Figure 2: Degenerate strategies produce block sensitivity in the temporal wagering task.

a. Schematic of behavioral paradigm. b. Block structure of task. c. Mean wait times on catch trials by reward in each block for one example, representative rat. d. Foraging theory-inspired behavioral model compares the value of the reward offer to the opportunity cost, or what the rat might miss out on by continuing to wait, which differs in each block. The starting point of the value function reflects the reward offer. e. Simulations of models that use different strategies to estimate the opportunity cost of time. The inference model infers the most likely block using Bayes’ Rule. The model-free reinforcement learning model estimates the opportunity cost as a running average of past rewards by a delta rule (simulation was with a learning rate of 0.3). The one trial back model treats the reward offer on the previous trial as the opportunity cost. The divisive normalization model divides the current offer by the average of past rewards. Units are arbitrary as scaling factors can determine the actual wait times in seconds.