Subjects initiated trials in a central nose-poke port, which was flanked by left and right poke ports (Figure S8A). Trial initiation caused the left and right pokes to light up, subjects then chose between them for the chance of obtaining a water reward. Reward probabilities changed in blocks, with three block types; left good (left = 0.75/right = 0.25), neutral (0.5/0.5) and right good (0.25/0.75). Block transitions from non-neutral blocks were triggered 10 trials after an exponential moving average (tau = 8 trials) crossed a 75% correct threshold. Block transitions from neutral blocks occurred with probability 0.1 on each trial after the 15th of the block to give an average neutral block length of 25 trials. Mice tracked the correct option (Figures S8B and S8C), choosing correctly at the ends of blocks with probability 0.80 ± 0.04 (mean ± SD), and adapting to reversals with a time constant of 3.57 trials (exponential fit tau). We assessed how previous trials affected the current choice using a logistic regression analysis with previous choices and outcomes as predictors (Figure S8D). Both previous choices and outcomes predicted the current choice with decreasing influence at increasing lag.
