Abstract
The current study investigated how task-relevant and task-irrelevant noises are incorporated into decision making and confidence computation. To answer this question, we manipulated the levels of white noises embedded in oriented Gabor patches while maintaining their RMS contrast equal. We presented eight Gabor patches with different orientations and asked participants to report the orientation of a target and their confidence simultaneously. The target was designated by a retro-cue. Our variables of interest were the level of noise added to a target (local noise) and the average level of noises added to all items including distractors (global noise). As participants were asked to judge a single target orientation, the local noise was task-relevant and the global noise was task-irrelevant. Overall, we found that confidence computation was more sensitive to noises than decisions at both local and global levels. First, when the level of global noise was low, decision accuracy did not vary depending on the level of local noise. However, the degree of confidence decreased when the target had a higher local noise level regardless of the actual correctness of the decision. Second, when the level of local noise was constant and only global noise increased, decision accuracy remained the same. Interestingly, the degree of confidence increased following the increased global noise regardless of the actual correctness of the decision, suggesting that confidence can become higher when distractors were noisier than the target. Finally, when both the level of local and global noises was high, both decision accuracy and confidence decreased. In sum, our results show that confidence incorporates both task-relevant and task-irrelevant noises whereas decision only partially incorporates task-relevant noise.