Dynamic Sources of Evidence Supporting Confidence Judgments and Error Detection

Our decisions are accompanied by a subjective sense of confidence about whether the choices we have made are correct or erroneous. Here we investigate the information on which these confidence judgments are based, and how they relate to the decision itself, by studying how fluctuations in perceptual information influence decisions and second-order metacognitive evaluations of confidence and accuracy. Human participants judged which of two dynamically changing stimuli contained more dots, under instructions emphasizing either speed or accuracy. Crucially, stimuli remained visible after the decision, before participants rated their confidence in their choice. We found that confidence and error detection depended on the balance of stimulus evidence accumulated in the periods both preceding and following the initial decision, regardless of speed-accuracy instruction. These findings suggest a shared computational basis for error detection and confidence judgments, with implications for current models of metacognitive evaluation of decision processes