Fig. 2: New associative learning led to sequential refinements of reward-seeking behaviors.

a Behavioral task. Rats entered the fixation port to initiate each trial, where they encountered three trial types (S^left; S^right; catch) with equal probabilities that respectively indicated water reward in the left or right port, or no reward in the case of catch trials (no stimulus). Task symbols were adapted from Avila and Lin²². b Rats initially learned an auditory discrimination task (old association phase). At the new learning phase, S^right was switched from the sound to a new house light, while other elements of the task remained the same. Symbols were adapted from Avila and Lin²². c The proportion of three types of reward-seeking behaviors toward the right reward port (light licks, catch licks, no-fixation licks) across sessions during new learning in individual animals (N = 7 rats), and their mean (±s.e.m.). No-fixation licks (cyan) refers to trials in which rats failed to first enter the fixation port before licking in the right reward port. Sessions were respectively aligned, in each column, at the D₀, D₁ or D₂ session of each animal. D₀ refers to the first new learning session with the new light stimulus; D₁ refers to the session when animals began to respond correctly in the new light trial; D₂ refers to the session when catch licks peaked. The D₀, D₁ and D₂ sessions in each animal were indicated by red, cyan, and black circles, respectively. Each row in top panels depicts behavior performance in one animal (#1–7). In the middle and right panels, only sessions in the new learning phase (starting from the D₀ session) were plotted. The emergence, as well as the following sequential elimination, of no-fixation licks and catch licks resembled the sequential refinement of reward-seeking behaviors under the stepwise learning strategy. Source data for this and all subsequent figures are provided as a Source Data file. d Behavioral sequences that animals might experience. Out of all possible sequences, only three types of rightward licking behaviors were consistently observed during new learning. Symbols were adapted from Avila and Lin²². e A stepwise learning model that accounts for the sequential refinement of the three types of rightward licking behaviors. e1 The behavioral events in the model arranged in the format as in Fig. 1b. e2 Behavioral sequences learned as reward predictors at the three discrete steps of learning, along with the compatible and incompatible behavioral sequences at each step. e3 Sequential refinement of the three types of rightward licking behaviors arranged in the format as in Fig. 1b.