UCLA

Adaptive cognition requires both persistence and flexibility: internal representations must remain stable enough to guide ongoing behavior, yet open enough to be revised when new evidence warrants it. This dissertation examines how this balance is achieved, advancing the view that the exchange between internal representations and external evidence is iterative rather than sequential. Internal states shape how incoming information is interpreted, and incoming information in turn prompts renewed processing — and revision — of the internal states themselves, in cycles that operate across timescales and levels of cognition and that are regulated by metacognitive signals of reliability. Three studies examine this iterative interplay in distinct settings. Study 1, using human intracranial recordings, shows that perception of a brief input unfolds through a temporally extended comparison with recently formed memory, which repeatedly informs the developing perceptual decision rather than concluding in a single match. Study 2 shows that the iteration continues beyond commitment: feedback that conflicts with a confident judgment triggers a re-examination of the internal evidence supporting the original choice, allowing false feedback to be partly resisted. Study 3 shows that the exchange reaches the representations directing behavior: under uncertainty, people revise beliefs about their own goals in light of the observed outcomes of their own actions. Together, the studies indicate that internal representations and external evidence meet repeatedly rather than once, that revision proceeds by returning to what is held rather than overwriting it, and that metacognition — monitoring the reliability of each side and controlling whether another cycle of revision runs — is what keeps this balance adaptive in dynamic environments.