Examining the Hyper-Correction of High Confidence Errors

Columbia Univ New York Morningside, New York NY

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): Erroneous responses to general information questions endorsed with higher confidence and given corrective feedback have been found to be more likely to be corrected at a subsequent surprise retest than are errors committed with less confidence and given corrective feedback. This "hyper-correction effect" runs counter to the predictions of the vast majority of memory models. It is proposed that two components underlie this hyper-correction of high confidence errors. The first is domain familiarity, and the second is surprise. The purpose of proposed Experiments 1-3 is to delineate the respective contributions of each component. In Experiment 1, retest is either is immediate, delayed a week, or delayed a month. It is hypothesized that, while the contribution of familiarity in the hyper-correction effect remains stable with retest delay, the contribution of the surprise component decreases. Thus, there should be an overall attenuation in the hyper-correction effect if retest is delayed. In Experiment 2, familiarity with and surprise are measured directly after feedback, and retest is either immediate or week-delayed. It is hypothesized that, while domain familiarity should be independently predictive of error correction at both immediate and week-delayed retest, surprise should be independently predictive at only immediate retest. In Experiment 3, newly learned materials are employed. It is hypothesized that these materials will evidence a weaker hyper-correction effect, as only the surprise and not the familiarity component will underlie the effect. Experiment 4 is an event-related fMRI study whose purpose is to localize the neural substrates of these components.