More liberal and less sensitive: Individual differences in visual working memory capacity predicts the metacognitive assessment of representational accuracy

Visual working memory (VWM) allows us to actively represent task-relevant visual information so that we can use it to guide our behavior. However, not only is its capacity limited, but its representational accuracy also varies. Thus, to avoid guiding our behaviors on inaccurate VWM representations, we need to metacognitively assess the accuracy of VWM representations. Here, across four experiments (total n = 663), we demonstrated that humans’ metacognitive assessment for VWM representations are suboptimal such that our VWM report is not always accurate even when endorsed with 100% confidence in its accuracy. Furthermore, the poor metacognitive assessment was particularly evident in low-capacity individuals, and it stemmed from two dissociable mechanisms, namely overly liberal confidence assignment and reduced metacognitive sensitivity to VWM representational accuracy. Taken together, by elucidating multiple mechanisms of overconfident errors, our results offer a novel insight into the nature of individual differences in VWM performance.