Fig. 1: Quantifying metacognitive accuracy within an evidence accumulation framework.

A Noisy sensory evidence accumulates over time, until the integrated evidence reaches one of two decision boundaries (a or 0). After the decision boundary is reached, evidence continues to accumulate. Model confidence is quantified as the integrated evidence after post-decisional evidence accumulation. B Histograms of model-predicted confidence for different levels of v-ratio (reflecting the ratio between post-decision drift rate and drift rate). Higher levels of v-ratio are associated with better dissociating corrects from errors. C Simulations from this dynamic evidence accumulation model show that v-ratio captures variation in M-ratio (r = 0.436; left panel), and critically, that M-ratio is also related to the differences in decision boundary (r = −0.552; middle panel). By design, decision boundary and v-ratio are unrelated to each other (r ∼ 0; right panel). Data are based on N = 100 simulations. Source data are provided as a Source Data file.