Many thanks again to all brilliant co-authors Andy McWilliams, Flora Chartier, @kndiaye.bsky.social, @smfleming.bsky.social and Marion Rouault!

To conclude, our study provides new insights into the architecture of global confidence across cognitive domains and paves the way for longitudinal studies of local and global metacognitive capacities in patient and developmental cohorts 🧠

Systematic cross-domains and cross-sessions comparison of local metacognitive metrics showed that local metacognitive bias was indeed more reliable between sessions than between domains!

Furthermore, considering a hierarchical organization of metacognition (from local to task-level to domain-level to self-evaluations), we expected that metacognitive estimates would be more stable across time points (test/retest) than across domains (memory/perception).

In memory 💭, we found that both local accuracy and confidence contributed to global confidence, whereas in perception 👁️, global confidence was predicted by local confidence alone, suggesting that it forms from a mixture of domain-specific and domain-general local contributors 🧐

In both domains and during test and retest, participants made decisions in blocks containing 2 item categories and selected the one they believed they performed better on, reflecting global confidence. With this we can estimate the individual contribution of each local cue to global confidence:

Yet, the mechanisms underlying global confidence have received little attention and it remains unknown whether it forms through similar or distinct processes across cognitive domains. We introduce a novel method to investigate self-performance estimates in memory and perceptual test/retest tasks.

While most research on metacognition focuses on local confidence, our ability to assess accuracy on a trial-by-trial basis, real-world decisions often rely on global confidence, a broader judgment of overall success, measured through self-performance estimates.