Dynamic combination of sensory and reward information under time pressure
(A) Average speed-accuracy tradeoff across subjects. The average performance (probability of a correct response) is plotted as a function of the average median response time (RT) across all subjects, separately for trials with a given level of time pressure. Schedule 1 to 6 correspond to trials with a deadline at 0.7s, 1.4s, 2.1s, 2.8s, 3.5s, and 4.2s, respectively. Averages for different schedules are color coded similarly in all panels. An asterisk above a line indicates that the difference between the performance values for the corresponding schedules was significant (two-sided sign test, p < 0.05). (B) Performance is plotted as a function of RT binned in ten equal intervals across all subjects. Performance increased with longer RT but plateaued at ~2.25s, indicating the leak in integration of sensory information over time. Note that revealed dots stayed on the screen for 2.0s. The solid curve shows the fit using a modified Weibull function (Eq 4 in Materials and Methods). The histogram shows the distribution of RT across the ten bins. (C) Performance reduced as the deadline approached. For each level of time pressure, we divided RTs into five bins and computed the probability of a correct response for trials in a given bin. The dashed lines indicate deadlines, and the asterisk next to each line indicates that performance for the corresponding schedule was significantly worse for longer RT (significant regression slope; two-sided t-test, p < 0.05). (D) Even with equal integration time, performance was worse under greater time pressure. Plotted is the average performance on RT-matched trials (based on mean) with adjacent levels of time pressure. The asterisk above each pair of histograms indicates that the difference between the performance values for the corresponding schedules was significant (two-sided sign test, p < 0.05).