Reporting confidence decreases response and change-of-mind accuracy in a perceptual decision task

Self-monitoring seems to be crucial for regulatory behavior, but it is not clear how it influences performance in simple cognitive tasks. Some studies suggest that increased monitoring improves metacognitive regulation and enhances performance, while others suggest it impairs learning, problem-solving or perceptual processes. We investigated whether the requirement to report confidence in perceptual decisions affects metacognitive regulation and response accuracy. Participants performed a visual discrimination task in which they provided two responses: initial and final. Depending on the condition, participants reported their confidence (either together with or following the initial decision), performed an additional task, or were asked to observe a blank screen between two responses. We expected that reporting decision confidence would induce efficient regulatory activity, which would benefit final accuracy. In three experiments, we did not find evidence that rating confidence improves regulatory processing or performance in perceptual tasks. Rather, when confidence ratings were retrospective, the final response improvement was smaller compared to the condition with no additional task, and changes of mind were less frequent and less corrective. Confidence ratings given jointly with the initial response generally decreased accuracy. The results suggest that deliberate monitoring might put additional strain on cognitive resources and impair lower-order task processing.