How does error correction occur during lexical learning?

We examined two theories of the mechanisms that enable error correction via corrective feedback. One theory focuses on enhancing the encoding of corrective feedback (corrective feedback-encoding facilitation account). The other is the recursive reminding theory, which considers memory integration between an initial event with error generation and a subsequent event involving correct answer feedback. The Japanese idiom pronunciation task was used in two experiments, in which it was manipulated whether the generated errors were visually presented, as well as corrective feedback. In an immediate retest after a five-minute retention interval, participants recalled their errors in the initial test and their correct answers. In addition, error trials fell into three ordinal confidence categories (low, medium, and high). First, a typical hypercorrection was replicated in which higher-confidence errors are more likely to be corrected. However, this was not observed when errors from the initial test were not recalled in the final test, which does not align with the corrective feedback-encoding facilitation account. The second issue was whether additional experience with the generated errors would enhance the error correction. Given the recursive reminding theory, the additional experience of errors should reinforce the mutual dependence between an error and the correct answer provided by feedback, improving cued recall performance later. This prediction is supported. The present findings suggest that the recursive reminding theory can explain the benefits of generating errors when learning through corrective feedback and can also be expanded to understand the hypercorrection effect.

Response Format, Not Semantic Activation, Influences the Failed Retrieval Effect.

In educational settings, tests are mainly used to measure the extent to which learners' knowledge and skill have been acquired. However, the act of taking a test also promotes learning itself. In particular, making errors on tests (i.e., searching for erroneous information) promotes learning. This is called the "failed retrieval effect" (FRE) and has been the subject of considerable study. Previous research shows that enhanced learning does not occur if feedback correcting an error is delayed. This is attributed to the relative absence of activated information. In this study, we manipulated both the amount of information to be retrieved prior to learning and the delay time until feedback is given to investigate their effects on learning. As a result, even when multiple incorrect answers were given to increase the degree of semantic activation, learning was not promoted beyond that found with traditional procedures that rely on only one incorrect response. The timing of feedback (immediate, short-delay, long-delay) also did not impact FRE. However, the manipulation of response format for erroneous information resulted in degraded performance when responses were typed and feedback was delayed. Based on this result, we suggested that the failed retrieval effect was not affected by semantic activation at the time of retrieval but was affected by response format. Moreover, the processing necessary for typing may affect FRE under the delayed feedback condition.

Electrophysiological decomposition of attentional factors on the hypercorrection effect of false lexical representations.

False memories endorsed with higher confidence are more likely to be corrected by feedback than those endorsed with lower confidence (hypercorrection effect). Errors made with high confidence and correct responses made with low confidence are both associated with large meta-memory mismatches. Therefore, they both represent a type of unexpected event which automatically captures participant attention, such that correct information provided via feedback is well-encoded. On the other hand, a study that measured participants' perceived practical value for items suggested that voluntary allocation of attention might involve the hypercorrection effect. The present study involved a lexical learning task with 28 undergraduate student participants and measurement of automatic and voluntary attentional allocations via P3a/novelty P3 and P3b respectively, both of which are event-related potentials (ERPs). Behavioral results replicated the hypercorrection effect in a lexical learning task and showed modulation of the effect with regard to perceived practical value. In addition, ERP measurement results demonstrated that both automatic and voluntary allocations of attentional resources were independently involved in the hypercorrection phenomenon.

Does delayed corrective feedback enhance acquisition of correct information?

Many studies concerned with misinformation correction during learning report that delayed corrective feedback is superior to immediate feedback. However, the mechanism for this effect has not been confirmed. The interference-perseveration theory predicts that immediate feedback following participants' wrong responses elicits proactive interference that deteriorates acquisition of feedback information. In contrast, delayed feedback following errors leads to participants' forgetting these errors during the delay period; consequently, in the latter, interference should decline leading to superior acquisition of corrective information. However, results of these studies have been inconsistent. The present study manipulated whether initial errors were visually cued before feedback (no error-cueing, error-cueing) along with the timing of the feedback (immediate, delayed). The interference-perseveration theory predicts that when errors are not cued, delayed feedback should result in superior acquisition of correct information compared to immediate feedback. When errors are cued, proactive interference should effect a deterioration in acquisition of corrective feedback. Results confirmed neither of these predictions, thus challenging the interference-perseveration hypothesis. Moreover, additional analysis suggested that memory for errors has the ability to enhance the retention of correct answers and it does not hinder recall.

Hypercorrection of high confidence errors in lexical representations.

Memory errors associated with higher confidence are more likely to be corrected than errors made with lower confidence, a phenomenon called the hypercorrection effect. This study investigated whether the hypercorrection effect occurs with phonological information of lexical representations. In Experiment 1, 15 participants performed a Japanese Kanji word-reading task, in which the words had several possible pronunciations. In the initial task, participants were required to read aloud each word and indicate their confidence in their response; this was followed by receipt of visual feedback of the correct response. A hypercorrection effect was observed, indicating generality of this effect beyond previous observations in memories based upon semantic or episodic representations. This effect was replicated in Experiment 2, in which 40 participants performed the same task as in Experiment 1. When the participant's ratings of the practical value of the words were controlled, a partial correlation between confidence and likelihood of later correcting the initial mistaken response was reduced. This suggests that the hypercorrection effect may be partially caused by an individual's recognition of the practical value of reading the words correctly.

Error-related negativity in a visual go/no-go task: children vs. adults.

Thirteen young adults (ages 21-25) and nine children (ages 7-11) were tested on a visual go/no-go task comparing response times (RTs), error rates, and amplitude and latency changes of error-related negativity (ERN). All experimental conditions were identical for both age groups. Results are consistent with the previous flanker task research showing an increase in ERN amplitude as children age. However, the present results indicate that the peak amplitude of ERN for 9-11 year old children is larger than that of 7-8 year old children, with no difference overall between young adults and children. ERN responses elicited by the flanker task continue to develop until late adolescence. Comparative results suggest that the visual go/no-go task may be more sensitive cognitive measure than the flanker task of mistakes made by children. Differences in time pressure to respond, complexity of the task, and feedback are discussed as possible explanations of differences in the two paradigms.

Error-related negativity in children: effect of an observer.

This research assessed the effect of an observer upon error-related negativity (ERN) in 20 children (ages 7-11): 9 worked alone and 11 worked under the observation of a friend of theirs (alone/audience groups). Erroneous motor responses were recorded to Go and No-Go visual stimuli (triangles in different orientations) at three brain sites (Fz, Cz, and Pz). Results for children were consistent with past adult studies in showing a significantly larger ERN only to erroneous responses in both the alone and the audience groups. Children in the audience group produced larger ERNs than those in the alone group. Furthermore, older children (ages 9-11) produced larger ERNs than younger children (ages 7-8) in both groups. These findings were discussed in the context of an action monitoring system that regulates responses, detects errors, and entrains affective responses associated with correct and incorrect responses.