Abstract
We investigated how humans trade off speed against accuracy in a modified random-dot motion task where the reward for making correct direction judgment decreased over time. To perform well in this task, the subjects needed to choose a response time that appropriately balanced the probability of making correct judgment (an increasing function of time) and the reward schedule (a decreasing function of time). We developed an optimal model of response time based on statistical decision theory. In this model, an ideal decision maker takes into account knowledge of his/her own speed-accuracy tradeoff (SAT) and the decreasing reward schedule to choose a response time so as to maximize expected reward. Method. Subjects were first trained to make direction judgment under 20 different conditions (5 different time constraints x 4 motion coherence levels). The subjects received a small and fixed monetary reward (1 cent) for making a correct judgment within the time limit and a relatively large penalty (5 cents) if s/he did not respond within time limit. This session allowed us to separately estimate SAT for each of the 4 coherence levels. In the following session, for each coherence level, we implemented 3 different decreasing reward schedules. The subjects could freely choose the time s/he preferred to make direction judgment but the amount of reward received if correct depended on the time the subjects made a button press to indicate his/her direction judgment. Results. Four subjects participated in the experiment. Each subject exhibited unique SAT profiles that varied as a function of coherence. All subjects adjusted mean response time when facing different reward schedules. Two subjects’ performance was indistinguishable from optimal in their timing, while the other two failed to adjust response time appropriately predicted by the optimal model.
Meeting abstract presented at VSS 2012