It is often said that we learn best from our mistakes, yet little is known about the neural and cognitive processes subserving the correction of errors in declarative memory tasks. Although recent years have seen some attempts to use cognitive strategies to suppress the high rates of false alarms found in tasks such as Deese-Roediger-McDermott (DRM) paradigm, few such strategies have proven successful. Although event-related potential (ERP) studies have been successful in measuring the rapid neural response to the detection and correction of errors in motor and perceptual judgment tasks, this technique has not yet been used to study these processes in declarative memory. In this proposal we describe a convergence of these cognitive and neuroscience approaches to error processing. Capitalizing on the superior temporal resolution of ERPs, we propose a series of behavioral and ERP experiments that measure the neural response to the detection of retrieval errors in semantic and episodic memory and determine whether this response predicts error correction on unexpected immediate and delayed memory retests. These studies focus on errors that individuals initially endorse as correct with high-confidence because they constitute not only errors in retrieval, but also a mismatch between subjective beliefs regarding one's memory accuracy (i.e., metamemory) and the actual accuracy of the response. Rather than being more difficult to revise, preliminary studies demonstrate that high-confidence errors are more likely to be corrected than low-confidence errors, at least when the information is of a factual nature. Within semantic memory, hypercorrection of high-confidence errors may be a combination of increased arousal consequent to detection of metamemory mismatch paired with the strong likelihood that subjects will be familiar with the correct answer. To disentangle the contribution of these factors to hypercorrection, the first 2 experiments examine this effect in situations where familiarity can facilitate error correction (Experiment 1: trivia question paradigm) and in situations where familiarity must be overcome in order for error correction to be successful (Experiment 2: DRM false-memory paradigm). To evaluate the hypothesis that the conceptual distinctiveness and emotional arousal of metamemory mismatch facilitate hypercorrection, Experiment 3 employs a von Restorff memory paradigm that compares and contrasts neural activity during the detection of perceptual, conceptual and emotional isolates, and evaluates the relationship of this activity to later memory performance. These studies will increase our understanding of the basic cognitive and neural principles underlying detection and correction of memory retrieval and monitoring errors, and have far reaching implications for populations that suffer memory and/or metacognitive deficits
