Fig 5. Subjects were faster on trials in which the difference between the estimated reward probabilities of the competing objects deviated more from zero.

(A-D) Panels A-D correspond to Experiments 1 to 4, respectively, and the error bars indicate s.e.m. (E-H) The same as in A-D but showing the average actual RT (median removed) over all subjects. Overall, the difference in estimated reward probabilities of the two alternative objects resulted in up to ~150 ms increase in RT when reward probabilities were close to each other relative to when they were very different.