In the repeating environment, mice exhibit a bias to repeat their previous successful choice as indicated by a higher probability to respond “right” when the previous response was “right” rather than “left”. Data points show group averages. Lines show predictions of the probabilistic choice model. e Difference between choice probabilities conditioned on the previous trial’s successful response (right minus left, gray area in (d)). Positive y values indicate a tendency to repeat the previous choice. Data points show group averages. Lines show predictions by the probabilistic choice model. f Choice accuracy on low contrast trials (0 and 6.25% contrast) in three different environments. Mice exhibited a gain in performance in the repeating and alternating over the neutral environment. Gray and black lines depict individual mice and the group average. For choice accuracy on all trials see Supplementary Fig. 1j. *p < 0.05, **p < 0.01, two-sided paired t-tests. Source data are provided as a Source Data file.