Figure 1.
Experimental task and behavioral results
(A) Animals tracked the number of dots on the screen and responded by touching a response pad at a time of their choice. The color of the frame and the dots represented the potential reward magnitude on that trial. The patterns on each side of the screen represented ITI duration.
(B) Timeline of one trial. An empty frame appeared on the left or right side of the screen. It was gradually filled with dots emerging from top to bottom appearing every 100, 200, or 300 ms, depending on trial type.
(C) The probability of getting reward increased as more dots appeared on the screen, following a sigmoid curve.
(D) Features of the immediate present and recent past context predicted animals’ actTime.
(E) Compared with a balanced design7 (left panel), there were more high reward magnitude and fast dot speed offers in the biased (right panel) design. The relative value of “medium offer” trials (thick box border) was therefore higher in the balanced design.
(F) Distribution of observed actTime across all trials in data collected using balanced and biased designs.
(G) The effect of the immediate context on actTime was not statistically different between environments (Figure S1 illustrates main effect of immediate context on actTime). The interaction coefficients from GLM1.1. Env. is the broader, general environment (balanced versus biased); Rew is reward magnitude; Speed is dot speed; ITI is inter-trial interval; PastRew is the reward obtained on the preceding trial; PastactTime is the observed actTime on the preceding trial. Error bars are standard error of the estimated coefficients.
(H) Monkeys waited longer before responding on “medium offer” trials in the biased compared with balanced environment, where the value of the “medium offer” was worth less than the average value of the environment. HRV, higher relative value; LRV, low relative value.
(I) There was no significant difference between the missed “medium offer” trials between the two environments. In (H) and (I), the gray columns are the mean across animals, error bars are the SEM across animals, and each line is data from individual animals. ∗p < 0.05.