Fig. 1: Impact of recently obtained rewards on action choice strategy.

a Schematic of a push/pull choice task performed by head-fixed rats based on probabilistic reward. Rats chose to either push or pull the lever depending on the block condition. Reward delivery or absence was instructed by different tones. The timeline represents a push block’s trial, indicating the center holding, lever movement, and outcome timing. The probability of high-tone sound indicating reward delivery was 70% after a correct choice and 10% after an incorrect choice. The tone probability was reversed after at least 30 correct trials and a 79% correct rate in the last 10 trials. See “Methods” section for details. b Representation of individual choices across blocks. Magenta and cyan vertical lines indicate individual choices in push or pull blocks, respectively (long, rewarded; short, non-rewarded). The gray line indicates the probability of a push choice (running average of the last five choices). The green line indicates the average rewarded choices in the past five trials (reward rate). c Number of trials needed to correctly change the choice after a block change. Magenta and cyan lines indicate the probability of a push choice after changing to a block with higher push or pull reward probability, respectively (20 rats; 58 sessions). Error bars represent standard deviation. Each dot represents one rat average. d Choice in upcoming trials was determined by the outcomes of previous trials. Contributions of rewarded outcomes (red) in the previous five trials and non-rewarded (blue) outcomes in the previous trials on choices in the current trial, as derived from logistic regression (58 sessions, 659 trials/session on average); *p < 0.05. Error bars represent standard deviation. On each box, the central mark indicates the median, the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers show the extreme data points not considered outliers. Notches show the 95% confidence interval. e Effect of the reward rate on the next choice. The probability of switching the choice in the next trial after either rewarded or non-rewarded outcome selection in the current trial, with statistical dependence on the reward rate (average ratio of rewards obtained in the last 5 trials). Error bars represent standard deviation. Each dot represents one rat average. f Reaction time and lever movement time correlated with reward rate. Error bars represent standard error of the mean (SEM).