Low-confidence Errors Research Articles

Neural correlates of metacognition across the adult lifespan

Metacognitive accuracy describes the degree of overlap between the subjective perception of one's decision accuracy (i.e. confidence) and objectively observed performance. With older age, the need for accurate metacognitive evaluation increases; however, error detection rates typically decrease. We investigated the effect of ageing on metacognitive accuracy using event-related potentials (ERPs) reflecting error detection and confidence: the error/correct negativity (Ne/c) and the error/correct positivity (Pe/c). Sixty-five healthy adults (20 to 76 years) completed a complex Flanker task and provided confidence ratings. We found that metacognitive accuracy declined with age beyond the expected decline in task performance, while the adaptive adjustment of behaviour was well preserved. Pe amplitudes following errors varied by confidence rating, but they did not mirror the reduction in metacognitive accuracy. Ne amplitudes decreased with age for low confidence errors. The results suggest that age-related difficulties in metacognitive evaluation could be related to an impaired integration of decision accuracy and confidence information processing. Ultimately, training the metacognitive evaluation of fundamental decisions in older adults might constitute a promising endeavour.

"A" for Effort: Rewarding Effortful Retrieval Attempts Improves Learning From General Knowledge Errors in Women.

Previous research has shown that the prospect of attaining a reward can promote task-engagement, up-regulate attention toward reward-relevant information, and facilitate enhanced encoding of new information into declarative memory. However, past research on reward-based enhancement of declarative memory has focused primarily on paradigms in which rewards are contingent upon accurate responses. Yet, findings from test-enhanced learning show that making errors can also be useful for learning if those errors represent effortful retrieval attempts and are followed by corrective feedback. Here, we used a challenging general knowledge task to examine the effects of explicitly rewarding retrieval effort, defined as a semantically plausible answer to a question (referenced to a semantic knowledge database www.mangelslab.org/bknorms), regardless of response accuracy. In particular, we asked whether intermittent rewards following effortful incorrect responses facilitated learning from corrective feedback as measured by incidental learning outcomes on a 24–48 h delayed retest. Given that effort-contingent extrinsic rewards represent the intersection between an internal locus of control and competency, we compared participants in this “Effort” group to three other groups in a between-subjects design: a Luck group that framed rewards as related to participant-chosen lottery numbers (reward with internal control, not competence-based), a random Award group that framed rewards as computer generated (no control, not competence-based), and a Control group with no reward, but matched on all other task features. Both men and women in the Effort group showed increased self-reports of concentration and positive feelings following the receipt of rewards, as well as subjective effort on the retest, compared to the Control group. However, only women additionally exhibited performance benefits of effort framing on error correction. These benefits were found for both rewarded and non-rewarded trials, but only for correction of low confidence errors, suggesting that effort-contingent rewards produced task-level changes in motivation to learn less familiar information in women, rather than trial-level influences in encoding or consolidation. The Luck and Award groups did not demonstrate significant motivational or behavioral benefits for either gender. These results suggest that both reward context and gender are important factors contributing to the effectiveness of rewards as tools to enhance learning from errors.

Beyond hypercorrection: remembering corrective feedback for low-confidence errors

ABSTRACTCorrecting errors based on corrective feedback is essential to successful learning. Previous studies have found that corrections to high-confidence errors are better remembered than low-confidence errors (the hypercorrection effect). The aim of this study was to investigate whether corrections to low-confidence errors can also be successfully retained in some cases. Participants completed an initial multiple-choice test consisting of control, trick and easy general-knowledge questions, rated their confidence after answering each question, and then received immediate corrective feedback. After a short delay, they were given a cued-recall test consisting of the same questions. In two experiments, we found high-confidence errors to control questions were better corrected on the second test compared to low-confidence errors – the typical hypercorrection effect. However, low-confidence errors to trick questions were just as likely to be corrected as high-confidence errors. Most surprisingly, we found that memory for the feedback and original responses, not confidence or surprise, were significant predictors of error correction. We conclude that for some types of material, there is an effortful process of elaboration and problem solving prior to making low-confidence errors that facilitates memory of corrective feedback.

Learning from Errors.

Although error avoidance during learning appears to be the rule in American classrooms, laboratory studies suggest that it may be a counterproductive strategy, at least for neurologically typical students. Experimental investigations indicate that errorful learning followed by corrective feedback is beneficial to learning. Interestingly, the beneficial effects are particularly salient when individuals strongly believe that their error is correct: Errors committed with high confidence are corrected more readily than low-confidence errors. Corrective feedback, including analysis of the reasoning leading up to the mistake, is crucial. Aside from the direct benefit to learners, teachers gain valuable information from errors, and error tolerance encourages students' active, exploratory, generative engagement. If the goal is optimal performance in high-stakes situations, it may be worthwhile to allow and even encourage students to commit and correct errors while they are in low-stakes learning situations rather than to assiduously avoid errors at all costs.

On Teaching Old Dogs New Tricks

Although older adults rarely outperform young adults on learning tasks, in the study reported here they surpassed their younger counterparts not only by answering more semantic-memory general-information questions correctly, but also by better correcting their mistakes. While both young and older adults exhibited a hypercorrection effect, correcting their high-confidence errors more than their low-confidence errors, the effect was larger for young adults. Whereas older adults corrected high-confidence errors to the same extent as did young adults, they outdid the young in also correcting their low-confidence errors. Their event-related potentials point to an attentional explanation: Both groups showed a strong attention-related P3a in conjunction with high-confidence-error feedback, but the older adults also showed strong P3as to low-confidence-error feedback. Indeed, the older adults were able to rally their attentional resources to learn the true answers regardless of their original confidence in the errors and regardless of their familiarity with the answers.

The role of prior knowledge in error correction for younger and older adults

Previous work has demonstrated that, when given feedback, younger adults are more likely to correct high-confidence errors compared with low-confidence errors, a finding termed the hypercorrection effect. Research examining the hypercorrection effect in both older and younger adults has demonstrated that the relationship between confidence and error correction was stronger for younger adults compared with older adults. However, recent work suggests that error correction is largely related to prior knowledge, while confidence may primarily serve as a proxy for prior knowledge. Prior knowledge generally remains stable or increases with age; thus, the current experiment explored how both confidence and prior knowledge contributed to error correction in younger and older adults. Participants answered general knowledge questions, rated how confident they were that their response was correct, received correct answer feedback, and rated their prior knowledge of the correct response. Overall, confidence was related to error correction for younger adults, but this relationship was much smaller for older adults. However, prior knowledge was strongly related to error correction for both younger and older adults. Confidence alone played little unique role in error correction after controlling for the role of prior knowledge. These data demonstrate that prior knowledge largely predicts error correction and suggests that both older and younger adults can use their prior knowledge to effectively correct errors in memory.

Hypercorrection of high confidence errors: Prior testing both enhances delayed performance and blocks the return of the errors

How people correct their mistakes and sustain those corrections over time is a problem of central interest to education. It might be thought that the erroneous beliefs that people hold with high confidence would be especially difficult to correct. Interestingly, people correct these high confidence errors more easily than low confidence errors, a phenomenon known as the ‘hypercorrection effect’. Unfortunately, though, with a delay in testing there is a tendency for some of these high confidence errors to reemerge – a finding with serious consequences for education. This study investigated the effect of intervening a test immediately after corrective feedback on preventing the return of the errors. It also investigated processing differences between prevention-focused and promotion-focused people. The most educationally important new finding was that testing immediately after corrective feedback not only greatly enhanced memory for the correct answers but also blocked the return of the errors.

Is there any electrophysiological evidence for subliminal error processing?

The role of error awareness in executive control and modification of behavior is not fully understood. In line with many recent studies showing that conscious awareness is unnecessary for numerous high-level processes such as strategic adjustments and decision making, it was suggested that error detection can also take place unconsciously. The Error Negativity (Ne) component, long established as a robust error-related component that differentiates between correct responses and errors, was a fine candidate to test this notion: if an Ne is elicited also by errors which are not consciously detected, it would imply a subliminal process involved in error monitoring that does not necessarily lead to conscious awareness of the error. Indeed, for the past decade, the repeated finding of a similar Ne for errors which became aware and errors that did not achieve awareness, compared to the smaller negativity elicited by correct responses (Correct Response Negativity; CRN), has lent the Ne the prestigious status of an index of subliminal error processing. However, there were several notable exceptions to these findings. The study in the focus of this review (Shalgi and Deouell, 2012) sheds new light on both types of previous results. We found that error detection as reflected by the Ne is correlated with subjective awareness: when awareness (or more importantly lack thereof) is more strictly determined using the wagering paradigm, no Ne is elicited without awareness. This result effectively resolves the issue of why there are many conflicting findings regarding the Ne and error awareness. The average Ne amplitude appears to be influenced by individual criteria for error reporting and therefore, studies containing different mixtures of participants who are more confident of their own performance or less confident, or paradigms that either encourage or don't encourage reporting low confidence errors will show different results. Based on this evidence, it is no longer possible to unquestioningly uphold the notion that the amplitude of the Ne is unrelated to subjective awareness, and therefore, that errors are detected without conscious awareness.

The hypercorrection effect in younger and older adults

ABSTRACTThe hypercorrection effect, which refers to the finding that errors committed with high confidence are more likely to be corrected than are low confidence errors, has been replicated many times, and with both young adults and children. In the present study, we contrasted older with younger adults. Participants answered general-information questions, made confidence ratings about their answers, were given corrective feedback, and then were retested on questions that they had gotten wrong. While younger adults showed the hypercorrection effect, older adults, despite higher overall accuracy on the general-information questions and excellent basic metacognitive ability, showed a diminished hypercorrection effect. Indeed, the correspondence between their confidence in their errors and the probability of correction was not significantly greater than zero, showing, for the first time, that a particular participant population is selectively impaired on this error correction task. These results potentially offer leverage both on the mechanisms underlying the hypercorrection effect and on reasons for older adults’ memory impairments, as well as on memory functions that are spared.

A Computational Model for Spatial Working Memory Deficits in Schizophrenia

Cognitive deficits in schizophrenia have been hypothesized to be caused by altered synaptic transmission in circuits of the prefrontal cortex. 2 main hypotheses have been put forward: reduced inhibition and hypofunctional NMDA receptors. Recently, Lee et al. (2008) found that spatial working memory deficits in schizophrenic patients include a disproportionately high incidence of high-confidence error responses. Here, we have studied what synaptic dysfunction can generate this specific behavioral deficit using a computational network model of spatial working memory. We developed quantitative behavioral readout from our network simulations, which reflected the qualitative properties of underlying neural dynamics. We then analyzed the behavioral effect of the GABAergic and glutamatergic hypotheses on our network simulations. We found that reduction in inhibitory transmission in the network caused a reduction in performance through an increase of high-confidence errors, as in the experimental data. In contrast, a concerted reduction in NMDA-receptor-dependent transmission reduced performance via increased low-confidence errors. Only when NMDA receptors were specifically depleted in interneurons did the behavioral read-out of our network mimic the behavioral results for schizophrenic patients. Thus, dynamics in our model network support a role of both global inhibition reduction and hypofunctional NMDA receptors in interneurons in generating the behavioral deficits of simple spatial working memory tasks in schizophrenia.

Hypercorrection of high confidence errors in children

Three experiments investigated whether the hypercorrection effect – the finding that errors committed with high confidence are easier, rather than more difficult, to correct than are errors committed with low confidence – occurs in grade school children as it does in young adults. All three experiments showed that Grade 3–6 children hypercorrected high confidence errors and the children also claimed that they ‘knew those answers all along.’ Experiment 2 included two second-guess tasks following error commission, one in which the children attempted to choose the correct answer from six options and the other in which they tried to generate a correct second response. Neither provided evidence that children actually knew high confidence corrections all along. Experiment 3, however, showed that the children had some preferential partial knowledge insofar as they needed fewer hints to guess the correct answers to high confidence than to low confidence errors.

The hypercorrection effect persists over a week, but high-confidence errors return

People's knowledge about the world often contains misconceptions that are well-learned and firmly believed. Although such misconceptions seem hard to correct, recent research has demonstrated that errors made with higher confidence are more likely to be corrected with feedback, a finding called the hypercorrection effect. We investigated whether this effect persists over a 1-week delay. Subjects answered general-knowledge questions about science, rated their confidence in each response, and received correct answer feedback. Half of the subjects reanswered the same questions immediately, while the other half reanswered them after a 1-week delay. The hypercorrection effect occurred on both the immediate and delayed final tests, but error correction decreased on the delayed test. When subjects failed to correct an error on the delayed test, they sometimes reproduced the same error from the initial test. Interestingly, high-confidence errors were more likely than low-confidence errors to be reproduced on the delayed test. These findings help to contextualize the hypercorrection effect within the broader memory literature by showing that high-confidence errors are more likely to be corrected, but they are also more likely to be reproduced if the correct answer is forgotten.

People's hypercorrection of high-confidence errors: did they know it all along?

This study investigated the "knew it all along" explanation of the hypercorrection effect. The hypercorrection effect refers to the finding that when people are given corrective feedback, errors that are committed with high confidence are easier to correct than low-confidence errors. Experiment 1 showed that people were more likely to claim that they knew it all along when they were given the answers to high-confidence errors as compared with low-confidence errors. Experiments 2 and 3 investigated whether people really did know the correct answers before being told or whether the claim in Experiment 1 was mere hindsight bias. Experiment 2 showed that (a) participants were more likely to choose the correct answer in a 2nd guess multiple-choice test when they had expressed an error with high rather than low confidence and (b) that they were more likely to generate the correct answers to high-confidence as compared with low-confidence errors after being told they were wrong and to try again. Experiment 3 showed that (c) people were more likely to produce the correct answer when given a 2-letter cue to high- rather than low-confidence errors and that (d) when feedback was scaffolded by presenting the target letters 1 by 1, people needed fewer such letter prompts to reach the correct answers when they had committed high- rather than low-confidence errors. These results converge on the conclusion that when people said that they knew it all along, they were right. This knowledge, no doubt, contributes to why they are able to correct those high-confidence errors so easily.

Surprising feedback improves later memory

The hypercorrection effect is the finding that high-confidence errors are more likely to be corrected after feedback than are low-confidence errors (Butterfield & Metcalfe, 2001). In two experiments, we explored the idea that the hypercorrection effect results from increased attention to surprising feedback. In Experiment 1, participants were more likely to remember the appearance of the presented feedback when the feedback did not match expectations. In Experiment 2, we replicated this effect using more distinctive sources and also demonstrated the hypercorrection effect in this modified paradigm. Overall, participants better remembered both the surface features and the content of surprising feedback.

The correction of errors committed with high confidence

Most theories predict that when people indicate that they are highly confident they are producing their strongest responses. Hence, if such a high confidence response is in error it should be overwritten only with great difficulty. In contrast to this prediction, we have found that people easily correct erroneous responses to general information questions endorsed as correct with high-confidence, so long as the correct answer is given as feedback. Three potential explanations for this unexpected hypercorrection effect are summarized. The explanation that is tested here, in two experiments, is that after a person commits a high-confidence error the correct answer feedback, being surprising or unexpected, is given more attention than is accorded to the feedback to low-confidence errors. This enhanced attentional capture leads to better memory. In both experiments, a tone detection task was presented concurrently with the corrective feedback to assess the attentional capture of feedback stimuli. In both, tone detection was selectively impaired during the feedback to high confidence errors. It was also negatively related to final performance, indicating that the attention not devoted to the tone detection was effectively engaged by the corrective feedback. These data support the attentional explanation of the high-confidence hypercorrection effect.

Neural correlates of error detection and correction in a semantic retrieval task

Event-related potentials (ERPs) were used to investigate the cognitive and neural substrates of immediate and 1-week delayed error correction in a semantic retrieval task. In particular, we pursued the basis for the ‘hypercorrection’ effect, the finding that erroneous responses endorsed as correct with high confidence are more likely than low-confidence errors to be corrected at retest. Presentation of negative, but not positive feedback about the accuracy of one’s response elicited a fronto-central negativity, similar to the ERN, which was somewhat sensitive to the degree to which negative feedback violated expectation. A fronto-central positivity, similar to the novelty-P3/P3a, more generally indexed detection of a metamemory error, given that it was larger in conditions of high metamemory mismatch than in conditions of low metamemory mismatch, irrespective of absolute task accuracy. For errors, amplitude of the fronto-central positivity, but not the preceding negativity, was correlated with correction on an immediate retest. Thus, to the extent that the fronto-central positivity indexes an orienting response, this response appears to facilitate initial encoding processes, but does not play a key role in memory consolidation. In contrast, a broad, inferior-temporal negativity occurring 300–600 ms after presentation of the correct answer was sensitive to subsequent memory performance at both immediate and delayed retests, but only for answers containing familiar semantic information. This negativity may reflect processes involved in the formation of an association between the question and pre-existing semantic information.