Extended Data Fig. 5: Comparing behavioral and neural discounting and decoding reward timing in single animals.

(a-c): mouse 3044. a, left panel. Normalized lick responses to the cues predicting reward delays across the population. For each neuron, the response was normalized to the highest response across the 4 possible delays. Neurons are sorted by the inferred behavioral discount factor. Right panel: Normalized neural responses to the cues predicting reward delays across the population (sorted by the behavioral discount factor). b, The behavioral and neural discount factors are not correlated (r = −0.29, P = 0.27, Spearman’s rank correlation, two-tailed Student’s t-test). c, Discount matrix for the neurons recorded in mouse 3044. This is the matrix used for decoding in panel g, top row. (d-f): same as panels (a-c) for mouse 3054. e, The behavioral and neural discount factors are not correlated in mouse 3054 (r = −0.029, P = 0.9, Spearman’s rank correlation, two-tailed Student’s t-test). f, Discount matrix for the neurons recorded in mouse 3054. This is the matrix used for decoding in panel g, bottom row. g, Decoding of reward timing at the single animal level for mouse 3044 (top row) and mouse 3054 (bottom row). The decoding is present but slightly less accurate as expected from the smaller number of neurons. h, discount factor inferred for neurons in mouse m3044 when dividing trials between low and high anticipatory lick rate. left panel, scatter plot of the value across neurons. right panel, the distribution across neurons of differences in inferred discount across the two conditions is not significant (mean = −0.0024, P = 0.96, two-tailed Student’s t-test). i, Same as panel h for mouse m3054. The difference in inferred value between low and high lick rate is significant (mean = 0.086, P = 0.0062, two-tailed Student’s t-test) but the mean effect is small compared to the standard deviation of inferred discount factors across neurons (s.d. = 0.19 for neurons in m3054).