Fig 7. An optimal model of RT for dynamic combination of sensory and reward information under time pressure and comparison between the optimal and actual RT in Experiment 2.

(A) Tradeoff between improved performance and less potential reward over time. The probability of a correct response (left axis) increases with longer integration time (the SAT curve in gray), while the reward harvested upon a correct response decreases over time (blue, right axis). Black circles show actual performance on trials with different levels of time pressure (based on data from an individual subject in Experiment 1), and the gray line shows the fit based on the modified Weibull function (Eq 1 in Materials and Methods). (B) The expected gain (EG), equal to the product of the SAT and reward function, is plotted as a function of time. An ideal decision maker uses the EG over time to pick the optimal RT. (C) Comparison between the actual RT and the optimal RT. Subjects systematically used less time than what is optimally prescribed to make decisions under time pressure for intermediate and slow schedules (schedules 3–6). Each circle marker represents the mean RT (± s.e.m.) of all subjects in a particular schedule against the optimal RT predicted by the ideal decision maker. Each histogram indicates the count distribution of median RT across subjects, and the horizontal dashed and solid lines represent the mean of each distribution and the optimal RT, respectively. The asterisk indicates statistical significance for the difference between the actual and optimal RT (two-sided sign test, p < 0.05).