Abstract
Most decisions are based on stimuli that appear in the midst of task-irrelevant information. Thus, accurate decision-making critically depends on a selection process that directs attention to task-relevant stimuli, while suppressing irrelevant information. It is often assumed that subjects can increase response accuracy by gathering more evidence before responding. However, the fact that the engagement of selective attention takes time, suggests the possibility of an alternative or additional mechanism for increasing accuracy: subjects may delay decision onset until selective attention has isolated task-relevant information. To test this, human subjects performed a parametric variant of a Stroop-like interference task. In our dots-interference task subjects reported the direction of motion of a set of target dots while ignoring the direction of distractor dots that either moved in the same direction, the opposite direction or orthogonal to the target dots. In different sessions subjects emphasized either speed or accuracy. In addition, subjects performed the same task but were required to synchronize their responses to a timing cue (response signal). We used the response signal paradigm to infer the time course of selective attention in the dot-interference task using an extended version of the standard drift-diffusion model. Data from the response signal paradigm suggests that it takes subjects approximately 120 ms to isolate information from the target dots in a winner-take-all fashion. Based on the time-course of selective attention we modeled reaction time distributions and accuracy in the speed and the accuracy condition of the reaction time version of the same dots-interference task. These results show that decision onset is to some degree under cognitive control and that human subjects not only raise response threshold, but also delay decision onset by ~50 ms to trade speed for accuracy in a Stroop-like interference task.
Meeting abstract presented at VSS 2014